Setdb2 Silences HIF-2α to Sustain the HIF-1α Metabolic Program in Trained Immunity

Setdb2 Silences HIF-2α to Sustain the HIF-1α Metabolic Program in Trained Immunity | 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 Research Article Setdb2 Silences HIF-2α to Sustain the HIF-1α Metabolic Program in Trained Immunity jeongsoon yong This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9391910/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract β-glucan-induced trained immunity requires sustained mTOR–HIF-1α signaling, yet how this metabolic program persists for days after the initial stimulus is removed remains unclear. Here, we show that the H3K9 methyltransferase Setdb2 resolves this question by selectively silencing HIF-2α ( Epas1 ) and multiple anti-inflammatory brake genes during the memory phase. In β-glucan-trained wild-type macrophages, the HIF-1α/HIF-2α mRNA ratio is 3-fold higher than in Setdb2 -deficient macrophages (35.4 vs. 12.0, p = 0.0003), reflecting selective silencing of the HIF-2α/M2 program. Consequently, trained wild-type macrophages exhibit significantly lower M2 pathway scores (p = 0.018) and stronger M1–M2 polarization bias toward M1 (p = 0.003) than Setdb2 knockouts. Upon LPS restimulation, this prepared imbalance produces an explosive HIF-1α transcriptional response in wild-type but not knockout macrophages (pathway score +0.93 vs. −0.02, p = 0.0009). We propose that Setdb2 functions as a metabolic memory gatekeeper : by closing the HIF-2α/M2 brake during the resting phase, it ensures that the HIF-1α glycolytic program remains poised for rapid re-engagement, thereby sustaining trained immunity. trained immunity HIF-1α HIF-2α/EPAS1 Setdb2 metabolic memory M1/M2 polarization immunometabolism β-glucan Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Trained immunity, the enhanced innate immune response following initial pathogen exposure, is driven by epigenetic and metabolic reprogramming of monocytes and macrophages (Netea et al., 2011 ). The mTOR–HIF-1α metabolic axis is central to this process: β-glucan activates mTOR, stabilizes HIF-1α, and shifts cellular metabolism toward aerobic glycolysis, generating metabolites such as fumarate and acetyl-CoA that sustain activating histone modifications at inflammatory gene loci (Cheng et al., 2014 ; Arts et al., 2016 ). A critical unresolved question is how the HIF-1α program persists during the memory phase (days 3–7), when the training stimulus has been removed and cells are in a resting state. HIF-1α protein is constitutively degraded under normoxic conditions via PHD–VHL–proteasome pathway, yet trained macrophages maintain HIF-1α transcriptional signatures days later (Cheng et al., 2014 ). This paradox suggests the existence of an epigenetic mechanism that maintains the competence for HIF-1α signaling independent of continuous metabolic input. HIF-2α, encoded by Epas1 , is a structurally related transcription factor that competes with HIF-1α for the shared dimerization partner ARNT/HIF-1β (Keith et al., 2012 ). Critically, HIF-1α and HIF-2α drive opposing macrophage programs: HIF-1α promotes M1 polarization and glycolysis, while HIF-2α promotes M2 polarization and anti-inflammatory gene expression including Arg1 , Mrc1 , and Vegfa (Takeda et al., 2010 ; Imtiyaz et al., 2010 ). This antagonism implies that silencing HIF-2α would remove a molecular brake on HIF-1α dominance. SET domain bifurcated 2 ( Setdb2 ) is an H3K9 trimethyltransferase recently shown to regulate trained immunity through both enzymatic and structural mechanisms (Hanten et al., 2025 ). We previously demonstrated that β-glucan training induces genome-wide chromatin closing dominance (open/close ratio 0.31–0.75), which is abolished in Setdb2 -deficient macrophages (Yong, 2025 ). However, the specific transcriptional targets of Setdb2-mediated closing and their functional consequences for metabolic programming remained undefined. Here, we identify Epas1 /HIF-2α as a critical Setdb2 target and demonstrate that Setdb2-mediated silencing of the HIF-2α/M2 program is the mechanism by which trained macrophages maintain HIF-1α metabolic competence during the memory phase. Results Setdb2 selectively silences anti-inflammatory brake genes during β-glucan training Re-analysis of RNA-seq data from wild-type (WT) and Setdb2 macrophage-specific knockout (KO) bone marrow-derived macrophages treated with β-glucan (GLU) or vehicle (CTRL) (GSE290872; Hanten et al., 2025 ) identified 209 genes suppressed (≥ 2-fold) in WT but not in KO during β-glucan training. These Setdb2-dependent targets included 11 transcription factors, 11 signaling kinases, 14 metabolic enzymes, 20 surface receptors, and 12 complement/defense genes. Among these, five anti-inflammatory signaling genes were consistently and significantly lower in WT than KO at the GLU condition: Epas1 /HIF-2α (2.85-fold, p = 0.008), Rarb /RAR-β (2.78-fold, p = 0.006), Ptgis /prostacyclin synthase (11.9-fold, p < 0.001), Adcy4 /adenylyl cyclase 4 (2.27-fold, p = 0.002), and Adrb1 /β1-adrenergic receptor (2.85-fold, p = 0.002). We termed these collectively as “metabolic brakes” because each encodes a negative regulator of inflammatory or glycolytic signaling. Setdb2 tilts the HIF-1α/HIF-2α balance during the memory phase Because HIF-1α and HIF-2α compete for ARNT binding, the ratio of their expression determines the dominant transcriptional program. At the GLU (memory) phase, Hif1a mRNA levels were comparable between WT and KO (160.9 vs. 152.0 CPM). However, Epas1 was dramatically lower in WT (4.6 CPM) than KO (13.0 CPM), producing a HIF-1α/HIF-2α ratio of 35.4 in WT versus 12.0 in KO (p = 0.0003). This 3-fold difference demonstrates that Setdb2 tilts the HIF balance by removing HIF-2α rather than increasing HIF-1α. This imbalance was already present during the memory phase (GLU), before any restimulation, establishing it as a preparatory rather than reactive event. Setdb2 suppresses the M2 transcriptional program To assess the functional consequence of HIF-2α silencing, we computed M1 and M2 pathway scores using curated gene sets of 34 M1 markers (inflammatory cytokines, glycolytic enzymes, M1 transcription factors) and 34 M2 markers (anti-inflammatory cytokines, oxidative phosphorylation genes, M2 surface receptors including HIF-2α-specific targets). At the GLU condition, WT macrophages showed significantly lower M2 scores than KO (0.088 vs. 0.260, p = 0.018) and a stronger M1–M2 bias (− 0.660 vs. −0.898, p = 0.003). Individual M2 genes confirmed this pattern: Mrc1 /CD206 (WT 277 vs. KO 362 CPM, p = 0.001), Ccl17 (843 vs. 1234, p = 0.006), Cd36 (327 vs. 657, p = 0.002), Pparg (17.3 vs. 24.6, p = 0.001), Fn1 (12.0 vs. 41.2, p = 0.009), and Clec7a (1019 vs. 1338, p = 0.005) were all significantly reduced in WT. Permutation testing of M1 versus M2 target fold-changes confirmed that M1 targets were significantly more induced than M2 targets during β-glucan training (observed difference + 1.557 log2FC, permutation p = 0.004). The prepared HIF imbalance enables explosive HIF-1α pathway activation upon restimulation To test whether the memory-phase HIF imbalance functionally impacts restimulation responses, we computed HIF-1α pathway activity scores using 26 established HIF-1α target genes across conditions. At the GLU (resting) phase, HIF-1α scores showed a non-significant trend (WT + 0.224 vs. KO + 0.128, p = 0.15), consistent with the pathway being primed but not yet activated. Upon LPS restimulation (GLULPS), WT macrophages showed dramatic HIF-1α pathway activation (score + 0.930) while KO macrophages failed to activate the pathway (score − 0.023, p = 0.0009). Similarly, mTORC1 pathway scores were significantly higher in WT (− 0.159) than KO (− 0.473, p < 0.0001) under restimulation. This pattern was reflected in individual cytokine and glycolytic gene responses. Upon restimulation, WT showed greater burst amplification than KO for Il6 (2545x vs. 1995x), Tnf (40.9x vs. 28.8x), Ccl4 (6.5x vs. 3.9x), Hk2 (3.9x vs. 3.0x), and Slc2a1 /GLUT1 (2.0x vs. 1.2x). Setdb2 deficiency reduces trained immunity cytokine output Consistent with the impaired HIF-1α response, Setdb2 KO macrophages produced significantly less cytokine mRNA upon restimulation: Tnf (− 45%), Il1rn (− 41%), Ccl2 (− 42%), Ccl4 (− 54%), and Il10 (− 54%). Il1b and Il6 were relatively preserved, suggesting selective rather than global impairment of cytokine production. Discussion Our findings reveal a previously unrecognized mechanism for trained immunity maintenance: Setdb2-mediated epigenetic silencing of HIF-2α and associated anti-inflammatory programs. We propose a model of metabolic memory through brake removal (Fig. 4). In the acute phase (0–24 hours), β-glucan activates the mTOR–HIF-1α axis via Dectin-1 signaling, as established by Cheng et al. ( 2014 ). Simultaneously, Setdb2 is induced and begins depositing H3K9me3 at the Epas1 locus and other anti-inflammatory gene promoters. In the memory phase (days 1–6), the training stimulus is removed, but the H3K9me3 marks persist, maintaining HIF-2α in a silenced state. This creates a permissive imbalance : HIF-1α retains exclusive access to ARNT, while competing M2 programs are epigenetically locked out. Upon secondary stimulation, this imbalance enables explosive HIF-1α pathway activation, glycolytic burst, and enhanced cytokine production. This model resolves the paradox of how HIF-1α-dependent glycolysis persists under normoxic conditions during the memory phase. The answer is not that HIF-1α itself is constitutively stabilized, but that its competitor, HIF-2α, is constitutively silenced. When HIF-1α is transiently stabilized by any secondary stimulus, it encounters no competition for ARNT binding, enabling a disproportionately strong transcriptional response. Our finding that Setdb2 silences not only HIF-2α but an entire network of anti-inflammatory brakes—including Rarb (retinoic acid signaling), Ptgis (prostacyclin synthesis), and Adcy4 (cAMP production)—suggests that