The proteasome activator subunits PA28α and PA28β have unique molecular roles in glioblastoma

The proteasome activator subunits PA28α and PA28β have unique molecular roles in glioblastoma | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article The proteasome activator subunits PA28α and PA28β have unique molecular roles in glioblastoma Samuel Weiss, Kyle Heemskerk, Ravinder Bahia, Samir Assaf, Xiaoguang Hao, and 7 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8682343/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract The proteasome is an essential protein complex that cancer cells rely on for amino acid recycling and protein degradation, making it a key therapeutic target. Its core particle can bind different activators that regulate its activity and substrate specificity, enabling significant functional diversity. However, the mechanisms by which different proteasome activators function across cell types and disease states are poorly understood. Here, we identified an enrichment of the non-constitutive activator PA28αβ in glioblastoma stem cells (GSCs) and tumours. Disruption of PA28αβ reduced the self-renewal of GSCs and improved survival in orthotopic xenograft and syngeneic models. Despite the canonical heteroheptameric composition of PA28αβ, we demonstrate that the individual subunits play dichotomous roles in glioblastoma. Analysis of proteasome activity and the immunopeptidome in response to PA28αβ perturbation revealed a unique role for the PA28β subunit in regulating immunoproteasome activity and antigen presentation. Interactome mapping identified the kinesin motor protein KIF11 as a binding partner of PA28αβ, and molecular profiling showed subunit-dependent alterations in protein trafficking pathways. Furthermore, PA28α displayed a conserved role in regulating KIF11 stability, promoting stemness in GSCs and tumour growth in a syngeneic glioma model. Collectively, our results reveal subunit-specific roles for PA28αβ with therapeutic potential in glioblastoma. Health sciences/Oncology/Cancer/Cancer stem cells Biological sciences/Cell biology/Proteolysis/Proteasome Biological sciences/Cancer/CNS cancer Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction The proteasome is a dynamic protein complex responsible for protein degradation and amino acid recycling 1 – 3 . Consequently, it regulates critical cellular functions, including the cell cycle, metabolism, protein homeostasis, and apoptosis 1 . Structurally, the proteasome consists of standard or immunoproteasome core particle subunits assembled into a ring-like conformation to which various activators can bind at either end 1 , 4 . The immunoproteasome core particle harbours alternative catalytic subunits induced by inflammatory cytokines and alters peptide production to enhance major histocompatibility complex (MHC) class I antigen presentation 1 . Proteasome activators, including the 19S regulatory particle, PA28αβ ( PSME1 & PSME2 ), PA28γ ( PSME3 ), and PA200 ( PSME4 ), modulate proteasome activity and function 1 . The core particles can degrade ubiquitinated and non-ubiquitinated proteins independently 5 , 6 . Alternatively, various activators can regulate their respective degradation targets and peptide processing 1 , 2 . The vast diversity of proteasome holo-complexes underscores the need for a comprehensive understanding of each activator’s specific role in different cellular and disease contexts 1 , 7 . Regulation of cell growth and proliferation, along with many oncogenic and tumour suppressor pathways, and overall proteostasis, makes the proteasome an attractive therapeutic target in cancer 1 , 8 . Inhibiting the proteasome pathway in cancer has traditionally focused on the proteasome core particle 8 . Targeting proteasome activator complexes offers an alternative approach to cancer therapy that could expand treatment options and improve patient survival. A few proteasome activators have been functionally investigated in specific cancers. For example, the proteasome activator PA200 has been identified as a target to enhance the response to immunotherapy in non-small cell lung cancer 7 . Furthermore, PA28γ has been shown to facilitate hepatitis C viral infection and the development of hepatocellular carcinoma 9 . In contrast, less is known about the activator PA28αβ (encoded by PSME1 & PSME2 ) in cancer. PA28αβ forms a heteroheptameric ring with four α and three β subunits 10 that bind to the proteasome core particle at either end 11 . Studies have described crystal structures of PA28αβ bound to either the standard or the immunoproteasome core particle, suggesting it has the capacity to regulate both species 11 – 13 . PA28αβ is canonically considered to alter peptide processing for MHC class I antigen presentation 14 , 15 ; however, there is conflicting evidence regarding its contribution to global peptide presentation and cytotoxic T cell responses. Knockout of the PA28β subunit in mice resulted in alterations in peptide processing, cytotoxic T cell responses, and immunoproteasome assembly 15 . Conversely, in other studies, PA28α and PA28β knockout mice displayed normal T cell responses but altered processing of only specific peptides for antigen presentation 16 , 17 . Therefore, it is unclear whether PA28αβ contributes to peptide processing for antigen presentation on its own or in conjunction with the immunoproteasome core particle, and whether this occurs in all cells. Further exploration of the role of PA28αβ in various cell types and disease contexts is essential to improve our understanding of its overall function. Glioblastoma (GBM) is the most common primary malignant brain tumour in adults, characterized by extensive heterogeneity 18 , 19 . Treatment options for GBM are limited, with a median overall survival of approximately 15 months 20 . GBM tumours contain malignant cells in many different states, including stem-like cells that harbour cancer stem cell qualities, such as the ability to initiate tumours in mice, self-renew, and undergo multilineage differentiation 18 , 19 . These cells, termed GBM stem cells (GSCs), have been shown to contribute to treatment resistance and can be isolated and cultured in neural stem cell conditions for mechanistic investigations 19 . The functional heterogeneity of the proteasome and its role in driving GBM and GSC growth and stem-like properties are not well understood. While proteasome core particle inhibitors revolutionized treatment strategies for multiple myeloma and mantle-cell lymphoma, comparable efficacy has not been demonstrated in other hematological malignancies or solid tumours 8 . The proteasome has been shown to promote GBM growth and GSC phenotypes 21 , but inhibition of the standard proteasome core particle with the brain-penetrant proteasome inhibitor marizomib did not provide a therapeutic benefit in a GBM phase III clinical trial 22 . It was also associated with an increased incidence of nervous system and psychiatric adverse events 22 . These adverse effects may be attributable to the proteasome’s vital role in normal cellular function 3 , as well as its noncanonical roles, such as a membrane-embedded core particle within neurons 6 and free 19S regulatory particles that regulate the synaptic proteome 23 . These studies underscore that proteasome subunits may have specialized functions in different cell types and tissues. Further understanding of proteasome biology across cell types and disease states is necessary to identify mechanisms with therapeutic potential. Investigating non-catalytic proteasome subunits or activators in GBM may reveal novel and targetable dependencies. Here, to find alternative proteasomal targets in GBM, we investigated proteasome activator expression in tumour tissue and GSCs. We identified PA28αβ as an enriched proteasome activator in GSCs and GBM tumours. Genetic or chemical targeting of PA28αβ decreased GSC self-renewal and extended survival in orthotopic xenograft and syngeneic models. Interestingly, we found a prominent role for the PA28β subunit in regulating immunoproteasome activity and antigen presentation in GSCs. We further uncovered that PA28αβ interacts with the kinesin motor protein KIF11, promoting its stability, primarily through the PA28α subunit. Overall, this study reveals the interaction between PA28αβ and KIF11 as a key regulator of GSCs, identifies PA28αβ as a potential therapeutic target in GBM, and expands our understanding of functional heterogeneity between the PA28αβ subunits. Results PA28αβ is upregulated in GSCs and GBM tumours To explore alternative strategies for targeting the proteasome in GBM, we analyzed proteasome gene expression utilizing single-cell RNA-seq of GSCs 24 and human NSPCs 25 . Furthermore, we compared proteasome gene expression in GBM tumours 26 with normal tissue 27 (Fig. 1 A). We integrated single-cell RNA expression data from 29 patient-derived GSCs and 4 NSPCs using Canonical Correlation Analysis and examined differences in proteasome gene expression between GSCs and NSPCs (Fig. 1 B, Supplemental Fig. 1A, Supplemental Table S1 ). Expression levels of the 20S core particle and 19S regulator particle gene families were generally elevated in GSCs relative to NSPCs (Fig. 1 B). Expression of PSME1 and PSME2 , which encode PA28αβ, was also upregulated in GSCs compared to NSPCs (Fig. 1 B; Pseudobulk Wald test; PSME1 : Padj < 1.87E-11, PSME2 : Padj < 3.6E-14). In contrast, PSME3 (PA28γ) and PSME4 (PA200) did not show significant upregulation in GSCs versus NSPCs (Fig. 1 B). Higher expression of the immunoproteasome core particle subunits PSMB8 and PSMB9 was observed in GSCs relative to NSPCs; however, PSMB10 did not show significant differential expression (Fig. 1 B; Pseudobulk Wald test; PSMB8 : Padj < 3.45E-4, PSMB9 : Padj < 1.4E-5). We next aggregated the proteasome core particle and activator complex gene families into module scores for comparison between GSCs and NSPCs (Supplemental Fig. 1B (single-cell scores)). To facilitate statistical comparison between the 29 GSCs and 4 NSPCs, we pseudo-bulked the expression module scores per sample and found that only the PA28αβ module score and the immunoproteasome module score were significantly elevated in GSCs relative to NSPCs (Fig. 1 C (pseudobulk score)). This finding is supported by a previous study that compared the proteome of GSCs with neural stem cells (NSCs), which showed enrichment of PSME1 (PA28α) and PSME2 (PA28β) in GSCs 28 . To validate these findings in tumours, we examined gene expression of proteasome family members in the Cancer Genome Atlas (TCGA)-GBM dataset (primary GBM: n = 372; recurrent GBM: n = 14; solid tissue normal: n = 5) 26 and compared it to the Genotype Tissue Expression (GTEx) human brain expression dataset (n = 300 donors; 2931 tissues) 27 (Supplemental Fig. 1C). The expression levels of PSME1 and PSME2 were significantly elevated in primary and recurrent GBM relative to normal tissue in the TCGA cohort, as well as compared to the GTEx normal brain tissue (Fig. 1 D, E). PSME3 and PSME4 were also significantly upregulated compared with GTEx normal brain but were less enriched than PSME1 and PSME2 relative to solid normal tissue (Supplemental Fig. 1C, Supplemental Table S2). As observed in the comparison between GSCs and NSPCs, PSMB8 and PSMB9 exhibited significant enrichment in GBM tumours, while PSMB10 exhibited a reduction in expression in GBM tumours, compared to normal tissue (Supplemental Fig. 1C, Supplemental Table S2). Lastly, we confirmed the protein level upregulation of PA28αβ in GBM tumours using the Clinical Proteomic Tumor Analysis Consortium (CPTAC) GBM dataset (n = 99, GBM; n = 10, normal) 29 (Fig. 1 F). Overall, these findings indicate that PSME1 and PSME2 are enriched in GSCs and GBM tumours compared to normal cells or tissues. In contrast, the immunoproteasome genes show variable expression in GSCs and GBM tumours compared to normal tissue. Given the consistent expression pattern of PA28αβ in GSCs and GBM tumours and the lack of prior studies examining its role in GBM and other malignancies, we further explored its role in GSCs. To identify functional implications of PA28αβ expression in GSCs, we examined RNA and protein expression of each subunit in our cohorts. As GSCs have been shown to exhibit transcriptional states along a Developmental-to-Injury Response axis 24 , we first investigated the correlation between PA28αβ expression and these states. The PA28αβ module score in GSCs and NSPCs was negatively correlated with the Developmental signature and positively correlated with the Injury Response signature (Fig. 1 G, H). Next, we assessed the protein expression levels of PA28α and PA28β in a cohort of 26 GSCs and observed variable expression among the samples (Fig. 1 I, Supplemental Fig. 1D). Our prior data included bulk RNA-seq for 25 of these GSCs 30 , which, upon further analysis, revealed a positive correlation between PSME1 and PSME2 RNA expression (Supplemental Fig. 1E); however, there was no correlation between PA28α and β protein expression levels in GSCs (Fig. 1 J). PA28α protein expression was correlated with PSME1 RNA expression, but PA28β protein expression was not correlated with PSME2 RNA expression (Supplemental Fig. 1F, G). These results suggest that PA28β expression may be post-transcriptionally regulated separately from PA28α in GSCs. PA28α protein expression showed a negative correlation with the Developmental signature and a positive correlation with the Injury Response signature (Supplemental Fig. 1H). Meanwhile, PA28β protein expression in GSCs did not correlate with Developmental or Injury Response signatures (Supplemental Fig. 1I). Taken together, our results indicate that PA28αβ is enriched in GSCs and GBM tumours compared to other proteasome activators and that its RNA expression correlates with the Injury Response signature in GSCs. Given that the injury response signature has been associated with early GBM development 31 , this suggests that PA28αβ may play a role in tumour initiation. However, PA28α and PA28β may have distinct protein-level regulatory mechanisms in GSCs. Genetic targeting of PA28αβ reduces GSC stemness and improves survival in vivo . As PA28αβ was upregulated in GSCs and GBM tumours, we sought to determine whether it promotes GSC features. First, we examined the expression of PA28αβ in GSCs cultured in stem cell-enriched (epidermal growth factor (EGF)/fibroblast growth factor (FGF) supplemented) and differentiation-promoting conditions (10% fetal bovine serum (FBS)). PA28αβ expression was increased in GSCs cultured in stem cell conditions (Fig. 2 A), suggesting it may be associated with stemness. Next, we genetically targeted PSME1 (PA28α) and PSME2 (PA28β) using CRISPR/Cas9 in two patient-derived GSC lines (BT67 and BT48) (Fig. 2 B, Supplemental Fig. 2A). Additionally, we used shRNA to target PSME1 (PA28α) in another GSC line (BT189) (Supplemental Fig. 2B). In all three GSCs, knockout (KO) or knockdown (KD) of PSME1 or PSME2 led to a reduction in both PA28α and PA28β protein expression (Fig. 2 B, Supplemental Fig. 2A, B). This finding aligns with previous studies utilizing PA28β KO mice or chronic myelogenous leukemia cells 15 , 32 and may be attributed to reciprocal stabilization during complex formation. However, residual expression of PA28β was detected in PSME1 KOs/KDs (Fig. 2 B, Supplemental Fig. 2B, C, D), and residual PA28α expression was detected in PSME2 KOs (Supplemental Fig. 2C, D). Considering the observation of a truncated PA28β for PSME2 guide RNA 1 in BT67 (Fig. 2 B, Supplemental Fig. 2C), most subsequent experiments were conducted using PSME1 guide RNA 1 and PSME2 guide RNA 2. First, we examined the effects of PA28αβ KO/KD on GSC growth, which showed minimal impact on in vitro growth patterns and doubling time (Fig. 2 C, D, Supplemental Fig. 2E, F). Additionally, cell cycle analysis of BT67 PA28αβ KOs demonstrated a modest yet significant reduction in S-phase cells compared to AAVS1 controls (Fig. 2 E). We further evaluated sphere formation as a proxy for self-renewal, finding that PA28αβ KO/KD reduced sphere-forming frequency in the GSCs (Fig. 2 F). Thus, targeting PA28αβ abrogates the self-renewing potential of GSCs. Next, to test whether these results translate to promotion of tumour growth in vivo , we orthotopically xenografted BT67 AAVS1 controls, PSME1 (PA28α) KOs, and PSME2 (PA28β) KOs into SCID mice. Mice xenografted with PSME1 KO and PSME2 KO BT67 GSCs exhibited significantly improved survival relative to controls (Fig. 2 G). We observed a similar result in BT189 PSME1 KDs compared to Scramble controls (Supplemental Fig. 2G). As the canonical function of PA28αβ is to regulate peptide processing for antigen presentation, we also examined its role in a syngeneic GBM model. shRNA-mediated KD of Psme1 or Psme2 in GL261 resulted in a reduction of both PA28α and PA28β protein levels, as was observed in human GSCs (Fig. 2 H). We engrafted GL261 Scramble, Psme1 shRNA1, and Psme2 shRNA2 into immunocompetent mice and, surprisingly, only the Psme1 KD improved survival in this model (Fig. 2 I). This result suggests that PA28α and PA28β may have distinct roles in immunoregulation, which could be explained by the observation of residual expression of non-targeted subunits. This may also be linked to the absence of protein-level correlation between PA28α and PA28β in GSCs (Fig. 1 J). It has previously been reported that PA28α and PA28β can form homoheptamers, albeit with lower stability, which could have individual functions 10 . Together, these results demonstrate that genetic targeting of PA28αβ in GSCs decreases sphere-forming capacity and improves survival in orthotopic xenograft models, whereas the individual subunits may have different effects on immunoregulation. Chemical targeting of PA28αβ abrogates stemness in GSCs. To validate our findings of reduced self-renewal in PA28αβ-targeted GSCs, we used a previously developed chemical probe that disrupts the PA28αβ complex via engagement of cysteine-22 on PA28α 32 (Fig. 3 A). We overexpressed (OE) FLAG-tagged PA28α and a FLAG-tagged PA28α with cysteine 22 mutated to alanine (C22A) as a control to ensure on-target inhibition in BT67 (Fig. 3 B) and BT48 (Supplemental Fig. 3A). We treated FLAG-PA28α-OE and FLAG-PA28α-C22A-OE GSCs with MY45A (enantiomer negative probe) or MY45B (targeting probe) and observed a reduction in PA28α and PA28β levels in MY45B-treated FLAG-PA28α-OE but not FLAG-PA28α-C22A-OE (Fig. 3 C, Supplemental Fig. 3B). Next, we examined the effect of MY45B on sphere formation. Treatment of FLAG-PA28α-OE GSCs with MY45B led to a decrease in sphere-forming frequency compared to MY45A (Fig. 3 D, Supplemental Fig. 3C). The effect was rescued in FLAG-PA28α-C22A-OE GSCs (Fig. 3 D, Supplemental Fig. 3C), indicating that specific disruption of the PA28αβ complex via MY45B engagement of C22 on PA28α reduces sphere formation in GSCs. As was observed with genetic targeting of PA28αβ, treatment of BT67 with MY45B did not affect cell viability (Fig. 3 E). We also examined the effect of MY45B on human fetal-derived NSCs (HF-NSCs) and found that MY45B reduced sphere formation compared to MY45A (Fig. 3 F), suggesting the mechanism may be stem-cell-associated. Treatment of human fetal NSC-derived astrocytes (HFA) with MY45B did not alter cell viability (Fig. 3 G), further supporting this hypothesis. Taken together, these results suggest that targeting PA28αβ in GSCs and NSCs reduces self-renewal. Differential immunoproteasome activity, antigen presentation, and tumour microenvironment regulation by PA28α and PA28β in GBM models. Next, we sought to determine whether PA28αβ-mediated regulation of stemness and tumour growth was attributable to alterations in standard or immunoproteasome activity. We used fluorogenic activity probes specific to each proteasome core particle to measure their respective activity 33 , 34 (Fig. 4 A). The probes have specificity for the chymotrypsin-like activity associated with the β5 standard and β5i immunoproteasome core particle subunits. In two of three GSC lines, genetic targeting of PA28αβ increased standard proteasome activity (Fig. 4 B, Supplemental Fig.A, C). In contrast, immunoproteasome activity was decreased in PSME2 KOs but not PSME1 KOs in both BT67 and BT48 (Fig. 4 C, Supplemental Fig. 4B). In BT189, PSME1 KD increased immunoproteasome activity (Supplemental Fig. 4D). The alterations in proteasome activity did not correlate with protein expression changes of the catalytic core subunits (Supplemental Fig. 4E-G). These findings highlight unique, GSC line-dependent alterations in proteasome activity upon targeting of PA28α or PA28β. Notably, PSME2 KO resulted in a conserved decrease in immunoproteasome activity, suggesting a specialized role for PA28β in immunoproteasome regulation. These results also indicate that the reduction in sphere-forming frequency observed in GSCs did not consistently correlate with a specific change in proteasome activity, implying the possibility of noncanonical functions. As the immunoproteasome core particle, along with PA28αβ, is known to regulate peptide processing for antigen presentation, we profiled the immunopeptidome in PSME1/2 KO BT67s (Fig. 4 D). First, we examined the total and cell surface expression of HLA class and did not observe any changes in PSME1 KO and PSME2 KOs (Supplemental Fig. 5A-C). These data suggest that PSME1 KO or PSME2 KO does not alter HLA class I expression or localization to the cell surface in BT67. For immunopeptidomic profiling, we captured HLA class I ABC antigens using a previously established protocol 35 on GSCs (Supplemental Fig. 5D). Analysis of the identified peptides meeting both identification and MHC binding prediction thresholds revealed similar profiles between AAVS1 controls and PSME1 KOs, with a reduced number of peptides identified in PSME2 KOs (Fig. 4 E, Supplemental Fig. 5E, F, Supplemental Table S3). Approximately 150 unique peptides were identified in AAVS1 and PSME1 KOs compared with fewer than 25 in PSME2 KOs (Fig. 4 F). Among the identified peptides, we examined the presentation and RNA expression of a peptide that mapped to melanoma-associated antigen family members, which are known tumour antigens 36 . The peptide KEIDKEEHL, mapping to MAGED1 and MAGED4, was detected exclusively in the immunopeptidomes of AAVS1 and PSME1 KO BT67, despite similar RNA expression across conditions (Fig. 4 G). We also examined a peptide that mapped to the tropomyosin gene family; another peptide from this family has been identified in GBM tumour tissue immunopeptidomics 37 . The peptide KEAETRAEF was detected in all AAVS1 and PSME1 KO samples, but only in one PSME2 KO sample, despite similar RNA expression (Supplemental Fig. 5G). These findings indicate that the resultant decrease in immunoproteasome activity due to PSME2 KO abrogates peptide processing for HLA class I antigen presentation in GSCs. The absence of antigen presentation changes in PSME1 KOs, again, could be due to residual expression of PA28β and/or the ability of either subunit to form homoheptamers 10 . To determine whether alterations in peptide processing for antigen presentation influence the immune microenvironment in tumours, we returned to the GL261 syngeneic GBM model (Fig. 2 H, I). We profiled Scramble, Psme1 KD, and Psme2 KD tumours using single-cell RNA-seq, which identified the presence of diverse cell types within the tumours (Fig. 4 H, I, Supplemental Fig. 5H-J). We identified cell clusters corresponding to different tumour cell states, immune cell types, and neural cell types (Fig. 4 I), which were represented across Scramble and KD tumours (Fig. 4 J). Differential gene expression analysis across these cell types confirmed downregulation of Psme1 and Psme2 in tumour cell clusters in each respective KD condition (Fig. 4 K, Supplemental Table S4). Analysis of myeloid cell populations showed a downregulation of M2-like markers, including Arg1 , Mrc1 , and Ccl24 38 , predominantly in Psme1 KD tumours (Fig. 4 L). Among the identified T cell populations, Socs1 , a regulator of inflammatory signalling 39 , was downregulated in both Psme1 and Psme2 KDs (Fig. 4 M). Conversely, Il31ra , a marker for an activated T cell expression signature in GBM 40 , was upregulated only in T cells from Psme1 KD tumours (Fig. 4 M). Taken together, these results suggest that Psme1 KD impairs tumour growth in GL261 by reducing immunosuppression in the microenvironment, while Psme2 KD does not have this effect. Decreased immunoproteasome activity and peptide processing for antigen presentation observed in PSME2 KO GSCs may be responsible for the lack of immune activation in Psme2 KD GL261 tumours. The improvement in survival and immune activation in Psme1 KD GL261, which follows the pattern of decreased stemness observed upon targeting PA28αβ, could be linked to a noncanonical function of this complex in GSCs. PA28αβ interactome and regulation of protein trafficking in GSCs. As a consistent alteration in proteasome activity was not observed upon targeting PA28αβ, we hypothesized that it may function noncanonically in GSCs. Interestingly, free 19S regulatory particles were recently found to bind and regulate the stability of AMPA receptors in neurons 23 . To determine if PA28αβ bound proteins other than the proteasome core particle in GSCs, we performed interactome analysis using FLAG-tagged PA28α and PA28β compared to a mCherry over-expression control pull-down (Fig. 5 A). First, we tested crosslinking followed by FLAG and mCherry pull-down, which showed robust enrichment of FLAG-PA28α, FLAG-PA28β, and mCherry (Supplemental Fig. 6A). Next, we performed crosslinking pull-down coupled with mass-spectrometry and label-free quantification (LFQ) on FLAG-PA28α, FLAG-PA28β, and mCherry BT67 (Supplemental Fig. 6B, Supplemental Table S5). Notably, proteins identified in the FLAG-PA28α and FLAG-PA28β pull-down, but not the mCherry pull-down were involved in protein trafficking (e.g., KIF11, TUBA1B, TUBA1C, DCTN1, Coatamer complex I proteins), ribosomal components (e.g., RPS29, RPS12), proteasome proteins ( PSME1 , PSME2 , PSMA2), and RNA splicing regulators (e.g., PRMT5, WDR77), among other protein complexes (Fig. 5 B, C, Supplemental Fig. 6C). Two of the top enriched proteins, PRMT5 and KIF11, have been previously identified as regulators of stemness in GSCs through their roles in RNA splicing regulation 41 and invasive/proliferative properties 42 , respectively. Thus, PA28αβ may regulate GSC stemness by interacting with these key mediators of GSC function. To determine how targeting PA28αβ disrupted the phenotype of GSCs, we transcriptionally profiled PA28αβ KO/KD GSCs using bulk-RNA-seq (Fig. 5 D). For BT67, PSME1 KOs and PSME2 KOs clustered closely together and exhibited very few differentially expressed genes (Supplemental Fig. 6D, E), suggesting they have a similar transcriptional state, despite the alterations in immunoproteasome activity and peptide processing we observed (Fig. 3 ). These results also explain the similar alteration in sphere formation and survival observed for orthotopic xenografts of BT67 PSME1 KOs and PSME2 KOs (Fig. 2 F, G). Gene set enrichment analysis of differentially expressed genes between PA28αβ-targeted GSCs and controls revealed significant pathway alterations, with the Endosomal/Vacuolar and peptide processing pathways exhibiting the highest enrichment upon PA28αβ targeting (Fig. 5 E-G, Supplemental Fig. 6F, Supplemental Table S6). Neuronal system and transfer RNA processing were identified as the most down-regulated pathways in PA28αβ-targeted GSCs (Fig. 5 F, Supplemental Fig. 6F, Supplemental Table S6). Several of the altered pathways are related to the identified interactors of PA28αβ (Fig. 5 C). For instance, KIF11 has been shown to regulate the neuronal microtubule array 43 , which could affect genes in the neuronal system. KIF11 also regulates trans-Golgi vesicle trafficking 44 , potentially contributing to alterations in the Endosomal/Vacuolar pathway. Moreover, PRMT5 is known to regulate RNA expression and processing 41 , 45 , and to localize to and modulate the trans-Golgi network 46 . We further performed immunofluorescence for PA28α in HF-NSCs and GSCs, demonstrating primary localization to endomembrane systems adjacent to the nucleus (Fig. 5 H; left and middle panels). To investigate if these membrane systems overlapped with the endoplasmic reticulum and the Golgi apparatus, we stained GSCs with concanavalin A, a lectin that binds to glycosylated proteins 47 . Indeed, concanavalin A staining was predominantly perinuclear and exhibited a similar distribution to that of PA28α (Fig. 5 H; right panel). PA28αβ’s interaction with essential mediators of GSC stemness, including PRMT5 and KIF11, likely contributes to the observed alterations in transcriptional state and promotion of self-renewal and tumour growth. To determine how targeting of PA28αβ alters the proteome of GSCs, we analyzed PA28αβ KO BT67s using mass-spectrometry-based LFQ proteomics (Fig. 5 I, Supplemental Fig. 7A-C, Supplemental Table S7). Quantification of PA28α and PA28β levels in their respective KOs showed significantly lower PA28α levels in PSME1 KO compared to AAVS1 controls and PSME2 KOs (Supplemental Fig. 7D). PA28β levels were not significantly different in PSME1 and PSME2 KO cultures, however PA28β was undetectable in all PSME2 KO replicates (Supplemental Fig. 7D). These results corroborate our findings of residual expression of PA28α and PA28β when genetically targeting the other subunit in GSCs. Differential protein expression analysis showed unique alterations in PSME1 KO and PSME2 KO GSCs compared to controls (Fig. 5 J, K, Supplemental Fig. 7E). Pathway enrichment analysis of altered proteins revealed changes in metabolism, extracellular matrix organization, and protein/small molecule trafficking in PSME1 KOs, whereas PSME2 KOs exhibited changes in Golgi-ER transport (Fig. 5 L, Supplemental Fig. 7F-J). PSME1 KOs compared to PSME2 KOs showed alterations related to protein trafficking, Golgi-related pathways, and metabolism (Fig. 7K). Taken