Epithelial states in colorectal cancer are co-determined by YAP associated fetal programming and WNT signaling

Epithelial states in colorectal cancer are co-determined by YAP associated fetal programming and WNT signaling | 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 Epithelial states in colorectal cancer are co-determined by YAP associated fetal programming and WNT signaling Yourae Hong, Susanti Susanti, Megan L. Mills, Kathryn L. Gilroy, and 24 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6812365/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 In the colonic epithelium, fetal programming is induced by Yes-associated protein (YAP) signaling, which is associated with tumorigenesis and progression in colorectal cancer (CRC). While CRC was long considered a WNT driven disease, here we show that fetal programmed cells with activated YAP signaling can arise independently of WNT activation. Fetal programmed cells, in the presence of hyper- or hypo-activated WNT, have different characteristics and are associated with poor relapse. Furthermore, using various mouse models, we demonstrated the conserved characteristics of two different fetal programmed cell states with different WNT activation, which can be switched from a hyperactivated to hypoactivated WNT state based on genetic alteration, particularly Kras mutation. Taken together, these data redefine the key determinants of epithelial cell states in colorectal cancer and integrate emerging preclinical biology into a robust landscape of states observed in human and mouse. Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction In recent years, a novel paradigm in colorectal cancer (CRC) has emerged questioning the dominant role of Wnt signaling in tumor progression and metastasis 1 – 4 . In the reigning paradigm, CRC can follow one of several molecular pathways in its development including WNT, chromosomal instability (CIN), and microsatellite instability (MSI) 5 . But most commonly, CRC pathogenesis is driven by mutations in the adenomatous polyposis coli ( APC ) gene, encoding APC protein that plays a pivotal role in tumor initiation and is a crucial component in the WNT signaling pathway 5 – 8 . Leucine-rich repeat-containing G protein-coupled receptor 5 ( LGR5 ) is an established marker of normal adult stem cells and is encoded by canonical WNT signaling 9 . The abundance of LGR5 during tumor initiation underscores the essential role played by WNT activation in tumor carcinogenesis 10 . However, cancers with weak expression of LGR5 3 or lacking mutations associated with WNT pathways 11 suggest that CRC progression is feasible even with weak or absent WNT activity. Furthermore, our previous work on consensus molecular subtypes (CMS) from bulk transcriptomic and intrinsic-CMS (iCMS) single-cell RNA sequencing analyses demonstrated varying levels of WNT signaling 12 , 13 . Following colonic injury, low levels of WNT with LGR5 deficiency are observed, driving transition from a typical adult stem cell to those expressing fetal stem cell markers 14 . Damage-induced epithelial cells show high expression of fetal/revival stem cell markers as well as YAP (Yes-associated protein) target genes 14 – 16 . This phenomenon suggests that YAP activation is critical in maintaining fetal characteristics during injury. Fetal programmed tumor cells were found to display active YAP signaling 16 , 17 . In tumors, a similar decrease of LGR5 has been observed with the emergence of fetal-like cells expressing ANXA1 , a widely used marker for identifying fetal programmed cells, accompanied by YAP signaling 3 . Regulation of YAP signaling during the development of CRC is complex and poorly understood. One postulated mechanism suggests that during both adenoma formation and tumorigenesis, WNT signaling activates YAP through an APC mutation, either independent of its involvement in the β-catenin complex 18 , or through the β-catenin destruction complex 19 . And the activated YAP suppresses the WNT homeostasis gene expression program 20 . In addition, over-expression of YAP induces the expression of WNT transcription targets LGR5 , as well as cyclin D1 ( CCND1 ), via its interaction with β-catenin in the nucleus 21 . Thus, the specific mechanisms by which YAP impacts WNT signaling can vary dependent on the cellular context, since both feedforward and negative feedback loops have been described to date. YAP activation is not only associated with the WNT signaling pathway, but is also regulated by the extracellular matrix (ECM), injury repairing status, oncogenic mutations such as BRAF V600E , and atypical protein kinase C (aPKC) deficiency 11 , 16 , 22 , 23 . To elucidate these states, we set out a study of human CRC patient samples in their full complexity and combined this with analysis of mouse models of CRC that have different genetic alterations in which the epithelial status is controlled, achieved through genetic manipulation. We provide an unbiased view of the diverse single-cell epithelial states in human tumors regarding different WNT activation and other key features such as fetal programming and YAP signaling in both human and mouse models and observed a conserved pattern of fetal programmed cells with high YAP activation. In particular, Kras mutation is relevant for driving the transition status of fetal programmed cells from WNT-high to WNT-low states, whereas Apc or Braf mutant models are not involved in the transition, and the characteristics of fetal programmed cells are conserved. We also recapitulated a conserved pattern of fetal programmed cells marked by high YAP activation. This study proposes a new paradigm for the cellular heterogeneity of CRC by exploiting functional status, which could provide new therapeutic targets for inhibiting tumor progression. Results Diverse cell states from colorectal cancer cells across different molecular subtypes To identify different cellular phenotypes based on status, we analyzed untreated primary CRC tissues from 81 patients, integrating previously published data from 63 patients with confirmed iCMS classifications 13 and 18 in-house patients (Figure 1a, Extended Figures 1a–b, Supplementary Table 1–2). To classify malignant cells in the 18 in-house tumor tissue samples, gradient boosting was employed using malignant cells from the previously defined cohort (Extended Figures 1c–d). The iCMS classification for the in-house cohort was applied only to the malignant cells, resulting in the identification of 10 patients as iCMS2 and seven patients as iCMS3, except for one patient who had an insufficient number of tumor cells (Extended Figures 2a–b). We also excluded two patients with low number of tumor cells, resulting in 78 patients’ tumor tissues samples being used for downstream analysis. To understand the states of malignant cells across the iCMS2 and iCMS3 groups, we performed hotspot analysis 24 using 78 patients and identified 18 out of 21 modules to be associated with specific biological characteristics (Figure 1b, Supplementary Table 3). Most of the 18 gene modules (n=16) were significantly enriched in either the iCMS2 groups or the iCMS3 groups, respectively (Figure 1c, Extended Figure 3a). Similar to a previous study 13 , the gene modules associating with metaplasia (Module 3) and goblet (Module 7) were enriched in iCMS3, and those associated with canonical WNT (Module 19, 20) were enriched in iCMS2 (Figure 1c). Three modules (Module 8, 11, and 15) were identified as fetal-associating modules. Of these, Module 15 (fetal-III) was enriched in iCMS2, while Modules 8 and 11 (fetal-I and fetal-II, respectively) were enriched in iCMS3. The implication is that fetal-associated characteristics can manifest in both iCMS2 (hyperactivated WNT status tumors) and iCMS3 (hypoactivated WNT status tumors) settings. Not only is YAP activation a key player in maintaining fetal characteristics during injury repair, but it also plays a crucial role within tumor systems 3,17 . The pivotal role of YAP is further substantiated by its consistent presence in a human tumor scRNA-seq dataset (Extended Figure 3b). Furthermore, Human colon panel Xenium spatial transcriptome data from two CRC patients showed the diversity of tumor cell states, including fetal programmed cell status (Extended Figure 4). The patient that identified as iCMS2 showed predominant expression of LGR5 in their tumor cells, while a tumor sample from the iCMS3 patient predominantly