trained immunity involves coordinated suppression of multiple resolution pathways. This is consistent with the clinical observation that trained immunity increases susceptibility to chronic inflammatory diseases such as atherosclerosis (Bekkering et al., 2014 ). An important distinction from our previous work on chromatin closing dominance (Yong, 2025 ) is the identification of mechanism. While the earlier study established that closing exceeds opening during trained immunity, the present work identifies the specific targets (HIF-2α/M2 program) and functional consequence (HIF-1α pathway dominance) of this closing. The two findings are complementary: genome-wide closing is the phenotype; HIF-2α silencing is the mechanism. Limitations of this study include the reliance on RNA-seq as a proxy for pathway activity; direct measurement of HIF-1α/HIF-2α protein levels and ARNT complex formation would strengthen the model. Additionally, the functional relevance of all 209 Setdb2 targets beyond the HIF-2α/M2 subset remains to be characterized. Therapeutically, pharmacological stabilization of HIF-2α could attenuate trained immunity by restoring HIF-1α/HIF-2α balance, though selective HIF-2α stabilizers remain to be developed. Conversely, Setdb2 inhibition could similarly dampen excessive inflammatory memory in chronic disease settings. STAR Methods Data sources RNA-seq count matrices and ATAC-seq BigWig files were obtained from GSE290872 (Hanten et al., 2025). This dataset comprises WT and Setdb2 macrophage-specific KO BMDMs under four conditions: CTRL (vehicle), GLU (β-glucan alone), LPS (LPS alone), and GLULPS (β-glucan + LPS restimulation), with 3 biological replicates per condition (24 RNA-seq + 24 ATAC-seq samples). Gene expression analysis Raw counts were normalized to counts per million (CPM). Differential expression was defined as ≥2-fold change between conditions with mean expression >10 counts. Setdb2-dependent genes were identified as those suppressed ≥2-fold in WT GLU vs. CTRL but not in KO GLU vs. KO CTRL. HIF-1α/HIF-2α ratio The ratio was computed as Hif1a CPM / Epas1 CPM per replicate. Statistical comparison was performed using two-sample t-test. Pathway activity scoring M1 (n=34) and M2 (n=34) gene sets were curated from MSigDB Hallmark collections and published macrophage polarization signatures. HIF-1α (n=26), mTORC1 (n=43), and AMPK (n=15) target gene sets were similarly curated. Pathway scores were computed as the mean z-score of member genes across all samples, calculated per replicate. Statistical comparisons between WT and KO used two-sample t-tests. Permutation test To test whether M1 and M2 targets show differential fold-change directions, log2 fold-changes (GLU/CTRL) were computed for M1 (n=8) and M2 (n=28) gene sets. The observed difference in means was compared against 10,000 random permutations of gene labels. Burst amplification Restimulation burst was computed as GLULPS CPM / GLU CPM per gene. Priming was defined as GLU CPM / CTRL CPM. Declarations Funding This research received no external funding. Author Contribution J.S.Y. conceived the study, performed all computational analyses, interpreted the data, and wrote the manuscript. Data Availability All data analysed in this study are publicly available. RNA-seq data from β-glucan-trained Setdb2 wild-type and knockout macrophages were obtained from the Gene Expression Omnibus under accession number GSE290872. No new data were generated. Analysis code is available from the corresponding author upon request. References Arts, R.J.W., et al. (2016). Glutaminolysis and fumarate accumulation integrate immunometabolic and epigenetic programs in trained immunity. Cell Metab. 24, 807–819. Bekkering, S., et al. (2014). Oxidized low-density lipoprotein induces long-term proinflammatory cytokine production and foam cell formation via epigenetic reprogramming of monocytes. Arterioscler. Thromb. Vasc. Biol. 34, 1731–1738. Cheng, S.C., et al. (2014). mTOR- and HIF-1α-mediated aerobic glycolysis as metabolic basis for trained immunity. Science 345, 1250684. Hanten, J.A., et al. (2025). Setdb2 Regulates Inflammatory Trigger-Induced Trained Immunity of Macrophages Through Two Different Epigenetic Mechanisms. Immunity (in press). Imtiyaz, H.Z., et al. (2010). Hypoxia-inducible factor 2α regulates macrophage function in mouse models of acute and tumor inflammation. J. Clin. Invest. 120, 2699–2714. Keith, B., Johnson, R.S., and Simon, M.C. (2012). HIF1α and HIF2α: sibling rivalry in hypoxic tumour growth and progression. Nat. Rev. Cancer 12, 9–22. Netea, M.G., Quintin, J., and van der Meer, J.W. (2011). Trained immunity: a memory for innate host defense. Cell Host Microbe 9, 355–361. Saeed, S., et al. (2014). Epigenetic programming during monocyte to macrophage differentiation and trained innate immunity. Science 345, 1251086. Takeda, N., et al. (2010). Differential activation and antagonistic function of HIF-α isoforms in macrophages are essential for NO homeostasis. Genes Dev. 24, 491–501. Yong, J.S. (2025). Trained immunity is defined by selective chromatin closing. Preprint. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted 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-9391910","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":635235882,"identity":"ba7de672-72ed-4d4c-8d99-74bb3fb01d9d","order_by":0,"name":"jeongsoon yong","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA20lEQVRIie3OMQrCMBTG8VcKcYl2TSh4hlcKxUHqVSKBTi5uDg4RQZeiq+DgFTxCpNCp6Cp09AIFQVwsOgkupm4O+Y0P/rwPwLL+kLO7aKwmfXxfmClxVSLGmyL5IWmpEV7bi+yHpK10sKfkFHp+dqhgGgPf6u8JnymJjJYRXyWSQS7B74jvSTCHHJGVfSwoMiAautQwbJA7y7vA4yvxrneoGyRO6gJqoaPXF8KchQbfmGwIBErLkKck6g1XkvLUmHhV+KjjYE3dy7m6xV1WGJJPAsA0y7Isy2riCc+FPZw1EZl4AAAAAElFTkSuQmCC","orcid":"","institution":"Korea University","correspondingAuthor":true,"prefix":"","firstName":"jeongsoon","middleName":"","lastName":"yong","suffix":""}],"badges":[],"createdAt":"2026-04-12 05:53:58","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9391910/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9391910/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108806608,"identity":"0bc158dd-6ea7-4ad5-b451-11c0f4a56b64","added_by":"auto","created_at":"2026-05-08 15:29:04","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":231704,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSetdb2 silences anti-inflammatory brake genes during β-glucan training.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(A) Identification of 209 Setdb2-dependent suppressed genes. (B) Functional categorization of 209 targets. (C) Expression of five brake genes (Epas1, Rarb, Ptgis, Adcy4, Adrb1) in WT vs. KO at GLU condition, all p \u0026lt; 0.01.\u003c/p\u003e","description":"","filename":"fig1.png","url":"https://assets-eu.researchsquare.com/files/rs-9391910/v1/36044473e210761bc9cf3a10.png"},{"id":108660775,"identity":"ffc378c6-671c-4d87-9b0b-de86483d2f4f","added_by":"auto","created_at":"2026-05-07 05:10:21","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":185775,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSetdb2 tilts the HIF-1α/HIF-2α balance and suppresses the M2 program.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(A) HIF-1α/HIF-2α mRNA ratio across conditions (WT GLU = 35.4, KO GLU = 12.0, p = 0.0003). (B) M1 and M2 pathway scores at GLU: WT M2 lower than KO (p = 0.018). (C) M1–M2 bias score at GLU: WT more M1-biased (p = 0.003). (D) Individual M2 genes lower in WT at GLU.\u003c/p\u003e","description":"","filename":"fig2.png","url":"https://assets-eu.researchsquare.com/files/rs-9391910/v1/7983951903bae5c7e3274246.png"},{"id":108660774,"identity":"fa824d49-6553-415c-9d95-32d0e7ab7885","added_by":"auto","created_at":"2026-05-07 05:10:21","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":230793,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePrepared HIF imbalance enables explosive HIF-1α activation upon restimulation.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(A) HIF-1α pathway scores across conditions: GLU shows priming, GLULPS shows WT-specific explosion (p = 0.0009). (B) mTORC1 pathway scores at GLULPS (p \u0026lt; 0.0001). (C) Burst amplification of cytokines and glycolytic genes: WT \u0026gt; KO for IL6, TNF, CCL4, HK2, SLC2A1.\u003c/p\u003e","description":"","filename":"fig3.png","url":"https://assets-eu.researchsquare.com/files/rs-9391910/v1/98c73148a627b802c6539adb.png"},{"id":108660772,"identity":"53111527-5e96-4bca-bf00-7f12bc4d2272","added_by":"auto","created_at":"2026-05-07 05:10:19","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":262791,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eModel: metabolic memory through brake removal.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSchematic illustration of the proposed mechanism. Acute phase: β-glucan activates mTOR–HIF-1α and induces Setdb2. Memory phase: Setdb2 deposits H3K9me3 at Epas1 and M2 gene loci, creating a permissive HIF-1α/HIF-2α imbalance. Restimulation: HIF-1α, encountering no HIF-2α competition, drives explosive glycolytic and cytokine responses. Setdb2 KO: HIF-2α remains expressed, competes with HIF-1α for ARNT, attenuates restimulation response.\u003c/p\u003e","description":"","filename":"fig4.png","url":"https://assets-eu.researchsquare.com/files/rs-9391910/v1/4438b550a740d44e2573f875.png"},{"id":108812331,"identity":"6ce9f29d-e64a-466d-8426-832708a8e819","added_by":"auto","created_at":"2026-05-08 16:10:23","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":998211,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9391910/v1/7578787a-ec1a-4ab5-a63d-8449a281dd8d.