together, these results suggest that targeting PA28αβ in GSCs modulates protein trafficking machinery among other pathways at both the RNA and protein level and that targeting PA28α or PA28β has unique effects on GSCs at the protein level. PA28αβ interacts with and stabilizes KIF11 in GSCs. As PRMT5 and KIF11 were identified as top-enriched proteins in our PA28αβ interactome analysis and have been previously associated with GSC stemness, we aimed to understand the function of these interactions. Initially, we used co-immunoprecipitation to validate the interaction of FLAG-tagged PA28α and PA28β with PRMT5 and KIF11 in BT48 (Fig. 6 A, Supplemental Fig. 8A). Subsequently, we used a proximity ligation assay (PLA) to assess the endogenous interaction of KIF11 and PRMT5 with PA28α in situ. PLA foci for KIF11 and PRMT5 were observed in BT189, BT67, and BT48 (Fig. 6 B), with no signals detected in IgG and no antibody controls (Supplemental Fig. 8B), thereby affirming the interaction of KIF11 and PRMT5 with PA28α in GSCs. Additionally, PLA foci for KIF11/PRMT5 and PA28α were also observed in induced pluripotent stem cell (iPSC) derived NSCs (Fig. 6 C). Quantification of PLA foci in GSCS and NSCs revealed significantly more interactions between KIF11and PA28α per nucleus in BT189 and BT67 compared to BT48 and NSCs (Fig. 6 D). Only BT67 harboured an increased number of PLA foci in comparison to the other GSCs and iPSC-NSCs for PRMT5 and PA28α (Fig. 6 E), suggesting the interaction with KIF11 may be more conserved and therefore, more likely to regulate stemness. BT48 exhibited lower baseline expression of KIF11 and PRMT5 compared to the other GSCs, indicating the relative level of interaction depends on overall expression (Supplemental Fig. 8C). Thus, PA28αβ interacts with KIF11 and PRMT5 in GSCs and iPSC-NSCs, suggesting a potential regulatory mechanism for stemness. To determine the functional implications of PA28αβ’s interaction with KIF11 and PRMT5, we combined genetic targeting of PA28αβ with chemical inhibition of each respective protein in GSCs and assessed cell viability along with sphere-forming frequency. In BT67 PSME1 and PSME2 KOs, and BT189 PSME1 KDs, no consistent alteration in viability responses to ispinesib, an inhibitor of KIF11, or GSK591, an inhibitory probe for PRMT5, was observed (Supplemental Fig. 8D-G). These findings concur with our previous observation of the absence of growth alterations in PA28αβ genetically targeted GSCs (Fig. 2 ). However, sub-IC50 doses of ispinesib and GSK591 further reduced sphere-forming frequency in BT67 PSME1 and PSME2 KOs (Fig. 6 F). In BT189 PSME1 KDs, ispinesib further decreased sphere formation, whereas GSK591 abrogated the difference between scramble and PSME1 KD (Supplemental Fig. 8H). Given the inconsistent effects of PRMT5 inhibition in different GSCs, we examined the downstream phenotypes of PRMT5 in our PA28αβ genetically targeted GSCs. As previously noted, PRMT5 is known to contribute to alternative splicing in GSCs 41 ; hence, we performed alternative splicing analysis on RNA-seq from BT67 and BT189 PSME1 /2 KO/KDs (Fig. 5 D, Supplemental Fig. 9). However, no significantly aberrant splicing events were detected between control and PSME1 /2 KO/KD GSCs (Supplemental Fig. 9). It remains a possibility that PRMT5 may regulate PA28αβ in GSCs, instead of PA28αβ regulating PRMT5. We focused our further mechanistic studies on KIF11, considering the consistent response observed in the two examined GSCs. First, we examined the basal expression levels of KIF11 in PA28αβ genetically targeted GSCs and did not observe a significant difference relative to controls (Supplemental Fig. 8I, J). As free 19S proteasome regulator particles have been shown to stabilize AMPA receptors at synapses 23 , we investigated the impact of PA28αβ genetic targeting on KIF11 stability. We used the cycloheximide chase assay 48 to examine the short-term stability of KIF11 following translation inhibition. PSME1 KO/KD and PSME2 KO resulted in decreased levels of KIF11 after 6 hours of cycloheximide treatment (Fig. 6 G, H, Supplemental Fig. 8K, L), indicating a decrease in KIF11 stability as a consequence of PA28αβ reduction. Next, to see whether we could rescue the effects of the reduction in KIF11 stability, we overexpressed KIF11 in control and PA28αβ genetically targeted GSCs (Fig. 6 I, Supplemental Fig. 8M-O). In BT67, exogenous KIF11 partially rescued the decreased sphere formation induced by PSME1 / 2 KO (Fig. 6 J). In BT189, KIF11 overexpression fully rescued the reduction in sphere formation as a result of PSME1 KD (Supplemental Fig. 8P). Taken together, these results suggest that PA28αβ binds to and stabilizes KIF11, thereby promoting stemness in GSCs. Interestingly, when we examined KIF11 expression in the syngeneic GBM model GL261 Psme1 and Psme2 KDs, we found that KIF11 protein expression was reduced only in the Psme1 KD condition (Fig. 6 K, L). This was not dependent on a reduction in RNA levels, as Psme1 KD did not reduce KIF11 RNA expression in GL261 (Fig. 6 M). These results suggest that targeting PA28α alone alters KIF11 stability in GL261. Thus, PA28α primarily stabilizes KIF11 in GSCs and a mouse model of GBM, thereby promoting stemness and tumour growth. Discussion In this study, we uncover a noncanonical role for PA28αβ in promoting GSC stemness through its interaction with and stabilization of KIF11. We explored proteasome gene expression in glioblastoma (GBM) and GSCs, identifying PA28αβ as the most enriched of the activator subunits. Targeting PA28αβ reduced GSC self-renewal and improved survival in orthotopic xenograft models. We identified a distinct role for the PA28β subunit in immunoproteasome regulation and peptide processing for HLA class I presentation in GSCs. Targeting of PA28αβ in a syngeneic GBM model highlighted a more conserved role for the PA28α subunit in regulating KIF11 stability and tumour growth. Overexpression of KIF11 partially rescued the loss of self-renewal in PSME1/2 (PA28α and PA28β) KO/KD GSCs, suggesting a molecular interplay that promotes GSC function. Here, we provide evidence that PA28αβ is a potential therapeutic target in GSCs and GBM, and that the α and β subunits can function heterogeneously. Our findings provide several advances in our understanding of PA28αβ and its role within the proteasome pathway in GBM. First, we observed that CRISPR/Cas9-mediated KO or shRNA-mediated KD of PSME1 or PSME2 decreased the protein levels of both PA28α and β, which was confirmed in multiple cell lines, including the mouse GBM model, GL261. This result has been observed in other models and mouse knockout studies 15 , 32 . It appears that PA28α and β mutually stabilize each other at the protein level, as similar effects were not detected at the RNA expression level, which was conserved in human and mouse cells. Conversely, we observed unique phenotypic changes in PSME1 or PSME2 KO/KDs; for example, decreased immunoproteasome activity was observed only in PSME2 KO GSCs, and in BT67, immmunopeptidomic analysis showed that only PA28β disruption appeared to affect peptide processing for MHC class I antigen presentation. Moreover, in GL261 tumours, PA28α individually promoted tumour growth and KIF11 stability. These findings may be explained by the observed residual expression of the opposite subunit in PSME1 or PSME2 KO/KDs and the ability of either subunit to form homoheptamers 10 . Indeed, it is peculiar that the protein levels of PA28α and PA28β did not correlate with each other in GSCs and showed distinct correlations with GSC transcriptional states. It is also possible that PA28α and PA28β, or the complex, function independently of the proteasome core particle. In future studies, it will be essential to consider that PA28α and PA28β may have distinct functions in GBM and in other cellular or disease contexts. Secondly, we provide new evidence regarding the role of PA28αβ in antigen presentation and immunoregulation in GBM. We profiled the immunopeptidome of PSME1 and PSME2 KO GSCs, which identified PA28β as the primary regulator of antigen presentation. This result correlated with decreased immunoproteasome activity only in PSME2 KOs. One limitation of our investigation was the reduced depth of peptide identification, possibly due to low HLA class I expression or limitations with our protocol. Nonetheless, there was a dramatic difference in the number of peptides identified in the controls and PSME1 KOs compared to PSME2 KOs. Coupled with our findings of decreased immunoproteasome activity in PSME2 KOs, these results implicate PA28β as the unique regulator of antigen processing in GSCs. With recent advances in cancer immunotherapies, these findings could have implications for mechanisms of treatment resistance to cancer vaccines, checkpoint blockade, or cell-based therapies. Thirdly, we found that targeting PA28αβ in GSCs had phenotypic consequences for self-renewal. We validated the resultant alteration in self-renewal using a chemical probe targeting the PA28αβ complex. The initial aim of our study was to find an alternative way to target the proteasome pathway in GBM; however, we found that higher PA28αβ expression in GSCs and GBM tumours did not necessarily dictate function. We observed a similar reduction of self-renewal in HF-NSCs upon chemical targeting of PA28αβ. Likewise, KIF11 and PA28αβ also interacted in iPSC-NSCs. It remains uncertain whether this associated phenotype is restricted to stem cells and whether differentiated neural cells retain essential PA28αβ functionality. Collectively, our results shed light on an important role for PA28αβ in GBM, but further investigation in multiple cell types is required to fully understand its function. Lastly, the functional role of PA28αβ in GSCs depended, at least in part, on interaction with and stabilization of KIF11, which may have implications for cancer therapy. Although KIF11 inhibitors demonstrated limited efficacy in clinical trials for solid cancers and melanoma 49 , 50 , there remains interest in targeting KIF11 in other tumours, including GBM 51 , and in developing new pharmacological agents 52 . Our results suggest that simultaneous targeting of PA28αβ and KIF11 represents a novel therapeutic approach for GBM. Consequently, developing protein-protein interaction inhibitors to destabilize KIF11 in co-expressing cells may be a promising strategy. The stabilization of KIF11 likely contributes to the downstream endomembrane system pathway alterations, considering its previously identified roles as a mediator of vesicle transport from the Golgi apparatus 44 and as a regulator of the microtubule array in neurons 43 . Several other interactors of PA28αβ were also identified in GSCs. Here, we focused on KIF11 due to its functional relevance; however, other interactors may be important in different contexts, such as interactions unique to the PA28α or β subunits, including RAB10 for PA28α and MYO5B for PA28β. Alternative interactors that influence PA28αβ function could be highly relevant in other cancers or cell types. In summary, we identified PA28αβ as a potential therapeutic target in GSCs that mediates stemness through its interaction with and stabilization of KIF11. Given the established relevance of both the proteasome pathway and KIF11 in cancer, this interaction warrants consideration for future therapeutic interventions. Moreover, we uncovered functional heterogeneity between the PA28α and PA28β subunits within GSCs. This suggests a paradigm in which PA28α and PA28β may possess alternative functions that operate independently of their canonical heteroheptameric form. Methods Ethics statement Experimental procedures were performed in accordance with the University of Calgary Ethics Review Board, Health Research Ethics Board of Alberta (HREBA), and the Animal Care Committee of the University of Calgary. Cell Culture Patient-derived GSC lines were established from tumour specimens obtained during surgical resection with written informed consent from patients and approval from the University of Calgary Ethics Review Board, HREBA (HREBA-CC-160762), Calgary, AB, Canada. Metadata, including sex and age, is provided in Supplemental Table S8. GSCs were cultured in stem cell enrichment media supplemented with EGF (20 ng/mL; STEMCELL Tech.), bFGF (20 ng/mL; STEMCELL Tech.), and heparin sulfate (2 µg/mL; STEMCELL Tech.). For stem cell differentiation culture conditions, EGF, bFGF, and heparin sulfate were omitted, and the media was supplemented with 10% fetal bovine serum (FBS). GSCs were authenticated to original parental tumours by short tandem repeat profiling (Calgary Laboratory Services and Department of Pathology and Laboratory Medicine, University of Calgary). Human fetal neural stem cells (HF-NSCs) were derived from normal human brain tissues obtained from 12- to 18-week-old fetuses from therapeutic abortions according to ethical guidelines, including written parental consent, approved by the institutional Review Board of the University of Calgary (REB14-1789). HF-NSCs were generated as previously described 53 . Briefly, fetal brain tissues were minced and treated with DNase (Roche) and trypsin (Invitrogen). The dissociated cells were filtered through a cell strainer (Millipore Sigma) and seeded in stem cell enrichment media supplemented with EGF, FGF, and heparin sulfate. The generated neurospheres were passaged, expanded, and cryopreserved for later use. HF-NSCs were differentiated into human fetal astrocytes (HFAs) in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% FBS and 2% penicillin-streptomycin. Human induced pluripotent stem cells (iPSCs) were obtained from ATCC (ATCC-ACS-1020) and cultured in mTeSR™1 media (STEMCELL Tech.) according to the manufacturer’s protocol. Neural stem cells were derived from iPSCs (iPSC-NSCs) by dual SMAD inhibition using the STEMdiff™ SMADi Neural Induction Kit (STEMCELL Tech.) following the manufacturer’s monolayer protocol and subsequently validated for marker expression. HEK293T/17 cells used for lentiviral preparation were obtained from ATCC (ATCC 293T/17) and cultured in DMEM with 10% FBS and 2% penicillin-streptomycin. GL261 mouse glioma cells were a gift from Dr. Stephen Robbins and cultured in DMEM with 10% FBS and 2% penicillin-streptomycin. All cells were routinely confirmed negative for mycoplasma contamination using the Universal Mycoplasma Detection Kit (ATCC) as per the manufacturer’s protocol. Analysis of human GSC and NSPC single-cell RNA-seq data Single-cell RNA-seq datasets encompassing 29 human patient-derived GSC cultures 24 and 4 human NSPC isolates 25 were analyzed using Seurat (v.5) 54 . Dataset integration was performed according to the anchor-based CCA integration method. Differential expression was performed on pseudo-bulked samples between GSCs and NSPCs using the DESeq2 model (v.1.42) 55 . Module scores for proteasome subunit gene sets were added to pseudo-bulked data for statistical comparison and to single-cell data for visualization in single-cell plots. Analysis of TCGA-GBM tumour RNA-seq and CPTAC GBM proteomics TCGA-GBM bulk RNA-seq 26 STAR Counts were accessed using the TCGAbiolinks (v.2.30) R package 56 . GTEx human brain tissue bulk RNA-seq 27 counts were accessed using the Recount3 (v.1.12) R package 57 . Gene expression counts were normalized, and differential expression analysis was performed using DESeq2 (v.1.42). CPTAC GBM tumour proteomics 29 data were accessed using the Python CPTAC package (v.1.5.14) 58 . PSME1 and PSME2 protein expression were compared between GBM tissue and GTEx normal tissue proteomics, as performed in the GBM CPTAC study. Immunoblotting Approximately 1–2 million cells were lysed in RIPA buffer with mild sonication. Isolated protein was quantified using the Bradford assay. 20 µg of protein per well was resolved by SDS-PAGE after denaturation (8 min; 95°C) in 1X Laemmli buffer and transferred to nitrocellulose membranes. Blots were probed overnight at 4°C with antibodies specific to the proteins examined, with β-tubulin or β-actin as loading controls (a complete list of antibodies used in this study is provided in Supplemental Table S9). Blots were washed and incubated with an HRP-conjugated secondary antibody, then imaged using ECL Select and an Amersham Imager 600 (General Electric). Quantification was performed using ImageJ 59 , measuring the mean grey value for bands and normalizing to the loading control. Analysis of in-house GSC bulk RNA-seq Our previously published bulk RNA-seq data on GSCs were used to assess correlations between PSME1 and PSME2 RNA and protein expression 30 . Briefly, RNA was extracted using the AllPrep DNA/RNA Universal kit (QIAGEN). library prep and sequencing were performed by the University of Calgary, Centre for Health Genomics and Informatics (CHGI). Reads were aligned to hg38, and DESeq2 was used for batch correction and count normalization 55 . Developmental and Injury Response signatures were derived from the top 100 genes in each signature 24 and computed using ssGSEA 60 . CRISPR/Cas9 KOs and shRNA KDs in GSCs CRISPR/Cas9 KOs and shRNA KD GSCs were generated as previously described 61 . Briefly, GSCs were transduced with a lentiviral Cas9-expressing vector and selected with blasticidin. Subsequently, GSCs were transduced with a gRNA expression vector targeting PSME1, PSME2 , or the AAVS1 safe harbour site as a cut control and subsequently selected with puromycin (gRNA sequences provided in Supplemental Table S10). For human shRNA-mediated knockdown, shRNA expression constructs targeting PSME1 and a scramble control (GeneCopoeia) were used for lentiviral transduction in GSCs as previously described 61 . In GL261, shRNA expression vectors targeting Psme1, Psme2 , and a scramble control (GeneCopoeia) were used for lentiviral transduction, following the same protocol as for human GSCs. KO and KD efficiency was examined by Western blotting. Cell growth assays, cell cycle analysis, and limiting dilution assays To monitor cell growth, GSCs were seeded in a 96-well plate, and cell viability was assessed daily using alamarBlue™ (ThermoFisher) according to the manufacturer’s protocol. An exponential growth model was fit to compare conditions. For cell cycle analysis, 2.5 x 10 5 cells were incubated with EdU (10 µM) for 2 hours at 37°C. Cells were fixed and stained using the Click-iT™ Plus EdU Alexa Fluor™ 647 Flow Cytometry Assay kit (ThermoFisher) and counterstained with propidium iodide using FxCycle™ PI/RNase Staining solution (Invitrogen). Analysis was performed using a CytoFLEX LX flow cytometer (Beckman Coulter) at the University of Calgary flow cytometry core facility. Data were analyzed using FACSDiva software version 6.1.3 (BD Biosciences). For limiting dilution assays, 512 cells were seeded in 6 wells of a 96-well plate and serially diluted to 256, 128, 64, 32, 16, 8, 4, 2, and 1 cell per well. A well with a sphere exceeding 100 µm in diameter was considered positive at different time points depending on the GSC or NSC line. Plates were scored as previously described using the ELDA web tool 62 , and experiments were repeated 2–3 times with similar results. Orthotopic xenografts and GL261 syngeneic tumour model All intracranial engraftments were performed as previously described 63 under our animal ethics protocol (AC21-0162) approved by the Animal Care Committee of the University of Calgary. For orthotopic xenografts, human GSCs were xenografted into female CB-17 SCID mice (6–8 weeks old). For the GL261 syngeneic brain tumour model, cells were implanted into female C57BL/6 mice (6–8 weeks old). Mice were housed in a Biohazard level 2 facility at 25°C, 45–55% humidity, and a 6 am to 8 pm light cycle. No sex-based analysis was performed. Mice were anesthetized and stereotactically injected with 100,000 GSCs or GL261 per condition into the right cerebral hemisphere at a depth of 3.5 mm. The experimental and humane endpoints were reached when animals exhibited any of the following signs: ataxia, > 15% weight loss, hunching, kyphosis, paresis, lethargy, poor oral intake, or domed heads. Mice were euthanized with a lethal dose of ketamine/xylazine (Ketamine 300–360 mg/kg & xylazine 30–40 mg/kg) followed by cervical dislocation. Exogenous protein expression in GSCs cDNAs encoding human PSME1 (BC000352), PSME2 (BC0004368) and KIF11 (BC136474) were acquired from the mammalian gene collection (Horizon Discovery). Site-directed mutagenesis was used to generate the PSME1C22A mutant cDNA. PSME1 , PSME1C22A , and PSME2 cDNAs were subcloned into pCDH-CMV-mCherry-EF1-Hygro (a gift from Oskar Laur (Addgene plasmid # 129440)), together with the sequences for a FLAG epitope tag and T2A-eGFP by excising the mCherry using the NheI and NotI restriction sites followed by NEBuilder HiFi assembly cloning (New England Biolabs) to generate pCDH-CMV-(PSME1, PSME1(C22A), PSME2)-FLAG-T2A-eGFP-EF1-Hygro. KIF11 cDNA was also subcloned into the same backbone to generate pCDH-CMV-KIF11-T2A-eGFP-EF1-Hygro. All lentivector constructs generated were packaged into lentiviral vector particles by the HBI Molecular Core Facility using a previously described protocol 64 . Briefly, 293FT cells (Thermofisher) were grown on 5 x 15 cm cell culture dishes to ~ 90% confluency and co-transfected with the relevant transfer vector, psPAX2 and pMD2.G (psPAX2 and pMD2.G were gifts from Didier Trono (Addgene plasmid # 12259, #12260). Between 48–72 hours post-transfection, the viral vector-containing media was harvested, clarified by centrifugation at 500 x g for 5 min and then 0.45 um filtered. The clarified media was underlaid with 10% sucrose in 1XPBS and centrifuged at 12000 x g for 4 hours. The lentiviral vector pellet was resuspended in ~ 200 µL sterile 1 XPBS and frozen at -80°C in 20 µL aliquots. The lentivirus titer was determined by real-time qPCR using the kit from Applied Biological Materials, and the functional titer was qualitatively confirmed by transducing HEK293 cells on 12-well plates with serial dilutions of lentivirus and monitoring eGFP fluorescence after 48 hours. GSCs were transduced with the cDNA-expressing vectors or the original mCherry construct and selected with hygromycin B (Roche). PA28αβ chemical probe MY45-B was previously published and validated as targeting cysteine 22 on PA28α 32 . For protein expression analysis in response to chemical treatment, GSCs were treated with MY45-A (0.5 µM – enantiomer control probe), MY45-B (0.5 µM), or DMSO for 48 hours, and pellets were collected for western blotting. For LDAs, GSCs were treated with one dose of MY45-A (0.5 µM) or MY45-B (0.5 µM) upon seeding, and plates were scored as described above at 21 days. Viability was assessed at 21 days using alamarBlue™ (ThermoFisher) according to the manufacturer’s protocol in wells seeded with 512 cells. HF-NSC LDAs were treated with MY45-A (0.5 µM) or MY45-B (0.5 µM) and scored as above at 21 days. HFAs (2,500 per well) in DMEM (10% FBS) were treated with MY45-A (0.5 µM), MY45-B (0.5 µM), or DMSO and assessed for viability at 7 days post-treatment. Proteasome activity assays The peptide-peptoid hybrid activity probes for the standard and immunoproteasome core particles have been previously described 33 , 34 , with probes synthesized for this study (Pepmic). GSCs (80,000) were seeded in duplicate technical replicate wells in a black 96-well plate with a clear bottom and incubated overnight. Proteasome activity probes (31 µM) were added to GSC and blank wells containing only GSC growth media. A first fluorescence reading (Excitation-485; Emission-535) was taken immediately after probe addition, and subsequent readings were acquired every 5 min for a total of 95 min. Each reading was normalized to the probe-specific blank by subtracting the blank value for each time point. Curves of normalized data were fit with a logistic growth model for comparison. Immunopeptidomics Basal and cell surface expression of HLA-ABC were assessed using western blotting and EZ-Link™ Sulfo-NHS-SS-Biotin (ThermoFisher) as previously described 65 . Briefly, cultured GSCs were rinsed twice with cold PBS, then incubated with sulfo-NHS-SS-biotin (1 mg/mL) for 10 minutes at 4°C with gentle rotation. The reaction was stopped using 100 mM glycine and 25 mM Tris-HCl, pH 7.4, for 5 minutes, followed by three washes with cold PBS. GSCs were subsequently lysed in RIPA buffer, and the input was set aside for western blotting. The remainder of the lysate was incubated with Dynabeads™ MyOne™ Streptavidin C1 Magnetic Beads (Invitrogen) overnight at 4°C to capture cell surface biotinylated proteins. Magnetic beads were separated from the supernatant and washed 4 times in cold PBS. Samples were resolved by SDS-PAGE and immunoblotted with HLA-ABC antibody (Abcam-ab70328) as described above. Immunopeptidomic analysis of BT67 AAVS1, PSME1KO, and PSME2KO was performed using a modified, previously described protocol 35 . Briefly, for each sample, 1x10 8 GSCs were collected, washed twice in PBS, and snap-frozen in liquid nitrogen for later use. CNBr-activated Sepharose beads (80 mg) (Cytiva) were rehydrated and coupled with Ultra-LEAF™ Purified anti-human HLA-A,B,C Antibody (1.5 mg) (W6/32 clone – BioLegend). Antibody-coupled beads were washed and blocked with glycine (0.2 M). The quality of antibody coupling was assessed by Coomassie staining of SDS-PAGE-resolved beads and wash fractions. GSC pellets were thawed rapidly and lysed in 1 ml lysis buffer (0.5% NP-40, 50 mM Tris, pH 8.0, 150 mM NaCl, 1 mM EDTA, with protease inhibitor (Roche)) using a tissue homogenizer to break up the pellet, followed by 1 h incubation with gentle rotation at 4°C. Lysates were centrifuged at 20000 x g for 20 min at 4°C, and the supernatant was collected. Lysates were incubated overnight at 4°C with gentle rotation with HLA-ABC antibody-coupled beads. HLA-ABC-bound beads were then washed as described in ref. 32, modified for washing in Micro Bio-Spin™ Chromatography Columns (Bio-Rad) with 1 mL of each buffer twice, with 30 sec centrifuges at 500 x g for buffer flow-through. HLA-ABC complexes were eluted with 200 µL of 1% Pierce™ Trifluoroacetic Acid, LC-MS Grade (ThermoFisher) by 500 x g centrifugation for 1 minute, repeated twice for a total eluate volume of 400 µL. HLA-ABC-bound beads, supernatant, and eluted HLA-ABC complexes were assessed by western blot for quality of pull-down and elution in one sample. Eluted HLA-ABC complexes were shipped to Biogenity (Denmark) for analysis. Eluate was up-concentrated on 3 kDa filters (Millipore) by wetting the filters with 400 µL 1% TFA and spun at 14000 x g. Then the samples were loaded and spun at 14000 x g with approximately 50 µL volume remaining. The flow-through was desalted using the SOLAµ SPE plate (ThermoFisher). Between applications, the solvents were spun through by centrifugation at 150 rpm. For each sample, the filters were activated with 200 µL of 100% Methanol, then 200 µL of 80% Acetonitrile containing 0.1% formic acid. The filters were subsequently equilibrated twice with 200 µL of 1% TFA, 3% Acetonitrile, after which the sample was loaded. After washing the tips twice with 200 µL of 0.1% formic acid, the peptides were eluted into clean 0.5 mL Eppendorf tubes using 40% Acetonitrile, 0.1% formic acid. The eluted peptides were concentrated in an Eppendorf Speedvac. The concentrated peptides were then reconstituted in 12 µL 2% acetonitrile, 1% trifluoroacetic acid for analysis. The peptides were centrifuged at 18,000 x g for 10 minutes to remove any residual particulate matter. Samples were analyzed using a Nanodrop to determine peptide concentration. For each sample, 500 ng of peptides were loaded onto a 2 cm C18 trap column (ThermoFisher), connected in-line to a 15 cm C18 reverse-phase EasySpray analytical column (ThermoFisher) using 100% Buffer A (0.1% formic acid in water) at 750 bar, on the EasyLC 1200 HPLC system (ThermoFisher), with the column oven set to 30°C. Peptides were eluted using a 70-minute gradient at a flow rate of 250 nL/min. The gradient began with a transition from 6% to 23% Buffer B (80% acetonitrile, 0.1% formic acid) and then increased to 38% Buffer B over 12 minutes. The final step involved ramping up to 95% Buffer B over 8 minutes, holding at this concentration for 7 minutes. The Orbitrap Exploris 480 (Thermo Fisher Scientific) was run in a DD-MS2 top 28 method. Full MS spectra were collected at a resolution of 60,000, with an AGC target of 300% or maximum injection time set to ‘auto’ and a scan range of 375–1500 m/z. The MS2 spectra were obtained at a resolution of 15.000, with an AGC target set to 75% or maximum injection time set to ‘auto’, a normalized collision energy of 28, and an intensity threshold of 1.0x10 4 . Dynamic exclusion was set to 60 s, and ions with a charge state < 2 or unknown were excluded. MS performance was evaluated by running complex cell lysate quality control standards. Analysis of the immunopeptidome data was performed using a previously described workflow 66 . HLA typing was performed using HLA-LA 67 on previously published whole-genome sequencing data from BT67 30 . Raw data were converted to mzML using msConvert 68 , and the database search was performed using Comet (2025.02) 69 with the human proteome (UP000005640) 70 . Search parameters for MS data included cut everywhere as the enzyme, a maximum missed cleavage of 2, and a mass error tolerance of 10 ppm. A fragment ion mass error tolerance of 1.5 Da was used. Search results were used as input for MHCvalidator 66 with the default configuration, using only MHCflurry for binding and processing score prediction 71 . Peptides with a database search q-value 95% were used for sample analysis. GL261 single-cell RNA-seq Scramble, Psme1 KD, and Psme2 KD GL261 (100,000) were implanted into the right cerebral hemisphere of 5 C57BL/6 mice per condition, as described above. The mice were euthanized 3 weeks post-engraftment, and tumours were dissected out of whole brains. Tumours were pooled for each condition, followed by manual dissociation in PBS containing 0.04% BSA and filtering through a 100 µm cell strainer (Millipore Sigma) to clear debris. Cells were washed 3 times in PBS containing 0.04% BSA and resuspended for viability check and counting. Each sample was confirmed to have > 80% viability, and a 7000-cell capture target was attempted based on the final cell number. Cells were transferred to the Hotchkiss Brain Institute NeuroOmics Core at the University of Calgary and subjected to single-cell transcriptomic profiling by the GEM-X Universal 3’ Gene Expression v4 (10X Genomics) following the manufacturer’s protocol. cDNA yields and libraries were confirmed to meet quality control standards for each sample. cDNA libraries were sequenced on a NextSeq P3 xleap (Illumina) 100-cycle, yielding 400 Mb per sample and ~ 20,000 reads/cell at the CHGI. GL261 single-cell RNA-seq data was analyzed using the standard Seurat (v.5) pipeline 54 . Briefly, BCL files were converted to fastq files using bcl2fastq (v.2.20–Illumina), followed by count matrix generation with cellranger (v.8.0.1–10X Genomics) using GRCm39 for alignment. Single-cell matrices were merged into a single Seurat object and integrated using the anchor-based CCA method. Cells were filtered for mitochondrial gene expression < 5%. Conserved markers (top 50) for the identified clusters were used to classify cell types with GPTCelltype 72 . GPTCelltype-identified cell types were manually confirmed and refined based on marker expression and clustering. Differential expression was assessed between conditions for all cell clusters using FindMarkers with default settings, and P adj < 0.05 was considered significant. PA28αβ interactome FLAG-tagged PA28α (PSME1) and PA28β (PSME2), as well as mCherry, were expressed in BT67 GSCs as described above. Anti-FLAG→ M2 magnetic beads (Millipore Sigma) (20 µL per sample) were used for the FLAG-PA28α and FLAG-PA28β pull-down. For the mCherry pull-down, anti-mCherry antibody (Abcam-ab183628) was cross-linked to Dynabeads™ Protein G (ThermoFisher) as per the manufacturer’s protocol with BS 3 (ThermoFisher) (20 µL per sample). Protein-protein crosslinking in BT67 GSCs was performed with Pierce™ DSP (ThermoFisher) (0.1 mM) in PBS for 30 min with gentle rotation, followed by two 5 min washes in cold PBS. Cells were then lysed by sonication in 50 mM Tris, pH 8.0, 150 mM NaCl, 1 mM EDTA, with protease inhibitor (Roche). 