expressed ANXA1 , a fetal marker (Extended Figure 4). Interestingly, in line with the single cell data, we found also one cluster expressing ANXA1 (Tumor subtype 11) in the iCMS2 patient, showing that some tumor cells in iCMS2 patients also undergo fetal programming (Figures 1d–e and Extended Figure 4). Two fetal programmed cell phenotypes in both iCMS groups in the presence of differing WNT levels To understand the molecular characteristics of hyperactivated and hypoactivated WNT signaling groups (iCMS2 and iCMS3 tumors, respectively), subpopulations of iCMS2 and iCMS3 tumor cells from 78 patients were identified (Figure 2a). Interestingly, we determined that the transcriptional properties of malignant cells are shared with the characteristics of normal epithelial cells, similar to previous research 25,26 (Figure 2b). In the iCMS2 group, there are six subtypes encompassing stem-like cells with high expression of LGR5 (LGR5 high stem), highly proliferative cells, two populations expressing enterocyte signatures (enterocyte-like type 1 and 2), goblet-like cells with MUC2 expression, and a small subset expressing tuft cell markers (Figures 2a–b and Extended Figure 5a). In the iCMS3 group, six subtypes were also identified, including an undifferentiated population expressing OLFM4 , highly proliferative cells, goblet-like cells with MUC2 expression, and metaplastic cells expressing metaplasia signatures ( ANXA10, CTSE, MUC5AC, and TFF2 ), recognized as gastric-type cell markers in CRC and serrated polyps 27 (Extended Figures 5a-b). Moreover, metaplastic cells exhibit shared signatures of both the absorptive lineage and goblet cell lineage (Figure 2b). Additionally, they display a pronounced metaplasia signature, as previously shown in serrated colorectal polyps 28 . To understand the association between malignant cell states and cell types, we analyzed relationships between gene module expression and cell type signatures (Figures 2c–d). Entero-like type 1 and 2 cell types in iCMS2 groups were positively correlated with fetal and WNT modules, while metaplastic cell types in iCMS3 groups were positively correlated with fetal, EMT, goblet, and metaplasia modules. The entero-like type 1 in the iCMS2 group, and the metaplasia population in the iCMS3 group both expressed high levels of fetal programming and YAP signatures, even in the presence of varying WNT levels (Figure 2e, Extended Figure 5c). This indicates that fetal-programmed tumor cells with YAP dependency give rise to two different phenotypes in different WNT levels: entero-like type 1 cells and metaplastic cells. Both the module approach and cell type identification approach demonstrate the single presence or combination of WNT and YAP states that characterize the CRC states. Additionally, pseudobulk analysis based on cell types from each patient identified that WNT activation and YAP, fetal reprogramming signaling are the important factors to distinguish cell types (Figures 2f-h, Extended Figure 5d) in iCMS2 and iCMS3 groups. In particular, principal component 1 (PC1) shows the positive correlation of WNT-association genes ( ASCL2, AXIN2 , and NKD1) , and PC1 can clarify the group from iCMS2 patients and iCMS3 patient. While YAP target ( F3, GADD45A, MSLN, and MYOF ) and fetal ( ANXA1, PLAUR, and EMP1 ) genes are negatively correlated with PC2 and distinguish the entero-like type 1 population in iCMS2 and metaplasia in iCMS3 from other cell types (Figures 2f-g). Within the cell types harboring fetal programming features, WNT levels are linked to distinct characteristics and genotypes Applying two WNT-associated modules and three fetal-associated modules to each cell type, we observed different activation profiles between WNT and fetal modules (Figures 3a–b). Within the three fetal modules, the entero-like type 1 group was enriched in fetal-III, and the metaplastic cell groups were enriched in fetal-I and II. We performed differential gene expression analysis between the two fetal programmed cell types. Results revealed that the enterocyte-like type 1 group exhibits higher expression of chemokine ligand groups, including CXCL1, CXCL2, CXCL3, and CCL20 (Figure 3c). In particular, the CXCL1 gene can recruit immune cell and stromal cells as well as promoting cancer progression and migration 29-31 . The metaplastic cell group showed higher expression of mucin genes ( MUC1, MUC2, and MUC5B ), known to contribute to the formation of mucosal layers and exert control over their microenvironment 32,33 . The distinctions between these two populations extend beyond the gene level to the pathway level, indicating distinct characteristics. GSEA results also showed enriched TNF-α signaling via NF-kB in the entero-like type 1 population, and glycolysis and fatty acid metabolism in metaplastic cells (Figures 3d–e). Interestingly, these pathways are associated with YAP activation level and fetal programming. Fatty acid metabolism is enriched in the metastatic process with high YAP activity 34 . TNF can activate YAP by inducing the nuclear translocation process of YAP 35 , and inducing a fetal state 36,37 . Additionally, entero-like type 1 and metaplasia cell populations are associated with different genotypes, KRAS and BRAF mutation, respectively (Extended Figure 6a). Furthermore, we performed a ligand-receptor analysis to understand whether fetal programmed cell populations can actively affect other tumor cells (Figure 3f) 38 . In both samples having hyper- and hypo-activated WNT, YAP-high/fetal programmed cell populations (entero-like type 1 and metaplasia) exhibited the highest number of significant interactions with other tumor groups, acting as both a sender and receiver. Using different significant ligand-receptor pairs from entero-like type 1 and metaplastic cells, we also identified specific interactions for each YAP-high/fetal programmed cell type (Figures 3f–h, Extended Figures 6b-e). The entero-like type 1 population interacted with other cell types by expressing L1CAM , indicating its potential ability to initiate metastasis in CRC 39 , while metaplastic cells interacted significantly with other cell types by expressing trefoil genes ( TFF1 and TFF2 ) and mucin genes ( MUC5AC and MUC6 ). Both cell types showed significant interaction with laminin ( LAMB3 and LAMC2 ) and integrin, which is associated with the epithelial–mesenchymal transition (EMT) 40,41 (Extended Figure 6b-d). Taken together, we revealed two different types of fetal programmed cell populations in both the WNT-high and WNT-low groups. Fetal programmed cells in the WNT-high group exhibited higher expression of TNF-α signaling including chemokines, while the WNT-low group showed elevated expression of metabolism related genes. Interestingly, both groups have higher expression of EMT genes, suggesting that these populations can play a critical role in tumor progression and metastasis. Mouse tumors recapitulate the diversity of cell states observed in human tumors and reveal the transition of fetal programmed cells according to genetic alteration To investigate the potential of pre-clinical mouse models in mimicking human CRC biology, we analyzed 28 mouse colonic tumor tissue samples from five mouse models having different genotypes and different WNT activation levels, VilCreER Apc fl/fl ( Apc fl/fl ) (A), VilCreER Apc fl/fl Kras G12D/+ (AK), VilCreER Apc fl/fl Kras G12D/+ Trp53 fl/fl Tgfbr1 fl/fl (AKPT), VilCreER Kras G12D/+ Trp53 fl/fl Notch1 Tg (KPN), and VilCreER Braf V600E/+ Alk5 fl/fl (BA) (Extended Figure 7a, Supplementary Table 5). Interestingly, mouse tumor cell types shared similar characteristics with human tumor cell types and the abundance of cell types depends on the various genotypic alterations (Figures 4a-b, Extended Figures 7b-c). For instance, Lgr5 high stem cells in the mouse model showed a positive correlation with human LGR5 high stem cells, and this was predominant in Apc fl/fl models. The mouse enterocyte-like population, more abundant in the Apc -drived and the KPN models, was positively correlated with human entero-like type 1 and LGR5 high cell types, especially, entero-like type 1. The metaplasia-like population was significantly dominant in the BA model, and most closely resembled with human metaplasia cell type. Notably, the AKPT and KPN models had more cell types and more diverse contributions for each cell type compared to the other models. Unique populations, such as the Tuft-like cells