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Setdb2 Silences HIF-2α to Sustain the HIF-1α Metabolic Program in Trained Immunity","fulltext":[{"header":"Introduction","content":"\u003cp\u003eTrained immunity, the enhanced innate immune response following initial pathogen exposure, is driven by epigenetic and metabolic reprogramming of monocytes and macrophages (Netea et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). The mTOR\u0026ndash;HIF-1α metabolic axis is central to this process: β-glucan activates mTOR, stabilizes HIF-1α, and shifts cellular metabolism toward aerobic glycolysis, generating metabolites such as fumarate and acetyl-CoA that sustain activating histone modifications at inflammatory gene loci (Cheng et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Arts et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eA critical unresolved question is how the HIF-1α program persists during the memory phase (days 3\u0026ndash;7), when the training stimulus has been removed and cells are in a resting state. HIF-1α protein is constitutively degraded under normoxic conditions via PHD\u0026ndash;VHL\u0026ndash;proteasome pathway, yet trained macrophages maintain HIF-1α transcriptional signatures days later (Cheng et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). This paradox suggests the existence of an epigenetic mechanism that maintains the competence for HIF-1α signaling independent of continuous metabolic input.\u003c/p\u003e \u003cp\u003eHIF-2α, encoded by \u003cem\u003eEpas1\u003c/em\u003e, is a structurally related transcription factor that competes with HIF-1α for the shared dimerization partner ARNT/HIF-1β (Keith et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). Critically, HIF-1α and HIF-2α drive opposing macrophage programs: HIF-1α promotes M1 polarization and glycolysis, while HIF-2α promotes M2 polarization and anti-inflammatory gene expression including \u003cem\u003eArg1\u003c/em\u003e, \u003cem\u003eMrc1\u003c/em\u003e, and \u003cem\u003eVegfa\u003c/em\u003e (Takeda et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2010\u003c/span\u003e; Imtiyaz et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). This antagonism implies that silencing HIF-2α would remove a molecular brake on HIF-1α dominance.\u003c/p\u003e \u003cp\u003eSET domain bifurcated 2 (\u003cem\u003eSetdb2\u003c/em\u003e) is an H3K9 trimethyltransferase recently shown to regulate trained immunity through both enzymatic and structural mechanisms (Hanten et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). We previously demonstrated that β-glucan training induces genome-wide chromatin closing dominance (open/close ratio 0.31\u0026ndash;0.75), which is abolished in \u003cem\u003eSetdb2\u003c/em\u003e-deficient macrophages (Yong, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). However, the specific transcriptional targets of Setdb2-mediated closing and their functional consequences for metabolic programming remained undefined.\u003c/p\u003e \u003cp\u003eHere, we identify \u003cem\u003eEpas1\u003c/em\u003e/HIF-2α as a critical Setdb2 target and demonstrate that Setdb2-mediated silencing of the HIF-2α/M2 program is the mechanism by which trained macrophages maintain HIF-1α metabolic competence during the memory phase.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eSetdb2 selectively silences anti-inflammatory brake genes during β-glucan training\u003c/h2\u003e \u003cp\u003eRe-analysis of RNA-seq data from wild-type (WT) and \u003cem\u003eSetdb2\u003c/em\u003e macrophage-specific knockout (KO) bone marrow-derived macrophages treated with β-glucan (GLU) or vehicle (CTRL) (GSE290872; Hanten et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) identified 209 genes suppressed (\u0026ge;\u0026thinsp;2-fold) in WT but not in KO during β-glucan training. These Setdb2-dependent targets included 11 transcription factors, 11 signaling kinases, 14 metabolic enzymes, 20 surface receptors, and 12 complement/defense genes.\u003c/p\u003e \u003cp\u003eAmong these, five anti-inflammatory signaling genes were consistently and significantly lower in WT than KO at the GLU condition: \u003cem\u003eEpas1\u003c/em\u003e/HIF-2α (2.85-fold, p\u0026thinsp;=\u0026thinsp;0.008), \u003cem\u003eRarb\u003c/em\u003e/RAR-β (2.78-fold, p\u0026thinsp;=\u0026thinsp;0.006), \u003cem\u003ePtgis\u003c/em\u003e/prostacyclin synthase (11.9-fold, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), \u003cem\u003eAdcy4\u003c/em\u003e/adenylyl cyclase 4 (2.27-fold, p\u0026thinsp;=\u0026thinsp;0.002), and \u003cem\u003eAdrb1\u003c/em\u003e/β1-adrenergic receptor (2.85-fold, p\u0026thinsp;=\u0026thinsp;0.002). We termed these collectively as \u0026ldquo;metabolic brakes\u0026rdquo; because each encodes a negative regulator of inflammatory or glycolytic signaling.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eSetdb2 tilts the HIF-1α/HIF-2α balance during the memory phase\u003c/h3\u003e\n\u003cp\u003eBecause HIF-1α and HIF-2α compete for ARNT binding, the ratio of their expression determines the dominant transcriptional program. At the GLU (memory) phase, \u003cem\u003eHif1a\u003c/em\u003e mRNA levels were comparable between WT and KO (160.9 vs. 152.0 CPM). However, \u003cem\u003eEpas1\u003c/em\u003e was dramatically lower in WT (4.6 CPM) than KO (13.0 CPM), producing a HIF-1α/HIF-2α ratio of 35.4 in WT versus 12.0 in KO (p\u0026thinsp;=\u0026thinsp;0.0003). This 3-fold difference demonstrates that Setdb2 tilts the HIF balance by removing HIF-2α rather than increasing HIF-1α.