500 µg of protein was incubated with anti-FLAG or anti-mCherry beads for 2 hours at room temperature with gentle rotation. Beads were subsequently washed 4 times in lysis buffer, resuspended in 1X Laemmli buffer, and proteins were eluted by denaturation (8 min; 95°C). The procedure was tested for immunoprecipitation efficiency prior to mass spectrometric analysis using SDS-PAGE and Western blotting. Samples were resolved through the stacking portion of an SDS-polyacrylamide gel and ran into a 12% SDS-polyacrylamide gel until the entire protein entered the lower gel. The gels were stained with Coomassie Blue to confirm total protein content in the lower gel. Total protein within the gel was then excised with a scalpel into an Eppendorf tube with 1 mL milliQ water. Samples were transferred to the Southern Alberta Mass Spectrometry Facility at the University of Calgary for analysis. For trypsin digestion, gel bands were sliced into small pieces of about 1 mm³ using a scalpel blade. Gel plugs were washed 3 times for 15 min in 50 mM ammonium bicarbonate/acetonitrile (50:50, v/v). Gel plugs were briefly rinsed in 100% acetonitrile and incubated for 20 min in fresh 100% acetonitrile. After being air dried, proteins were reduced with dithiothreitol (DTT; 10 mM in 100 mM ammonium bicarbonate) for 30 min at 56°C and alkylated with iodoacetamide (IAA; 50 mM in 100 mM ammonium bicarbonate) for 30 mins in the dark at room temperature. Gel plugs were washed with 50 mM ammonium bicarbonate/acetonitrile (50:50, v/v) for 15 min and dehydrated with acetonitrile as described above. Gel plugs were then rehydrated with a trypsin solution (Promega; 0.02 µg/ul in 40mM ammonium bicarbonate 10% acetonitrile) and incubated for 2 h on ice. The excess of trypsin solution was removed, and trypsin buffer (40 mM ammonium bicarbonate 10% acetonitrile) was added to cover the gel pieces. Trypsin digestion was performed for 16 h at 37°C. After the digestion, the tryptic peptides were transferred into a new tube containing 5 µl of extraction solution (acetonitrile/water/ 10% trifluoroacetic acid (60:30:10, v/v). Tryptic peptides were extracted twice from the gel plugs by incubation in the extraction solution and vortexing for 10 min. The extracted peptides were combined into the same Eppendorf tube. Samples were then lyophilized and resuspended in 15 µl of 1% formic acid in water. For LC-MS/MS, tryptic peptides were analyzed on an Orbitrap Fusion Lumos Tribrid (ThermoFisher) operated with Xcalibur (v.4.4.16.14) and coupled to an Easy-nLC 1200 system. Tryptic peptides were loaded onto a C18 trap (75 um x 2 cm; Acclaim PepMap 100, P/N 164946; ThermoFisher) at a flow rate of 2 µL/min of solvent A (0.1% formic acid in LC-MS grade water). Peptides were eluted using a 120 min gradient from 5 to 40% (5% to 28% in 22 min followed by an increase to 40% B in 3 min) of solvent B (0.1% formic acid in 80% LC-MS grade acetonitrile) at a flow rate of 0.3 µL/min and separated on a C18 analytical column (75 um x 50 cm; PepMap RSLC C18; P/N ES803; ThermoFisher). Peptides were then electrosprayed at 2.1 kV into the ion transfer tube (300°C) of the Orbitrap Lumos operating in positive mode. The Orbitrap first performed a full MS scan at a resolution of 120,000 FWHM to detect the precursor ion having a m/z between 375 and 1575 and a + 2 to + 7 charge. The Orbitrap AGC and the maximum injection time were set at 4x10 5 and 50 ms, respectively. The Orbitrap was operated using the top speed mode with a 3 sec cycle time for precursor selection. The most intense precursor ions presenting a peptidic isotopic profile and having an intensity threshold of at least 5,000 were isolated using the quadrupole and fragmented with HCD (30% collision energy) in the ion routing multipole. The fragment ions (MS2) were analyzed in the ion trap at a rapid scan rate. The AGC and the maximum injection time were set at 1x10 4 and 35 ms, respectively, for the ion trap. Dynamic exclusion was enabled for 30 sec to avoid of the acquisition of same precursor ion having a similar m/z (plus or minus 10 ppm). For data analysis, Lumos raw data files were converted to MGF files using RawConverter (v.1.1.0.18–Scripps) operating in a data-dependent mode. Monoisotopic precursors having a charge state of + 2 to + 7 were selected for conversion. MGF files were used to search the human proteome database using Mascot algorithm (v.2.7–Matrix Sciences). Search parameters for MS data included trypsin as the enzyme, a maximum number of missed cleavages of 1, a peptide charge of 2 or higher, cysteine carbamidomethylation as fixed modification, methionine oxidation as variable modification and a mass error tolerance of 10 ppm. A mass error tolerance of 0.6 Da was selected for the fragment ions. Only peptides with scores exceeding 95% confidence were retained for further analysis. The Mascot data files were imported into Scaffold (v.5.3.2–Proteome Software Inc) for comparison of samples based on mass spectral counting. Bulk RNA-seq in PA28αβ genetically targeted GSCs RNA was isolated using the Amersham RNAspin Mini Kit (Cytiva) following the manufacturer’s protocol. Library prep and sequencing were performed by the CHGI-UCalgary. Reads were aligned to GRCh38 using Rsubread 73 . Normalization, clustering, and differential expression were performed with DESeq2 55 . A P adj |1| was considered significant. Pathway analysis was performed using gene set enrichment analysis within the ReactomePA R package (v.1.46.0) 74 . Immunofluorescence and concanavalin A staining Cells were plated in 8- or 16-well chamber slides coated with poly-ornithine and laminin. After 72 h, cells were fixed with 4% paraformaldehyde. For immunofluorescent staining of PA28α, cells were blocked/permeabilized using goat serum (10%), BSA (2%), and Triton-X100 (0.3%) in PBS. Cells were incubated with primary antibodies overnight, then incubated with fluorophore-conjugated secondary antibody (Invitrogen) for 1 hour. Cells were counterstained with DAPI and mounted with coverslips. Images were acquired with a Leica SP8 confocal microscope. For Concanavalin A staining, cells were washed in Hank’s Balanced Salt Solution (HBSS) 3 times, followed by incubation with 50 µg/mL of Concanavalin A – Alexa Fluor 647 (Invitrogen) in HBSS for 30 minutes. Cells were counterstained with mounting media containing DAPI and imaged as above. These images were acquired at a single z-plane. Proteomics Cell pellets were frozen in liquid nitrogen and stored at -80°C. Cells were lysed by sonication in 50 mM Tris, pH 8.0, 150 mM NaCl, 1 mM EDTA, with protease inhibitor (Roche). Lysates were centrifuged at 15,000 x g for 10 min at 4°C, and the supernatant was collected. Protein (200 µg per sample) was precipitated using trichloroacetic acid (20%). The protein precipitate was washed 3 times with cold acetone, then air-dried for 2 min. Protein was resuspended in Urea 8M in 100 mM Tris-HCl pH 8, followed by centrifugation at 10000 x g for 1 min. Dithiothreitol (10 mM) was added to the protein suspension, and the mixture was incubated for 30 min at 37°C. Protein was transferred into Microcon YM-30 (Millipore) and centrifuged for 15 min at 14000 x g. Membranes were washed with 200 µL Urea 8M in 100 mM Tris-HCl pH 8, and centrifuged for 15 min at 14000 x g. Membranes were incubated with Iodoacetamide (100 µL – 50 mM in Urea 8M in 100 mM Tris-HCl pH 8) at room temp for 20 min in the dark, then centrifuged for 10 min at 14000 x g. Membranes were washed with Urea 8M in 100 mM Tris-HCl pH 8 and centrifuged for 15 min at 14000 x g 3 times. Membranes were then washed with Ammonium Bicarbonate (50 mM) and centrifuged for 15 min at 14000 x g 3 times. Proteins on the membrane were digested with trypsin (Promega) (0.22 µg/µL in Ammonium Bicarbonate (50 mM)) overnight at 37°C in the dark. Membranes were then centrifuged for 5 sec at 100 x g to collect condensation and incubated at room temp for 10 min. Membranes were transferred to a new collection tube and centrifuged for 10 min at 14000 x g. Ammonium Bicarbonate (50 mM – 40 µL) was added to the membranes and incubated for 3 min followed by centrifugation for 10 min at 14000 x g. The previous step was repeated, and then NaCl (0.5 M – 40 µL) was added to the membrane and incubated for 3 min followed by centrifugation for 15 min at 14000 x g. Digested peptide flow-through was then transferred to a new tube and freeze-dried. Dried peptides were resuspended in Formic acid (1%) for desalting on UPLC. Peptides were analyzed on an Orbitrap Fusion Lumos Tribrid mass spectrometer (ThermoFisher) operated with Xcalibur (v.4.4.16.14) and coupled to an Easy-nLC 1200 system (ThermoFisher). Peptides were loaded onto a C18 trap (75 um x 2 cm; Acclaim PepMap 100, ThermoFisher) at a flow rate of 2 µL/min of solvent A (0.1% formic acid in LC-MS grade water). Peptides were eluted using a 120 min gradient from 5 to 40% (5% to 28% in 105 min followed by an increase to 40% B in 15 min) of solvent B (0.1% formic acid in 80% LC-MS grade acetonitrile) at a flow rate of 0.3 µL/min and separated on a C18 analytical column (75 um x 50 cm; PepMap RSLC C18; ThermoFisher). Peptides were then electrosprayed using 2.1 kV voltage into the ion transfer tube (300°C) of the Orbitrap Lumos operating in positive mode. The Orbitrap first performed a full MS scan at a resolution of 120000 FWHM to detect precursor ions with m / z between 375 and 1575, and charge states + 2 to + 7. The Orbitrap AGC (Auto Gain Control) and the maximum injection time were set to 4x10 5 and 50 ms, respectively. The Orbitrap was operated in top-speed mode with a 3 sec cycle time for precursor selection. The most intense precursor ions exhibiting a peptidic isotopic profile and an intensity threshold of at least 5000 were isolated using the quadrupole and fragmented with HCD (30% collision energy) in the ion-routing multipole. The fragment ions (MS 2 ) were analyzed in the ion trap at a rapid scan rate. The AGC and the maximum injection time were set at 1x10 4 and 35 ms, respectively, for the ion trap. Dynamic exclusion was enabled for 45 sec to prevent the acquisition of the same precursor ion with a similar m/z (within ± 10 ppm). Raw data files were converted to Mascot Generic Format using RawConverter (Scripps-v.1.1.0.18) operating in a data-dependent mode. Monoisotopic precursors with charge states of + 2 to + 7 were selected for conversion. The database search was performed using the human proteome with the Mascot algorithm (Matrix Sciences; version 2.7). Search parameters for MS data included trypsin as the enzyme, a maximum of 1 missed cleavage, a peptide charge of 2 or higher, cysteine carbamidomethylation as a fixed modification, methionine oxidation as a variable modification, and a mass error tolerance of 10 ppm. A mass error tolerance of 0.6 Da was selected for the fragment ions. Only peptides identified with a confidence score > 95% were retained for further analysis. The Mascot files were imported into Scaffold (Proteome Software, Inc.) for comparison of different samples using Top 3 Total Ion Chromatograms (TICs). Proteins with 3 or greater peptides identified were analyzed for cross-sample comparison using limma 75 with a cutoff of P 1. Pathway analysis was performed using GSEA within the ReactomePA R package 74 . Co-Immunoprecipitation GSCs were collected 10 days post-seeding. Protein-protein crosslinking was performed with Pierce™ DSP (ThermoFisher) (0.1 mM) in PBS for 30 min with gentle rotation, followed by two 5 min washes in cold PBS. Cells were then lysed by sonication in 50 mM Tris, pH 8.0, 150 mM NaCl, 1 mM EDTA, 0.5% NP-40 with protease inhibitor (Roche). 500 µg of protein was incubated with anti-FLAG or IgG isotype beads overnight at 4°C with gentle rotation. Beads were subsequently washed 4 times in lysis buffer, resuspended in 1X Laemmli buffer, and proteins were eluted by denaturation (8 min; 95°C). Immunoprecipitated samples, alongside input (5%) and immunodepleted fractions (5%), were resolved by SDS-PAGE, transferred to nitrocellulose membranes, and immunoblotted for specific proteins. Proximity Ligation Assay Cells were plated as described above for immunofluorescence and fixed with 4% PFA. Cells were washed 3 times with PBS, then permeabilized with PBS, Triton-X100 (0.3%) for 10 min, followed by 1 h block in Duolink → In Situ PLA Ⓡ 1X Blocking Solution (Millipore Sigma). Cells were then incubated with primary antibody mixtures, IgG controls, or no-antibody controls for 2 h at 37°C, 5% CO 2 . Cells were then subjected to the Duolink Ⓡ In Situ PLA Ⓡ anti-rabbit PLUS and anti-Mouse MINUS staining (Millipore Sigma), followed by ligation and amplification according to the manufacturer’s guidelines. Images were acquired on a Leica SP8 confocal microscope with 20 z-planes collapsed into a maximum intensity projection using ImageJ. Three regions of interest were captured for quantification per condition. PLA foci were quantified using CellProfiler (v.4.2.8) 76 . GSC viability dose response to KIF11 and PRMT5 inhibitors and LDAs Cells were plated at a 2,500 cell/well density in a 96-well plate and incubated overnight. Escalating doses of the KIF11 inhibitor ispinisib (Selleck) or the PRMT5 inhibitory probe GSK591 (Structural Genomics Consortium) were added to 3 technical replicate wells per dose. Viability was assessed 13 days after inhibitor addition using alamarBlue™ (ThermoFisher) according to the manufacturer’s protocol. Viability readings were averaged across technical replicates and normalized to vehicle controls. For LDAs, sub-IC50 doses were determined by viability readings. Ispinisib (BT67: 2 nM; BT189: 1 nM) or GSK591 (BT67: 100 nM; BT189: 100 nM) were added to LDA plates 1 d after seeding and scored as described above. Scores were normalized to AAVS1 controls for BT67 and Scramble controls for BT189. Alternative splicing analysis Alternative splicing analysis was performed using the SpliceWiz R package (v.1.4.1) 77 using default settings. Bulk RNA-seq on PA28αβ genetically targeted GSCs, aligned as described above, were used as input for SpliceWiz. Cycloheximide chase assay GSCs (1 x 10 6 ) were seeded in 6-well plates and incubated for 72 hours. DMSO or cycloheximide (Selleck) (300 µg/mL) was added to the GSC wells, followed by immediate collection of the 0 h timepoint. Cell pellets were washed twice in PBS, flash frozen in liquid nitrogen, and stored at -80°C. Additional pellets were collected at 3 and 6 h time points. The DMSO vehicle pellets were collected at 6 h. Cell pellets were lysed and immunoblotted as described above. Quantification was performed as described above for immunoblotting. Quantitative Polymerase Chain Reaction (qPCR) Cell pellets were frozen in liquid nitrogen and stored at -80°C. RNA was extracted using the RNAspin Mini Kit (Cytiva) as per the manufacturer’s protocol and quantified by nanodrop. cDNA was synthesized from RNA (500 ng) using qScript™ cDNA SuperMix (Quantabio) as per the manufacturer’s protocol. qPCR was performed with cDNA in duplicate technical replicate wells with primers specific to mouse Kif11 and Actb (Supplemental Table S9) and FastStart Essential DNA Green Master (Roche). Cycling conditions were 95°C (3 min), 95°C (15 sec), 60°C (45 sec with signal capture), return to step 2 and repeat 50X, followed by a melting curve. Statistical Analysis All experiments were performed with three replicates unless otherwise stated. Statistical analyses were performed using GraphPad Prism Version 10 or R Version 4.3. Specific tests used are described in the figure captions or directly in the text. Data are reported as mean ± SEM unless otherwise stated, and P < 0.05 was considered statistically significant. Median survival in mouse tumour experiments was estimated using the Kaplan-Meier method, and the Log-Rank Mantel-Cox test was used to assess statistical significance. Declarations Data availability: Single-cell RNA-seq data for GSCs is available from the Broad Institute Single-Cell Portal ( https://singlecell.broadinstitute.org/single_cell/study/SCP503 ). Single-cell RNA-seq for NSPCs is available through the NCBI BioProject database (PRJNA798712). TCGA-GBM RNA-seq is available through the TCGA-Biolinks R package. GTEx normal tissue RNA-seq is available through the recount3 R package. CPTAC-GBM proteomics data can be accessed through the CPTAC Python package. RNA-seq for our in-house cohort of GSCs are available at the European Genome-phenome Archive (EGAS00001002709). Immunopeptidomic data and interactome data have been deposited in the ProteomeExchange Database (#######(Available to reviewer/editor upon request)). RNA sequencing data on PSME1/2 KO and KD GSCs generated for this project have been deposited in SRA (####### (Available to reviewer/editor upon request)). GL261 single-cell RNA-seq has been deposited at the Broad Institute Single-Cell Portal (##### (Available to reviewer/editor upon request)). All other data and code used for analysis are available upon request. Competing interests: None. Author Contributions: Conceptualization: K.M.H, R.K.B, S.A, H.A.L, S.W; Methodology: K.M.H, R.K.B, S.A, X.H, F.H, G.D, R.H, O.C, B.M, B.F.C, H.A.L; Formal Analysis and Visualization: K.M.H, R.K.B, F.H, G.D, X.H, O.C; Supervision: H.A.L., S.W; Writing: K.M.H; Review and Editing: S.A, R.K.B, H.A.L, S.W; Funding Acquisition: H.A.L, S.W. Acknowledgements: This research project was supported by a grant from the Canadian Institutes of Health Research #153246 to H.A.L. and S.W. We would like to thank Dr. Darci J. Trader for providing advice regarding proteasome activity probes. We thank Dr. Wee Yong’s lab at the University of Calgary for providing human fetal brain tissue. We thank Drs. Alisha Poole and Stephen Robbins for providing GL261 glioma cells. We thank Dr. Yiping Liu at The Flow Cytometry Core Facility (University of Calgary) for cell cycle analysis. We thank Dr. Frank Visser at the HBI Molecular Core Facility (University of Calgary) for molecular cloning and lentiviral preparation. We thank Dr. Charlotte D’Mello at the HBI Neuro Omics Core (University of Calgary) for the single-cell library preparation. We thank the Centre for Health Genomics and Informatics (University of Calgary) for performing library preparation and sequencing. We thank the Southern Alberta Mass-Spectrometry Facility (University of Calgary) for running the interactome samples. We thank the HBI-AMP (University of Calgary) for use of microscopes. We thank Biogenity (Denmark) for running the immunopeptidomics samples. We thank Dr. Danielle A. Bozek, Emilie Cutts, Rory Mulloy, Dr. Sorana Morrissy, and Dr. Marco Gallo for scientific or editorial input. References Rousseau, A., Bertolotti, A.: Regulation of proteasome assembly and activity in health and disease. Nat. Rev. Mol. Cell. Biol. 19 , 697–712 (2018) Fabre, B., et al.: Deciphering preferential interactions within supramolecular protein complexes: the proteasome case. Mol. Syst. Biol. 11 , 771 (2015) Suraweera, A., Münch, C., Hanssum, A., Bertolotti, A.: Failure of Amino Acid Homeostasis Causes Cell Death following Proteasome Inhibition. Mol. Cell. 48 , 242–253 (2012) De La Peña, A.H., Goodall, E.A., Gates, S.N., Lander, G.C., Martin, A.: Substrate-engaged 26 S proteasome structures reveal mechanisms for ATP-hydrolysis–driven translocation. Science. 362 , eaav0725 (2018) Sahu, I., et al.: The 20S as a stand-alone proteasome in cells can degrade the ubiquitin tag. Nat. Commun. 12 , 6173 (2021) Ramachandran, K.V., Margolis, S.S.: A mammalian nervous-system-specific plasma membrane proteasome complex that modulates neuronal function. Nat. Struct. Mol. Biol. 24 , 419–430 (2017) Javitt, A., et al.: The proteasome regulator PSME4 modulates proteasome activity and antigen diversity to abrogate antitumor immunity in NSCLC. Nat. Cancer. 4 , 629–647 (2023) Manasanch, E.E., Orlowski, R.Z.: Proteasome inhibitors in cancer therapy. Nat. Rev. Clin. Oncol. 14 , 417–433 (2017) Moriishi, K., et al.: Critical role of PA28γ in hepatitis C virus-associated steatogenesis and hepatocarcinogenesis. Proc. Natl. Acad. Sci. 104, 1661–1666 (2007) Huber, E.M., Groll, M.: The Mammalian Proteasome Activator PA28 Forms an Asymmetric α4β3 Complex. Structure. 25 , 1473–1480e3 (2017) Chen, J., et al.: Cryo-EM of mammalian PA28αβ-iCP immunoproteasome reveals a distinct mechanism of proteasome activation by PA28αβ. Nat. Commun. 12 , 739 (2021) Zhao, J., et al.: Structural insights into the human PA28–20S proteasome enabled by efficient tagging and purification of endogenous proteins. Proc. Natl. Acad. Sci. 119, e2207200119 (2022) Lesne, J., et al.: Conformational maps of human 20S proteasomes reveal PA28- and immuno-dependent inter-ring crosstalks. Nat. Commun. 11 , 6140 (2020) Raule, M., et al.: PA28αβ Reduces Size and Increases Hydrophilicity of 20S Immunoproteasome Peptide Products. Chem. Biol. 21 , 470–480 (2014) Preckel, T., et al.: Impaired Immunoproteasome Assembly and Immune Responses in PA28 . Mice Sci. 286 , 2162–2165 (1999) De Graaf, N., et al.: PA28 and the proteasome immunosubunits play a central and independent role in the production of MHC class I-binding peptides in vivo. Eur. J. Immunol. 41 , 926–935 (2011) Murata, S., et al.: Immunoproteasome assembly and antigen presentation in mice lacking both PA28alpha and PA28beta. EMBO J. 20 , 5898–5907 (2001) Nomura, M., et al.: The multilayered transcriptional architecture of glioblastoma ecosystems. Nat. Genet. 57 , 1155–1167 (2025) Sloan, A.R., Silver, D.J., Kint, S., Gallo, M., Lathia, J.D.: Cancer stem cell hypothesis 2.0 in glioblastoma: Where are we now and where are we going? Neuro-Oncol. 26 , 785–795 (2024) Schaff, L.R., Mellinghoff, I.K.: Glioblastoma and Other Primary Brain Malignancies in Adults: A Review. JAMA. 329 , 574 (2023) Di, K., et al.: Marizomib activity as a single agent in malignant gliomas: ability to cross the blood-brain barrier. Neuro-Oncol. 18 , 840–848 (2016) Roth, P., et al.: Marizomib for patients with newly diagnosed glioblastoma: A randomized phase 3 trial. Neuro-Oncol. 26 , 1670–1682 (2024) Sun, C., et al.: An abundance of free regulatory (19 S ) proteasome particles regulates neuronal synapses. Science. 380 , eadf2018 (2023) Richards, L.M., et al.: Gradient of Developmental and Injury Response transcriptional states defines functional vulnerabilities underpinning glioblastoma heterogeneity. Nat. Cancer. 2 , 157–173 (2021) Liu, D.D., et al.: Purification and characterization of human neural stem and progenitor cells. Cell. 186 , 1179–1194e15 (2023) Verhaak, R.G.W., et al.: Integrated Genomic Analysis Identifies Clinically Relevant Subtypes of Glioblastoma Characterized by Abnormalities in PDGFRA, IDH1, EGFR, and NF1. Cancer Cell. 17 , 98–110 (2010) The GTEx Consortium: The GTEx Consortium atlas of genetic regulatory effects across human tissues. Science. 369 , 1318–1330 (2020) Thirant, C., et al.: Differential Proteomic Analysis of Human Glioblastoma and Neural Stem Cells Reveals HDGF as a Novel Angiogenic Secreted Factor. STEM CELLS. 30 , 845–853 (2012) Wang, L.-B., et al.: Proteogenomic and metabolomic characterization of human glioblastoma. Cancer Cell. 39 , 509–528e20 (2021) Shen, Y., et al.: Comprehensive genomic profiling of glioblastoma tumors, BTICs, and xenografts reveals stability and adaptation to growth environments. Proc. Natl. Acad. Sci. 116, 19098–19108 (2019) Hamed, A.A., et al.: Gliomagenesis mimics an injury response orchestrated by neural crest-like cells. Nature. 638 , 499–509 (2025) Lazear, M.R., et al.: Proteomic discovery of chemical probes that perturb protein complexes in human cells. Mol. Cell. 83 , 1725–1742e12 (2023) Zerfas, B.L., Trader, D.J.: Monitoring the Immunoproteasome in Live Cells Using an Activity-Based Peptide–Peptoid Hybrid Probe. J. Am. Chem. Soc. 141 , 5252–5260 (2019) Zerfas, B.L., Coleman, R.A., Salazar-Chaparro, A.F., Macatangay, N.J., Trader, D.J.: Fluorescent Probes with Unnatural Amino Acids to Monitor Proteasome Activity in Real-Time. ACS Chem. Biol. 15 , 2588–2596 (2020) Sirois, I., Isabelle, M., Duquette, J.D., Saab, F., Caron, E., Immunopeptidomics: Isolation of Mouse and Human MHC Class I- and II-Associated Peptides for Mass Spectrometry Analysis. J. Vis. Exp. 63052 (2021). 10.3791/63052 Xiao, J.: Biological functions of melanoma-associated antigens. World J. Gastroenterol. 10 , 1849 (2004) Shraibman, B., et al.: Identification of Tumor Antigens Among the HLA Peptidomes of Glioblastoma Tumors and Plasma. Mol. Cell. Proteom. 18 , 1255–1268 (2019) Lee, S.H., et al.: M2-like, dermal macrophages are maintained via IL-4/CCL24–mediated cooperative interaction with eosinophils in cutaneous leishmaniasis. Sci. Immunol. 5 , eaaz4415 (2020) Ilangumaran, S., Bobbala, D., Ramanathan, S.: SOCS1: Regulator of T Cells in Autoimmunity and Cancer. In: Yoshimura, A. vol (ed.) Emerging Concepts Targeting Immune Checkpoints in Cancer and Autoimmunity, vol. 410, pp. 159–189. Springer International Publishing, Cham (2017) Skadborg, S.K., et al.: Nivolumab Reaches Brain Lesions in Patients with Recurrent Glioblastoma and Induces T-cell Activity and Upregulation of Checkpoint Pathways. Cancer Immunol. Res. 12 , 1202–1220 (2024) Sachamitr, P., et al.: PRMT5 inhibition disrupts splicing and stemness in glioblastoma. Nat. Commun. 12 , 979 (2021) Venere, M., et al.: The mitotic kinesin KIF11 is a driver of invasion, proliferation, and self-renewal in glioblastoma. Sci. Transl Med. 7 , (2015) Myers, K.A., Baas, P.W.: Kinesin-5 regulates the growth of the axon by acting as a brake on its microtubule array. J. Cell. Biol. 178 , 1081–1091 (2007) Wakana, Y., et al.: Kinesin-5/Eg5 is important for transport of CARTS from the trans-Golgi network to the cell surface. J. Cell. Biol. 202 , 241–250 (2013) Zhao, Y.: SMN and symmetric arginine dimethylation of RNA polymerase II C-terminal domain control termination. Nature. 529 , 48–53 (2016) Zhou, Z., et al.: PRMT5 regulates Golgi apparatus structure through methylation of the golgin GM130. Cell. Res. 20 , 1023–1033 (2010) Mandal, D.K., Brewer, C.F.: Differences in the binding affinities of dimeric concanavalin A (including acetyl and succinyl derivatives) and tetrameric concanavalin A with large oligomannose-type glycopeptides. Biochemistry. 32 , 5116–5120 (1993) Kao, S.-H., et al.: Analysis of Protein Stability by the Cycloheximide Chase Assay. BIO-Protoc 5 , (2015) Blagden, S.P., et al.: A phase I trial of ispinesib, a kinesin spindle protein inhibitor, with docetaxel in patients with advanced solid tumours. Br. J. Cancer. 98 , 894–899 (2008) Lee, C.W., et al.: A phase II study of ispinesib (SB-715992) in patients with metastatic or recurrent malignant melanoma: a National Cancer Institute of Canada Clinical Trials Group trial. Invest. New. Drugs. 26 , 249–255 (2008) Cheng, Y.L., et al.: Multiplexed single-cell lineage tracing of mitotic kinesin inhibitor resistance in glioblastoma. Cell. Rep. 43 , 114139 (2024) Garcia-Saez, I., Skoufias, D.A.: Eg5 targeting agents: From new anti-mitotic based inhibitor discovery to cancer therapy and resistance. Biochem. Pharmacol. 184 , 114364 (2021) Chojnacki, A., Weiss, S.: Production of neurons, astrocytes and oligodendrocytes from mammalian CNS stem cells. Nat. Protoc. 3 , 935–940 (2008) Hao, Y., et al.: Dictionary learning for integrative, multimodal and scalable single-cell analysis. Nat. Biotechnol. 42 , 293–304 (2024) Love, M.I., Huber, W., Anders, S.: Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15 , 550 (2014) Colaprico, A., et al.: TCGAbiolinks: an R/Bioconductor package for integrative analysis of TCGA data. Nucleic Acids Res. 44 , e71–e71 (2016) Wilks, C., et al.: recount3: summaries and queries for large-scale RNA-seq expression and splicing. Genome Biol. 22 , 323 (2021) Lindgren, C.M., et al.: Simplified and Unified Access to Cancer Proteogenomic Data. J. Proteome Res. 20 , 1902–1910 (2021) Schneider, C.A., Rasband, W.S., Eliceiri, K.: W. NIH Image to ImageJ: 25 years of image analysis. Nat. Methods. 