and the EE-like cells, were enriched in AKPT and KPN models (Extended Figure 7d). Among the 7 mouse tumor cell types, we identified two cell types that have a higher fetal module score than other cell types (Figure 4c). The enterocyte-like population and metaplasia-like population both exhibited high fetal module and YAP signaling scores, as well as a high iCMS3 score. However, we observed that the WNT activation level differed between the two fetal programmed cell groups (Figure 4c-d). Specifically, the enterocyte-like population exhibited high WNT and iCMS2 scores, whereas the metaplasia-like population had a high score only for the fetal modules and YAP signaling, but not for WNT activation. Furthermore, a comparative analysis between the enterocyte-like and metaplasia-like mouse cell type supports that both groups demonstrated a similar pattern to human-derived cell types. The metaplasia-like group, resembling human metaplasia cells, showed higher expression of the mucin genes, fetal and YAP-related genes ( Muc13 , Cd55, Emp1, Nt5e, and Asap1 ). Conversely, the enterocyte-like population was associated with a higher expression of chemokine ligand group, WNT, fetal and YAP-related genes ( Cxcl1, Cxcl2, Axin2, Sox4, Ly6a, F3, and Ctgf) (Figure 4e). Both gene expression and pathway signaling were shown to have close similarities among human and mouse cell-types. TNF-α and WNT-β catenin signaling were predominant in the enterocyte-like group, whereas fatty acid metabolism was enriched in the metaplasia-like group (Figure 4f). The fetal programmed population is dynamically variable by genotypic alterations 42 . To identify the dynamics of fetal programmed cells across different genotypic alterations, we performed the CellRank trajectory analysis 43 using only the two fetal programmed tumor cell populations (Figures 4g-i). Trajectory analysis combining with CytoTRACE algorithm 44 revealed that enterocyte-like cells are less differentiated than metaplasia-like cells, and these cells can be converted into the metaplasia-like population (Figures 4g-h). Interestingly, we identified that this transition was facilitated by the Kras mutation (Figure 4i). While the Apc fl /fl model is distinguished only by the enterocyte-like population, AK, AKPT, and KPN cells are more likely to become metaplasia-like population. In particular, AKPT and KPN fetal programmed cells show more heterogeneity than other mouse models, suggesting that AKPT and KPN samples exhibit higher plasticity than others. This suggests that while the Apc and Braf mutations are only associated with the two distinct fetal programmed cells respectively, Kras mutation can facilitate the switch between these two cell types. In addition to the mouse model, fetal programmed cells from human tumor samples showed the different differentiated states between two fetal cell subgroups (Extended Figures 7e-g). Trajectory analysis using entero-like type 1 and metaplasia populations revealed that metaplasia cells are more differentiated than entero-like type 1 cells, which is consistent with the mouse data (Extended Figure 7e). Overall, these results demonstrate different fetal programmed cell populations in diverse mouse models, and their transition associated with Kras mutational status. Apc and Braf mutations may initiate the acquisition of fetal characteristics, as shown in several studies 20,22 , but these mutations are not sufficient to promote the transition to other states. Kras mutation can facilitate the transition from the WNT-high to the WNT-low state in fetal programmed cells. Tumor-specific immune and stromal populations are actively linked to the fetal-programmed phenotype Fetal programmed cells are induced by mutations, but non-genetic factors can also force cells into a fetal state. Especially, the extracellular matrix (ECM), which is one of the key factors in the transition of epithelial cells into the fetal programmed cell state 14,45 . To understand the interaction between fetal programmed tumor cells and the tumor microenvironment (TME), we performed cell-cell interaction analysis using a curated ligand and receptor database 38 . Significant ligand-receptor interaction pairs among tumor and tumor microenvironment cells showed that entero-like type 1 cells and metaplasia cells had particularly strong interactions with TME cells, especially the stromal cell population (Figure 5a, Extended Figure 8a). The interaction of tumor and TME cells, especially with regard to ECM genes, showed the two fetal programmed populations in iCMS2 and iCMS3 (entero-like type 1 and metaplasia) could influence and be influenced by the stromal cell population (Figure 5b). Interestingly, CXC chemokine interactions between tumor and stromal cells were enriched in the iCMS2 tumor fetal programmed group. We identified CXC chemokine expression was predominant in the iCMS2 fetal programmed cells (entero-like type 1) (Figure 3c-d). Furthermore, the number of interactions between CXC chemokines ( CXCL1 , CXCL2 , CXCL3 , and CXCL8 ) and atypical chemokine receptor 1 ( ACKR1 ) was higher in entero-like type 1 than in other groups (Figure 5c-d). These findings suggests that the two subtypes of fetal programmed cells can interact with stromal cells in different ways. In addition to identifying stromal cells, we also confirmed that fetal programmed cells actively interact with T/NK lymphocytes (Extended Figure 8a). Among the significant interactions between tumor and T/NK cells, we found that the number of interactions for the immune checkpoint receptors, including HAVCR2 (TIM-3), LAG3 , and TIGIT, was higher in fetal programmed cells from each iCMS group than in other tumor cell populations (Figure 5e). Higher expression of immune checkpoint ligands was observed in the fetal-programmed population in each iCMS group, particularly metaplastic cells (Figure 5f). This suggests that fetal programmed cells could be a potential target for immunotherapy. Fetal programmed subtypes are prognostic biomarker for relapse in CRC As above, using single-cell RNA-sequencing data, we identified two groups of fetal reprogrammed cells with high YAP signaling (entero-like type 1 and metaplasia) that are affected by genetic and non-genetic factors. To understand whether these populations are prognostic, we used bulk gene expression data and clinical information from a previously published cohort 46 (Figures 6a-e). We first typed them for iCMS and confirmed that WNT activation is only enriched in iCMS2 samples (Extended Figure 9a), as seen in the single-cell data. The subgroups within each iCMS group are identified by estimating the proportion of cell types using the CIBERSORTx deconvolution method 47 (Figures 6a and 6c, Extended Figures 9b-c). Groups with high proportions of entero-like type 1 or metaplasia cells displayed higher expression of the fetal/YAP geneset, including three fetal modules (Figures 6b and 6d). Notably, these two groups are correlated with poor relapse and progression than other groups (Figure 6e). This suggests that the fetal programmed cell population is one of the most aggressive components in CRC tumors, accelerating cancer progression, and relapse. Moreover, compared with the CMS classification, CMS1 and CMS4 samples are more abundant in Entero-like type 1 high and Metaplasia high groups (Extended Figure 9d), suggesting that these populations engage in active crosstalk with the immune and stromal compartments. Discussion An exciting new paradigm is entering the field of colorectal cancer. Consistently, fetal reprogramming in CRC, from primary to metastasis, has been identified from diverse research 17 , 42 , 48 , 49 . These discoveries and definitions are fragmented with underlying mechanisms largely unknown. However, the consistency of the data forms an important contrast to the WNT-signaling dominated disease paradigm. Interestingly, fetal reprogramming feature can be affected by genetic alteration in mouse organoid 42 or tissue samples 22 , as well as tumor microenvironment from immunocompetent mice 49 . Much of the current data comes from mouse models, making it crucial to have a clear picture of what is going on in human tumors. To contribute to these new paradigms, we set out to take an unbiased view of the epithelial cell states in CRC tumors, integrating all emerging pre-clinical concepts in our analysis and data interpretation. In the iCMS classification system we had previously unveiled two