\u003c/p\u003e \u003cp\u003eThis imbalance was already present during the memory phase (GLU), before any restimulation, establishing it as a \u003cb\u003epreparatory\u003c/b\u003e rather than reactive event.\u003c/p\u003e\n\u003ch3\u003eSetdb2 suppresses the M2 transcriptional program\u003c/h3\u003e\n\u003cp\u003eTo assess the functional consequence of HIF-2α silencing, we computed M1 and M2 pathway scores using curated gene sets of 34 M1 markers (inflammatory cytokines, glycolytic enzymes, M1 transcription factors) and 34 M2 markers (anti-inflammatory cytokines, oxidative phosphorylation genes, M2 surface receptors including HIF-2α-specific targets).\u003c/p\u003e \u003cp\u003eAt the GLU condition, WT macrophages showed significantly lower M2 scores than KO (0.088 vs. 0.260, p\u0026thinsp;=\u0026thinsp;0.018) and a stronger M1\u0026ndash;M2 bias (\u0026minus;\u0026thinsp;0.660 vs. \u0026minus;0.898, p\u0026thinsp;=\u0026thinsp;0.003). Individual M2 genes confirmed this pattern: \u003cem\u003eMrc1\u003c/em\u003e/CD206 (WT 277 vs. KO 362 CPM, p\u0026thinsp;=\u0026thinsp;0.001), \u003cem\u003eCcl17\u003c/em\u003e (843 vs. 1234, p\u0026thinsp;=\u0026thinsp;0.006), \u003cem\u003eCd36\u003c/em\u003e (327 vs. 657, p\u0026thinsp;=\u0026thinsp;0.002), \u003cem\u003ePparg\u003c/em\u003e (17.3 vs. 24.6, p\u0026thinsp;=\u0026thinsp;0.001), \u003cem\u003eFn1\u003c/em\u003e (12.0 vs. 41.2, p\u0026thinsp;=\u0026thinsp;0.009), and \u003cem\u003eClec7a\u003c/em\u003e (1019 vs. 1338, p\u0026thinsp;=\u0026thinsp;0.005) were all significantly reduced in WT.\u003c/p\u003e \u003cp\u003ePermutation testing of M1 versus M2 target fold-changes confirmed that M1 targets were significantly more induced than M2 targets during β-glucan training (observed difference\u0026thinsp;+\u0026thinsp;1.557 log2FC, permutation p\u0026thinsp;=\u0026thinsp;0.004).\u003c/p\u003e\n\u003ch3\u003eThe prepared HIF imbalance enables explosive HIF-1α pathway activation upon restimulation\u003c/h3\u003e\n\u003cp\u003eTo test whether the memory-phase HIF imbalance functionally impacts restimulation responses, we computed HIF-1α pathway activity scores using 26 established HIF-1α target genes across conditions. At the GLU (resting) phase, HIF-1α scores showed a non-significant trend (WT\u0026thinsp;+\u0026thinsp;0.224 vs. KO\u0026thinsp;+\u0026thinsp;0.128, p\u0026thinsp;=\u0026thinsp;0.15), consistent with the pathway being primed but not yet activated.\u003c/p\u003e \u003cp\u003eUpon LPS restimulation (GLULPS), WT macrophages showed dramatic HIF-1α pathway activation (score\u0026thinsp;+\u0026thinsp;0.930) while KO macrophages failed to activate the pathway (score\u0026thinsp;\u0026minus;\u0026thinsp;0.023, p\u0026thinsp;=\u0026thinsp;0.0009). Similarly, mTORC1 pathway scores were significantly higher in WT (\u0026minus;\u0026thinsp;0.159) than KO (\u0026minus;\u0026thinsp;0.473, p\u0026thinsp;\u0026lt;\u0026thinsp;0.0001) under restimulation.\u003c/p\u003e \u003cp\u003eThis pattern was reflected in individual cytokine and glycolytic gene responses. Upon restimulation, WT showed greater burst amplification than KO for \u003cem\u003eIl6\u003c/em\u003e (2545x vs. 1995x), \u003cem\u003eTnf\u003c/em\u003e (40.9x vs. 28.8x), \u003cem\u003eCcl4\u003c/em\u003e (6.5x vs. 3.9x), \u003cem\u003eHk2\u003c/em\u003e (3.9x vs. 3.0x), and \u003cem\u003eSlc2a1\u003c/em\u003e/GLUT1 (2.0x vs. 1.2x).\u003c/p\u003e\n\u003ch3\u003eSetdb2 deficiency reduces trained immunity cytokine output\u003c/h3\u003e\n\u003cp\u003eConsistent with the impaired HIF-1α response, \u003cem\u003eSetdb2\u003c/em\u003e KO macrophages produced significantly less cytokine mRNA upon restimulation: \u003cem\u003eTnf\u003c/em\u003e (\u0026minus;\u0026thinsp;45%), \u003cem\u003eIl1rn\u003c/em\u003e (\u0026minus;\u0026thinsp;41%), \u003cem\u003eCcl2\u003c/em\u003e (\u0026minus;\u0026thinsp;42%), \u003cem\u003eCcl4\u003c/em\u003e (\u0026minus;\u0026thinsp;54%), and \u003cem\u003eIl10\u003c/em\u003e (\u0026minus;\u0026thinsp;54%). \u003cem\u003eIl1b\u003c/em\u003e and \u003cem\u003eIl6\u003c/em\u003e were relatively preserved, suggesting selective rather than global impairment of cytokine production.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eOur findings reveal a previously unrecognized mechanism for trained immunity maintenance: Setdb2-mediated epigenetic silencing of HIF-2α and associated anti-inflammatory programs. We propose a model of \u003cb\u003emetabolic memory through brake removal\u003c/b\u003e (Fig.\u0026nbsp;4).