9 , 671–675 (2012) Barbie, D.A., et al.: Systematic RNA interference reveals that oncogenic KRAS-driven cancers require TBK1. Nature. 462 , 108–112 (2009) MacLeod, G., et al.: Genome-Wide CRISPR-Cas9 Screens Expose Genetic Vulnerabilities and Mechanisms of Temozolomide Sensitivity in Glioblastoma Stem Cells. Cell. Rep. 27 , 971–986e9 (2019) Hu, Y., Smyth, G.K., ELDA: Extreme limiting dilution analysis for comparing depleted and enriched populations in stem cell and other assays. J. Immunol. Methods. 347 , 70–78 (2009) Cusulin, C., et al.: Precursor States of Brain Tumor Initiating Cell Lines Are Predictive of Survival in Xenografts and Associated with Glioblastoma Subtypes. Stem Cell. Rep. 5 , 1–9 (2015) Jiang, W., et al.: An optimized method for high-titer lentivirus preparations without ultracentrifugation. Sci. Rep. 5 , 13875 (2015) Taylor, K.R., et al.: Glioma synapses recruit mechanisms of adaptive plasticity. Nature. 623 , 366–374 (2023) Kovalchik, K.A., et al.: Machine learning-enhanced immunopeptidomics applied to T-cell epitope discovery for COVID-19 vaccines. Nat. Commun. 15 , 10316 (2024) Dilthey, A.T., et al.: HLA*LA—HLA typing from linearly projected graph alignments. Bioinformatics. 35 , 4394–4396 (2019) Adusumilli, R., Mallick, P.: Data Conversion with ProteoWizard msConvert. in Proteomics (eds Comai, L., Katz, J. E. & Mallick, P.) vol. 1550 339–368Springer New York, New York, NY, (2017) Eng, J.K., Jahan, T.A., Hoopmann, M.R., Comet: An open-source MS / MS sequence database search tool. PROTEOMICS. 13 , 22–24 (2013) The UniProt Consortium: UniProt: the Universal Protein Knowledgebase in 2025. Nucleic Acids Res. 53 , D609–D617 (2025) O’Donnell, T.J., Rubinsteyn, A., Laserson, U.: MHCflurry 2.0: Improved Pan-Allele Prediction of MHC Class I-Presented Peptides by Incorporating Antigen Processing. Cell. Syst. 11 , 42–48e7 (2020) Hou, W., Ji, Z.: Assessing GPT-4 for cell type annotation in single-cell RNA-seq analysis. Nat. Methods. 21 , 1462–1465 (2024) Liao, Y., Smyth, G.K., Shi, W.: The R package Rsubread is easier, faster, cheaper and better for alignment and quantification of RNA sequencing reads. Nucleic Acids Res. 47 , e47–e47 (2019) Yu, G., He, Q.-Y.: ReactomePA: an R/Bioconductor package for reactome pathway analysis and visualization. Mol. Biosyst. 12 , 477–479 (2016) Ritchie, M.E., et al.: limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 43 , e47 (2015) Stirling, D.R., et al.: CellProfiler 4: improvements in speed, utility and usability. BMC Bioinform. 22 , 433 (2021) Wong, A.C.H., Wong, J.J.-L., Rasko, J.E.J., Schmitz, U.: SpliceWiz: interactive analysis and visualization of alternative splicing in R. Brief. Bioinform. 25 , bbad468 (2023) Additional Declarations There is NO Competing Interest. Supplementary Files Heemskerketal2026supplementalfigures.docx Supplemental Figures and Captions Supplementaltables.xlsx Supplemental Tables Cite Share Download PDF Status: Under Review Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8682343","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":628234690,"identity":"7ec36fa5-ae64-4e5f-ae64-2019cc42f565","order_by":0,"name":"Samuel Weiss","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA0UlEQVRIiWNgGAWjYFACxgYGhgogzU6aljNAmpk0i9pI0WJwvLn54895h+X5m9kvfq5sY5DnbyCk5czBNmnebYcNZxzmKZY828ZgOOMAAS1mNxLbmBm3HU5gOMyTINnYxpDAQFDL/YdAh805nCB/mCf5J0iLPGFbGBskeBsOJxgcZj8GtsWAkBb7M4lt0jzH0g03HuZhs2w4J2G4kZAWyfbjjz/+qLGWlzve/vhmQ5mNvBwhLUiAx4CBkU2CePVAwP6AgeEPSTpGwSgYBaNghAAApYxDhjv9WysAAAAASUVORK5CYII=","orcid":"https://orcid.org/0009-0003-9190-563X","institution":"University of Calgary","correspondingAuthor":true,"prefix":"","firstName":"Samuel","middleName":"","lastName":"Weiss","suffix":""},{"id":628234691,"identity":"aeb2e1f9-f9d9-4bd8-80e5-07c9bfbfdb83","order_by":1,"name":"Kyle Heemskerk","email":"","orcid":"","institution":"University of Calgary","correspondingAuthor":false,"prefix":"","firstName":"Kyle","middleName":"","lastName":"Heemskerk","suffix":""},{"id":628234692,"identity":"b4303964-f069-44a5-be87-44378e6ddf85","order_by":2,"name":"Ravinder Bahia","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Ravinder","middleName":"","lastName":"Bahia","suffix":""},{"id":628234694,"identity":"29a3f16f-af64-4ced-9cf1-bb6511fa41ec","order_by":3,"name":"Samir Assaf","email":"","orcid":"","institution":"University of Calgary","correspondingAuthor":false,"prefix":"","firstName":"Samir","middleName":"","lastName":"Assaf","suffix":""},{"id":628234696,"identity":"86652c26-53b2-4094-978d-11ad7c9fd844","order_by":4,"name":"Xiaoguang Hao","email":"","orcid":"","institution":"University of Calgary","correspondingAuthor":false,"prefix":"","firstName":"Xiaoguang","middleName":"","lastName":"Hao","suffix":""},{"id":628234698,"identity":"6d6bb14a-480c-451f-ae5a-ca901a923b94","order_by":5,"name":"Fatima Hamood","email":"","orcid":"","institution":"University of Calgary","correspondingAuthor":false,"prefix":"","firstName":"Fatima","middleName":"","lastName":"Hamood","suffix":""},{"id":628234699,"identity":"0640d627-15d5-4f5f-9f9c-db583ed2a2d2","order_by":6,"name":"Gayatri Dronamraju","email":"","orcid":"","institution":"University of Calgary","correspondingAuthor":false,"prefix":"","firstName":"Gayatri","middleName":"","lastName":"Dronamraju","suffix":""},{"id":628234701,"identity":"56ea8c63-e4a3-4a80-a883-0f75571f1900","order_by":7,"name":"Rozina Hassam","email":"","orcid":"","institution":"University of Calgary","correspondingAuthor":false,"prefix":"","firstName":"Rozina","middleName":"","lastName":"Hassam","suffix":""},{"id":628234704,"identity":"9ed31dc4-c68e-497a-bf57-041565332ce4","order_by":8,"name":"Orsolya Cseh","email":"","orcid":"","institution":"University of Calgary","correspondingAuthor":false,"prefix":"","firstName":"Orsolya","middleName":"","lastName":"Cseh","suffix":""},{"id":628234708,"identity":"5f9e8008-b36b-4f35-be43-a1a3bb935ad1","order_by":9,"name":"Bruno Melillo","email":"","orcid":"https://orcid.org/0000-0002-9708-5287","institution":"The Scripps Research Institute","correspondingAuthor":false,"prefix":"","firstName":"Bruno","middleName":"","lastName":"Melillo","suffix":""},{"id":628234711,"identity":"8b27547a-2e04-43e3-a8de-9f5efb1dff1a","order_by":10,"name":"Benjamin Cravatt","email":"","orcid":"https://orcid.org/0000-0001-5330-3492","institution":"Scripps Research Institute","correspondingAuthor":false,"prefix":"","firstName":"Benjamin","middleName":"","lastName":"Cravatt","suffix":""},{"id":628234713,"identity":"02e77b92-2646-4cf2-b3b8-70aecff2aacc","order_by":11,"name":"Artee Luchman","email":"","orcid":"https://orcid.org/0000-0002-6663-9929","institution":"University of Calgary","correspondingAuthor":false,"prefix":"","firstName":"Artee","middleName":"","lastName":"Luchman","suffix":""}],"badges":[],"createdAt":"2026-01-23 20:30:31","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8682343/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8682343/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108495098,"identity":"19c17304-8928-4daa-9bfb-04fdbcc3e063","added_by":"auto","created_at":"2026-05-05 10:08:41","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":840159,"visible":true,"origin":"","legend":"\u003cp\u003eProteasome subunit expression and PA28abexpression in GBM stem cells and tumours compared to NSPCs and normal tissue.\u003c/p\u003e\n\u003cp\u003eA) Schematic representation of sample cohorts used for proteasome subunit analyses. Created in BioRender.\u003c/p\u003e\n\u003cp\u003eB) Proteasome subunit expression in single-cell RNA-seq data from GBM stem cells (GSCs) and neural stem \u0026amp; progenitor cells (NSPC) isolates segregated by sample type (n=29 GSC, n=4 NSPCs).\u003c/p\u003e\n\u003cp\u003eC) Proteasome species gene family module score in single-cell RNA-seq data from GSCs and NSPC isolates (Mann-Whitney test with Bonferroni correction \u003cem\u003eP\u003c/em\u003e value; n=29 GSC, n=4 NSPCs).\u003c/p\u003e\n\u003cp\u003eD) \u003cem\u003ePSME1\u003c/em\u003e RNA expression in GBM tumours compared to normal solid tissue and brain tissue (Wald test; ** \u003cem\u003ePadj\u003c/em\u003e\u0026lt;0.01, **** \u003cem\u003ePadj\u003c/em\u003e\u0026lt;0.0001; primary GBM: n=372, recurrent GBM: n =14, solid tissue normal: n=5, normal brain (GTEx): n = 300, 2931 tissues).\u003c/p\u003e\n\u003cp\u003eE) \u003cem\u003ePSME2\u003c/em\u003e RNA expression in GBM tumours compared to normal solid tissue and brain tissue (Wald test; * \u003cem\u003ePadj\u003c/em\u003e\u0026lt;0.05, **** \u003cem\u003ePadj\u003c/em\u003e\u0026lt;0.0001; primary GBM: n=372, recurrent GBM: n =14, solid tissue normal: n=5, normal brain (GTEx): n = 300, 2931 tissues).\u003c/p\u003e\n\u003cp\u003eF) PA28a (left) and PA28b (right) protein expression in the CPTAC GBM cohort compared to normal brain controls (Mann-Whitney test, ****\u003cem\u003eP\u0026lt;\u003c/em\u003e0.0001; GBM: n=99; normal: n=10)\u003c/p\u003e\n\u003cp\u003eG) Correlation between PA28abexpression score and the Developmental signature score in GSC and NSPC single cells (Pearson correlation \u003cem\u003eP \u003c/em\u003evalue; n=29 GSC, n=4 NSPCs).\u003c/p\u003e\n\u003cp\u003eH) Correlation between PA28abexpression score and the Injury Response signature score in GSC and NSPC single cells (Pearson correlation \u003cem\u003eP \u003c/em\u003evalue; n=29 GSC, n=4 NSPCs).\u003c/p\u003e\n\u003cp\u003eI) PA28a and b protein levels in GSCs (n=14).\u003c/p\u003e\n\u003cp\u003eJ) Lack of correlation between relative PA28a and b protein levels in GSCs (Pearson correlation \u003cem\u003eP \u003c/em\u003evalue; n=25).\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-8682343/v1/72b95cc406045cad3ec3e488.png"},{"id":108494445,"identity":"60aa77f1-83fd-4df4-8ea9-86bf1d889fd0","added_by":"auto","created_at":"2026-05-05 10:05:18","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":486038,"visible":true,"origin":"","legend":"\u003cp\u003eGenetic targeting of PA28ab decreases sphere formation and reduces tumour growth in GSCs and disrupts syngeneic GBM growth.\u003c/p\u003e\n\u003cp\u003eA)\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp; PA28ab protein levels in a panel of GSCs cultured in EGF/FGF or FBS (representative of n=3).\u003c/p\u003e\n\u003cp\u003eB)\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp; PA28ab protein levels in BT67 CRISPR/Cas9 KOs with two guide RNAs per gene (\u003cem\u003ePSME1\u003c/em\u003e and \u003cem\u003ePSME2\u003c/em\u003e genes encode PA28a and b proteins, respectively) (Representative of n=3).\u003c/p\u003e\n\u003cp\u003eC)\u0026nbsp;\u0026nbsp;\u0026nbsp; Growth curves of BT67 PA28ab KOs with two guide RNAs per gene (n=3; bars indicate mean +/- SEM).\u003c/p\u003e\n\u003cp\u003eD)\u0026nbsp;\u0026nbsp;\u0026nbsp; Doubling time in PA28ab KO/KD GSCs compared to controls (n=3; bars indicate best fit value +/- 95% confidence interval).\u003c/p\u003e\n\u003cp\u003eE)\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp; Cell cycle analysis of BT67 PA28ab KO GSCs (n=3; bars indicate mean +/- SEM; 2-way ANOVA; *\u003cem\u003eP\u003c/em\u003e\u0026lt;0.05).\u003c/p\u003e\n\u003cp\u003eF)\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp; Sphere-forming frequency of PA28ab KO/KD GSCs compared to controls (6 limiting dilutions per GSC line; bars indicate estimate of sphere-forming frequency +/- upper and lower limits; c\u003csup\u003e2\u003c/sup\u003e; *\u003cem\u003eP\u003c/em\u003e\u0026lt;0.05, **\u003cem\u003eP\u003c/em\u003e\u0026lt;0.01, ****\u003cem\u003eP\u003c/em\u003e\u0026lt;0.0001).\u003c/p\u003e\n\u003cp\u003eG)\u0026nbsp;\u0026nbsp;\u0026nbsp; Kaplan–Meier survival curve of mice orthotopically xenografted with BT67 PA28ab KO GSCs compared to AAVS1 controls (n=8 mice per group; Log-rank test; \u003cem\u003eP\u003c/em\u003e\u0026lt;0.0001).\u003c/p\u003e\n\u003cp\u003eH)\u0026nbsp;\u0026nbsp;\u0026nbsp; PA28a and PA28b protein levels in GL261 transduced with \u003cem\u003ePsme1\u003c/em\u003e, \u003cem\u003ePsme2\u003c/em\u003e, and Scramble shRNAs (representative of n=3).\u003c/p\u003e\n\u003cp\u003eI)\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp; Kaplan–Meier survival curve of mice orthotopically engrafted with GL261 Scramble control, \u003cem\u003ePsme1\u003c/em\u003e KD, and \u003cem\u003ePsme2\u003c/em\u003e KD (n=10 mice per group; Log-rank test;\u003cem\u003e P\u003c/em\u003e\u0026lt;0.0001).\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-8682343/v1/ed94714a7dcd3028bcbc1321.png"},{"id":108494447,"identity":"33dceb57-349e-4ad1-85ee-ca6e1c00a1c1","added_by":"auto","created_at":"2026-05-05 10:05:20","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":327035,"visible":true,"origin":"","legend":"\u003cp\u003eSmall molecule targeting of PA28ab reduces sphere formation in GSCs and HF-NSCs.\u003c/p\u003e\n\u003cp\u003eA)\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp; Schematic representation of MY45B targeting of PA28ab. Created in BioRender.\u003c/p\u003e\n\u003cp\u003eB)\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp; PA28ab expression in FLAG-\u003cem\u003ePSME1\u003c/em\u003e and FLAG-\u003cem\u003ePSME1\u003c/em\u003e-C22A transduced BT67 compared to mCherry and untransduced controls (representative of n=3).\u003c/p\u003e\n\u003cp\u003eC)\u0026nbsp;\u0026nbsp;\u0026nbsp; PA28ab expression in MY45A (negative probe control) and MY45B treated FLAG-\u003cem\u003ePSME1\u003c/em\u003e-OE and FLAG-\u003cem\u003ePSME1\u003c/em\u003e-C22A-OE BT67 (representative of n=2).\u003c/p\u003e\n\u003cp\u003eD)\u0026nbsp;\u0026nbsp;\u0026nbsp; Sphere-forming frequency of \u003cem\u003ePSME1\u003c/em\u003e-OE and FLAG-\u003cem\u003ePSME1\u003c/em\u003e-C22A-OE BT67 treated with MY45A (0.5 mM) or MY45B (0.5 mM) (6 limiting dilutions per condition; bars indicate estimate of sphere-forming frequency +/- upper and lower limits; c\u003csup\u003e2\u003c/sup\u003e; *\u003cem\u003eP\u003c/em\u003e\u0026lt;0.05, **\u003cem\u003eP\u003c/em\u003e\u0026lt;0.01).\u003c/p\u003e\n\u003cp\u003eE)\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp; Cell viability of \u003cem\u003ePSME1\u003c/em\u003e-OE and FLAG-\u003cem\u003ePSME1\u003c/em\u003e-C22A-OE BT67 treated with MY45A (0.5 mM) or MY45B (0.5 mM) (n=4).\u003c/p\u003e\n\u003cp\u003eF)\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp; Sphere-forming frequency of HF-NSCs treated with MY45A (0.5 mM) or MY45B (0.5 mM) (6 limiting dilutions per condition; bars indicate estimate of sphere-forming frequency +/- upper and lower limits; c\u003csup\u003e2\u003c/sup\u003e; ****\u003cem\u003eP\u003c/em\u003e\u0026lt;0.0001).\u003c/p\u003e\n\u003cp\u003eG)\u0026nbsp;\u0026nbsp;\u0026nbsp; Cell viability of HFAs treated with DMSO, MY45A (0.5 mM), or MY45B (0.5 mM) (n=3).\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-8682343/v1/f9fd4b622c0a91704c52ab55.png"},{"id":108494451,"identity":"7e01cfed-7efa-462d-98bd-9f9a798207ff","added_by":"auto","created_at":"2026-05-05 10:05:24","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":846215,"visible":true,"origin":"","legend":"\u003cp\u003eDifferential immunoproteasome, antigen presentation, and tumour microenvironment regulation by PA28a and PA28b in GBM models.\u003c/p\u003e\n\u003cp\u003eA) Schematic representation of standard and immunoproteasome core particle activity monitoring in live cells. Created in BioRender.\u003c/p\u003e\n\u003cp\u003eB) Standard 20S core particle proteasome activity measured over time in BT67 \u003cem\u003ePSME1 \u003c/em\u003eKO and \u003cem\u003ePSME2\u003c/em\u003eKO GSCs (n=3; bars indicate mean +/- SEM; Extra sum-of-squares F-test; logistic growth model).\u003c/p\u003e\n\u003cp\u003eC) Immunoproteasome core particle proteasome activity measured over time in BT67 \u003cem\u003ePSME1 \u003c/em\u003eKO and \u003cem\u003ePSME2\u003c/em\u003eKO GSCs (n=3; bars indicate mean +/- SEM; Extra sum-of-squares F-test; logistic growth model).\u003c/p\u003e\n\u003cp\u003eD) Schematic representation of the immunopeptidomics workflow. Created in BioRender.\u003c/p\u003e\n\u003cp\u003eE) Identified peptide counts from HLA class I ABC immunopeptide elution in AAVS1, \u003cem\u003ePSME1\u003c/em\u003e KO, and \u003cem\u003ePSME2 \u003c/em\u003eKO BT67 (n=3 per condition; Kruskal-Wallis with Dunn multiple comparison; *\u003cem\u003eP\u003c/em\u003e\u0026lt;0.05).\u003c/p\u003e\n\u003cp\u003eF) Number of unique peptides in AAVS1, \u003cem\u003ePSME1\u003c/em\u003e KO, and \u003cem\u003ePSME2\u003c/em\u003e KO BT67 from the three pooled replicates.\u003c/p\u003e\n\u003cp\u003eG) Peptide KEIDKEEHL BLAST E values and percent identity for Melanoma-associated antigen gene D1 (MAGED1) and D4 (MAGED4), with the peptide’s detection in immunopetidomics and RNA expression of the genes in AAVS1, \u003cem\u003ePSME1\u003c/em\u003eKO, and \u003cem\u003ePSME2\u003c/em\u003e KO BT67.\u003c/p\u003e\n\u003cp\u003eH) Schematic representation of a single-cell RNA-seq experiment on GL261 Scramble, \u003cem\u003ePSME1\u003c/em\u003e, and \u003cem\u003ePSME2\u003c/em\u003eKD orthotopic tumours. \u003cem\u003ePSME1\u003c/em\u003e shRNA1 and \u003cem\u003ePSME2\u003c/em\u003e shRNA2 were used to generate orthotopic tumours. Created in BioRender.\u003c/p\u003e\n\u003cp\u003eI) Dimensionality reduction of single-cell RNA-seq on GL261 Scramble, \u003cem\u003ePSME1 \u003c/em\u003eKD, and \u003cem\u003ePSME2\u003c/em\u003e KD orthotopic tumours with annotated cell types (n=5 mice pooled per condition).\u003c/p\u003e\n\u003cp\u003eJ) Frequencies of the identified cell types present in GL261 Scramble, \u003cem\u003ePSME1 \u003c/em\u003eKD, and \u003cem\u003ePSME2\u003c/em\u003e KD orthotopic tumours.\u003c/p\u003e\n\u003cp\u003eK) Expression of \u003cem\u003ePsme1 \u003c/em\u003eand \u003cem\u003ePsme2 \u003c/em\u003ein single cells from different tumour cell populations from GL261 Scramble, \u003cem\u003ePSME1 \u003c/em\u003eKD, and \u003cem\u003ePSME2\u003c/em\u003e KD orthotopic tumours.\u003c/p\u003e\n\u003cp\u003eL) Expression of \u003cem\u003eArg1, Mrc1, \u003c/em\u003eand \u003cem\u003eCcl24 \u003c/em\u003ein single cells from different myeloid cell populations from GL261 Scramble, \u003cem\u003ePSME1 \u003c/em\u003eKD, and \u003cem\u003ePSME2\u003c/em\u003e KD orthotopic tumours.\u003c/p\u003e\n\u003cp\u003eM) Expression of \u003cem\u003eSocs1 \u003c/em\u003eand \u003cem\u003eIl31ra \u003c/em\u003ein single cells from different T cell populations from GL261 Scramble, \u003cem\u003ePSME1 \u003c/em\u003eKD, and \u003cem\u003ePSME2\u003c/em\u003e KD orthotopic tumours.\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-8682343/v1/27cd58c41035c6931b31c5cc.png"},{"id":108493533,"identity":"6fea4103-158b-4938-a58d-7f75ebc4b083","added_by":"auto","created_at":"2026-05-05 10:00:53","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1290360,"visible":true,"origin":"","legend":"\u003cp\u003eInteractome of the PA28ab complex and transcriptional alterations in PA28ab KO/KD GSCs.\u003c/p\u003e\n\u003cp\u003eA) Schematic representation of PA28ab interactome analysis in BT67 FLAG-tagged PA28a and b subunits compared to mCherry controls (n=3 per condition). Created in BioRender.\u003c/p\u003e\n\u003cp\u003eB) Differentially enriched proteins in FLAG-PA28a pull-down compared to mCherry pull-down (top proteins labelled) (n=3 per condition).\u003c/p\u003e\n\u003cp\u003eC) STRING-DB interaction network showing proteins and protein complexes pulled down with both PA28a and b subunits and not pulled down with mCherry.\u003c/p\u003e\n\u003cp\u003eD) Schematic representation of mRNA-seq analysis of PA28ab KO/KD BT189 and BT67 GSCs. Created in BioRender.\u003c/p\u003e\n\u003cp\u003eE) Differentially expressed genes in PA28ab KO/KD GSCs, with the top 10 up- and down-regulated genes labelled (n=3 per condition; Wald test; \u003cem\u003eP\u003c/em\u003eadj\u0026lt;0.05, |Log2FC|\u0026gt;1 cutoff).\u003c/p\u003e\n\u003cp\u003eF) Top 10 altered GSEA pathways in PA28ab KO/KD GSCs compared to controls (up- and down-regulated) (GSEA \u003cem\u003ePadj\u003c/em\u003e\u0026lt;0.05 cutoff).\u003c/p\u003e\n\u003cp\u003eG) GSEA pathway enrichment plot of the Endosomal/Vacuolar pathway in PA28ab KO/KD GSCs compared to controls (upregulated left, downregulated right) (GSEA \u003cem\u003eP\u003c/em\u003eadj=2.448692e-02).\u003c/p\u003e\n\u003cp\u003eH) Immunofluorescent staining of PA28a in HF-NSCs (left) and BT67 (middle), and concanavalin A staining in BT67 (right) (representative of n=2 for HF-NSCs, n=3 for BT67 per condition).\u003c/p\u003e\n\u003cp\u003eI) Schematic representation of proteomics (label-free-quantification (LFQ)) analysis of PA28ab KO BT67 GSCs. Created in BioRender.\u003c/p\u003e\n\u003cp\u003eJ) Differentially expressed proteins in PA28a KO compared to AAVS1 BT67 (n=3 per condition; limma moderated t-test; \u003cem\u003eP\u003c/em\u003e\u0026lt;0.05, |Log2FC|\u0026gt;1 cutoff).\u003c/p\u003e\n\u003cp\u003eK) Differentially expressed proteins in PA28b KO compared to AAVS1 BT67 (n=3 per condition; limma moderated t-test; \u003cem\u003eP\u003c/em\u003e\u0026lt;0.05, |Log2FC|\u0026gt;1 cutoff).\u003c/p\u003e\n\u003cp\u003eL) GSEA enrichment network of differentially enriched pathways and proteins in PA28ab KO BT67 GSCs. Large nodes represent pathways and small nodes represent differentially expressed proteins within the pathways (GSEA FDR \u003cem\u003eq\u003c/em\u003e\u0026lt;0.1 cutoff).\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-8682343/v1/44022b586132094ff13f90a7.png"},{"id":108494666,"identity":"0db904de-3935-44a3-a806-156a724401a2","added_by":"auto","created_at":"2026-05-05 10:06:25","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":648954,"visible":true,"origin":"","legend":"\u003cp\u003ePA28ab interacts with and stabilizes KIF11 in GSCs.\u003c/p\u003e\n\u003cp\u003eA)\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp; Immunoprecipitation of FLAG-PA28a and western blotting for KIF11, PRMT5, PA28b, and FLAG in BT48-FLAG-PA28a-OE (representative of n=3).\u003c/p\u003e\n\u003cp\u003eB)\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp; Proximity ligation assay (PLA) of endogenous PA28a \u0026amp; KIF11 (upper) and PA28a \u0026amp; PRMT5 (lower) in three GSC lines (representative of n=2 per line).\u003c/p\u003e\n\u003cp\u003eC)\u0026nbsp;\u0026nbsp;\u0026nbsp; Proximity ligation assay (PLA) of endogenous PA28a \u0026amp; KIF11 (upper) and PA28a \u0026amp; PRMT5 (lower) in iPSC-NSCs (representative of n=2).\u003c/p\u003e\n\u003cp\u003eD)\u0026nbsp;\u0026nbsp;\u0026nbsp; Quantification of PA28a \u0026amp; KIF11 PLA foci adjacent to each nucleus in GSCs and iPSC-NSCs (n=2 per condition; three ROIs per replicate; ANOVA; ****\u003cem\u003eP\u003c/em\u003e\u0026lt;0.0001)\u003c/p\u003e\n\u003cp\u003eE)\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp; Quantification of PA28a \u0026amp; PRMT5 PLA foci adjacent to each nucleus in GSCs and iPSC-NSCs (n=2 per condition; three ROIs per replicate; ANOVA; ***\u003cem\u003eP\u003c/em\u003e\u0026lt;0.001, ****\u003cem\u003eP\u003c/em\u003e\u0026lt;0.0001)\u003c/p\u003e\n\u003cp\u003eF)\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp; Sphere-forming frequency of PA28ab KO and AAVS1 control BT67s treated with DMSO, GSK591, or Ispinesib (6 limiting dilutions per condition; bars indicate estimate of sphere-forming frequency +/- upper and lower limits; c\u003csup\u003e2\u003c/sup\u003e; **\u003cem\u003eP\u003c/em\u003e\u0026lt;0.01, ***\u003cem\u003eP\u003c/em\u003e\u0026lt;0.001, ****\u003cem\u003eP\u003c/em\u003e\u0026lt;0.0001).\u003c/p\u003e\n\u003cp\u003eG)\u0026nbsp;\u0026nbsp;\u0026nbsp; KIF11, PA28a, PA28b protein levels in cycloheximide (CHX) treated PA28ab KO and AAVS1 control BT67 (representative of n=3).\u003c/p\u003e\n\u003cp\u003eH)\u0026nbsp;\u0026nbsp;\u0026nbsp; Quantification of KIF11 protein levels in cycloheximide (CHX) treated PA28ab KO and AAVS1 control BT67s (n=3; bars represent mean +/- SEM; ANOVA; *\u003cem\u003eP\u003c/em\u003e\u0026lt;0.05).\u003c/p\u003e\n\u003cp\u003eI)\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp; Quantification of KIF11 protein level in KIF11OE PA28ab KO and AAVS1 control BT67s compared to parental untransduced BT67s (n=3, bars represent mean +/- SEM; ANOVA; *\u003cem\u003eP\u003c/em\u003e\u0026lt;0.05, **\u003cem\u003eP\u003c/em\u003e\u0026lt;0.01).\u003c/p\u003e\n\u003cp\u003eJ)\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp; Sphere-forming frequency of KIF11OE PA28ab KO and AAVS1 control BT67s and parental untransduced BT67s normalized to AAVS1 (6 limiting dilutions per condition; bars indicate estimate of sphere-forming frequency +/- upper and lower limits; c\u003csup\u003e2\u003c/sup\u003e \u003cem\u003eP \u003c/em\u003evalues depicted).\u003c/p\u003e\n\u003cp\u003eK)\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp; PA28a, PA28b, and KIF11 protein levels in GL261 transduced with \u003cem\u003ePSME1\u003c/em\u003e, \u003cem\u003ePSME2\u003c/em\u003e, and Scramble shRNAs (representative of n=3).\u003c/p\u003e\n\u003cp\u003eL)\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp; Quantification of KIF11 protein levels in Scramble, \u003cem\u003ePSME1\u003c/em\u003e KDs, and \u003cem\u003ePSME2\u003c/em\u003e KDs normalized to Scramble (n=3; bars represent mean +/- SEM; ANOVA; *\u003cem\u003eP\u003c/em\u003e\u0026lt;0.05, **\u003cem\u003eP\u003c/em\u003e\u0026lt;0.01).\u003c/p\u003e\n\u003cp\u003eM)\u0026nbsp;\u0026nbsp;\u0026nbsp; Expression of \u003cem\u003eKif11 \u003c/em\u003emRNA in Scramble, \u003cem\u003ePSME1\u003c/em\u003e KD, and \u003cem\u003ePSME2\u003c/em\u003e KD GL161 normalized to Scramble and \u003cem\u003eActin\u003c/em\u003e (n=3; bars represent mean +/- SEM; ANOVA; non-significant)\u003c/p\u003e","description":"","filename":"image6.png","url":"https://assets-eu.researchsquare.com/files/rs-8682343/v1/43df3337e4b114deed30d49b.png"},{"id":108497959,"identity":"40c58c31-7a18-406b-815f-e3cb091900b7","added_by":"auto","created_at":"2026-05-05 10:14:01","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4794759,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8682343/v1/bde24913-cadc-4989-b8d2-88480bc928ef.pdf"},{"id":108494438,"identity":"93fde4ef-081a-4c41-9fb6-a3e9cba0adac","added_by":"auto","created_at":"2026-05-05 10:05:13","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":6287057,"visible":true,"origin":"","legend":"Supplemental Figures and Captions","description":"","filename":"Heemskerketal2026supplementalfigures.docx","url":"https://assets-eu.researchsquare.com/files/rs-8682343/v1/70176ebc75496fc596c24056.docx"},{"id":108496061,"identity":"19a3fd4c-dcf3-4883-a1bd-b59320c2a745","added_by":"auto","created_at":"2026-05-05 10:11:14","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":43138936,"visible":true,"origin":"","legend":"Supplemental Tables","description":"","filename":"Supplementaltables.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-8682343/v1/96f1db23ca3a19bad55b7cfd.xlsx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"The proteasome activator subunits PA28α and PA28β have unique molecular roles in glioblastoma","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe proteasome is a dynamic protein complex responsible for protein degradation and amino acid recycling\u003csup\u003e\u003cspan additionalcitationids=\"CR2\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e. Consequently, it regulates critical cellular functions, including the cell cycle, metabolism, protein homeostasis, and apoptosis\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. Structurally, the proteasome consists of standard or immunoproteasome core particle subunits assembled into a ring-like conformation to which various activators can bind at either end\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. The immunoproteasome core particle harbours alternative catalytic subunits induced by inflammatory cytokines and alters peptide production to enhance major histocompatibility complex (MHC) class I antigen presentation\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. Proteasome activators, including the 19S regulatory particle, PA28αβ (\u003cem\u003ePSME1\u003c/em\u003e \u0026amp; \u003cem\u003ePSME2\u003c/em\u003e), PA28γ (\u003cem\u003ePSME3\u003c/em\u003e), and PA200 (\u003cem\u003ePSME4\u003c/em\u003e), modulate proteasome activity and function\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. The core particles can degrade ubiquitinated and non-ubiquitinated proteins independently\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e,\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e. Alternatively, various activators can regulate their respective degradation targets and peptide processing\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. The vast diversity of proteasome holo-complexes underscores the need for a comprehensive understanding of each activator\u0026rsquo;s specific role in different cellular and disease contexts\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eRegulation of cell growth and proliferation, along with many oncogenic and tumour suppressor pathways, and overall proteostasis, makes the proteasome an attractive therapeutic target in cancer\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. Inhibiting the proteasome pathway in cancer has traditionally focused on the proteasome core particle\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. Targeting proteasome activator complexes offers an alternative approach to cancer therapy that could expand treatment options and improve patient survival. A few proteasome activators have been functionally investigated in specific cancers. For example, the proteasome activator PA200 has been identified as a target to enhance the response to immunotherapy in non-small cell lung cancer\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e. Furthermore, PA28γ has been shown to facilitate hepatitis C viral infection and the development of hepatocellular carcinoma\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIn contrast, less is known about the activator PA28αβ (encoded by \u003cem\u003ePSME1\u003c/em\u003e \u0026amp; \u003cem\u003ePSME2\u003c/em\u003e) in cancer. PA28αβ forms a heteroheptameric ring with four α and three β subunits\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e that bind to the proteasome core particle at either end\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. Studies have described crystal structures of PA28αβ bound to either the standard or the immunoproteasome core particle, suggesting it has the capacity to regulate both species\u003csup\u003e\u003cspan additionalcitationids=\"CR12\" citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e. PA28αβ is canonically considered to alter peptide processing for MHC class I antigen presentation\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e,\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e; however, there is conflicting evidence regarding its contribution to global peptide presentation and cytotoxic T cell responses. Knockout of the PA28β subunit in mice resulted in alterations in peptide processing, cytotoxic T cell responses, and immunoproteasome assembly\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. Conversely, in other studies, PA28α and PA28β knockout mice displayed normal T cell responses but altered processing of only specific peptides for antigen presentation\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e,\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e. Therefore, it is unclear whether PA28αβ contributes to peptide processing for antigen presentation on its own or in conjunction with the immunoproteasome core particle, and whether this occurs in all cells. Further exploration of the role of PA28αβ in various cell types and disease contexts is essential to improve our understanding of its overall function.