distinct tumor cell groups that convey diverse genetic and transcriptomic profiles to elucidate the unique characteristics inherent in each subtype 13 . We now embarked on looking for further subclasses within the iCMS groups, aiming to identify the distribution of WNT activation and the novel fetal programming with YAP states across a comprehensive cohort of CRC samples. Within iCMS2 patients, we found predominant populations of LGR5 high stem cells that strongly express WNT signaling genes while iCMS3 patients lack in LGR5 gene expression and display predominance of a goblet-like cell population. In both iCMS2 and iCMS3 patients, we observed a fetal-programmed tumor cell population—entero-like type 1 in the iCMS2 group and metaplasia in the iCMS3 group—both displaying high levels of YAP signaling in different WNT activation status. These cells have been suggested to play an important role in CRC relapse. Additionally, we described the distinct characteristics of these aggressive cells with different WNT levels. The expression profile of the entero-like type 1 population is more enriched in the TNF-α signaling pathway, in contrast to metaplastic cells, which are more enriched in a metabolic pathway that includes fatty acid metabolism. This suggests that different WNT status (WNT-high vs. WNT-low) may influence their different properties. These findings are also supported by a variety of mouse models that have observed different WNT levels. In the mouse models, two fetal programmed cell populations exhibited similar characteristics to human fetal programmed cells. Specifically, the enterocyte-like group, had a high fetal module score as well as a high WNT score, whereas the metaplasia-like group only had a high fetal module score. In addition to the fetal and WNT scores, we confirmed that the two cell types had characteristics similar to human fetal programmed cell types. The enterocyte-like group had higher score of TNF-a signaling pathway enrichment, whereas the metaplasia-like group had a higher score of fatty acid metabolism enrichment. In primary cancers, fetal-like tumor cells having high YAP activity are thought to be associated with tumorigenesis, poor prognosis, chemotherapy resistance, and metastatic colonization 20 , 48 , 50 – 52 , and interestingly these cells are plastic during progression and metastasis 51 . Using human and mouse scRNA-seq data, we revealed that the distinct characteristics of fetal programmed cells could be regulated by genetic and non-genetic factors. Both human and mouse model exhibited fetal programmed cells, which were abundant in RAS/RAF mutation groups. Specifically, Kras mutation could facilitate the transition from the hyper- to the hypo-activated WNT state in fetal programmed cells, unlike Braf mutation which could only facilitate to be the hypo-activated WNT state only. Among various mouse models, AKPT and KPN mouse models showed more heterogenous patterns in fetal programmed cell population than other models, again advocating Kras mutation as an important factor to promote plasticity. In addition to genetic factors, we also revealed the supportive evidence that non-genetic factors can influence and be influenced by the fetal programmed cells. Using ligand receptor interaction analysis, the number of significant ligand receptor interactions between the two fetal programmed subpopulations and TME cells are higher than for the other tumor cell types, suggesting these cells can actively interact with microenvironment cells, especially, CAFs and ECs. Collectively, our study provides a comprehensive map of epithelial cell states at single-cell resolution in both human and mouse, highlighting the critical role of fetal programmed cells in colorectal tumors (Fig. 6f). Using scRNA-seq data, we demonstrated two different fetal programmed cell groups with different WNT activation states. Fetal programmed cells exhibit plasticity and can be modulated by genetic and non-genetic factors. Notably, we revealed that Kras mutation can facilitate the transition of fetal programmed cells from the WNT-high to WNT-low state. These conserved patterns of fetal characteristics in both human and mouse ensure the availability of appropriate pre-clinical models for each colorectal subtype. Online Methods Collection of patient tissue samples and scRNA-seq sequencing for colorectal cancer patients This study was approved by the institutional review boards of Commissie Medische Ethiek UZ KU Leuven/Onderzoek (approval no. S66460 and S63391) for the new in-house cohort. This study was carried out in accordance with ethical guidelines and all patients provided written informed consent. Tumor and adjacent normal tissue samples were rinsed with PBS and minced into pieces measuring <1mm 3 . The minced pieces were transferred to a digestion solution containing Collagenase P (Roche), DNAse I (Roche), and DMEM (Thermofisher Scientific), and incubated for 15 minutes at 37°C, shaking 3 times every 5 minutes. After incubation, samples were vortexed for 10 seconds and pipetted up and down for 1 minute. Samples were filtered using a 40µm nylon mesh (ThermoFisher Scientific), then 30ml of PBS was added. After centrifuging for 5 minutes at 4°C, the supernatant was discarded. Cell pellets were resuspended with RBC lysis (Roche) and incubated for 5 minutes at room temperature. They were centrifuged again for 5 minutes at 4 °C, then supernatant was discarded. The cell pellet was resuspended with buffer containing BSA (Invitrogen) and PBS, then filtered using a Flowmi Tip Strainer 40µm (Fisher Scientific) to remove debris. The fresh single-cell suspensions were loaded into the Chromium Chip G, following the manufacturer’s instructions. The Chromium Single Cell 5’ v1.1 kit was used for the generation of libraries, which were sequenced using NovaSeq6000, and processed using CellRanger v6.0.2. Pre-processing and quality control of scRNA-seq data from patient tissue samples The raw gene expression matrix was filtered using the following criteria. DropletUtils (1.14.2) 53 was used to remove empty droplets, and low-quality cells were removed using Unique Molecular Identifier (UMI) and number of genes. We only selected cells with >200 genes, >400 UMIs, and <25% of mitochondrial genes. DoubletFinder 54 (v2.0.3) was used for estimating doublets using the principal component from 1 to 10, and optimal pK value. After filtering out low-quality cells and doublets, the expression matrix was normalized and processed using Seurat’s standard pipeline (v4.1) 55 , 176,120 cells remained. Patient specificity was corrected using reciprocal principal component analysis (RPCA), and dimensional reduction and clustering was performed using RPCA output. The major cell types were annotated by the known marker gene expression, which was confirmed by CellTypist 56 , using the ‘Cells_Intestinal_Tract’ model with default parameters (Supplementary Table 2). As the proliferating cluster was mixed with multiple cell types having proliferative characteristics, we defined each cell type after extracting the proliferative cluster only. Assigned cell types were subclustered again by the same process, but for tumor epithelial cells, we filtered out those with <1,000 genes, which is the recommended approach for detecting malignant cells 57 . Classifying malignant cells from epithelial cells: In-house cohort We performed a gradient boosting method to exclude non-malignant populations from tumor samples from the in-house cohort (n=18), using a classifier that identifies non-malignant and malignant cells. Tumor and normal epithelial cells 13 were used to train the classifier to distinguish malignant cells. Epithelial cells in tumor samples with a prediction score exceeding 0.5 from the xgboost model were identified as malignant cells. It was confirmed that the remaining cells had patient specificity using a Chi-squared test (Extended Figure 1c–d). Intrinsic CMS (iCMS) scoring in malignant cells and identification of iCMS for each patient We identified iCMS labels using pseudobulk data from tumor cells for each patient. To calculate patient-level iCMS, we constructed a pseudobulk dataset by combining the UMI count matrix of tumor cells from each patient, except 1 patient who had 50 tumor cells. The count matrix of 17 patients was normalized using DESeq2 (v1.34.0) 58 , and PCA was performed using 715 iCMS signature genes with prcomp function in R. iCMS for each patient was identified