\u003c/p\u003e \u003cp\u003eIn the acute phase (0\u0026ndash;24 hours), β-glucan activates the mTOR\u0026ndash;HIF-1α axis via Dectin-1 signaling, as established by Cheng et al. (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Simultaneously, Setdb2 is induced and begins depositing H3K9me3 at the \u003cem\u003eEpas1\u003c/em\u003e locus and other anti-inflammatory gene promoters. In the memory phase (days 1\u0026ndash;6), the training stimulus is removed, but the H3K9me3 marks persist, maintaining HIF-2α in a silenced state. This creates a \u003cb\u003epermissive imbalance\u003c/b\u003e: HIF-1α retains exclusive access to ARNT, while competing M2 programs are epigenetically locked out. Upon secondary stimulation, this imbalance enables explosive HIF-1α pathway activation, glycolytic burst, and enhanced cytokine production.\u003c/p\u003e \u003cp\u003eThis model resolves the paradox of how HIF-1α-dependent glycolysis persists under normoxic conditions during the memory phase. The answer is not that HIF-1α itself is constitutively stabilized, but that its competitor, HIF-2α, is constitutively silenced. When HIF-1α is transiently stabilized by any secondary stimulus, it encounters no competition for ARNT binding, enabling a disproportionately strong transcriptional response.\u003c/p\u003e \u003cp\u003eOur finding that Setdb2 silences not only HIF-2α but an entire network of anti-inflammatory brakes\u0026mdash;including \u003cem\u003eRarb\u003c/em\u003e (retinoic acid signaling), \u003cem\u003ePtgis\u003c/em\u003e (prostacyclin synthesis), and \u003cem\u003eAdcy4\u003c/em\u003e (cAMP production)\u0026mdash;suggests that trained immunity involves coordinated suppression of multiple resolution pathways. This is consistent with the clinical observation that trained immunity increases susceptibility to chronic inflammatory diseases such as atherosclerosis (Bekkering et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2014\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAn important distinction from our previous work on chromatin closing dominance (Yong, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) is the identification of mechanism. While the earlier study established that closing exceeds opening during trained immunity, the present work identifies the specific targets (HIF-2α/M2 program) and functional consequence (HIF-1α pathway dominance) of this closing. The two findings are complementary: genome-wide closing is the phenotype; HIF-2α silencing is the mechanism.\u003c/p\u003e \u003cp\u003eLimitations of this study include the reliance on RNA-seq as a proxy for pathway activity; direct measurement of HIF-1α/HIF-2α protein levels and ARNT complex formation would strengthen the model. Additionally, the functional relevance of all 209 Setdb2 targets beyond the HIF-2α/M2 subset remains to be characterized.\u003c/p\u003e \u003cp\u003eTherapeutically, pharmacological stabilization of HIF-2α could attenuate trained immunity by restoring HIF-1α/HIF-2α balance, though selective HIF-2α stabilizers remain to be developed. Conversely, Setdb2 inhibition could similarly dampen excessive inflammatory memory in chronic disease settings.\u003c/p\u003e"},{"header":"STAR Methods","content":"\u003cp\u003eData sources\u003c/p\u003e\n\u003cp\u003eRNA-seq count matrices and ATAC-seq BigWig files were obtained from GSE290872 (Hanten et al., 2025). This dataset comprises WT and Setdb2 macrophage-specific KO BMDMs under four conditions: CTRL (vehicle), GLU (\u0026beta;-glucan alone), LPS (LPS alone), and GLULPS (\u0026beta;-glucan + LPS restimulation), with 3 biological replicates per condition (24 RNA-seq + 24 ATAC-seq samples).\u003c/p\u003e\n\u003cp\u003eGene expression analysis\u003c/p\u003e\n\u003cp\u003eRaw counts were normalized to counts per million (CPM). Differential expression was defined as \u0026ge;2-fold change between conditions with mean expression \u0026gt;10 counts. Setdb2-dependent genes were identified as those suppressed \u0026ge;2-fold in WT GLU vs. CTRL but not in KO GLU vs. KO CTRL.\u003c/p\u003e\n\u003cp\u003eHIF-1\u0026alpha;/HIF-2\u0026alpha; ratio\u003c/p\u003e\n\u003cp\u003eThe ratio was computed as Hif1a CPM / Epas1 CPM per replicate. Statistical comparison was performed using two-sample t-test.\u003c/p\u003e\n\u003cp\u003ePathway activity scoring\u003c/p\u003e\n\u003cp\u003eM1 (n=34) and M2 (n=34) gene sets were curated from MSigDB Hallmark collections and published macrophage polarization signatures. HIF-1\u0026alpha; (n=26), mTORC1 (n=43), and AMPK (n=15) target gene sets were similarly curated. Pathway scores were computed as the mean z-score of member genes across all samples, calculated per replicate. Statistical comparisons between WT and KO used two-sample t-tests.\u003c/p\u003e\n\u003cp\u003ePermutation test\u003c/p\u003e\n\u003cp\u003eTo test whether M1 and M2 targets show differential fold-change directions, log2 fold-changes (GLU/CTRL) were computed for M1 (n=8) and M2 (n=28) gene sets. The observed difference in means was compared against 10,000 random permutations of gene labels.\u003c/p\u003e\n\u003cp\u003eBurst amplification\u003c/p\u003e\n\u003cp\u003eRestimulation burst was computed as GLULPS CPM / GLU CPM per gene. Priming was defined as GLU CPM / CTRL CPM.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eFunding\u003c/h2\u003e \u003cp\u003eThis research received no external funding.