\u003c/p\u003e \u003cp\u003eGlioblastoma (GBM) is the most common primary malignant brain tumour in adults, characterized by extensive heterogeneity\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e,\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e. Treatment options for GBM are limited, with a median overall survival of approximately 15 months\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e. GBM tumours contain malignant cells in many different states, including stem-like cells that harbour cancer stem cell qualities, such as the ability to initiate tumours in mice, self-renew, and undergo multilineage differentiation\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e,\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e. These cells, termed GBM stem cells (GSCs), have been shown to contribute to treatment resistance and can be isolated and cultured in neural stem cell conditions for mechanistic investigations\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e. The functional heterogeneity of the proteasome and its role in driving GBM and GSC growth and stem-like properties are not well understood.\u003c/p\u003e \u003cp\u003eWhile proteasome core particle inhibitors revolutionized treatment strategies for multiple myeloma and mantle-cell lymphoma, comparable efficacy has not been demonstrated in other hematological malignancies or solid tumours\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. The proteasome has been shown to promote GBM growth and GSC phenotypes\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e, but inhibition of the standard proteasome core particle with the brain-penetrant proteasome inhibitor marizomib did not provide a therapeutic benefit in a GBM phase III clinical trial\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e. It was also associated with an increased incidence of nervous system and psychiatric adverse events\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e. These adverse effects may be attributable to the proteasome\u0026rsquo;s vital role in normal cellular function\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e, as well as its noncanonical roles, such as a membrane-embedded core particle within neurons\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e and free 19S regulatory particles that regulate the synaptic proteome\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e. These studies underscore that proteasome subunits may have specialized functions in different cell types and tissues. Further understanding of proteasome biology across cell types and disease states is necessary to identify mechanisms with therapeutic potential.\u003c/p\u003e \u003cp\u003eInvestigating non-catalytic proteasome subunits or activators in GBM may reveal novel and targetable dependencies. Here, to find alternative proteasomal targets in GBM, we investigated proteasome activator expression in tumour tissue and GSCs. We identified PA28αβ as an enriched proteasome activator in GSCs and GBM tumours. Genetic or chemical targeting of PA28αβ decreased GSC self-renewal and extended survival in orthotopic xenograft and syngeneic models. Interestingly, we found a prominent role for the PA28β subunit in regulating immunoproteasome activity and antigen presentation in GSCs. We further uncovered that PA28αβ interacts with the kinesin motor protein KIF11, promoting its stability, primarily through the PA28α subunit. Overall, this study reveals the interaction between PA28αβ and KIF11 as a key regulator of GSCs, identifies PA28αβ as a potential therapeutic target in GBM, and expands our understanding of functional heterogeneity between the PA28αβ subunits.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003ePA28αβ is upregulated in GSCs and GBM tumours\u003c/h2\u003e \u003cp\u003eTo explore alternative strategies for targeting the proteasome in GBM, we analyzed proteasome gene expression utilizing single-cell RNA-seq of GSCs\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e and human NSPCs\u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e. Furthermore, we compared proteasome gene expression in GBM tumours\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e with normal tissue\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA). We integrated single-cell RNA expression data from 29 patient-derived GSCs and 4 NSPCs using Canonical Correlation Analysis and examined differences in proteasome gene expression between GSCs and NSPCs (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB, Supplemental Fig.\u0026nbsp;1A, Supplemental Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). Expression levels of the 20S core particle and 19S regulator particle gene families were generally elevated in GSCs relative to NSPCs (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB). Expression of \u003cem\u003ePSME1\u003c/em\u003e and \u003cem\u003ePSME2\u003c/em\u003e, which encode PA28αβ, was also upregulated in GSCs compared to NSPCs (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB; Pseudobulk Wald test; \u003cem\u003ePSME1\u003c/em\u003e: \u003cem\u003ePadj\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;1.87E-11, \u003cem\u003ePSME2\u003c/em\u003e: \u003cem\u003ePadj\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;3.6E-14). In contrast, \u003cem\u003ePSME3\u003c/em\u003e (PA28γ) and \u003cem\u003ePSME4\u003c/em\u003e (PA200) did not show significant upregulation in GSCs versus NSPCs (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB). Higher expression of the immunoproteasome core particle subunits \u003cem\u003ePSMB8\u003c/em\u003e and \u003cem\u003ePSMB9\u003c/em\u003e was observed in GSCs relative to NSPCs; however, \u003cem\u003ePSMB10\u003c/em\u003e did not show significant differential expression (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB; Pseudobulk Wald test; \u003cem\u003ePSMB8\u003c/em\u003e: \u003cem\u003ePadj\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;3.45E-4, \u003cem\u003ePSMB9\u003c/em\u003e: \u003cem\u003ePadj\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;1.4E-5). We next aggregated the proteasome core particle and activator complex gene families into module scores for comparison between GSCs and NSPCs (Supplemental Fig.\u0026nbsp;1B (single-cell scores)). To facilitate statistical comparison between the 29 GSCs and 4 NSPCs, we pseudo-bulked the expression module scores per sample and found that only the PA28αβ module score and the immunoproteasome module score were significantly elevated in GSCs relative to NSPCs (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eC (pseudobulk score)). This finding is supported by a previous study that compared the proteome of GSCs with neural stem cells (NSCs), which showed enrichment of \u003cem\u003ePSME1\u003c/em\u003e (PA28α) and \u003cem\u003ePSME2\u003c/em\u003e (PA28β) in GSCs\u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo validate these findings in tumours, we examined gene expression of proteasome family members in the Cancer Genome Atlas (TCGA)-GBM dataset (primary GBM: n\u0026thinsp;=\u0026thinsp;372; recurrent GBM: n\u0026thinsp;=\u0026thinsp;14; solid tissue normal: n\u0026thinsp;=\u0026thinsp;5)\u003csup\u003e26\u003c/sup\u003e and compared it to the Genotype Tissue Expression (GTEx) human brain expression dataset (n\u0026thinsp;=\u0026thinsp;300 donors; 2931 tissues)\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e (Supplemental Fig.\u0026nbsp;1C). The expression levels of \u003cem\u003ePSME1\u003c/em\u003e and \u003cem\u003ePSME2\u003c/em\u003e were significantly elevated in primary and recurrent GBM relative to normal tissue in the TCGA cohort, as well as compared to the GTEx normal brain tissue (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eD, E). \u003cem\u003ePSME3\u003c/em\u003e and \u003cem\u003ePSME4\u003c/em\u003e were also significantly upregulated compared with GTEx normal brain but were less enriched than \u003cem\u003ePSME1\u003c/em\u003e and \u003cem\u003ePSME2\u003c/em\u003e relative to solid normal tissue (Supplemental Fig.\u0026nbsp;1C, Supplemental Table S2). As observed in the comparison between GSCs and NSPCs, \u003cem\u003ePSMB8\u003c/em\u003e and \u003cem\u003ePSMB9\u003c/em\u003e exhibited significant enrichment in GBM tumours, while \u003cem\u003ePSMB10\u003c/em\u003e exhibited a reduction in expression in GBM tumours, compared to normal tissue (Supplemental Fig.\u0026nbsp;1C, Supplemental Table S2). Lastly, we confirmed the protein level upregulation of PA28αβ in GBM tumours using the Clinical Proteomic Tumor Analysis Consortium (CPTAC) GBM dataset (n\u0026thinsp;=\u0026thinsp;99, GBM; n\u0026thinsp;=\u0026thinsp;10, normal)\u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eF). Overall, these findings indicate that \u003cem\u003ePSME1\u003c/em\u003e and \u003cem\u003ePSME2\u003c/em\u003e are enriched in GSCs and GBM tumours compared to normal cells or tissues. In contrast, the immunoproteasome genes show variable expression in GSCs and GBM tumours compared to normal tissue. Given the consistent expression pattern of PA28αβ in GSCs and GBM tumours and the lack of prior studies examining its role in GBM and other malignancies, we further explored its role in GSCs.\u003c/p\u003e \u003cp\u003eTo identify functional implications of PA28αβ expression in GSCs, we examined RNA and protein expression of each subunit in our cohorts. As GSCs have been shown to exhibit transcriptional states along a Developmental-to-Injury Response axis\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e, we first investigated the correlation between PA28αβ expression and these states. The PA28αβ module score in GSCs and NSPCs was negatively correlated with the Developmental signature and positively correlated with the Injury Response signature (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eG, H). Next, we assessed the protein expression levels of PA28α and PA28β in a cohort of 26 GSCs and observed variable expression among the samples (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eI, Supplemental Fig.\u0026nbsp;1D). Our prior data included bulk RNA-seq for 25 of these GSCs\u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e, which, upon further analysis, revealed a positive correlation between \u003cem\u003ePSME1\u003c/em\u003e and \u003cem\u003ePSME2\u003c/em\u003e RNA expression (Supplemental Fig.\u0026nbsp;1E); however, there was no correlation between PA28α and β protein expression levels in GSCs (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eJ). PA28α protein expression was correlated with \u003cem\u003ePSME1\u003c/em\u003e RNA expression, but PA28β protein expression was not correlated with \u003cem\u003ePSME2\u003c/em\u003e RNA expression (Supplemental Fig.\u0026nbsp;1F, G). These results suggest that PA28β expression may be post-transcriptionally regulated separately from PA28α in GSCs. PA28α protein expression showed a negative correlation with the Developmental signature and a positive correlation with the Injury Response signature (Supplemental Fig.\u0026nbsp;1H). Meanwhile, PA28β protein expression in GSCs did not correlate with Developmental or Injury Response signatures (Supplemental Fig.\u0026nbsp;1I). Taken together, our results indicate that PA28αβ is enriched in GSCs and GBM tumours compared to other proteasome activators and that its RNA expression correlates with the Injury Response signature in GSCs. Given that the injury response signature has been associated with early GBM development\u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e, this suggests that PA28αβ may play a role in tumour initiation. However, PA28α and PA28β may have distinct protein-level regulatory mechanisms in GSCs.\u003c/p\u003e \u003cp\u003e \u003cb\u003eGenetic targeting of PA28αβ reduces GSC stemness and improves survival\u003c/b\u003e \u003cb\u003ein vivo\u003c/b\u003e.\u003c/p\u003e \u003cp\u003eAs PA28αβ was upregulated in GSCs and GBM tumours, we sought to determine whether it promotes GSC features. First, we examined the expression of PA28αβ in GSCs cultured in stem cell-enriched (epidermal growth factor (EGF)/fibroblast growth factor (FGF) supplemented) and differentiation-promoting conditions (10% fetal bovine serum (FBS)). PA28αβ expression was increased in GSCs cultured in stem cell conditions (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA), suggesting it may be associated with stemness. Next, we genetically targeted \u003cem\u003ePSME1\u003c/em\u003e (PA28α) and \u003cem\u003ePSME2\u003c/em\u003e (PA28β) using CRISPR/Cas9 in two patient-derived GSC lines (BT67 and BT48) (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB, Supplemental Fig.\u0026nbsp;2A). Additionally, we used shRNA to target \u003cem\u003ePSME1\u003c/em\u003e (PA28α) in another GSC line (BT189) (Supplemental Fig.\u0026nbsp;2B). In all three GSCs, knockout (KO) or knockdown (KD) of \u003cem\u003ePSME1\u003c/em\u003e or \u003cem\u003ePSME2\u003c/em\u003e led to a reduction in both PA28α and PA28β protein expression (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB, Supplemental Fig.\u0026nbsp;2A, B). This finding aligns with previous studies utilizing PA28β KO mice or chronic myelogenous leukemia cells\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e,\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e and may be attributed to reciprocal stabilization during complex formation. However, residual expression of PA28β was detected in \u003cem\u003ePSME1\u003c/em\u003e KOs/KDs (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB, Supplemental Fig.\u0026nbsp;2B, C, D), and residual PA28α expression was detected in \u003cem\u003ePSME2\u003c/em\u003e KOs (Supplemental Fig.\u0026nbsp;2C, D). Considering the observation of a truncated PA28β for \u003cem\u003ePSME2\u003c/em\u003e guide RNA 1 in BT67 (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB, Supplemental Fig.\u0026nbsp;2C), most subsequent experiments were conducted using \u003cem\u003ePSME1\u003c/em\u003e guide RNA 1 and \u003cem\u003ePSME2\u003c/em\u003e guide RNA 2. First, we examined the effects of PA28αβ KO/KD on GSC growth, which showed minimal impact on \u003cem\u003ein vitro\u003c/em\u003e growth patterns and doubling time (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eC, D, Supplemental Fig.\u0026nbsp;2E, F). Additionally, cell cycle analysis of BT67 PA28αβ KOs demonstrated a modest yet significant reduction in S-phase cells compared to AAVS1 controls (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eE). We further evaluated sphere formation as a proxy for self-renewal, finding that PA28αβ KO/KD reduced sphere-forming frequency in the GSCs (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eF). Thus, targeting PA28αβ abrogates the self-renewing potential of GSCs.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eNext, to test whether these results translate to promotion of tumour growth \u003cem\u003ein vivo\u003c/em\u003e, we orthotopically xenografted BT67 AAVS1 controls, \u003cem\u003ePSME1\u003c/em\u003e (PA28α) KOs, and \u003cem\u003ePSME2\u003c/em\u003e (PA28β) KOs into SCID mice. Mice xenografted with \u003cem\u003ePSME1\u003c/em\u003e KO and \u003cem\u003ePSME2\u003c/em\u003e KO BT67 GSCs exhibited significantly improved survival relative to controls (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eG). We observed a similar result in BT189 \u003cem\u003ePSME1\u003c/em\u003e KDs compared to Scramble controls (Supplemental Fig.\u0026nbsp;2G). As the canonical function of PA28αβ is to regulate peptide processing for antigen presentation, we also examined its role in a syngeneic GBM model. shRNA-mediated KD of \u003cem\u003ePsme1\u003c/em\u003e or \u003cem\u003ePsme2\u003c/em\u003e in GL261 resulted in a reduction of both PA28α and PA28β protein levels, as was observed in human GSCs (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eH). We engrafted GL261 Scramble, \u003cem\u003ePsme1\u003c/em\u003e shRNA1, and \u003cem\u003ePsme2\u003c/em\u003e shRNA2 into immunocompetent mice and, surprisingly, only the \u003cem\u003ePsme1\u003c/em\u003e KD improved survival in this model (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eI). This result suggests that PA28α and PA28β may have distinct roles in immunoregulation, which could be explained by the observation of residual expression of non-targeted subunits. This may also be linked to the absence of protein-level correlation between PA28α and PA28β in GSCs (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eJ). It has previously been reported that PA28α and PA28β can form homoheptamers, albeit with lower stability, which could have individual functions\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. Together, these results demonstrate that genetic targeting of PA28αβ in GSCs decreases sphere-forming capacity and improves survival in orthotopic xenograft models, whereas the individual subunits may have different effects on immunoregulation.\u003c/p\u003e \u003cp\u003e \u003cb\u003eChemical targeting of PA28αβ abrogates stemness in GSCs.\u003c/b\u003e \u003c/p\u003e \u003cp\u003eTo validate our findings of reduced self-renewal in PA28αβ-targeted GSCs, we used a previously developed chemical probe that disrupts the PA28αβ complex via engagement of cysteine-22 on PA28α\u003csup\u003e32\u003c/sup\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA). We overexpressed (OE) FLAG-tagged PA28α and a FLAG-tagged PA28α with cysteine 22 mutated to alanine (C22A) as a control to ensure on-target inhibition in BT67 (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB) and BT48 (Supplemental Fig.\u0026nbsp;3A). We treated FLAG-PA28α-OE and FLAG-PA28α-C22A-OE GSCs with MY45A (enantiomer negative probe) or MY45B (targeting probe) and observed a reduction in PA28α and PA28β levels in MY45B-treated FLAG-PA28α-OE but not FLAG-PA28α-C22A-OE (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC, Supplemental Fig.\u0026nbsp;3B). Next, we examined the effect of MY45B on sphere formation. Treatment of FLAG-PA28α-OE GSCs with MY45B led to a decrease in sphere-forming frequency compared to MY45A (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eD, Supplemental Fig.\u0026nbsp;3C). The effect was rescued in FLAG-PA28α-C22A-OE GSCs (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eD, Supplemental Fig.\u0026nbsp;3C), indicating that specific disruption of the PA28αβ complex via MY45B engagement of C22 on PA28α reduces sphere formation in GSCs. As was observed with genetic targeting of PA28αβ, treatment of BT67 with MY45B did not affect cell viability (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eE). We also examined the effect of MY45B on human fetal-derived NSCs (HF-NSCs) and found that MY45B reduced sphere formation compared to MY45A (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eF), suggesting the mechanism may be stem-cell-associated. Treatment of human fetal NSC-derived astrocytes (HFA) with MY45B did not alter cell viability (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eG), further supporting this hypothesis. Taken together, these results suggest that targeting PA28αβ in GSCs and NSCs reduces self-renewal.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cb\u003eDifferential immunoproteasome activity, antigen presentation, and tumour microenvironment regulation by PA28α and PA28β in GBM models.\u003c/b\u003e \u003c/p\u003e \u003cp\u003eNext, we sought to determine whether PA28αβ-mediated regulation of stemness and tumour growth was attributable to alterations in standard or immunoproteasome activity. We used fluorogenic activity probes specific to each proteasome core particle to measure their respective activity\u003csup\u003e\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e,\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA). The probes have specificity for the chymotrypsin-like activity associated with the β5 standard and β5i immunoproteasome core particle subunits. In two of three GSC lines, genetic targeting of PA28αβ increased standard proteasome activity (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eB, Supplemental Fig.A, C). In contrast, immunoproteasome activity was decreased in \u003cem\u003ePSME2\u003c/em\u003e KOs but not \u003cem\u003ePSME1\u003c/em\u003e KOs in both BT67 and BT48 (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eC, Supplemental Fig.\u0026nbsp;4B). In BT189, \u003cem\u003ePSME1\u003c/em\u003e KD increased immunoproteasome activity (Supplemental Fig.\u0026nbsp;4D). The alterations in proteasome activity did not correlate with protein expression changes of the catalytic core subunits (Supplemental Fig.\u0026nbsp;4E-G). These findings highlight unique, GSC line-dependent alterations in proteasome activity upon targeting of PA28α or PA28β. Notably, \u003cem\u003ePSME2\u003c/em\u003e KO resulted in a conserved decrease in immunoproteasome activity, suggesting a specialized role for PA28β in immunoproteasome regulation. These results also indicate that the reduction in sphere-forming frequency observed in GSCs did not consistently correlate with a specific change in proteasome activity, implying the possibility of noncanonical functions.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eAs the immunoproteasome core particle, along with PA28αβ, is known to regulate peptide processing for antigen presentation, we profiled the immunopeptidome in \u003cem\u003ePSME1/2\u003c/em\u003e KO BT67s (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eD). First, we examined the total and cell surface expression of HLA class and did not observe any changes in \u003cem\u003ePSME1\u003c/em\u003e KO and \u003cem\u003ePSME2\u003c/em\u003e KOs (Supplemental Fig.\u0026nbsp;5A-C). These data suggest that \u003cem\u003ePSME1\u003c/em\u003e KO or \u003cem\u003ePSME2\u003c/em\u003e KO does not alter HLA class I expression or localization to the cell surface in BT67. For immunopeptidomic profiling, we captured HLA class I ABC antigens using a previously established protocol\u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e on GSCs (Supplemental Fig.\u0026nbsp;5D). Analysis of the identified peptides meeting both identification and MHC binding prediction thresholds revealed similar profiles between AAVS1 controls and \u003cem\u003ePSME1\u003c/em\u003e KOs, with a reduced number of peptides identified in \u003cem\u003ePSME2\u003c/em\u003e KOs (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eE, Supplemental Fig.\u0026nbsp;5E, F, Supplemental Table S3). Approximately 150 unique peptides were identified in AAVS1 and \u003cem\u003ePSME1\u003c/em\u003e KOs compared with fewer than 25 in \u003cem\u003ePSME2\u003c/em\u003e KOs (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eF). Among the identified peptides, we examined the presentation and RNA expression of a peptide that mapped to melanoma-associated antigen family members, which are known tumour antigens\u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u003c/sup\u003e. The peptide KEIDKEEHL, mapping to MAGED1 and MAGED4, was detected exclusively in the immunopeptidomes of AAVS1 and \u003cem\u003ePSME1\u003c/em\u003e KO BT67, despite similar RNA expression across conditions (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eG). We also examined a peptide that mapped to the tropomyosin gene family; another peptide from this family has been identified in GBM tumour tissue immunopeptidomics\u003csup\u003e\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e. The peptide KEAETRAEF was detected in all AAVS1 and \u003cem\u003ePSME1\u003c/em\u003e KO samples, but only in one \u003cem\u003ePSME2\u003c/em\u003e KO sample, despite similar RNA expression (Supplemental Fig.\u0026nbsp;5G). These findings indicate that the resultant decrease in immunoproteasome activity due to \u003cem\u003ePSME2\u003c/em\u003e KO abrogates peptide processing for HLA class I antigen presentation in GSCs. The absence of antigen presentation changes in \u003cem\u003ePSME1\u003c/em\u003e KOs, again, could be due to residual expression of PA28β and/or the ability of either subunit to form homoheptamers\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eTo determine whether alterations in peptide processing for antigen presentation influence the immune microenvironment in tumours, we returned to the GL261 syngeneic GBM model (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eH, I). We profiled Scramble, \u003cem\u003ePsme1\u003c/em\u003e KD, and \u003cem\u003ePsme2\u003c/em\u003e KD tumours using single-cell RNA-seq, which identified the presence of diverse cell types within the tumours (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eH, I, Supplemental Fig.\u0026nbsp;5H-J). We identified cell clusters corresponding to different tumour cell states, immune cell types, and neural cell types (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eI), which were represented across Scramble and KD tumours (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eJ). Differential gene expression analysis across these cell types confirmed downregulation of \u003cem\u003ePsme1\u003c/em\u003e and \u003cem\u003ePsme2\u003c/em\u003e in tumour cell clusters in each respective KD condition (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eK, Supplemental Table S4). Analysis of myeloid cell populations showed a downregulation of M2-like markers, including \u003cem\u003eArg1\u003c/em\u003e, \u003cem\u003eMrc1\u003c/em\u003e, and \u003cem\u003eCcl24\u003c/em\u003e\u003csup\u003e38\u003c/sup\u003e, predominantly in \u003cem\u003ePsme1\u003c/em\u003e KD tumours (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eL). Among the identified T cell populations, \u003cem\u003eSocs1\u003c/em\u003e, a regulator of inflammatory signalling\u003csup\u003e\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u003c/sup\u003e, was downregulated in both \u003cem\u003ePsme1\u003c/em\u003e and \u003cem\u003ePsme2\u003c/em\u003e KDs (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eM). Conversely, \u003cem\u003eIl31ra\u003c/em\u003e, a marker for an activated T cell expression signature in GBM\u003csup\u003e\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e, was upregulated only in T cells from \u003cem\u003ePsme1\u003c/em\u003e KD tumours (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eM). Taken together, these results suggest that \u003cem\u003ePsme1\u003c/em\u003e KD impairs tumour growth in GL261 by reducing immunosuppression in the microenvironment, while \u003cem\u003ePsme2\u003c/em\u003e KD does not have this effect. Decreased immunoproteasome activity and peptide processing for antigen presentation observed in \u003cem\u003ePSME2\u003c/em\u003e KO GSCs may be responsible for the lack of immune activation in \u003cem\u003ePsme2\u003c/em\u003e KD GL261 tumours. The improvement in survival and immune activation in \u003cem\u003ePsme1\u003c/em\u003e KD GL261, which follows the pattern of decreased stemness observed upon targeting PA28αβ, could be linked to a noncanonical function of this complex in GSCs.\u003c/p\u003e \u003cp\u003e \u003cb\u003ePA28αβ interactome and regulation of protein trafficking in GSCs.\u003c/b\u003e \u003c/p\u003e \u003cp\u003eAs a consistent alteration in proteasome activity was not observed upon targeting PA28αβ, we hypothesized that it may function noncanonically in GSCs. Interestingly, free 19S regulatory particles were recently found to bind and regulate the stability of AMPA receptors in neurons\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e. To determine if PA28αβ bound proteins other than the proteasome core particle in GSCs, we performed interactome analysis using FLAG-tagged PA28α and PA28β compared to a mCherry over-expression control pull-down (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA). First, we tested crosslinking followed by FLAG and mCherry pull-down, which showed robust enrichment of FLAG-PA28α, FLAG-PA28β, and mCherry (Supplemental Fig.\u0026nbsp;6A). Next, we performed crosslinking pull-down coupled with mass-spectrometry and label-free quantification (LFQ) on FLAG-PA28α, FLAG-PA28β, and mCherry BT67 (Supplemental Fig.