using hierarchical clustering of iCMS signature genes and PCA analysis (Extended Figure 1e). Identification of shared programs from each gene module The auto-correlated gene modules were identified using Hotspot packages 24 with negative binomial models. The Top2000 genes were selected based on the Z-score of auto-correlations with significance (FDR < 0.01), and pair-wise gene associations were calculated. The modules were identified from a gene-gene affinity matrix, using the create_modules function, with min_genes_threshold = 40, core_only=False, and fdr_threshold=0.01, to recruit enough modules. Making the reference map using previously published cohort and label transfer to in-house cohort We performed sub-clustering analysis of multiple cell compartments from a previously published cohort that already identified major cell types (syn26844071), and only used samples from tumor or normal tissues. Specifically in tumor epithelial cells, only cells that identified as iCMS2 or iCMS3 were collected to eliminate the non-malignant population in tumor tissues. After extracting each cell type, RPCA was used to correct patient specificity. Before performing RPCA, we excluded 2 patients that had 50 cells. Using RPCA output, we performed non-linear dimensional reduction and clustering using Seurat functions. Clusters expressing genes that cannot possibly coexist were labelled as doublets and eliminated from the analysis. Data integration between the published cohort and in-house cohort was performed using label transfer from Seurat. Through FindTransferAnchors , TransferData , and MapQuery function, cell type and UMAP axes were projected from reference to in-house cohort samples. In total 51,444 malignant cells from 78 patients were used for the downstream analysis. Pseudobulk analysis in malignant human cells To reduce the variance between each patient 59 and find the strongest component affecting epithelial subpopulation in the iCMS2 and iCMS3 groups, we converted the UMI count matrix to pseudobulk RNA count by combining UMI counts per gene for each patient and for each major cell type with a count of >50 cells. The aggregated count matrix for 235 pseudobulk samples was transformed to DESeq2 format to perform normalization. Principal component analysis (PCA) was applied on scaled data using 2,000 variable genes using the prcomp function in R. The significantly correlated genes with each PC were selected by absolute of Pearson correlation coefficient (PCC) > 0.4 and P -value < 0.05. Ligand–receptor interaction within malignant cells, and between malignant cells and stromal or immune cells Cell–cell communication analysis was run using LIANA (v0.1.12) 38 , providing the aggregated rank from various methods and multiple databases. The aggregated rank was obtained from 5 different methods (NATMI, LogFC mean, Connectome, SingleCellSignalR, and cellphonedb) with consensus resource. The significant crosstalk lists were filtered using the value of aggregated rank as it can be interpreted as P- value (aggregate_rank < 0.05). Identification of signature genes Specifically expressed gene groups for each cell type were identified using MAST, implemented to FindAllMarker function in Seurat package. Significant genes were selected from the Top100 genes after passing the following criteria: log2FC > 0.25 and Adjusted P-value < 0.05. Gene signature scoring for each subtype The scoring for each signature was performed by calculating the average expression level of each cluster then subtracting the aggregated expression of control gene score sets. The control gene score was calculated by the average expression of randomly selected 100 genes, then replicated 10 times. The signatures that were used in this paper are summarized in Supplementary Table 4. Xenium sample preparation and analysis Xenium assays were run using the 322-gene pre-designed 10x Genomics Human Colon panel (1000642). Tissue sections of 5 µm were placed on the 10x Genomics Xenium slides, followed by drying for 30 min at room temperature, 3h at 42°C, and overnight in a desiccator at room temperature, according to the demonstrated protocol of the Xenium In Situ for FFPE - Tissue Preparation Guide (CG000578 – Rev C). Next, slides were processed within 2 days for deparaffinization and decrosslinking according to the demonstrated protocol of Xenium In Situ for FFPE – Deparaffinization & Decrosslinking (CG000580 – Rev C), followed by immediate proceeding to Xenium probe hybridization, ligation, amplification, and autofluorescence quenching according to the demonstrated protocol of Xenium In Situ for FFPE – Probe Hybridization, Ligation & Amplification User Guide (CG000582). Subsequently, slides were placed inside the 10x Genomics Xenium Analyzer instrument (LISCO, KU Leuven Institute for Single Cell Omics) for automated readout cycles and fluorescent image acquisition, on-board image pre-processing, signal decoding and cell segmentation. The Xenium output files were used for downstream analysis off-instrument. After the Xenium runs, slides were additionally processed for post-Xenium H&E staining according to the demonstrated protocol of the Xenium In Situ Gene Expression – Post-Xenium Analyzer H&E Staining (CG000613). Stained slides were imaged using a custom Nikon Ti2 8-stage system (LISCO). Analysis of spatial transcriptomics Xenium data were filtered using scanpy (v1.9.3) 60 . Cells expressing less than 10 transcripts were filtered, and genes expressed in less than 5 cells were filtered. The counts for each cell were normalized using scanpy.pp.normalize_total with default parameters, and then logarithmized. We computed principal component analysis and applied Uniform Manifold Approximation and Projection (UMAP) to perform dimensionality reduction. Cells were clustered into subgroups using Leiden clustering, implemented in scanpy.tl.leiden . To improve the accuracy of cell type annotation, we combined cell annotation based leiden clustering and label transfer based on single-cell RNA sequencing data. Leiden clusters were annotated using marker genes (Extended Figure 4). Subsequently, Robust Cell Type Decomposition (RCTD) was performed to identify cell type in Xenium. Single-cell RNA sequencing data with major cell type annotation was used as reference. Cells with low confidence in the status of "reject" and "doublet_uncertain" were removed, and only cells with robustness based on manual annotation and RCTD label transfer were retained to recalculate the principal components, neighborhood graphs, and visualized using UMAP graphs. Mouse tumor models Mouse experiments were carried out in accordance with the UK Home Office regulations, project licences PP3908577 and 70/9112, with the approval of the Animal Welfare and Ethical Review Board of the University of Glasgow. Mice were genotyped by Transnetyx (Tennessee, USA). Mice were housed in a specific pathogen-free facility in individual ventilated and conventional open top cages with a 12-hours light/dark cycle and free access to standard chow diet and drinking water. Five different intestinal cancer models were used in this study. Supplementary Table 5 summarizes the number, sex age, and genotypes used in this study. VilCreER Apc fl/fl (A, n=4)) and VilCreER Apc fl/fl Lsl-Kras G12D/+ (AK, n=3) models were induced with a single injection of 70 µL of 100 µM 4-hydroxytamoxifen (Sigma, H7904), via colonoscope-guided injection into the sub-mucosa wall of the mid colon using a Karl Storz TELE PACK VET X LED endoscopic video unit 61-64 . A and AK mice were sampled when they had a colonic tumor burden. VilCreER Braf V600E/+ Alk5 fl/fl (BA) were induced with a single IP injection of 2 mg tamoxifen (T5648) 65,66 . Then aged until showing clinical signs of intestinal tumorigenesis as previously described 22 . All mice were induced between the age of 7-20 weeks. Organoids used for the KPN transplant model (n=9) were derived from VillinCreER Kras G12D/+ Trp53 fl/fl R26 N1icd/+ mice 1 induced with 2 mg Tamoxifen intraperitoneal and aged to clinical end point. Tumors were taken for culture either from the small intestine (BVKPN RKAC13.1e, n=2) and GFVKPND GFSD106.2e, n=5) or from liver metastasis 1 (BVKPN RKAC3.2f, n=2). Organoids used for the AKPT transplant model (n=9) were derived from spontaneous small intestine tumors from VillinCreER Apc fl/fl Kras G12D/+ Trp53 fl/fl Tgfbr1 fl/fl mice: VAKPT RJV42.1a (n=4), VAKPT RJV51.1c (n=2) and VAKPT RJV6.2a (n=3) 1 . These organoid lines were injected by colonoscope-guided injections into the sub-mucosa of the mid colon of immune competent C5BL/6J mice (Charles River). Tumors were confirmed by colonoscopy and harvested for single cell RNA sequencing. Tumor Organoid Culture Organoids were grown in growth factor reduced Matrigel (Corning, Catalog No. 356231) in a 3D system. Advanced DMEM/F12 (Invitrogen, Catalog No. 12634–028) was supplemented with 2 mmol/L glutamine, 10 mmol/L HEPES, and 100 U/mL penicillin/streptomycin (ThermoFisher Scientific, 15140122), B27 (Invitrogen, Catalog No. 12587–010), and N2-supplement (Thermo Fisher Scientific, 17502001). For culturing organoids, 100 ng/mL Noggin (Peprotech, Catalog No. 250–38), and 50 ng/mL EGF (Peprotech, Catalog No. AF-100–15) were added. Organoids were cultured in six-well plates (BD Falcon) at 37°C in an atmosphere containing 5% CO2 and 95% air, with passaging every 2–3 days. Prior to injection, organoids from a single confluent well of 6-well cell culture plates (Corning, Catalog No. 353046) were harvested and dissociated by fiercely pipetting and were washed twice with PBS. Organoids were resuspended in 70 μl PBS and injected into a mouse in a single injection. Mouse tissue processing for scRNA-seq Mouse colonic tumors were excised into PBS and then chopped with a Mcllwan Tissue Chopper. Chopped paste was transferred to GentleMACS C tubes (Miltenyi Biotec, 130-093-237) containing enzyme mix (Miltenyi Biotec, 130-096-730; 2.35 mL of RPMI1640, 100 µL Enzyme D, 50 µL Enzyme R, and 12.5 µL Enzyme A). GentleMACS program (37C_m_TDK_1) was run on the GentleMACS Octo Dissociator with Heaters (Miltenyi Biotec, 130-096-427). The digested samples were spun and filtered through a 70 µm strainer with adding 10 mL of RPMI-10%FBS-2 mmol/L Ethylenediaminetetraacetic acid (EDTA). The supernatant from the suspension was discarded after centrifugation at 400 RPM for 3 minutes at 4°C. The cell pellet was resuspended with 0.5mL buffer containing 2% FBS, 25mM HEPES, 2mM EDTA in PBS, and transferred to a FACS collection tube on ice. Dissociated cells were sorted using a BD FACSAria (BD Biosciences) and DAPI (Thermo Fisher Scientific, 62248) to remove dead cells. Sorted cells were loaded into the Chromium Chip G, according to the manufacturer’s instructions. The reagents from 10x Chromium Single Cell 3’ v3 kit (10x Genomics) were used for the generation of libraries. Libraries were sequenced on a NovaSeq6000. Pre-processing and quality control of scRNA sequencing data from diverse mouse models Mouse sequences were aligned using Cellranger version 6.1.2 with reference genome mm10 version 3.0.0 using the developer’s standard workflow. Data were then analysed using the R package Seurat, and samples were filtered to include cells with > 100 genes, > 400 UMI counts, < 5% mitochondrial genes and < 10% haemoglobin genes. Samples were integrated using RPCA before performing data scaling, dimension reduction with PCA and clustering using the standard Seurat workflow. The resulting cells were annotated using CellTypist with the ‘Adult_Mouse_Gut’ reference, with cell types representing epithelial subtypes labelled as epithelial. To obtain high-quality epithelial cells, we applied tumor cell-specific cutoff (> 500 genes), similar to the criteria used for human tumor cells and subclustering using RPCA was performed. Additionally, sample-specific clusters or clusters that did not express epithelial cell markers were discarded. Analysis of tumor epithelial cells from mouse models Subclustering using RPCA was performed on identified malignant tumor cells. To apply the human gene signature to mice, we converted human genes to mouse orthologs using the babelgene R package (v22.9). The scoring for each signature was identical with the human signature scoring method previously mentioned. The gene set enrichment analysis (GSEA) was performed using a list of the differentially expressed genes using MAST, performed using FindMarkers function in Seurat with a threshold set to >10% of cells. The significance was determined by Benjamini-Hochberg adjusted P -value. Analysis of public bulk gene expression data and deconvolution based on CIBERSORTx Normalized gene expression data for CRC patient samples was obtained from GSE39582 46 and excluded patients without survival information. In total, 557 patients from GSE39582 were used for the downstream analysis. We classified iCMS based on the iCMS signature score from a previous paper 13 . The cell abundance from bulk gene expression was estimated using CIBERSORTx 47 with default parameters. First, a signature matrix was created using an internal single-cell reference matrix. To improve profile specificity across cell types, we only used signature genes for each cell type. Next, we imputed cell fractions from bulk gene expression using the signature matrix. GSEA was conducted by pre-ranked genes that were identified by a ordered list of differentially expressed genes with specific groups and others. Statistics The statistical analyses were performed using R (v4.1). To determine significance, results were compared using the two-sided Wilcoxon rank-sum exact test, the one-way ANOVA, and Pearson’s chi-squared test. For the analysis of differential gene expression for each group, P-values were adjusted using the Bonferroni correction. Detailed descriptions are in the Methods and Figure legends. Declarations Acknowledgements The IMMUcan project has received funding from the Innovative Medicines Initiative 2 Joint Undertaking under grant agreement No 821558. This Joint Undertaking receives support from the European Union’s Horizon 2020 research and innovation programme and EFPIA. IMI.europa.eu. S.T. is supported by the Belgian Foundation Against Cancer (FAF-C/2018/1301) and a BOF-Fundamental Clinical Research mandate (FKO) from KU Leuven. Y.H. is supported by a FWO Junior Postdoctoral fellowship (12D5823N). S.T. and S.V. are supported by the Research Foundation Flanders (G0C9620N, G067821N). S.T is supported by Stichting Tegan Kanker grant (2020-082). S.S., M.L.M., K.L.G., T.R.M.L., M.W., R.A.R., A.D.C., and O.J.S. were supported by Cancer Research UK core funding to the CRUK Scotland Institute (A31287) and a CRUK Core programme award to O.J.S (DRCQQR-May21\100002). M.L.M., K.L.G., T.R.M.L., A.D.C., and O.J.S. were supported by a Cancer Research UK Accelerator Award (A26825). S.S., M.L.M., K.L.G., A.D.C., and O.J.S. were supported by Mark Foundation ASPIRE I Award. S.S. and O.J.S. were supported by a CRUK Programme Grant (DRCNPG-Jun22\100002). M.W was supported by CRUK Clinical Academic Training Programme (A29706) and Chief Scientist Office (CSO PCL/24/01) lectureships. T.V and K.V are supported by KU Leuven (C14/22/125), the Research Foundation Flanders (I001818N, I009724N) and Vlaamse Veerkracht (VV023-06 “PRISMO”). K.V and N.V. are supported by the Research Foundation Flanders (G005923N). The authors would like to thank the Research Services at the CRUK Scotland Institute (CRUK SI): The Biological Service Unit (BSU), Histology, Molecular Technologies and Central Services, funded by CRUK core funding to the CRUK Scotland Institute (A31287). The authors also thank Catherine Winchester (CRUK SI) for critical review of the manuscript. Author contribution Conceptualization: Y.H., S.S., O.J.S., S.T. Data Curation: Y.H., K.L.G Formal analysis: Y.H., K.L.G., Z.H., F.D.V.B., A.O., S.V Funding Acquisition: O.J.S., S.T. Investigation: S.S., M.L.M., T.R.M.L., M.W., R.A.R., B.V.B., Y.M., L.L., V.P., A.M.P., N.V. Methodology: Y.H., S.S., M.L.M., K.L.G., Z.H. Resource: S.S., M.L.M., T.R.M.L., M.W., R.A.R., K.V., G.R., G.D.H., X.S., G.B., A.D.H., H.P., T.V. Supervision: A.D.C., O.J.S., S.T. Visualization: Y.H., Z.H., F.D.V.B., A.O. Writing – Original Draft: Y.H., S.S., O.J.S., S.T. Data availability The publicly available scRNA-seq and bulk gene expression dataset is referred to in the Methods (syn26844071 and GSE39582). Mouse scRNA-seq data for five mouse model is available under Gene Expression Omnibus (GEO), with accession number GSE280631. All other materials are available from the corresponding authors upon reasonable request. Competing interests O.J.S. has received research funding from AstraZeneca, Boehringer Ingelheim, Novartis. The other authors declare no competing interests. References Jackstadt, R. et al. Epithelial NOTCH Signaling Rewires the Tumor Microenvironment of Colorectal Cancer to Drive