\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eJ.S.Y. conceived the study, performed all computational analyses, interpreted the data, and wrote the manuscript.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eAll data analysed in this study are publicly available. RNA-seq data from β-glucan-trained Setdb2 wild-type and knockout macrophages were obtained from the Gene Expression Omnibus under accession number GSE290872. No new data were generated. Analysis code is available from the corresponding author upon request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eArts, R.J.W., et al. (2016). Glutaminolysis and fumarate accumulation integrate immunometabolic and epigenetic programs in trained immunity. Cell Metab. 24, 807\u0026ndash;819.\u003c/li\u003e\n \u003cli\u003eBekkering, S., et al. (2014). Oxidized low-density lipoprotein induces long-term proinflammatory cytokine production and foam cell formation via epigenetic reprogramming of monocytes. Arterioscler. Thromb. Vasc. Biol. 34, 1731\u0026ndash;1738.\u003c/li\u003e\n \u003cli\u003eCheng, S.C., et al. (2014). mTOR- and HIF-1\u0026alpha;-mediated aerobic glycolysis as metabolic basis for trained immunity. Science 345, 1250684.\u003c/li\u003e\n \u003cli\u003eHanten, J.A., et al. (2025). Setdb2 Regulates Inflammatory Trigger-Induced Trained Immunity of Macrophages Through Two Different Epigenetic Mechanisms. Immunity (in press).\u003c/li\u003e\n \u003cli\u003eImtiyaz, H.Z., et al. (2010). Hypoxia-inducible factor 2\u0026alpha; regulates macrophage function in mouse models of acute and tumor inflammation. J. Clin. Invest. 120, 2699\u0026ndash;2714.\u003c/li\u003e\n \u003cli\u003eKeith, B., Johnson, R.S., and Simon, M.C. (2012). HIF1\u0026alpha; and HIF2\u0026alpha;: sibling rivalry in hypoxic tumour growth and progression. Nat. Rev. Cancer 12, 9\u0026ndash;22.\u003c/li\u003e\n \u003cli\u003eNetea, M.G., Quintin, J., and van der Meer, J.W. (2011). Trained immunity: a memory for innate host defense. Cell Host Microbe 9, 355\u0026ndash;361.\u003c/li\u003e\n \u003cli\u003eSaeed, S., et al. (2014). Epigenetic programming during monocyte to macrophage differentiation and trained innate immunity. Science 345, 1251086.\u003c/li\u003e\n \u003cli\u003eTakeda, N., et al. (2010). Differential activation and antagonistic function of HIF-\u0026alpha; isoforms in macrophages are essential for NO homeostasis. Genes Dev. 24, 491\u0026ndash;501.\u003c/li\u003e\n \u003cli\u003eYong, J.S. (2025). Trained immunity is defined by selective chromatin closing. Preprint.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"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":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"trained immunity, HIF-1α, HIF-2α/EPAS1, Setdb2, metabolic memory, M1/M2 polarization, immunometabolism, β-glucan","lastPublishedDoi":"10.21203/rs.3.rs-9391910/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9391910/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eβ-glucan-induced trained immunity requires sustained mTOR–HIF-1α signaling, yet how this metabolic program persists for days after the initial stimulus is removed remains unclear. Here, we show that the H3K9 methyltransferase Setdb2 resolves this question by selectively silencing HIF-2α (\u003cem\u003eEpas1\u003c/em\u003e) and multiple anti-inflammatory brake genes during the memory phase. In β-glucan-trained wild-type macrophages, the HIF-1α/HIF-2α mRNA ratio is 3-fold higher than in \u003cem\u003eSetdb2\u003c/em\u003e-deficient macrophages (35.4 vs. 12.0, p = 0.0003), reflecting selective silencing of the HIF-2α/M2 program. Consequently, trained wild-type macrophages exhibit significantly lower M2 pathway scores (p = 0.018) and stronger M1–M2 polarization bias toward M1 (p = 0.003) than \u003cem\u003eSetdb2\u003c/em\u003e knockouts. Upon LPS restimulation, this prepared imbalance produces an explosive HIF-1α transcriptional response in wild-type but not knockout macrophages (pathway score +0.93 vs. −0.02, p = 0.0009). We propose that Setdb2 functions as a \u003cstrong\u003emetabolic memory gatekeeper\u003c/strong\u003e: by closing the HIF-2α/M2 brake during the resting phase, it ensures that the HIF-1α glycolytic program remains poised for rapid re-engagement, thereby sustaining trained immunity.\u003c/p\u003e","manuscriptTitle":"Setdb2 Silences HIF-2α to Sustain the HIF-1α Metabolic Program in Trained Immunity","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-07 05:09:31","doi":"10.21203/rs.3.rs-9391910/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"712945f9-f0bf-4834-83ef-3d20d294e124","owner":[],"postedDate":"May 7th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-05-07T05:09:31+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-07 05:09:31","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9391910","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9391910","identity":"rs-9391910","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}