\u0026nbsp;6B, Supplemental Table S5). Notably, proteins identified in the FLAG-PA28α and FLAG-PA28β pull-down, but not the mCherry pull-down were involved in protein trafficking (e.g., KIF11, TUBA1B, TUBA1C, DCTN1, Coatamer complex I proteins), ribosomal components (e.g., RPS29, RPS12), proteasome proteins (\u003cem\u003ePSME1\u003c/em\u003e, \u003cem\u003ePSME2\u003c/em\u003e, PSMA2), and RNA splicing regulators (e.g., PRMT5, WDR77), among other protein complexes (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB, C, Supplemental Fig.\u0026nbsp;6C). Two of the top enriched proteins, PRMT5 and KIF11, have been previously identified as regulators of stemness in GSCs through their roles in RNA splicing regulation\u003csup\u003e\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e and invasive/proliferative properties\u003csup\u003e\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e, respectively. Thus, PA28αβ may regulate GSC stemness by interacting with these key mediators of GSC function.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo determine how targeting PA28αβ disrupted the phenotype of GSCs, we transcriptionally profiled PA28αβ KO/KD GSCs using bulk-RNA-seq (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eD). For BT67, \u003cem\u003ePSME1\u003c/em\u003e KOs and \u003cem\u003ePSME2\u003c/em\u003e KOs clustered closely together and exhibited very few differentially expressed genes (Supplemental Fig.\u0026nbsp;6D, E), suggesting they have a similar transcriptional state, despite the alterations in immunoproteasome activity and peptide processing we observed (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). These results also explain the similar alteration in sphere formation and survival observed for orthotopic xenografts of BT67 \u003cem\u003ePSME1\u003c/em\u003e KOs and \u003cem\u003ePSME2\u003c/em\u003e KOs (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eF, G). Gene set enrichment analysis of differentially expressed genes between PA28αβ-targeted GSCs and controls revealed significant pathway alterations, with the Endosomal/Vacuolar and peptide processing pathways exhibiting the highest enrichment upon PA28αβ targeting (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eE-G, Supplemental Fig.\u0026nbsp;6F, Supplemental Table S6). Neuronal system and transfer RNA processing were identified as the most down-regulated pathways in PA28αβ-targeted GSCs (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eF, Supplemental Fig.\u0026nbsp;6F, Supplemental Table S6). Several of the altered pathways are related to the identified interactors of PA28αβ (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eC). For instance, KIF11 has been shown to regulate the neuronal microtubule array\u003csup\u003e\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e, which could affect genes in the neuronal system. KIF11 also regulates trans-Golgi vesicle trafficking\u003csup\u003e\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e, potentially contributing to alterations in the Endosomal/Vacuolar pathway. Moreover, PRMT5 is known to regulate RNA expression and processing\u003csup\u003e\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e,\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e\u003c/sup\u003e, and to localize to and modulate the trans-Golgi network\u003csup\u003e\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eWe further performed immunofluorescence for PA28α in HF-NSCs and GSCs, demonstrating primary localization to endomembrane systems adjacent to the nucleus (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eH; left and middle panels). To investigate if these membrane systems overlapped with the endoplasmic reticulum and the Golgi apparatus, we stained GSCs with concanavalin A, a lectin that binds to glycosylated proteins\u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e. Indeed, concanavalin A staining was predominantly perinuclear and exhibited a similar distribution to that of PA28α (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eH; right panel). PA28αβ\u0026rsquo;s interaction with essential mediators of GSC stemness, including PRMT5 and KIF11, likely contributes to the observed alterations in transcriptional state and promotion of self-renewal and tumour growth.\u003c/p\u003e \u003cp\u003eTo determine how targeting of PA28αβ alters the proteome of GSCs, we analyzed PA28αβ KO BT67s using mass-spectrometry-based LFQ proteomics (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eI, Supplemental Fig.\u0026nbsp;7A-C, Supplemental Table S7). Quantification of PA28α and PA28β levels in their respective KOs showed significantly lower PA28α levels in \u003cem\u003ePSME1\u003c/em\u003e KO compared to AAVS1 controls and \u003cem\u003ePSME2\u003c/em\u003e KOs (Supplemental Fig.\u0026nbsp;7D). PA28β levels were not significantly different in \u003cem\u003ePSME1\u003c/em\u003e and \u003cem\u003ePSME2\u003c/em\u003e KO cultures, however PA28β was undetectable in all \u003cem\u003ePSME2\u003c/em\u003e KO replicates (Supplemental Fig.\u0026nbsp;7D). These results corroborate our findings of residual expression of PA28α and PA28β when genetically targeting the other subunit in GSCs. Differential protein expression analysis showed unique alterations in \u003cem\u003ePSME1\u003c/em\u003e KO and \u003cem\u003ePSME2\u003c/em\u003e KO GSCs compared to controls (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eJ, K, Supplemental Fig.\u0026nbsp;7E). Pathway enrichment analysis of altered proteins revealed changes in metabolism, extracellular matrix organization, and protein/small molecule trafficking in \u003cem\u003ePSME1\u003c/em\u003e KOs, whereas \u003cem\u003ePSME2\u003c/em\u003e KOs exhibited changes in Golgi-ER transport (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eL, Supplemental Fig.\u0026nbsp;7F-J). \u003cem\u003ePSME1\u003c/em\u003e KOs compared to \u003cem\u003ePSME2\u003c/em\u003e KOs showed alterations related to protein trafficking, Golgi-related pathways, and metabolism (Fig.\u0026nbsp;7K). Taken together, these results suggest that targeting PA28αβ in GSCs modulates protein trafficking machinery among other pathways at both the RNA and protein level and that targeting PA28α or PA28β has unique effects on GSCs at the protein level.\u003c/p\u003e \u003cp\u003e \u003cb\u003ePA28αβ interacts with and stabilizes KIF11 in GSCs.\u003c/b\u003e \u003c/p\u003e \u003cp\u003eAs PRMT5 and KIF11 were identified as top-enriched proteins in our PA28αβ interactome analysis and have been previously associated with GSC stemness, we aimed to understand the function of these interactions. Initially, we used co-immunoprecipitation to validate the interaction of FLAG-tagged PA28α and PA28β with PRMT5 and KIF11 in BT48 (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eA, Supplemental Fig.\u0026nbsp;8A). Subsequently, we used a proximity ligation assay (PLA) to assess the endogenous interaction of KIF11 and PRMT5 with PA28α \u003cem\u003ein situ.\u003c/em\u003e PLA foci for KIF11 and PRMT5 were observed in BT189, BT67, and BT48 (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eB), with no signals detected in IgG and no antibody controls (Supplemental Fig.\u0026nbsp;8B), thereby affirming the interaction of KIF11 and PRMT5 with PA28α in GSCs. Additionally, PLA foci for KIF11/PRMT5 and PA28α were also observed in induced pluripotent stem cell (iPSC) derived NSCs (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eC). Quantification of PLA foci in GSCS and NSCs revealed significantly more interactions between KIF11and PA28α per nucleus in BT189 and BT67 compared to BT48 and NSCs (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eD). Only BT67 harboured an increased number of PLA foci in comparison to the other GSCs and iPSC-NSCs for PRMT5 and PA28α (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eE), suggesting the interaction with KIF11 may be more conserved and therefore, more likely to regulate stemness. BT48 exhibited lower baseline expression of KIF11 and PRMT5 compared to the other GSCs, indicating the relative level of interaction depends on overall expression (Supplemental Fig.\u0026nbsp;8C). Thus, PA28αβ interacts with KIF11 and PRMT5 in GSCs and iPSC-NSCs, suggesting a potential regulatory mechanism for stemness.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo determine the functional implications of PA28αβ\u0026rsquo;s interaction with KIF11 and PRMT5, we combined genetic targeting of PA28αβ with chemical inhibition of each respective protein in GSCs and assessed cell viability along with sphere-forming frequency. In BT67 \u003cem\u003ePSME1\u003c/em\u003e and \u003cem\u003ePSME2\u003c/em\u003e KOs, and BT189 \u003cem\u003ePSME1\u003c/em\u003e KDs, no consistent alteration in viability responses to ispinesib, an inhibitor of KIF11, or GSK591, an inhibitory probe for PRMT5, was observed (Supplemental Fig.\u0026nbsp;8D-G). These findings concur with our previous observation of the absence of growth alterations in PA28αβ genetically targeted GSCs (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). However, sub-IC50 doses of ispinesib and GSK591 further reduced sphere-forming frequency in BT67 \u003cem\u003ePSME1\u003c/em\u003e and \u003cem\u003ePSME2\u003c/em\u003e KOs (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eF). In BT189 \u003cem\u003ePSME1\u003c/em\u003e KDs, ispinesib further decreased sphere formation, whereas GSK591 abrogated the difference between scramble and \u003cem\u003ePSME1\u003c/em\u003e KD (Supplemental Fig.\u0026nbsp;8H). Given the inconsistent effects of PRMT5 inhibition in different GSCs, we examined the downstream phenotypes of PRMT5 in our PA28αβ genetically targeted GSCs. As previously noted, PRMT5 is known to contribute to alternative splicing in GSCs\u003csup\u003e\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e; hence, we performed alternative splicing analysis on RNA-seq from BT67 and BT189 \u003cem\u003ePSME1\u003c/em\u003e/2 KO/KDs (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eD, Supplemental Fig.\u0026nbsp;9). However, no significantly aberrant splicing events were detected between control and \u003cem\u003ePSME1\u003c/em\u003e/2 KO/KD GSCs (Supplemental Fig.\u0026nbsp;9). It remains a possibility that PRMT5 may regulate PA28αβ in GSCs, instead of PA28αβ regulating PRMT5. We focused our further mechanistic studies on KIF11, considering the consistent response observed in the two examined GSCs.\u003c/p\u003e \u003cp\u003eFirst, we examined the basal expression levels of KIF11 in PA28αβ genetically targeted GSCs and did not observe a significant difference relative to controls (Supplemental Fig.\u0026nbsp;8I, J). As free 19S proteasome regulator particles have been shown to stabilize AMPA receptors at synapses\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e, we investigated the impact of PA28αβ genetic targeting on KIF11 stability. We used the cycloheximide chase assay\u003csup\u003e\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e to examine the short-term stability of KIF11 following translation inhibition. \u003cem\u003ePSME1\u003c/em\u003e KO/KD and \u003cem\u003ePSME2\u003c/em\u003e KO resulted in decreased levels of KIF11 after 6 hours of cycloheximide treatment (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eG, H, Supplemental Fig.\u0026nbsp;8K, L), indicating a decrease in KIF11 stability as a consequence of PA28αβ reduction. Next, to see whether we could rescue the effects of the reduction in KIF11 stability, we overexpressed KIF11 in control and PA28αβ genetically targeted GSCs (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eI, Supplemental Fig.\u0026nbsp;8M-O). In BT67, exogenous KIF11 partially rescued the decreased sphere formation induced by \u003cem\u003ePSME1\u003c/em\u003e/\u003cem\u003e2\u003c/em\u003e KO (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eJ). In BT189, KIF11 overexpression fully rescued the reduction in sphere formation as a result of \u003cem\u003ePSME1\u003c/em\u003e KD (Supplemental Fig.\u0026nbsp;8P). Taken together, these results suggest that PA28αβ binds to and stabilizes KIF11, thereby promoting stemness in GSCs. Interestingly, when we examined KIF11 expression in the syngeneic GBM model GL261 \u003cem\u003ePsme1 and Psme2\u003c/em\u003e KDs, we found that KIF11 protein expression was reduced only in the \u003cem\u003ePsme1\u003c/em\u003e KD condition (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eK, L). This was not dependent on a reduction in RNA levels, as \u003cem\u003ePsme1\u003c/em\u003e KD did not reduce KIF11 RNA expression in GL261 (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eM). These results suggest that targeting PA28α alone alters KIF11 stability in GL261. Thus, PA28α primarily stabilizes KIF11 in GSCs and a mouse model of GBM, thereby promoting stemness and tumour growth.\u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this study, we uncover a noncanonical role for PA28αβ in promoting GSC stemness through its interaction with and stabilization of KIF11. We explored proteasome gene expression in glioblastoma (GBM) and GSCs, identifying PA28αβ as the most enriched of the activator subunits. Targeting PA28αβ reduced GSC self-renewal and improved survival in orthotopic xenograft models. We identified a distinct role for the PA28β subunit in immunoproteasome regulation and peptide processing for HLA class I presentation in GSCs. Targeting of PA28αβ in a syngeneic GBM model highlighted a more conserved role for the PA28α subunit in regulating KIF11 stability and tumour growth. Overexpression of KIF11 partially rescued the loss of self-renewal in \u003cem\u003ePSME1/2\u003c/em\u003e (PA28α and PA28β) KO/KD GSCs, suggesting a molecular interplay that promotes GSC function. Here, we provide evidence that PA28αβ is a potential therapeutic target in GSCs and GBM, and that the α and β subunits can function heterogeneously.\u003c/p\u003e \u003cp\u003eOur findings provide several advances in our understanding of PA28αβ and its role within the proteasome pathway in GBM. First, we observed that CRISPR/Cas9-mediated KO or shRNA-mediated KD of \u003cem\u003ePSME1\u003c/em\u003e or \u003cem\u003ePSME2\u003c/em\u003e decreased the protein levels of both PA28α and β, which was confirmed in multiple cell lines, including the mouse GBM model, GL261. This result has been observed in other models and mouse knockout studies\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e,\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e. It appears that PA28α and β mutually stabilize each other at the protein level, as similar effects were not detected at the RNA expression level, which was conserved in human and mouse cells. Conversely, we observed unique phenotypic changes in \u003cem\u003ePSME1\u003c/em\u003e or \u003cem\u003ePSME2\u003c/em\u003e KO/KDs; for example, decreased immunoproteasome activity was observed only in \u003cem\u003ePSME2\u003c/em\u003e KO GSCs, and in BT67, immmunopeptidomic analysis showed that only PA28β disruption appeared to affect peptide processing for MHC class I antigen presentation. Moreover, in GL261 tumours, PA28α individually promoted tumour growth and KIF11 stability. These findings may be explained by the observed residual expression of the opposite subunit in \u003cem\u003ePSME1\u003c/em\u003e or \u003cem\u003ePSME2\u003c/em\u003e KO/KDs and the ability of either subunit to form homoheptamers\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. Indeed, it is peculiar that the protein levels of PA28α and PA28β did not correlate with each other in GSCs and showed distinct correlations with GSC transcriptional states. It is also possible that PA28α and PA28β, or the complex, function independently of the proteasome core particle. In future studies, it will be essential to consider that PA28α and PA28β may have distinct functions in GBM and in other cellular or disease contexts.\u003c/p\u003e \u003cp\u003eSecondly, we provide new evidence regarding the role of PA28αβ in antigen presentation and immunoregulation in GBM. We profiled the immunopeptidome of \u003cem\u003ePSME1\u003c/em\u003e and \u003cem\u003ePSME2\u003c/em\u003e KO GSCs, which identified PA28β as the primary regulator of antigen presentation. This result correlated with decreased immunoproteasome activity only in \u003cem\u003ePSME2\u003c/em\u003e KOs. One limitation of our investigation was the reduced depth of peptide identification, possibly due to low HLA class I expression or limitations with our protocol. Nonetheless, there was a dramatic difference in the number of peptides identified in the controls and \u003cem\u003ePSME1\u003c/em\u003e KOs compared to \u003cem\u003ePSME2\u003c/em\u003e KOs. Coupled with our findings of decreased immunoproteasome activity in \u003cem\u003ePSME2\u003c/em\u003e KOs, these results implicate PA28β as the unique regulator of antigen processing in GSCs. With recent advances in cancer immunotherapies, these findings could have implications for mechanisms of treatment resistance to cancer vaccines, checkpoint blockade, or cell-based therapies.\u003c/p\u003e \u003cp\u003eThirdly, we found that targeting PA28αβ in GSCs had phenotypic consequences for self-renewal. We validated the resultant alteration in self-renewal using a chemical probe targeting the PA28αβ complex. The initial aim of our study was to find an alternative way to target the proteasome pathway in GBM; however, we found that higher PA28αβ expression in GSCs and GBM tumours did not necessarily dictate function. We observed a similar reduction of self-renewal in HF-NSCs upon chemical targeting of PA28αβ. Likewise, KIF11 and PA28αβ also interacted in iPSC-NSCs. It remains uncertain whether this associated phenotype is restricted to stem cells and whether differentiated neural cells retain essential PA28αβ functionality. Collectively, our results shed light on an important role for PA28αβ in GBM, but further investigation in multiple cell types is required to fully understand its function.\u003c/p\u003e \u003cp\u003eLastly, the functional role of PA28αβ in GSCs depended, at least in part, on interaction with and stabilization of KIF11, which may have implications for cancer therapy. Although KIF11 inhibitors demonstrated limited efficacy in clinical trials for solid cancers and melanoma\u003csup\u003e\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e,\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e, there remains interest in targeting KIF11 in other tumours, including GBM\u003csup\u003e\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e\u003c/sup\u003e, and in developing new pharmacological agents\u003csup\u003e\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u003c/sup\u003e. Our results suggest that simultaneous targeting of PA28αβ and KIF11 represents a novel therapeutic approach for GBM. Consequently, developing protein-protein interaction inhibitors to destabilize KIF11 in co-expressing cells may be a promising strategy. The stabilization of KIF11 likely contributes to the downstream endomembrane system pathway alterations, considering its previously identified roles as a mediator of vesicle transport from the Golgi apparatus\u003csup\u003e\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e and as a regulator of the microtubule array in neurons\u003csup\u003e\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e. Several other interactors of PA28αβ were also identified in GSCs. Here, we focused on KIF11 due to its functional relevance; however, other interactors may be important in different contexts, such as interactions unique to the PA28α or β subunits, including RAB10 for PA28α and MYO5B for PA28β. Alternative interactors that influence PA28αβ function could be highly relevant in other cancers or cell types.\u003c/p\u003e \u003cp\u003eIn summary, we identified PA28αβ as a potential therapeutic target in GSCs that mediates stemness through its interaction with and stabilization of KIF11. Given the established relevance of both the proteasome pathway and KIF11 in cancer, this interaction warrants consideration for future therapeutic interventions. Moreover, we uncovered functional heterogeneity between the PA28α and PA28β subunits within GSCs. This suggests a paradigm in which PA28α and PA28β may possess alternative functions that operate independently of their canonical heteroheptameric form.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003eEthics statement\u003c/h2\u003e \u003cp\u003eExperimental procedures were performed in accordance with the University of Calgary Ethics Review Board, Health Research Ethics Board of Alberta (HREBA), and the Animal Care Committee of the University of Calgary.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eCell Culture\u003c/h3\u003e\n\u003cp\u003e Patient-derived GSC lines were established from tumour specimens obtained during surgical resection with written informed consent from patients and approval from the University of Calgary Ethics Review Board, HREBA (HREBA-CC-160762), Calgary, AB, Canada. Metadata, including sex and age, is provided in Supplemental Table S8. GSCs were cultured in stem cell enrichment media supplemented with EGF (20 ng/mL; STEMCELL Tech.), bFGF (20 ng/mL; STEMCELL Tech.), and heparin sulfate (2 \u0026micro;g/mL; STEMCELL Tech.). For stem cell differentiation culture conditions, EGF, bFGF, and heparin sulfate were omitted, and the media was supplemented with 10% fetal bovine serum (FBS). GSCs were authenticated to original parental tumours by short tandem repeat profiling (Calgary Laboratory Services and Department of Pathology and Laboratory Medicine, University of Calgary). Human fetal neural stem cells (HF-NSCs) were derived from normal human brain tissues obtained from 12- to 18-week-old fetuses from therapeutic abortions according to ethical guidelines, including written parental consent, approved by the institutional Review Board of the University of Calgary (REB14-1789). HF-NSCs were generated as previously described\u003csup\u003e\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e. Briefly, fetal brain tissues were minced and treated with DNase (Roche) and trypsin (Invitrogen). The dissociated cells were filtered through a cell strainer (Millipore Sigma) and seeded in stem cell enrichment media supplemented with EGF, FGF, and heparin sulfate. The generated neurospheres were passaged, expanded, and cryopreserved for later use. HF-NSCs were differentiated into human fetal astrocytes (HFAs) in Dulbecco\u0026rsquo;s Modified Eagle Medium (DMEM) supplemented with 10% FBS and 2% penicillin-streptomycin. Human induced pluripotent stem cells (iPSCs) were obtained from ATCC (ATCC-ACS-1020) and cultured in mTeSR\u0026trade;1 media (STEMCELL Tech.) according to the manufacturer\u0026rsquo;s protocol. Neural stem cells were derived from iPSCs (iPSC-NSCs) by dual SMAD inhibition using the STEMdiff\u0026trade; SMADi Neural Induction Kit (STEMCELL Tech.) following the manufacturer\u0026rsquo;s monolayer protocol and subsequently validated for marker expression. HEK293T/17 cells used for lentiviral preparation were obtained from ATCC (ATCC 293T/17) and cultured in DMEM with 10% FBS and 2% penicillin-streptomycin. GL261 mouse glioma cells were a gift from Dr. Stephen Robbins and cultured in DMEM with 10% FBS and 2% penicillin-streptomycin. All cells were routinely confirmed negative for mycoplasma contamination using the Universal Mycoplasma Detection Kit (ATCC) as per the manufacturer\u0026rsquo;s protocol.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eAnalysis of human GSC and NSPC single-cell RNA-seq data\u003c/h2\u003e \u003cp\u003eSingle-cell RNA-seq datasets encompassing 29 human patient-derived GSC cultures\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e and 4 human NSPC isolates\u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e were analyzed using Seurat (v.5)\u003csup\u003e54\u003c/sup\u003e. Dataset integration was performed according to the anchor-based CCA integration method. Differential expression was performed on pseudo-bulked samples between GSCs and NSPCs using the DESeq2 model (v.1.42)\u003csup\u003e\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e\u003c/sup\u003e. Module scores for proteasome subunit gene sets were added to pseudo-bulked data for statistical comparison and to single-cell data for visualization in single-cell plots.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eAnalysis of TCGA-GBM tumour RNA-seq and CPTAC GBM proteomics\u003c/h3\u003e\n\u003cp\u003eTCGA-GBM bulk RNA-seq\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e STAR Counts were accessed using the TCGAbiolinks (v.2.30) R package\u003csup\u003e\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e\u003c/sup\u003e. GTEx human brain tissue bulk RNA-seq\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e counts were accessed using the Recount3 (v.1.12) R package\u003csup\u003e\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e\u003c/sup\u003e. Gene expression counts were normalized, and differential expression analysis was performed using DESeq2 (v.1.42). CPTAC GBM tumour proteomics\u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e data were accessed using the Python CPTAC package (v.1.5.14)\u003csup\u003e\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e\u003c/sup\u003e. PSME1 and PSME2 protein expression were compared between GBM tissue and GTEx normal tissue proteomics, as performed in the GBM CPTAC study.\u003c/p\u003e\n\u003ch3\u003eImmunoblotting\u003c/h3\u003e\n\u003cp\u003eApproximately 1\u0026ndash;2\u0026nbsp;million cells were lysed in RIPA buffer with mild sonication. Isolated protein was quantified using the Bradford assay. 20 \u0026micro;g of protein per well was resolved by SDS-PAGE after denaturation (8 min; 95\u0026deg;C) in 1X Laemmli buffer and transferred to nitrocellulose membranes. Blots were probed overnight at 4\u0026deg;C with antibodies specific to the proteins examined, with β-tubulin or β-actin as loading controls (a complete list of antibodies used in this study is provided in Supplemental Table S9). Blots were washed and incubated with an HRP-conjugated secondary antibody, then imaged using ECL Select and an Amersham Imager 600 (General Electric). Quantification was performed using ImageJ\u003csup\u003e\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e, measuring the mean grey value for bands and normalizing to the loading control.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eAnalysis of in-house GSC bulk RNA-seq\u003c/h2\u003e \u003cp\u003eOur previously published bulk RNA-seq data on GSCs were used to assess correlations between PSME1 and PSME2 RNA and protein expression\u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e. Briefly, RNA was extracted using the AllPrep DNA/RNA Universal kit (QIAGEN). library prep and sequencing were performed by the University of Calgary, Centre for Health Genomics and Informatics (CHGI). Reads were aligned to hg38, and DESeq2 was used for batch correction and count normalization\u003csup\u003e\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e\u003c/sup\u003e. Developmental and Injury Response signatures were derived from the top 100 genes in each signature\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e and computed using ssGSEA\u003csup\u003e\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eCRISPR/Cas9 KOs and shRNA KDs in GSCs\u003c/h2\u003e \u003cp\u003eCRISPR/Cas9 KOs and shRNA KD GSCs were generated as previously described\u003csup\u003e\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e\u003c/sup\u003e. Briefly, GSCs were transduced with a lentiviral Cas9-expressing vector and selected with blasticidin. Subsequently, GSCs were transduced with a gRNA expression vector targeting \u003cem\u003ePSME1, PSME2\u003c/em\u003e, or the AAVS1 safe harbour site as a cut control and subsequently selected with puromycin (gRNA sequences provided in Supplemental Table S10). For human shRNA-mediated knockdown, shRNA expression constructs targeting \u003cem\u003ePSME1\u003c/em\u003e and a scramble control (GeneCopoeia) were used for lentiviral transduction in GSCs as previously described\u003csup\u003e\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e\u003c/sup\u003e. In GL261, shRNA expression vectors targeting \u003cem\u003ePsme1, Psme2\u003c/em\u003e, and a scramble control (GeneCopoeia) were used for lentiviral transduction, following the same protocol as for human GSCs. KO and KD efficiency was examined by Western blotting.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eCell growth assays, cell cycle analysis, and limiting dilution assays\u003c/h2\u003e \u003cp\u003eTo monitor cell growth, GSCs were seeded in a 96-well plate, and cell viability was assessed daily using alamarBlue\u0026trade; (ThermoFisher) according to the manufacturer\u0026rsquo;s protocol. An exponential growth model was fit to compare conditions. For cell cycle analysis, 2.5 x 10\u003csup\u003e5\u003c/sup\u003e cells were incubated with EdU (10 \u0026micro;M) for 2 hours at 37\u0026deg;C. Cells were fixed and stained using the Click-iT\u0026trade; Plus EdU Alexa Fluor\u0026trade; 647 Flow Cytometry Assay kit (ThermoFisher) and counterstained with propidium iodide using FxCycle\u0026trade; PI/RNase Staining solution (Invitrogen). Analysis was performed using a CytoFLEX LX flow cytometer (Beckman Coulter) at the University of Calgary flow cytometry core facility. Data were analyzed using FACSDiva software version 6.1.3 (BD Biosciences). For limiting dilution assays, 512 cells were seeded in 6 wells of a 96-well plate and serially diluted to 256, 128, 64, 32, 16, 8, 4, 2, and 1 cell per well. A well with a sphere exceeding 100 \u0026micro;m in diameter was considered positive at different time points depending on the GSC or NSC line. Plates were scored as previously described using the ELDA web tool\u003csup\u003e\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e\u003c/sup\u003e, and experiments were repeated 2\u0026ndash;3 times with similar results.