Poor-Prognosis Subtypes and Metastasis. Cancer Cell 36 , 319-336 e7 (2019). Canellas-Socias, A. et al. Metastatic recurrence in colorectal cancer arises from residual EMP1(+) cells. Nature 611 , 603-613 (2022). Vasquez, E.G. et al. Dynamic and adaptive cancer stem cell population admixture in colorectal neoplasia. Cell Stem Cell 29 , 1213-1228 e8 (2022). Chen, G.T. et al. Disruption of beta-Catenin-Dependent Wnt Signaling in Colon Cancer Cells Remodels the Microenvironment to Promote Tumor Invasion. Mol Cancer Res 20 , 468-484 (2022). Nguyen, L.H., Goel, A. & Chung, D.C. Pathways of Colorectal Carcinogenesis. Gastroenterology 158 , 291-302 (2020). Cancer Genome Atlas, N. Comprehensive molecular characterization of human colon and rectal cancer. Nature 487 , 330-7 (2012). Schmitt, M. & Greten, F.R. The inflammatory pathogenesis of colorectal cancer. Nat Rev Immunol 21 , 653-667 (2021). Dow, L.E. et al. Apc Restoration Promotes Cellular Differentiation and Reestablishes Crypt Homeostasis in Colorectal Cancer. Cell 161 , 1539-1552 (2015). Schepers, A.G. et al. Lineage tracing reveals Lgr5+ stem cell activity in mouse intestinal adenomas. Science 337 , 730-5 (2012). Barker, N. et al. Crypt stem cells as the cells-of-origin of intestinal cancer. Nature 457 , 608-11 (2009). Martinez-Ordonez, A. et al. Hyaluronan driven by epithelial aPKC deficiency remodels the microenvironment and creates a vulnerability in mesenchymal colorectal cancer. Cancer Cell 41 , 252-271 e9 (2023). Guinney, J. et al. The consensus molecular subtypes of colorectal cancer. Nat Med 21 , 1350-6 (2015). Joanito, I. et al. Single-cell and bulk transcriptome sequencing identifies two epithelial tumor cell states and refines the consensus molecular classification of colorectal cancer. Nat Genet 54 , 963-975 (2022). Yui, S. et al. YAP/TAZ-Dependent Reprogramming of Colonic Epithelium Links ECM Remodeling to Tissue Regeneration. Cell Stem Cell 22 , 35-49 e7 (2018). Chen, L. et al. TGFB1 induces fetal reprogramming and enhances intestinal regeneration. Cell Stem Cell 30 , 1520-1537 e8 (2023). Ayyaz, A. et al. Single-cell transcriptomes of the regenerating intestine reveal a revival stem cell. Nature 569 , 121-125 (2019). Han, T. et al. Lineage Reversion Drives WNT Independence in Intestinal Cancer. Cancer Discov 10 , 1590-1609 (2020). Cai, J., Maitra, A., Anders, R.A., Taketo, M.M. & Pan, D. beta-Catenin destruction complex-independent regulation of Hippo-YAP signaling by APC in intestinal tumorigenesis. Genes Dev 29 , 1493-506 (2015). Azzolin, L. et al. YAP/TAZ incorporation in the beta-catenin destruction complex orchestrates the Wnt response. Cell 158 , 157-70 (2014). Gregorieff, A., Liu, Y., Inanlou, M.R., Khomchuk, Y. & Wrana, J.L. Yap-dependent reprogramming of Lgr5(+) stem cells drives intestinal regeneration and cancer. Nature 526 , 715-8 (2015). Deng, F. et al. YAP triggers the Wnt/beta-catenin signalling pathway and promotes enterocyte self-renewal, regeneration and tumorigenesis after DSS-induced injury. Cell Death Dis 9 , 153 (2018). Leach, J.D.G. et al. Oncogenic BRAF, unrestrained by TGFbeta-receptor signalling, drives right-sided colonic tumorigenesis. Nat Commun 12 , 3464 (2021). Ogasawara, N. et al. Discovery of non-genomic drivers of YAP signaling modulating the cell plasticity in CRC tumor lines. iScience 27 , 109247 (2024). DeTomaso, D. & Yosef, N. Hotspot identifies informative gene modules across modalities of single-cell genomics. Cell Syst 12 , 446-456 e9 (2021). Li, H. et al. Reference component analysis of single-cell transcriptomes elucidates cellular heterogeneity in human colorectal tumors. Nat Genet 49 , 708-718 (2017). Pelka, K. et al. Spatially organized multicellular immune hubs in human colorectal cancer. Cell 184 , 4734-4752 e20 (2021). Kim, J.H. et al. Gastric-type expression signature in serrated pathway-associated colorectal tumors. Hum Pathol 46 , 643-56 (2015). Chen, B. et al. Differential pre-malignant programs and microenvironment chart distinct paths to malignancy in human colorectal polyps. Cell 184 , 6262-6280 e26 (2021). Wang, D. et al. CXCL1 induced by prostaglandin E2 promotes angiogenesis in colorectal cancer. J Exp Med 203 , 941-51 (2006). Zhuo, C., Ruan, Q., Zhao, X., Shen, Y. & Lin, R. CXCL1 promotes colon cancer progression through activation of NF-kappaB/P300 signaling pathway. Biol Direct 17 , 34 (2022). Yang, X. et al. Collagen 1-mediated CXCL1 secretion in tumor cells activates fibroblasts to promote radioresistance of esophageal cancer. Cell Rep 42 , 113270 (2023). Hollingsworth, M.A. & Swanson, B.J. Mucins in cancer: protection and control of the cell surface. Nat Rev Cancer 4 , 45-60 (2004). Sheng, Y.H. et al. MUC1-mediated Macrophage Activation Promotes Colitis-associated Colorectal Cancer via Activating the Interleukin-6/ Signal Transducer and Activator of Transcription 3 Axis. Cell Mol Gastroenterol Hepatol 14 , 789-811 (2022). Lee, C.K. et al. Tumor metastasis to lymph nodes requires YAP-dependent metabolic adaptation. Science 363 , 644-649 (2019). Caire, R. et al. YAP Transcriptional Activity Dictates Cell Response to TNF In Vitro. Front Immunol 13 , 856247 (2022). Shimomura, K. et al. Sleeping Beauty transposon mutagenesis identified genes and pathways involved in inflammation-associated colon tumor development. Nat Commun 14 , 6514 (2023). Fey, S.K., Vaquero-Siguero, N. & Jackstadt, R. Dark force rising: Reawakening and targeting of fetal-like stem cells in colorectal cancer. Cell Rep 43 , 114270 (2024). Dimitrov, D. et al. Comparison of methods and resources for cell-cell communication inference from single-cell RNA-Seq data. Nat Commun 13 , 3224 (2022). Ganesh, K. et al. L1CAM defines the regenerative origin of metastasis-initiating cells in colorectal cancer. Nat Cancer 1 , 28-45 (2020). Puram, S.V. et al. Single-Cell Transcriptomic Analysis of Primary and Metastatic Tumor Ecosystems in Head and Neck Cancer. Cell 171 , 1611-1624 e24 (2017). Malagoli Tagliazucchi, G., Wiecek, A.J., Withnell, E. & Secrier, M. Genomic and microenvironmental heterogeneity shaping epithelial-to-mesenchymal trajectories in cancer. Nat Commun 14 , 789 (2023). Mzoughi, S. et al. A Mutation-driven oncofetal regression fuels phenotypic plasticity in colorectal cancer. bioRxiv , 2023.12.10.570854 (2023). Lange, M. et al. CellRank for directed single-cell fate mapping. Nat Methods 19 , 159-170 (2022). Gulati, G.S. et al. Single-cell transcriptional diversity is a hallmark of developmental potential. Science 367 , 405-411 (2020). Qin, X. et al. An oncogenic phenoscape of colonic stem cell polarization. Cell 186 , 5554-5568 e18 (2023). Marisa, L. et al. Gene expression classification of colon cancer into molecular subtypes: characterization, validation, and prognostic value. PLoS Med 10 , e1001453 (2013). Newman, A.M. et al. Determining cell type abundance and expression from bulk tissues with digital cytometry. Nat Biotechnol 37 , 773-782 (2019). Moorman, A. et al. Progressive plasticity during colorectal cancer metastasis. bioRxiv , 2023.08.18.553925 (2023). Goto, N. et al. SOX17 enables immune evasion of early colorectal adenomas and cancers. Nature 627 , 636-645 (2024). Touil, Y. et al. Colon cancer cells escape 5FU chemotherapy-induced cell death by entering stemness and quiescence associated with the c-Yes/YAP axis. Clin Cancer Res 20 , 837-46 (2014). Heinz, M.C. et al. Liver Colonization by Colorectal Cancer Metastases Requires YAP-Controlled Plasticity at the Micrometastatic Stage. Cancer Res 82 , 1953-1968 (2022). Roulis, M. et al. Paracrine orchestration of intestinal tumorigenesis by a mesenchymal niche. Nature 580 , 524-529 (2020). Lun, A.T.L. et al. EmptyDrops: distinguishing cells from empty droplets in droplet-based single-cell RNA sequencing data. Genome Biol 20 , 63 (2019). McGinnis, C.S., Murrow, L.M. & Gartner, Z.J. DoubletFinder: Doublet Detection in Single-Cell RNA Sequencing Data Using Artificial Nearest Neighbors. Cell Syst 8 , 329-337 e4 (2019). Hao, Y. et al. Integrated analysis of multimodal single-cell data. Cell 184 , 3573-3587 e29 (2021). Dominguez Conde, C. et al. Cross-tissue immune cell analysis reveals tissue-specific features in humans. Science 376 , eabl5197 (2022). Gavish, A. et al. Hallmarks of transcriptional intratumour heterogeneity across a thousand tumours. Nature 618 , 598-606 (2023). 