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eOrthotopic xenografts and GL261 syngeneic tumour model\u003c/h2\u003e \u003cp\u003eAll intracranial engraftments were performed as previously described\u003csup\u003e\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e\u003c/sup\u003e under our animal ethics protocol (AC21-0162) approved by the Animal Care Committee of the University of Calgary. For orthotopic xenografts, human GSCs were xenografted into female CB-17 SCID mice (6\u0026ndash;8 weeks old). For the GL261 syngeneic brain tumour model, cells were implanted into female C57BL/6 mice (6\u0026ndash;8 weeks old). Mice were housed in a Biohazard level 2 facility at 25\u0026deg;C, 45\u0026ndash;55% humidity, and a 6 am to 8 pm light cycle. No sex-based analysis was performed. Mice were anesthetized and stereotactically injected with 100,000 GSCs or GL261 per condition into the right cerebral hemisphere at a depth of 3.5 mm. The experimental and humane endpoints were reached when animals exhibited any of the following signs: ataxia, \u0026gt;\u0026thinsp;15% weight loss, hunching, kyphosis, paresis, lethargy, poor oral intake, or domed heads. Mice were euthanized with a lethal dose of ketamine/xylazine (Ketamine 300\u0026ndash;360 mg/kg \u0026amp; xylazine 30\u0026ndash;40 mg/kg) followed by cervical dislocation.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eExogenous protein expression in GSCs\u003c/h2\u003e \u003cp\u003ecDNAs encoding human \u003cem\u003ePSME1\u003c/em\u003e (BC000352), \u003cem\u003ePSME2\u003c/em\u003e (BC0004368) and \u003cem\u003eKIF11\u003c/em\u003e (BC136474) were acquired from the mammalian gene collection (Horizon Discovery). Site-directed mutagenesis was used to generate the \u003cem\u003ePSME1C22A\u003c/em\u003e mutant cDNA. \u003cem\u003ePSME1\u003c/em\u003e, \u003cem\u003ePSME1C22A\u003c/em\u003e, and \u003cem\u003ePSME2\u003c/em\u003e cDNAs were subcloned into pCDH-CMV-mCherry-EF1-Hygro (a gift from Oskar Laur (Addgene plasmid # 129440)), together with the sequences for a FLAG epitope tag and T2A-eGFP by excising the mCherry using the NheI and NotI restriction sites followed by NEBuilder HiFi assembly cloning (New England Biolabs) to generate pCDH-CMV-(PSME1, PSME1(C22A), PSME2)-FLAG-T2A-eGFP-EF1-Hygro. \u003cem\u003eKIF11\u003c/em\u003e cDNA was also subcloned into the same backbone to generate pCDH-CMV-KIF11-T2A-eGFP-EF1-Hygro. All lentivector constructs generated were packaged into lentiviral vector particles by the HBI Molecular Core Facility using a previously described protocol\u003csup\u003e\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e\u003c/sup\u003e. Briefly, 293FT cells (Thermofisher) were grown on 5 x 15 cm cell culture dishes to ~\u0026thinsp;90% confluency and co-transfected with the relevant transfer vector, psPAX2 and pMD2.G (psPAX2 and pMD2.G were gifts from Didier Trono (Addgene plasmid # 12259, #12260). Between 48\u0026ndash;72 hours post-transfection, the viral vector-containing media was harvested, clarified by centrifugation at 500 x g for 5 min and then 0.45 um filtered. The clarified media was underlaid with 10% sucrose in 1XPBS and centrifuged at 12000 x g for 4 hours. The lentiviral vector pellet was resuspended in ~\u0026thinsp;200 \u0026micro;L sterile 1 XPBS and frozen at -80\u0026deg;C in 20 \u0026micro;L aliquots. The lentivirus titer was determined by real-time qPCR using the kit from Applied Biological Materials, and the functional titer was qualitatively confirmed by transducing HEK293 cells on 12-well plates with serial dilutions of lentivirus and monitoring eGFP fluorescence after 48 hours. GSCs were transduced with the cDNA-expressing vectors or the original mCherry construct and selected with hygromycin B (Roche).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003ePA28αβ chemical probe\u003c/h2\u003e \u003cp\u003eMY45-B was previously published and validated as targeting cysteine 22 on PA28α\u003csup\u003e32\u003c/sup\u003e. For protein expression analysis in response to chemical treatment, GSCs were treated with MY45-A (0.5 \u0026micro;M \u0026ndash; enantiomer control probe), MY45-B (0.5 \u0026micro;M), or DMSO for 48 hours, and pellets were collected for western blotting. For LDAs, GSCs were treated with one dose of MY45-A (0.5 \u0026micro;M) or MY45-B (0.5 \u0026micro;M) upon seeding, and plates were scored as described above at 21 days. Viability was assessed at 21 days using alamarBlue\u0026trade; (ThermoFisher) according to the manufacturer\u0026rsquo;s protocol in wells seeded with 512 cells. HF-NSC LDAs were treated with MY45-A (0.5 \u0026micro;M) or MY45-B (0.5 \u0026micro;M) and scored as above at 21 days. HFAs (2,500 per well) in DMEM (10% FBS) were treated with MY45-A (0.5 \u0026micro;M), MY45-B (0.5 \u0026micro;M), or DMSO and assessed for viability at 7 days post-treatment.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eProteasome activity assays\u003c/h2\u003e \u003cp\u003eThe peptide-peptoid hybrid activity probes for the standard and immunoproteasome core particles have been previously described\u003csup\u003e\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e,\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e, with probes synthesized for this study (Pepmic). GSCs (80,000) were seeded in duplicate technical replicate wells in a black 96-well plate with a clear bottom and incubated overnight. Proteasome activity probes (31 \u0026micro;M) were added to GSC and blank wells containing only GSC growth media. A first fluorescence reading (Excitation-485; Emission-535) was taken immediately after probe addition, and subsequent readings were acquired every 5 min for a total of 95 min. Each reading was normalized to the probe-specific blank by subtracting the blank value for each time point. Curves of normalized data were fit with a logistic growth model for comparison.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eImmunopeptidomics\u003c/h2\u003e \u003cp\u003eBasal and cell surface expression of HLA-ABC were assessed using western blotting and EZ-Link\u0026trade; Sulfo-NHS-SS-Biotin (ThermoFisher) as previously described\u003csup\u003e\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e\u003c/sup\u003e. Briefly, cultured GSCs were rinsed twice with cold PBS, then incubated with sulfo-NHS-SS-biotin (1 mg/mL) for 10 minutes at 4\u0026deg;C with gentle rotation. The reaction was stopped using 100 mM glycine and 25 mM Tris-HCl, pH 7.4, for 5 minutes, followed by three washes with cold PBS. GSCs were subsequently lysed in RIPA buffer, and the input was set aside for western blotting. The remainder of the lysate was incubated with Dynabeads\u0026trade; MyOne\u0026trade; Streptavidin C1 Magnetic Beads (Invitrogen) overnight at 4\u0026deg;C to capture cell surface biotinylated proteins. Magnetic beads were separated from the supernatant and washed 4 times in cold PBS. Samples were resolved by SDS-PAGE and immunoblotted with HLA-ABC antibody (Abcam-ab70328) as described above.\u003c/p\u003e \u003cp\u003eImmunopeptidomic analysis of BT67 AAVS1, PSME1KO, and PSME2KO was performed using a modified, previously described protocol\u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e. Briefly, for each sample, 1x10\u003csup\u003e8\u003c/sup\u003e GSCs were collected, washed twice in PBS, and snap-frozen in liquid nitrogen for later use. CNBr-activated Sepharose beads (80 mg) (Cytiva) were rehydrated and coupled with Ultra-LEAF\u0026trade; Purified anti-human HLA-A,B,C Antibody (1.5 mg) (W6/32 clone \u0026ndash; BioLegend). Antibody-coupled beads were washed and blocked with glycine (0.2 M). The quality of antibody coupling was assessed by Coomassie staining of SDS-PAGE-resolved beads and wash fractions. GSC pellets were thawed rapidly and lysed in 1 ml lysis buffer (0.5% NP-40, 50 mM Tris, pH 8.0, 150 mM NaCl, 1 mM EDTA, with protease inhibitor (Roche)) using a tissue homogenizer to break up the pellet, followed by 1 h incubation with gentle rotation at 4\u0026deg;C. Lysates were centrifuged at 20000 x g for 20 min at 4\u0026deg;C, and the supernatant was collected. Lysates were incubated overnight at 4\u0026deg;C with gentle rotation with HLA-ABC antibody-coupled beads. HLA-ABC-bound beads were then washed as described in ref. 32, modified for washing in Micro Bio-Spin\u0026trade; Chromatography Columns (Bio-Rad) with 1 mL of each buffer twice, with 30 sec centrifuges at 500 x g for buffer flow-through. HLA-ABC complexes were eluted with 200 \u0026micro;L of 1% Pierce\u0026trade; Trifluoroacetic Acid, LC-MS Grade (ThermoFisher) by 500 x g centrifugation for 1 minute, repeated twice for a total eluate volume of 400 \u0026micro;L. HLA-ABC-bound beads, supernatant, and eluted HLA-ABC complexes were assessed by western blot for quality of pull-down and elution in one sample. Eluted HLA-ABC complexes were shipped to Biogenity (Denmark) for analysis.\u003c/p\u003e \u003cp\u003eEluate was up-concentrated on 3 kDa filters (Millipore) by wetting the filters with 400 \u0026micro;L 1% TFA and spun at 14000 x g. Then the samples were loaded and spun at 14000 x g with approximately 50 \u0026micro;L volume remaining. The flow-through was desalted using the SOLA\u0026micro; SPE plate (ThermoFisher). Between applications, the solvents were spun through by centrifugation at 150 rpm. For each sample, the filters were activated with 200 \u0026micro;L of 100% Methanol, then 200 \u0026micro;L of 80% Acetonitrile containing 0.1% formic acid. The filters were subsequently equilibrated twice with 200 \u0026micro;L of 1% TFA, 3% Acetonitrile, after which the sample was loaded. After washing the tips twice with 200 \u0026micro;L of 0.1% formic acid, the peptides were eluted into clean 0.5 mL Eppendorf tubes using 40% Acetonitrile, 0.1% formic acid. The eluted peptides were concentrated in an Eppendorf Speedvac. The concentrated peptides were then reconstituted in 12 \u0026micro;L 2% acetonitrile, 1% trifluoroacetic acid for analysis. The peptides were centrifuged at 18,000 x g for 10 minutes to remove any residual particulate matter. Samples were analyzed using a Nanodrop to determine peptide concentration. For each sample, 500 ng of peptides were loaded onto a 2 cm C18 trap column (ThermoFisher), connected in-line to a 15 cm C18 reverse-phase EasySpray analytical column (ThermoFisher) using 100% Buffer A (0.1% formic acid in water) at 750 bar, on the EasyLC 1200 HPLC system (ThermoFisher), with the column oven set to 30\u0026deg;C. Peptides were eluted using a 70-minute gradient at a flow rate of 250 nL/min. The gradient began with a transition from 6% to 23% Buffer B (80% acetonitrile, 0.1% formic acid) and then increased to 38% Buffer B over 12 minutes. The final step involved ramping up to 95% Buffer B over 8 minutes, holding at this concentration for 7 minutes.\u003c/p\u003e \u003cp\u003eThe Orbitrap Exploris 480 (Thermo Fisher Scientific) was run in a DD-MS2 top 28 method. Full MS spectra were collected at a resolution of 60,000, with an AGC target of 300% or maximum injection time set to \u0026lsquo;auto\u0026rsquo; and a scan range of 375\u0026ndash;1500 m/z. The MS2 spectra were obtained at a resolution of 15.000, with an AGC target set to 75% or maximum injection time set to \u0026lsquo;auto\u0026rsquo;, a normalized collision energy of 28, and an intensity threshold of 1.0x10\u003csup\u003e4\u003c/sup\u003e. Dynamic exclusion was set to 60 s, and ions with a charge state\u0026thinsp;\u0026lt;\u0026thinsp;2 or unknown were excluded. MS performance was evaluated by running complex cell lysate quality control standards.\u003c/p\u003e \u003cp\u003eAnalysis of the immunopeptidome data was performed using a previously described workflow\u003csup\u003e\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e\u003c/sup\u003e. HLA typing was performed using HLA-LA\u003csup\u003e67\u003c/sup\u003e on previously published whole-genome sequencing data from BT67\u003csup\u003e30\u003c/sup\u003e. Raw data were converted to mzML using msConvert\u003csup\u003e\u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e\u003c/sup\u003e, and the database search was performed using Comet (2025.02)\u003csup\u003e\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e\u003c/sup\u003e with the human proteome (UP000005640)\u003csup\u003e\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e\u003c/sup\u003e. Search parameters for MS data included cut everywhere as the enzyme, a maximum missed cleavage of 2, and a mass error tolerance of 10 ppm. A fragment ion mass error tolerance of 1.5 Da was used. Search results were used as input for MHCvalidator\u003csup\u003e\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e\u003c/sup\u003e with the default configuration, using only MHCflurry for binding and processing score prediction\u003csup\u003e\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e\u003c/sup\u003e. Peptides with a database search q-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 and a presentation percentile\u0026thinsp;\u0026gt;\u0026thinsp;95% were used for sample analysis.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003eGL261 single-cell RNA-seq\u003c/h2\u003e \u003cp\u003eScramble, \u003cem\u003ePsme1\u003c/em\u003e KD, and \u003cem\u003ePsme2\u003c/em\u003e KD GL261 (100,000) were implanted into the right cerebral hemisphere of 5 C57BL/6 mice per condition, as described above. The mice were euthanized 3 weeks post-engraftment, and tumours were dissected out of whole brains. Tumours were pooled for each condition, followed by manual dissociation in PBS containing 0.04% BSA and filtering through a 100 \u0026micro;m cell strainer (Millipore Sigma) to clear debris. Cells were washed 3 times in PBS containing 0.04% BSA and resuspended for viability check and counting. Each sample was confirmed to have \u0026gt;\u0026thinsp;80% viability, and a 7000-cell capture target was attempted based on the final cell number. Cells were transferred to the Hotchkiss Brain Institute NeuroOmics Core at the University of Calgary and subjected to single-cell transcriptomic profiling by the GEM-X Universal 3\u0026rsquo; Gene Expression v4 (10X Genomics) following the manufacturer\u0026rsquo;s protocol. cDNA yields and libraries were confirmed to meet quality control standards for each sample. cDNA libraries were sequenced on a NextSeq P3 xleap (Illumina) 100-cycle, yielding 400 Mb per sample and ~\u0026thinsp;20,000 reads/cell at the CHGI.\u003c/p\u003e \u003cp\u003eGL261 single-cell RNA-seq data was analyzed using the standard Seurat (v.5) pipeline\u003csup\u003e\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e\u003c/sup\u003e. Briefly, BCL files were converted to fastq files using bcl2fastq (v.2.20\u0026ndash;Illumina), followed by count matrix generation with cellranger (v.8.0.1\u0026ndash;10X Genomics) using GRCm39 for alignment. Single-cell matrices were merged into a single Seurat object and integrated using the anchor-based CCA method. Cells were filtered for mitochondrial gene expression\u0026thinsp;\u0026lt;\u0026thinsp;5%. Conserved markers (top 50) for the identified clusters were used to classify cell types with GPTCelltype\u003csup\u003e\u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e72\u003c/span\u003e\u003c/sup\u003e. GPTCelltype-identified cell types were manually confirmed and refined based on marker expression and clustering. Differential expression was assessed between conditions for all cell clusters using FindMarkers with default settings, and \u003cem\u003eP\u003c/em\u003eadj\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered significant.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003ePA28αβ interactome\u003c/h2\u003e \u003cp\u003eFLAG-tagged PA28α (PSME1) and PA28β (PSME2), as well as mCherry, were expressed in BT67 GSCs as described above. Anti-FLAG\u0026rarr; M2 magnetic beads (Millipore Sigma) (20 \u0026micro;L per sample) were used for the FLAG-PA28α and FLAG-PA28β pull-down. For the mCherry pull-down, anti-mCherry antibody (Abcam-ab183628) was cross-linked to Dynabeads\u0026trade; Protein G (ThermoFisher) as per the manufacturer\u0026rsquo;s protocol with BS\u003csup\u003e3\u003c/sup\u003e (ThermoFisher) (20 \u0026micro;L per sample). Protein-protein crosslinking in BT67 GSCs was performed with Pierce\u0026trade; DSP (ThermoFisher) (0.1 mM) in PBS for 30 min with gentle rotation, followed by two 5 min washes in cold PBS. Cells were then lysed by sonication in 50 mM Tris, pH 8.0, 150 mM NaCl, 1 mM EDTA, with protease inhibitor (Roche). 500 \u0026micro;g of protein was incubated with anti-FLAG or anti-mCherry beads for 2 hours at room temperature with gentle rotation. Beads were subsequently washed 4 times in lysis buffer, resuspended in 1X Laemmli buffer, and proteins were eluted by denaturation (8 min; 95\u0026deg;C). The procedure was tested for immunoprecipitation efficiency prior to mass spectrometric analysis using SDS-PAGE and Western blotting. Samples were resolved through the stacking portion of an SDS-polyacrylamide gel and ran into a 12% SDS-polyacrylamide gel until the entire protein entered the lower gel. The gels were stained with Coomassie Blue to confirm total protein content in the lower gel. Total protein within the gel was then excised with a scalpel into an Eppendorf tube with 1 mL milliQ water. Samples were transferred to the Southern Alberta Mass Spectrometry Facility at the University of Calgary for analysis.\u003c/p\u003e \u003cp\u003eFor trypsin digestion, gel bands were sliced into small pieces of about 1 mm\u0026sup3; using a scalpel blade. Gel plugs were washed 3 times for 15 min in 50 mM ammonium bicarbonate/acetonitrile (50:50, v/v). Gel plugs were briefly rinsed in 100% acetonitrile and incubated for 20 min in fresh 100% acetonitrile. After being air dried, proteins were reduced with dithiothreitol (DTT; 10 mM in 100 mM ammonium bicarbonate) for 30 min at 56\u0026deg;C and alkylated with iodoacetamide (IAA; 50 mM in 100 mM ammonium bicarbonate) for 30 mins in the dark at room temperature. Gel plugs were washed with 50 mM ammonium bicarbonate/acetonitrile (50:50, v/v) for 15 min and dehydrated with acetonitrile as described above. Gel plugs were then rehydrated with a trypsin solution (Promega; 0.02 \u0026micro;g/ul in 40mM ammonium bicarbonate 10% acetonitrile) and incubated for 2 h on ice. The excess of trypsin solution was removed, and trypsin buffer (40 mM ammonium bicarbonate 10% acetonitrile) was added to cover the gel pieces. Trypsin digestion was performed for 16 h at 37\u0026deg;C. After the digestion, the tryptic peptides were transferred into a new tube containing 5 \u0026micro;l of extraction solution (acetonitrile/water/ 10% trifluoroacetic acid (60:30:10, v/v). Tryptic peptides were extracted twice from the gel plugs by incubation in the extraction solution and vortexing for 10 min. The extracted peptides were combined into the same Eppendorf tube. Samples were then lyophilized and resuspended in 15 \u0026micro;l of 1% formic acid in water.\u003c/p\u003e \u003cp\u003eFor LC-MS/MS, tryptic peptides were analyzed on an Orbitrap Fusion Lumos Tribrid (ThermoFisher) operated with Xcalibur (v.4.4.16.14) and coupled to an Easy-nLC 1200 system. Tryptic peptides were loaded onto a C18 trap (75 um x 2 cm; Acclaim PepMap 100, P/N 164946; ThermoFisher) at a flow rate of 2 \u0026micro;L/min of solvent A (0.1% formic acid in LC-MS grade water). Peptides were eluted using a 120 min gradient from 5 to 40% (5% to 28% in 22 min followed by an increase to 40% B in 3 min) of solvent B (0.1% formic acid in 80% LC-MS grade acetonitrile) at a flow rate of 0.3 \u0026micro;L/min and separated on a C18 analytical column (75 um x 50 cm; PepMap RSLC C18; P/N ES803; ThermoFisher). Peptides were then electrosprayed at 2.1 kV into the ion transfer tube (300\u0026deg;C) of the Orbitrap Lumos operating in positive mode. The Orbitrap first performed a full MS scan at a resolution of 120,000 FWHM to detect the precursor ion having a m/z between 375 and 1575 and a\u0026thinsp;+\u0026thinsp;2 to +\u0026thinsp;7 charge. The Orbitrap AGC and the maximum injection time were set at 4x10\u003csup\u003e5\u003c/sup\u003e and 50 ms, respectively. The Orbitrap was operated using the top speed mode with a 3 sec cycle time for precursor selection. The most intense precursor ions presenting a peptidic isotopic profile and having an intensity threshold of at least 5,000 were isolated using the quadrupole and fragmented with HCD (30% collision energy) in the ion routing multipole. The fragment ions (MS2) were analyzed in the ion trap at a rapid scan rate. The AGC and the maximum injection time were set at 1x10\u003csup\u003e4\u003c/sup\u003e and 35 ms, respectively, for the ion trap. Dynamic exclusion was enabled for 30 sec to avoid of the acquisition of same precursor ion having a similar m/z (plus or minus 10 ppm).\u003c/p\u003e \u003cp\u003eFor data analysis, Lumos raw data files were converted to MGF files using RawConverter (v.1.1.0.18\u0026ndash;Scripps) operating in a data-dependent mode. Monoisotopic precursors having a charge state of +\u0026thinsp;2 to +\u0026thinsp;7 were selected for conversion. MGF files were used to search the human proteome database using Mascot algorithm (v.2.7\u0026ndash;Matrix Sciences). Search parameters for MS data included trypsin as the enzyme, a maximum number of missed cleavages of 1, a peptide charge of 2 or higher, cysteine carbamidomethylation as fixed modification, methionine oxidation as variable modification and a mass error tolerance of 10 ppm. A mass error tolerance of 0.6 Da was selected for the fragment ions. Only peptides with scores exceeding 95% confidence were retained for further analysis. The Mascot data files were imported into Scaffold (v.5.3.2\u0026ndash;Proteome Software Inc) for comparison of samples based on mass spectral counting.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003eBulk RNA-seq in PA28αβ genetically targeted GSCs\u003c/h2\u003e \u003cp\u003eRNA was isolated using the Amersham RNAspin Mini Kit (Cytiva) following the manufacturer\u0026rsquo;s protocol. Library prep and sequencing were performed by the CHGI-UCalgary. Reads were aligned to GRCh38 using Rsubread\u003csup\u003e\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e\u003c/sup\u003e. Normalization, clustering, and differential expression were performed with DESeq2\u003csup\u003e55\u003c/sup\u003e. A \u003cem\u003eP\u003c/em\u003eadj\u0026thinsp;\u0026lt;\u0026thinsp;0.05 and Log2FC\u0026gt;|1| was considered significant. Pathway analysis was performed using gene set enrichment analysis within the ReactomePA R package (v.1.46.0)\u003csup\u003e74\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003eImmunofluorescence and concanavalin A staining\u003c/h2\u003e \u003cp\u003eCells were plated in 8- or 16-well chamber slides coated with poly-ornithine and laminin. After 72 h, cells were fixed with 4% paraformaldehyde. For immunofluorescent staining of PA28α, cells were blocked/permeabilized using goat serum (10%), BSA (2%), and Triton-X100 (0.3%) in PBS. Cells were incubated with primary antibodies overnight, then incubated with fluorophore-conjugated secondary antibody (Invitrogen) for 1 hour. Cells were counterstained with DAPI and mounted with coverslips. Images were acquired with a Leica SP8 confocal microscope. For Concanavalin A staining, cells were washed in Hank\u0026rsquo;s Balanced Salt Solution (HBSS) 3 times, followed by incubation with 50 \u0026micro;g/mL of Concanavalin A \u0026ndash; Alexa Fluor 647 (Invitrogen) in HBSS for 30 minutes. Cells were counterstained with mounting media containing DAPI and imaged as above. These images were acquired at a single z-plane.\u003c/p\u003e \u003cdiv id=\"Sec23\" class=\"Section3\"\u003e \u003ch2\u003eProteomics\u003c/h2\u003e \u003cp\u003eCell pellets were frozen in liquid nitrogen and stored at -80\u0026deg;C. Cells were lysed by sonication in 50 mM Tris, pH 8.0, 150 mM NaCl, 1 mM EDTA, with protease inhibitor (Roche). Lysates were centrifuged at 15,000 x g for 10 min at 4\u0026deg;C, and the supernatant was collected. Protein (200 \u0026micro;g per sample) was precipitated using trichloroacetic acid (20%). The protein precipitate was washed 3 times with cold acetone, then air-dried for 2 min. Protein was resuspended in Urea 8M in 100 mM Tris-HCl pH 8, followed by centrifugation at 10000 x g for 1 min. Dithiothreitol (10 mM) was added to the protein suspension, and the mixture was incubated for 30 min at 37\u0026deg;C. Protein was transferred into Microcon YM-30 (Millipore) and centrifuged for 15 min at 14000 x g. Membranes were washed with 200 \u0026micro;L Urea 8M in 100 mM Tris-HCl pH 8, and centrifuged for 15 min at 14000 x g. Membranes were incubated with Iodoacetamide (100 \u0026micro;L \u0026ndash; 50 mM in Urea 8M in 100 mM Tris-HCl pH 8) at room temp for 20 min in the dark, then centrifuged for 10 min at 14000 x g. Membranes were washed with Urea 8M in 100 mM Tris-HCl pH 8 and centrifuged for 15 min at 14000 x g 3 times. Membranes were then washed with Ammonium Bicarbonate (50 mM) and centrifuged for 15 min at 14000 x g 3 times. Proteins on the membrane were digested with trypsin (Promega) (0.22 \u0026micro;g/\u0026micro;L in Ammonium Bicarbonate (50 mM)) overnight at 37\u0026deg;C in the dark. Membranes were then centrifuged for 5 sec at 100 x g to collect condensation and incubated at room temp for 10 min. Membranes were transferred to a new collection tube and centrifuged for 10 min at 14000 x g. Ammonium Bicarbonate (50 mM \u0026ndash; 40 \u0026micro;L) was added to the membranes and incubated for 3 min followed by centrifugation for 10 min at 14000 x g. The previous step was repeated, and then NaCl (0.5 M \u0026ndash; 40 \u0026micro;L) was added to the membrane and incubated for 3 min followed by centrifugation for 15 min at 14000 x g. Digested peptide flow-through was then transferred to a new tube and freeze-dried. Dried peptides were resuspended in Formic acid (1%) for desalting on UPLC.\u003c/p\u003e \u003cp\u003ePeptides were analyzed on an Orbitrap Fusion Lumos Tribrid mass spectrometer (ThermoFisher) operated with Xcalibur (v.4.4.16.14) and coupled to an Easy-nLC 1200 system (ThermoFisher). Peptides were loaded onto a C18 trap (75 um x 2 cm; Acclaim PepMap 100, ThermoFisher) at a flow rate of 2 \u0026micro;L/min of solvent A (0.1% formic acid in LC-MS grade water). Peptides were eluted using a 120 min gradient from 5 to 40% (5% to 28% in 105 min followed by an increase to 40% B in 15 min) of solvent B (0.1% formic acid in 80% LC-MS grade acetonitrile) at a flow rate of 0.3 \u0026micro;L/min and separated on a C18 analytical column (75 um x 50 cm; PepMap RSLC C18; ThermoFisher). Peptides were then electrosprayed using 2.1 kV voltage into the ion transfer tube (300\u0026deg;C) of the Orbitrap Lumos operating in positive mode. The Orbitrap first performed a full MS scan at a resolution of 120000 FWHM to detect precursor ions with \u003cem\u003em\u003c/em\u003e/\u003cem\u003ez\u003c/em\u003e between 375 and 1575, and charge states\u0026thinsp;+\u0026thinsp;2 to +\u0026thinsp;7. The Orbitrap AGC (Auto Gain Control) and the maximum injection time were set to 4x10\u003csup\u003e5\u003c/sup\u003e and 50 ms, respectively. The Orbitrap was operated in top-speed mode with a 3 sec cycle time for precursor selection. The most intense precursor ions exhibiting a peptidic isotopic profile and an intensity threshold of at least 5000 were isolated using the quadrupole and fragmented with HCD (30% collision energy) in the ion-routing multipole. The fragment ions (MS\u003csup\u003e2\u003c/sup\u003e) were analyzed in the ion trap at a rapid scan rate. The AGC and the maximum injection time were set at 1x10\u003csup\u003e4\u003c/sup\u003e and 35 ms, respectively, for the ion trap. Dynamic exclusion was enabled for 45 sec to prevent the acquisition of the same precursor ion with a similar m/z (within \u0026plusmn;\u0026thinsp;10 ppm).\u003c/p\u003e \u003cp\u003eRaw data files were converted to Mascot Generic Format using RawConverter (Scripps-v.1.1.0.18) operating in a data-dependent mode. Monoisotopic precursors with charge states of +\u0026thinsp;2 to +\u0026thinsp;7 were selected for conversion. The database search was performed using the human proteome with the Mascot algorithm (Matrix Sciences; version 2.7). Search parameters for MS data included trypsin as the enzyme, a maximum of 1 missed cleavage, a peptide charge of 2 or higher, cysteine carbamidomethylation as a fixed modification, methionine oxidation as a variable modification, and a mass error tolerance of 10 ppm. A mass error tolerance of 0.6 Da was selected for the fragment ions. Only peptides identified with a confidence score\u0026thinsp;\u0026gt;\u0026thinsp;95% were retained for further analysis. The Mascot files were imported into Scaffold (Proteome Software, Inc.) for comparison of different samples using Top 3 Total Ion Chromatograms (TICs). Proteins with 3 or greater peptides identified were analyzed for cross-sample comparison using limma\u003csup\u003e\u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e75\u003c/span\u003e\u003c/sup\u003e with a cutoff of \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 and |LogFC|\u0026gt;1. Pathway analysis was performed using GSEA within the ReactomePA R package\u003csup\u003e\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section2\"\u003e \u003ch2\u003eCo-Immunoprecipitation\u003c/h2\u003e \u003cp\u003eGSCs were collected 10 days post-seeding. Protein-protein crosslinking was performed with Pierce\u0026trade; DSP (ThermoFisher) (0.1 mM) in PBS for 30 min with gentle rotation, followed by two 5 min washes in cold PBS. Cells were then lysed by sonication in 50 mM Tris, pH 8.0, 150 mM NaCl, 1 mM EDTA, 0.5% NP-40 with protease inhibitor (Roche). 500 \u0026micro;g of protein was incubated with anti-FLAG or IgG isotype beads overnight at 4\u0026deg;C with gentle rotation. Beads were subsequently washed 4 times in lysis buffer, resuspended in 1X Laemmli buffer, and proteins were eluted by denaturation (8 min; 95\u0026deg;C). Immunoprecipitated samples, alongside input (5%) and immunodepleted fractions (5%), were resolved by SDS-PAGE, transferred to nitrocellulose membranes, and immunoblotted for specific proteins.\u003c/p\u003e \u003cdiv id=\"Sec25\" class=\"Section3\"\u003e \u003ch2\u003eProximity Ligation Assay\u003c/h2\u003e \u003cp\u003eCells were plated as described above for immunofluorescence and fixed with 4% PFA. Cells were washed 3 times with PBS, then permeabilized with PBS, Triton-X100 (0.3%) for 10 min, followed by 1 h block in Duolink\u003csup\u003e\u0026rarr;\u003c/sup\u003e In Situ PLA\u003csup\u003eⓇ\u003c/sup\u003e 1X Blocking Solution (Millipore Sigma). Cells were then incubated with primary antibody mixtures, IgG controls, or no-antibody controls for 2 h at 37\u0026deg;C, 5% CO\u003csub\u003e2\u003c/sub\u003e. Cells were then subjected to the Duolink\u003csup\u003eⓇ\u003c/sup\u003e In Situ PLA\u003csup\u003eⓇ\u003c/sup\u003e anti-rabbit PLUS and anti-Mouse MINUS staining (Millipore Sigma), followed by ligation and amplification according to the manufacturer\u0026rsquo;s guidelines. Images were acquired on a Leica SP8 confocal microscope with 20 z-planes collapsed into a maximum intensity projection using ImageJ. Three regions of interest were captured for quantification per condition. PLA foci were quantified using CellProfiler (v.4.2.8)\u003csup\u003e76\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec26\" class=\"Section3\"\u003e \u003ch2\u003eGSC viability dose response to KIF11 and PRMT5 inhibitors and LDAs\u003c/h2\u003e \u003cp\u003eCells were plated at a 2,500 cell/well density in a 96-well plate and incubated overnight. Escalating doses of the KIF11 inhibitor ispinisib (Selleck) or the PRMT5 inhibitory probe GSK591 (Structural Genomics Consortium) were added to 3 technical replicate wells per dose. Viability was assessed 13 days after inhibitor addition using alamarBlue\u0026trade; (ThermoFisher) according to the manufacturer\u0026rsquo;s protocol. Viability readings were averaged across technical replicates and normalized to vehicle controls. For LDAs, sub-IC50 doses were determined by viability readings. Ispinisib (BT67: 2 nM; BT189: 1 nM) or GSK591 (BT67: 100 nM; BT189: 100 nM) were added to LDA plates 1 d after seeding and scored as described above. Scores were normalized to AAVS1 controls for BT67 and Scramble controls for BT189.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec27\" class=\"Section3\"\u003e \u003ch2\u003eAlternative splicing analysis\u003c/h2\u003e \u003cp\u003eAlternative splicing analysis was performed using the SpliceWiz R package (v.1.4.1)\u003csup\u003e77\u003c/sup\u003e using default settings. Bulk RNA-seq on PA28αβ genetically targeted GSCs, aligned as described above, were used as input for SpliceWiz.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec28\" class=\"Section2\"\u003e \u003ch2\u003eCycloheximide chase assay\u003c/h2\u003e \u003cp\u003eGSCs (1 x 10\u003csup\u003e6\u003c/sup\u003e) were seeded in 6-well plates and incubated for 72 hours. DMSO or cycloheximide (Selleck) (300 \u0026micro;g/mL) was added to the GSC wells, followed by immediate collection of the 0 h timepoint. Cell pellets were washed twice in PBS, flash frozen in liquid nitrogen, and stored at -80\u0026deg;C. Additional pellets were collected at 3 and 6 h time points. The DMSO vehicle pellets were collected at 6 h. Cell pellets were lysed and immunoblotted as described above. Quantification was performed as described above for immunoblotting.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec29\" class=\"Section2\"\u003e \u003ch2\u003eQuantitative Polymerase Chain Reaction (qPCR)\u003c/h2\u003e \u003cp\u003eCell pellets were frozen in liquid nitrogen and stored at -80\u0026deg;C. RNA was extracted using the RNAspin Mini Kit (Cytiva) as per the manufacturer\u0026rsquo;s protocol and quantified by nanodrop. cDNA was synthesized from RNA (500 ng) using qScript\u0026trade; cDNA SuperMix (Quantabio) as per the manufacturer\u0026rsquo;s protocol. qPCR was performed with cDNA in duplicate technical replicate wells with primers specific to mouse \u003cem\u003eKif11\u003c/em\u003e and \u003cem\u003eActb\u003c/em\u003e (Supplemental Table S9) and FastStart Essential DNA Green Master (Roche). Cycling conditions were 95\u0026deg;C (3 min), 95\u0026deg;C (15 sec), 60\u0026deg;C (45 sec with signal capture), return to step 2 and repeat 50X, followed by a melting curve.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec30\" class=\"Section2\"\u003e \u003ch2\u003eStatistical Analysis\u003c/h2\u003e \u003cp\u003eAll experiments were performed with three replicates unless otherwise stated. Statistical analyses were performed using GraphPad Prism Version 10 or R Version 4.3. Specific tests used are described in the figure captions or directly in the text. Data are reported as mean \u0026plusmn; SEM unless otherwise stated, and \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant. Median survival in mouse tumour experiments was estimated using the Kaplan-Meier method, and the Log-Rank Mantel-Cox test was used to assess statistical significance.\u003c/p\u003e \u003c/div\u003e "},{"header":"Declarations","content":"\u003cdiv id=\"Sec31\" class=\"Section2\"\u003e \u003ch2\u003eData availability:\u003c/h2\u003e \u003cp\u003eSingle-cell RNA-seq data for GSCs is available from the Broad Institute Single-Cell Portal (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://singlecell.broadinstitute.org/single_cell/study/SCP503\u003c/span\u003e\u003cspan address=\"https://singlecell.broadinstitute.org/single_cell/study/SCP503\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). Single-cell RNA-seq for NSPCs is available through the NCBI BioProject database (PRJNA798712). TCGA-GBM RNA-seq is available through the TCGA-Biolinks R package. GTEx normal tissue RNA-seq is available through the recount3 R package. CPTAC-GBM proteomics data can be accessed through the CPTAC Python package. RNA-seq for our in-house cohort of GSCs are available at the European Genome-phenome Archive (EGAS00001002709). Immunopeptidomic data and interactome data have been deposited in the ProteomeExchange Database (#######(Available to reviewer/editor upon request)). RNA sequencing data on \u003cem\u003ePSME1/2\u003c/em\u003e KO and KD GSCs generated for this project have been deposited in SRA (####### (Available to reviewer/editor upon request)). GL261 single-cell RNA-seq has been deposited at the Broad Institute Single-Cell Portal (##### (Available to reviewer/editor upon request)). All other data and code used for analysis are available upon request.\u003c/p\u003e \u003c/div\u003e\u003cp\u003e \u003ch2\u003eCompeting interests:\u003c/h2\u003e \u003cp\u003eNone.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eAuthor Contributions:\u003c/h2\u003e \u003cp\u003eConceptualization: K.M.H, R.K.B, S.A, H.A.L, S.W; Methodology: K.M.H, R.K.B, S.A, X.H, F.H, G.D, R.H, O.C, B.M, B.F.C, H.A.L; Formal Analysis and Visualization: K.M.H, R.K.B, F.H, G.D, X.H, O.C; Supervision: H.A.L., S.W; Writing: K.M.H; Review and Editing: S.A, R.K.B, H.A.L, S.W; Funding Acquisition: H.A.L, S.W.\u003c/p\u003e\u003ch2\u003eAcknowledgements:\u003c/h2\u003e \u003cp\u003eThis research project was supported by a grant from the Canadian Institutes of Health Research #153246 to H.A.L. and S.W. We would like to thank Dr. Darci J. Trader for providing advice regarding proteasome activity probes. We thank Dr. Wee Yong\u0026rsquo;s lab at the University of Calgary for providing human fetal brain tissue. We thank Drs. Alisha Poole and Stephen Robbins for providing GL261 glioma cells. We thank Dr. Yiping Liu at The Flow Cytometry Core Facility (University of Calgary) for cell cycle analysis. We thank Dr. Frank Visser at the HBI Molecular Core Facility (University of Calgary) for molecular cloning and lentiviral preparation. We thank Dr. Charlotte D\u0026rsquo;Mello at the HBI Neuro Omics Core (University of Calgary) for the single-cell library preparation. We thank the Centre for Health Genomics and Informatics (University of Calgary) for performing library preparation and sequencing. We thank the Southern Alberta Mass-Spectrometry Facility (University of Calgary) for running the interactome samples. We thank the HBI-AMP (University of Calgary) for use of microscopes. We thank Biogenity (Denmark) for running the immunopeptidomics samples. We thank Dr. Danielle A. Bozek, Emilie Cutts, Rory Mulloy, Dr. Sorana Morrissy, and Dr. Marco Gallo for scientific or editorial input.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eRousseau, A., Bertolotti, A.: Regulation of proteasome assembly and activity in health and disease. Nat. Rev. Mol. Cell. Biol. \u003cb\u003e19\u003c/b\u003e, 697\u0026ndash;712 (2018)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFabre, B., et al.: Deciphering preferential interactions within supramolecular protein complexes: the proteasome case. Mol. Syst. Biol. \u003cb\u003e11\u003c/b\u003e, 771 (2015)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSuraweera, A., M\u0026uuml;nch, C., Hanssum, A., Bertolotti, A.: Failure of Amino Acid Homeostasis Causes Cell Death following Proteasome Inhibition. Mol. Cell. \u003cb\u003e48\u003c/b\u003e, 242\u0026ndash;253 (2012)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDe La Pe\u0026ntilde;a, A.H., Goodall, E.A., Gates, S.N., Lander, G.C., Martin, A.: Substrate-engaged 26 \u003cem\u003eS\u003c/em\u003e proteasome structures reveal mechanisms for ATP-hydrolysis\u0026ndash;driven translocation. Science. \u003cb\u003e362\u003c/b\u003e, eaav0725 (2018)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSahu, I., et al.: The 20S as a stand-alone proteasome in cells can degrade the ubiquitin tag. Nat. Commun. \u003cb\u003e12\u003c/b\u003e, 6173 (2021)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRamachandran, K.V., Margolis, S.S.: A mammalian nervous-system-specific plasma membrane proteasome complex that modulates neuronal function. Nat. Struct. Mol. Biol. \u003cb\u003e24\u003c/b\u003e, 419\u0026ndash;430 (2017)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJavitt, A., et al.: The proteasome regulator PSME4 modulates proteasome activity and antigen diversity to abrogate antitumor immunity in NSCLC. Nat. Cancer. \u003cb\u003e4\u003c/b\u003e, 629\u0026ndash;647 (2023)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eManasanch, E.E., Orlowski, R.Z.: Proteasome inhibitors in cancer therapy. Nat. Rev. Clin. Oncol. \u003cb\u003e14\u003c/b\u003e, 417\u0026ndash;433 (2017)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMoriishi, K., et al.: Critical role of PA28γ in hepatitis C virus-associated steatogenesis and hepatocarcinogenesis. \u003cem\u003eProc. Natl. Acad. Sci.\u003c/em\u003e 104, 1661\u0026ndash;1666 (2007)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuber, E.M., Groll, M.: The Mammalian Proteasome Activator PA28 Forms an Asymmetric α4β3 Complex. Structure. \u003cb\u003e25\u003c/b\u003e, 1473\u0026ndash;1480e3 (2017)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen, J., et al.: Cryo-EM of mammalian PA28αβ-iCP immunoproteasome reveals a distinct mechanism of proteasome activation by PA28αβ. Nat. Commun. \u003cb\u003e12\u003c/b\u003e, 739 (2021)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao, J., et al.: Structural insights into the human PA28\u0026ndash;20S proteasome enabled by efficient tagging and purification of endogenous proteins. \u003cem\u003eProc. Natl. Acad. Sci.\u003c/em\u003e 119, e2207200119 (2022)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLesne, J., et al.: Conformational maps of human 20S proteasomes reveal PA28- and immuno-dependent inter-ring crosstalks. Nat. Commun. \u003cb\u003e11\u003c/b\u003e, 6140 (2020)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRaule, M., et al.: PA28αβ Reduces Size and Increases Hydrophilicity of 20S Immunoproteasome Peptide Products. Chem. Biol. \u003cb\u003e21\u003c/b\u003e, 470\u0026ndash;480 (2014)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePreckel, T., et al.: Impaired Immunoproteasome Assembly and Immune Responses in \u003cem\u003ePA28\u003c/em\u003e. Mice Sci. \u003cb\u003e286\u003c/b\u003e, 2162\u0026ndash;2165 (1999)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDe Graaf, N., et al.: PA28 and the proteasome immunosubunits play a central and independent role in the production of MHC class I-binding peptides in vivo. Eur. J. Immunol. \u003cb\u003e41\u003c/b\u003e, 926\u0026ndash;935 (2011)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMurata, S., et al.: Immunoproteasome assembly and antigen presentation in mice lacking both PA28alpha and PA28beta. EMBO J. \u003cb\u003e20\u003c/b\u003e, 5898\u0026ndash;5907 (2001)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNomura, M., et al.: The multilayered transcriptional architecture of glioblastoma ecosystems. Nat. Genet. \u003cb\u003e57\u003c/b\u003e, 1155\u0026ndash;1167 (2025)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSloan, A.R., Silver, D.J., Kint, S., Gallo, M., Lathia, J.D.: Cancer stem cell hypothesis 2.0 in glioblastoma: Where are we now and where are we going? Neuro-Oncol. \u003cb\u003e26\u003c/b\u003e, 785\u0026ndash;795 (2024)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchaff, L.R., Mellinghoff, I.K.: Glioblastoma and Other Primary Brain Malignancies in Adults: A Review. JAMA. \u003cb\u003e329\u003c/b\u003e, 574 (2023)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDi, K., et al.: Marizomib activity as a single agent in malignant gliomas: ability to cross the blood-brain barrier. Neuro-Oncol. \u003cb\u003e18\u003c/b\u003e, 840\u0026ndash;848 (2016)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRoth, P., et al.: Marizomib for patients with newly diagnosed glioblastoma: A randomized phase 3 trial. Neuro-Oncol. \u003cb\u003e26\u003c/b\u003e, 1670\u0026ndash;1682 (2024)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSun, C., et al.: An abundance of free regulatory (19 \u003cem\u003eS\u003c/em\u003e) proteasome particles regulates neuronal synapses. Science. \u003cb\u003e380\u003c/b\u003e, eadf2018 (2023)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRichards, L.M., et al.: Gradient of Developmental and Injury Response transcriptional states defines functional vulnerabilities underpinning glioblastoma heterogeneity. Nat. Cancer. \u003cb\u003e2\u003c/b\u003e, 157\u0026ndash;173 (2021)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu, D.D., et al.: Purification and characterization of human neural stem and progenitor cells. Cell. \u003cb\u003e186\u003c/b\u003e, 1179\u0026ndash;1194e15 (2023)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVerhaak, R.G.W., et al.: Integrated Genomic Analysis Identifies Clinically Relevant Subtypes of Glioblastoma Characterized by Abnormalities in PDGFRA, IDH1, EGFR, and NF1. Cancer Cell. \u003cb\u003e17\u003c/b\u003e, 98\u0026ndash;110 (2010)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThe GTEx Consortium: The GTEx Consortium atlas of genetic regulatory effects across human tissues. Science. \u003cb\u003e369\u003c/b\u003e, 1318\u0026ndash;1330 (2020)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThirant, C., et al.: Differential Proteomic Analysis of Human Glioblastoma and Neural Stem Cells Reveals HDGF as a Novel Angiogenic Secreted Factor. STEM CELLS. \u003cb\u003e30\u003c/b\u003e, 845\u0026ndash;853 (2012)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, L.-B., et al.: Proteogenomic and metabolomic characterization of human glioblastoma. Cancer Cell. \u003cb\u003e39\u003c/b\u003e, 509\u0026ndash;528e20 (2021)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShen, Y., et al.: Comprehensive genomic profiling of glioblastoma tumors, BTICs, and xenografts reveals stability and adaptation to growth environments. \u003cem\u003eProc. Natl. Acad. Sci.\u003c/em\u003e 116, 19098\u0026ndash;19108 (2019)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHamed, A.A., et al.: Gliomagenesis mimics an injury response orchestrated by neural crest-like cells. Nature. \u003cb\u003e638\u003c/b\u003e, 499\u0026ndash;509 (2025)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLazear, M.R., et al.: Proteomic discovery of chemical probes that perturb protein complexes in human cells. Mol. Cell. \u003cb\u003e83\u003c/b\u003e, 1725\u0026ndash;1742e12 (2023)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZerfas, B.L., Trader, D.J.: Monitoring the Immunoproteasome in Live Cells Using an Activity-Based Peptide\u0026ndash;Peptoid Hybrid Probe. J. Am. Chem. Soc. \u003cb\u003e141\u003c/b\u003e, 5252\u0026ndash;5260 (2019)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZerfas, B.L., Coleman, R.A., Salazar-Chaparro, A.F., Macatangay, N.J., Trader, D.J.: Fluorescent Probes with Unnatural Amino Acids to Monitor Proteasome Activity in Real-Time. ACS Chem. Biol. \u003cb\u003e15\u003c/b\u003e, 2588\u0026ndash;2596 (2020)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSirois, I., Isabelle, M., Duquette, J.D., Saab, F., Caron, E., Immunopeptidomics: Isolation of Mouse and Human MHC Class I- and II-Associated Peptides for Mass Spectrometry Analysis. J. Vis. Exp. \u003cb\u003e63052\u003c/b\u003e (2021). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3791/63052\u003c/span\u003e\u003cspan address=\"10.3791/63052\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXiao, J.: Biological functions of melanoma-associated antigens. World J. Gastroenterol. \u003cb\u003e10\u003c/b\u003e, 1849 (2004)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShraibman, B., et al.: Identification of Tumor Antigens Among the HLA Peptidomes of Glioblastoma Tumors and Plasma. Mol. Cell. Proteom. \u003cb\u003e18\u003c/b\u003e, 1255\u0026ndash;1268 (2019)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee, S.H., et al.: M2-like, dermal macrophages are maintained via IL-4/CCL24\u0026ndash;mediated cooperative interaction with eosinophils in cutaneous leishmaniasis. Sci. Immunol. \u003cb\u003e5\u003c/b\u003e, eaaz4415 (2020)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIlangumaran, S., Bobbala, D., Ramanathan, S.: SOCS1: Regulator of T Cells in Autoimmunity and Cancer. In: Yoshimura, A. vol (ed.) Emerging Concepts Targeting Immune Checkpoints in Cancer and Autoimmunity, vol. 410, pp. 159\u0026ndash;189. Springer International Publishing, Cham (2017)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSkadborg, S.K., et al.: Nivolumab Reaches Brain Lesions in Patients with Recurrent Glioblastoma and Induces T-cell Activity and Upregulation of Checkpoint Pathways. Cancer Immunol. Res. \u003cb\u003e12\u003c/b\u003e, 1202\u0026ndash;1220 (2024)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSachamitr, P., et al.: PRMT5 inhibition disrupts splicing and stemness in glioblastoma. Nat. Commun. \u003cb\u003e12\u003c/b\u003e, 979 (2021)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVenere, M., et al.: The mitotic kinesin KIF11 is a driver of invasion, proliferation, and self-renewal in glioblastoma. Sci. Transl Med. \u003cb\u003e7\u003c/b\u003e, (2015)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMyers, K.A., Baas, P.W.: Kinesin-5 regulates the growth of the axon by acting as a brake on its microtubule array. J. Cell. Biol. \u003cb\u003e178\u003c/b\u003e, 1081\u0026ndash;1091 (2007)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWakana, Y., et al.: Kinesin-5/Eg5 is important for transport of CARTS from the trans-Golgi network to the cell surface. J. Cell. Biol. \u003cb\u003e202\u003c/b\u003e, 241\u0026ndash;250 (2013)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao, Y.: SMN and symmetric arginine dimethylation of RNA polymerase II C-terminal domain control termination. Nature. \u003cb\u003e529\u003c/b\u003e, 48\u0026ndash;53 (2016)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou, Z., et al.: PRMT5 regulates Golgi apparatus structure through methylation of the golgin GM130. Cell. Res. \u003cb\u003e20\u003c/b\u003e, 1023\u0026ndash;1033 (2010)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMandal, D.K., Brewer, C.F.: Differences in the binding affinities of dimeric concanavalin A (including acetyl and succinyl derivatives) and tetrameric concanavalin A with large oligomannose-type glycopeptides. Biochemistry. \u003cb\u003e32\u003c/b\u003e, 5116\u0026ndash;5120 (1993)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKao, S.-H., et al.: Analysis of Protein Stability by the Cycloheximide Chase Assay. BIO-Protoc \u003cb\u003e5\u003c/b\u003e, (2015)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBlagden, S.P., et al.: A phase I trial of ispinesib, a kinesin spindle protein inhibitor, with docetaxel in patients with advanced solid tumours. Br. J. Cancer. \u003cb\u003e98\u003c/b\u003e, 894\u0026ndash;899 (2008)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee, C.W., et al.: A phase II study of ispinesib (SB-715992) in patients with metastatic or recurrent malignant melanoma: a National Cancer Institute of Canada Clinical Trials Group trial. Invest. New. Drugs. \u003cb\u003e26\u003c/b\u003e, 249\u0026ndash;255 (2008)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCheng, Y.L., et al.: Multiplexed single-cell lineage tracing of mitotic kinesin inhibitor resistance in glioblastoma. Cell. Rep. \u003cb\u003e43\u003c/b\u003e, 114139 (2024)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGarcia-Saez, I., Skoufias, D.A.: Eg5 targeting agents: From new anti-mitotic based inhibitor discovery to cancer therapy and resistance. Biochem. Pharmacol. \u003cb\u003e184\u003c/b\u003e, 114364 (2021)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChojnacki, A., Weiss, S.: Production of neurons, astrocytes and oligodendrocytes from mammalian CNS stem cells. Nat. Protoc. \u003cb\u003e3\u003c/b\u003e, 935\u0026ndash;940 (2008)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHao, Y., et al.: Dictionary learning for integrative, multimodal and scalable single-cell analysis. Nat. Biotechnol. \u003cb\u003e42\u003c/b\u003e, 293\u0026ndash;304 (2024)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLove, M.I., Huber, W., Anders, S.: Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. \u003cb\u003e15\u003c/b\u003e, 550 (2014)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eColaprico, A., et al.: TCGAbiolinks: an R/Bioconductor package for integrative analysis of TCGA data. Nucleic Acids Res. \u003cb\u003e44\u003c/b\u003e, e71\u0026ndash;e71 (2016)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWilks, C., et al.: recount3: summaries and queries for large-scale RNA-seq expression and splicing. Genome Biol. \u003cb\u003e22\u003c/b\u003e, 323 (2021)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLindgren, C.M., et al.: Simplified and Unified Access to Cancer Proteogenomic Data. J. Proteome Res. \u003cb\u003e20\u003c/b\u003e, 1902\u0026ndash;1910 (2021)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchneider, C.A., Rasband, W.S., Eliceiri, K.: W. NIH Image to ImageJ: 25 years of image analysis. Nat. Methods. \u003cb\u003e9\u003c/b\u003e, 671\u0026ndash;675 (2012)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarbie, D.A., et al.: Systematic RNA interference reveals that oncogenic KRAS-driven cancers require TBK1. Nature. \u003cb\u003e462\u003c/b\u003e, 108\u0026ndash;112 (2009)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMacLeod, G., et al.: Genome-Wide CRISPR-Cas9 Screens Expose Genetic Vulnerabilities and Mechanisms of Temozolomide Sensitivity in Glioblastoma Stem Cells. Cell. Rep. \u003cb\u003e27\u003c/b\u003e, 971\u0026ndash;986e9 (2019)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu, Y., Smyth, G.K., ELDA: Extreme limiting dilution analysis for comparing depleted and enriched populations in stem cell and other assays. J. Immunol. Methods. \u003cb\u003e347\u003c/b\u003e, 70\u0026ndash;78 (2009)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCusulin, C., et al.: Precursor States of Brain Tumor Initiating Cell Lines Are Predictive of Survival in Xenografts and Associated with Glioblastoma Subtypes. Stem Cell. Rep. \u003cb\u003e5\u003c/b\u003e, 1\u0026ndash;9 (2015)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJiang, W., et al.: An optimized method for high-titer lentivirus preparations without ultracentrifugation. Sci. Rep. \u003cb\u003e5\u003c/b\u003e, 13875 (2015)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTaylor, K.R., et al.: Glioma synapses recruit mechanisms of adaptive plasticity. Nature. \u003cb\u003e623\u003c/b\u003e, 366\u0026ndash;374 (2023)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKovalchik, K.A., et al.: Machine learning-enhanced immunopeptidomics applied to T-cell epitope discovery for COVID-19 vaccines. Nat. Commun. \u003cb\u003e15\u003c/b\u003e, 10316 (2024)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDilthey, A.T., et al.: HLA*LA\u0026mdash;HLA typing from linearly projected graph alignments. Bioinformatics. \u003cb\u003e35\u003c/b\u003e, 4394\u0026ndash;4396 (2019)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAdusumilli, R., Mallick, P.: Data Conversion with ProteoWizard msConvert. in \u003cem\u003eProteomics\u003c/em\u003e (eds Comai, L., Katz, J. E. \u0026amp; Mallick, P.) vol. 1550 339\u0026ndash;368Springer New York, New York, NY, (2017)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEng, J.K., Jahan, T.A., Hoopmann, M.R., Comet: An open-source MS / MS sequence database search tool. PROTEOMICS. \u003cb\u003e13\u003c/b\u003e, 22\u0026ndash;24 (2013)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThe UniProt Consortium: UniProt: the Universal Protein Knowledgebase in 2025. Nucleic Acids Res. \u003cb\u003e53\u003c/b\u003e, D609\u0026ndash;D617 (2025)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eO\u0026rsquo;Donnell, T.J., Rubinsteyn, A., Laserson, U.: MHCflurry 2.0: Improved Pan-Allele Prediction of MHC Class I-Presented Peptides by Incorporating Antigen Processing. Cell. Syst. \u003cb\u003e11\u003c/b\u003e, 42\u0026ndash;48e7 (2020)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHou, W., Ji, Z.: Assessing GPT-4 for cell type annotation in single-cell RNA-seq analysis. Nat. Methods. \u003cb\u003e21\u003c/b\u003e, 1462\u0026ndash;1465 (2024)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiao, Y., Smyth, G.K., Shi, W.: The R package Rsubread is easier, faster, cheaper and better for alignment and quantification of RNA sequencing reads. Nucleic Acids Res. \u003cb\u003e47\u003c/b\u003e, e47\u0026ndash;e47 (2019)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYu, G., He, Q.-Y.: ReactomePA: an R/Bioconductor package for reactome pathway analysis and visualization. Mol. Biosyst. \u003cb\u003e12\u003c/b\u003e, 477\u0026ndash;479 (2016)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRitchie, M.E., et al.: limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. \u003cb\u003e43\u003c/b\u003e, e47 (2015)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStirling, D.R., et al.: CellProfiler 4: improvements in speed, utility and usability. BMC Bioinform. \u003cb\u003e22\u003c/b\u003e, 433 (2021)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWong, A.C.H., Wong, J.J.-L., Rasko, J.E.J., Schmitz, U.: SpliceWiz: interactive analysis and visualization of alternative splicing in R. Brief. Bioinform. \u003cb\u003e25\u003c/b\u003e, bbad468 (2023)\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-8682343/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8682343/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe proteasome is an essential protein complex that cancer cells rely on for amino acid recycling and protein degradation, making it a key therapeutic target. Its core particle can bind different activators that regulate its activity and substrate specificity, enabling significant functional diversity. However, the mechanisms by which different proteasome activators function across cell types and disease states are poorly understood. Here, we identified an enrichment of the non-constitutive activator PA28αβ in glioblastoma stem cells (GSCs) and tumours. Disruption of PA28αβ reduced the self-renewal of GSCs and improved survival in orthotopic xenograft and syngeneic models. Despite the canonical heteroheptameric composition of PA28αβ, we demonstrate that the individual subunits play dichotomous roles in glioblastoma. Analysis of proteasome activity and the immunopeptidome in response to PA28αβ perturbation revealed a unique role for the PA28β subunit in regulating immunoproteasome activity and antigen presentation. Interactome mapping identified the kinesin motor protein KIF11 as a binding partner of PA28αβ, and molecular profiling showed subunit-dependent alterations in protein trafficking pathways. Furthermore, PA28α displayed a conserved role in regulating KIF11 stability, promoting stemness in GSCs and tumour growth in a syngeneic glioma model. Collectively, our results reveal subunit-specific roles for PA28αβ with therapeutic potential in glioblastoma.\u003c/p\u003e","manuscriptTitle":"The proteasome activator subunits PA28α and PA28β have unique molecular roles in glioblastoma","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-04 08:57:45","doi":"10.21203/rs.3.rs-8682343/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"communications-biology","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"commsbio","sideBox":"Learn more about [Communications Biology](http://www.nature.com/commsbio/)","snPcode":"","submissionUrl":"","title":"Communications Biology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Communications Series","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"47eb8b9b-3326-4a6d-a84d-a84b1901e991","owner":[],"postedDate":"May 4th, 2026","published":true,"recentEditorialEvents":[{"type":"reviewerAgreed","content":"This content is not available.","date":"2026-05-07T11:34:12+00:00","index":2,"fulltext":"This content is not available."}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":66866249,"name":"Health sciences/Oncology/Cancer/Cancer stem cells"},{"id":66866250,"name":"Biological sciences/Cell biology/Proteolysis/Proteasome"},{"id":66866251,"name":"Biological sciences/Cancer/CNS cancer"}],"tags":[],"updatedAt":"2026-05-04T08:57:45+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-04 08:57:45","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8682343","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8682343","identity":"rs-8682343","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}