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). Zhang, M. et al. IDEAS: individual level differential expression analysis for single-cell RNA-seq data. Genome Biol 23 , 33 (2022). Wolf, F.A., Angerer, P. & Theis, F.J. SCANPY: large-scale single-cell gene expression data analysis. Genome Biol 19 , 15 (2018). Additional Declarations The authors declare potential competing interests as follows: O.J.S. has received research funding from AstraZeneca, Boehringer Ingelheim, Novartis. The other authors declare no competing interests. Supplementary Files ExtendedFigures.pdf 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-6812365","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":466550594,"identity":"7e09a235-8925-4dc5-8baa-6ec1dcd3929f","order_by":0,"name":"Yourae Hong","email":"","orcid":"https://orcid.org/0000-0001-6683-433X","institution":"KU Leuven","correspondingAuthor":false,"prefix":"","firstName":"Yourae","middleName":"","lastName":"Hong","suffix":""},{"id":466550595,"identity":"2b0fbf94-6006-4f53-a116-359ae8b950cc","order_by":1,"name":"Susanti Susanti","email":"","orcid":"https://orcid.org/0000-0002-6570-7117","institution":"Cancer Research UK Scotland Institute","correspondingAuthor":false,"prefix":"","firstName":"Susanti","middleName":"","lastName":"Susanti","suffix":""},{"id":466550596,"identity":"4596d98e-7bb9-4b6a-9497-bbfcbd443d0a","order_by":2,"name":"Megan L. Mills","email":"","orcid":"https://orcid.org/0009-0000-0922-7912","institution":"Cancer Research UK Scotland Institute","correspondingAuthor":false,"prefix":"","firstName":"Megan","middleName":"L.","lastName":"Mills","suffix":""},{"id":466550597,"identity":"dce8c11e-d083-4aaf-a01f-0b3cf2b8136f","order_by":3,"name":"Kathryn L. Gilroy","email":"","orcid":"https://orcid.org/0000-0003-4607-194X","institution":"Cancer Research UK Scotland Institute","correspondingAuthor":false,"prefix":"","firstName":"Kathryn","middleName":"L.","lastName":"Gilroy","suffix":""},{"id":466550598,"identity":"00f6a8e3-e191-40cf-be32-50fab1f3c8d5","order_by":4,"name":"Zedong Hu","email":"","orcid":"https://orcid.org/0009-0003-1410-6041","institution":"KU Leuven","correspondingAuthor":false,"prefix":"","firstName":"Zedong","middleName":"","lastName":"Hu","suffix":""},{"id":466550599,"identity":"1d628051-f104-4922-bd83-21af44a8aef9","order_by":5,"name":"Fausto Velez Bravo","email":"","orcid":"https://orcid.org/0000-0002-8596-3943","institution":"KU Leuven","correspondingAuthor":false,"prefix":"","firstName":"Fausto","middleName":"Velez","lastName":"Bravo","suffix":""},{"id":466550600,"identity":"8d1a9c06-c353-44e5-9937-808cfb185237","order_by":6,"name":"Aleksandra Ostapiuk","email":"","orcid":"https://orcid.org/0000-0002-9629-4773","institution":"KU Leuven","correspondingAuthor":false,"prefix":"","firstName":"Aleksandra","middleName":"","lastName":"Ostapiuk","suffix":""},{"id":466550601,"identity":"efe7ed62-4891-43c9-b9cc-4fbf31d6d835","order_by":7,"name":"Tamsin R. M. Lannagan","email":"","orcid":"https://orcid.org/0000-0002-8206-8898","institution":"Cancer Research UK Scotland Institute","correspondingAuthor":false,"prefix":"","firstName":"Tamsin","middleName":"R. M.","lastName":"Lannagan","suffix":""},{"id":466550602,"identity":"c0fe7411-e402-4ac4-a181-2982418dc81c","order_by":8,"name":"Mark White","email":"","orcid":"https://orcid.org/0000-0002-6309-3254","institution":"Cancer Research UK Scotland Institute","correspondingAuthor":false,"prefix":"","firstName":"Mark","middleName":"","lastName":"White","suffix":""},{"id":466550603,"identity":"2f2ee592-a3f9-4da5-8c27-74bc9d395aca","order_by":9,"name":"Rachel A. Ridgway","email":"","orcid":"https://orcid.org/0000-0003-0198-8244","institution":"Cancer Research UK Scotland Institute","correspondingAuthor":false,"prefix":"","firstName":"Rachel","middleName":"A.","lastName":"Ridgway","suffix":""},{"id":466550604,"identity":"2b2b1a65-6307-4369-9127-b5d3655b5131","order_by":10,"name":"Ben Van den Bosch","email":"","orcid":"","institution":"KU Leuven","correspondingAuthor":false,"prefix":"","firstName":"Ben","middleName":"Van den","lastName":"Bosch","suffix":""},{"id":466550605,"identity":"3596db0a-222e-4912-ad2d-36aa12305b39","order_by":11,"name":"Yasmine Morsa","email":"","orcid":"","institution":"KU Leuven","correspondingAuthor":false,"prefix":"","firstName":"Yasmine","middleName":"","lastName":"Morsa","suffix":""},{"id":466550606,"identity":"f079fa3d-6ac1-4f6e-831c-dc992f5be3e2","order_by":12,"name":"Lore Liekens","email":"","orcid":"","institution":"KU Leuven","correspondingAuthor":false,"prefix":"","firstName":"","middleName":"","lastName":"Lore Liekens","suffix":""},{"id":466550607,"identity":"8f821436-22fe-49a9-a7c3-fa22914024f1","order_by":13,"name":"Valentina Pomella","email":"","orcid":"","institution":"KU Leuven","correspondingAuthor":false,"prefix":"","firstName":"Valentina","middleName":"","lastName":"Pomella","suffix":""},{"id":466550608,"identity":"4b4786fd-854f-462b-bb30-e44908ef27a2","order_by":14,"name":"Allyson Moraig Peddle","email":"","orcid":"https://orcid.org/0000-0001-7597-5468","institution":"KU Leuven","correspondingAuthor":false,"prefix":"","firstName":"Allyson","middleName":"Moraig","lastName":"Peddle","suffix":""},{"id":466550609,"identity":"e57e763c-5a58-4464-8c20-2cce5ed7733c","order_by":15,"name":"Niels Vandermeulen","email":"","orcid":"","institution":"KU Leuven","correspondingAuthor":false,"prefix":"","firstName":"Niels","middleName":"","lastName":"Vandermeulen","suffix":""},{"id":466550610,"identity":"07879ec9-3723-44b7-8e71-36e7f28fcee1","order_by":16,"name":"Sara Verbandt","email":"","orcid":"https://orcid.org/0000-0003-1242-3716","institution":"KU Leuven","correspondingAuthor":false,"prefix":"","firstName":"Sara","middleName":"","lastName":"Verbandt","suffix":""},{"id":466550611,"identity":"d516b858-dcef-4c74-8750-87a50c149f86","order_by":17,"name":"Katy Vandereyken","email":"","orcid":"https://orcid.org/0000-0002-4477-5866","institution":"KU Leuven","correspondingAuthor":false,"prefix":"","firstName":"Katy","middleName":"","lastName":"Vandereyken","suffix":""},{"id":466550612,"identity":"d0b7977f-7921-4178-89eb-036cdc16a9c8","order_by":18,"name":"Gertjan Rasschaert","email":"","orcid":"","institution":"KU Leuven","correspondingAuthor":false,"prefix":"","firstName":"Gertjan","middleName":"","lastName":"Rasschaert","suffix":""},{"id":466550613,"identity":"88951aca-7456-4a05-a872-1daf5951858c","order_by":19,"name":"Gert De Hertogh","email":"","orcid":"https://orcid.org/0000-0001-8494-7725","institution":"UZ Leuven","correspondingAuthor":false,"prefix":"","firstName":"Gert","middleName":"","lastName":"De Hertogh","suffix":""},{"id":466550614,"identity":"431e8e39-3552-4395-b129-d439a1c1ebda","order_by":20,"name":"Xavier Sagaert","email":"","orcid":"https://orcid.org/0000-0003-3532-9984","institution":"UZ Leuven","correspondingAuthor":false,"prefix":"","firstName":"Xavier","middleName":"","lastName":"Sagaert","suffix":""},{"id":466550615,"identity":"a7e9f996-08fe-494d-b86c-d75cad9f6451","order_by":21,"name":"Gabriele Bislenghi","email":"","orcid":"https://orcid.org/0000-0002-9212-9287","institution":"UZ Leuven","correspondingAuthor":false,"prefix":"","firstName":"Gabriele","middleName":"","lastName":"Bislenghi","suffix":""},{"id":466550616,"identity":"be3bf378-0593-4e62-8677-25dd9efe7525","order_by":22,"name":"André D’Hoore","email":"","orcid":"https://orcid.org/0000-0002-6978-809X","institution":"UZ Leuven","correspondingAuthor":false,"prefix":"","firstName":"André","middleName":"","lastName":"D’Hoore","suffix":""},{"id":466550617,"identity":"d51aae1e-d715-4f76-8692-25a10d715aca","order_by":23,"name":"Hubert Piessevaux","email":"","orcid":"https://orcid.org/0000-0003-1193-0601","institution":"UCLouvain","correspondingAuthor":false,"prefix":"","firstName":"Hubert","middleName":"","lastName":"Piessevaux","suffix":""},{"id":466550618,"identity":"198dd32e-9a5f-43fa-a99f-1e893ca434e1","order_by":24,"name":"Thierry Voet","email":"","orcid":"https://orcid.org/0000-0003-1204-9963","institution":"KU Leuven","correspondingAuthor":false,"prefix":"","firstName":"Thierry","middleName":"","lastName":"Voet","suffix":""},{"id":466550619,"identity":"9e19992a-a4c3-47a1-a340-62971893e770","order_by":25,"name":"Andrew D. Campbell","email":"","orcid":"https://orcid.org/0000-0003-3930-1276","institution":"Cancer Research UK Scotland Institute","correspondingAuthor":false,"prefix":"","firstName":"Andrew","middleName":"D.","lastName":"Campbell","suffix":""},{"id":466550620,"identity":"8f87877a-2880-4a8b-a565-86443756e3cc","order_by":26,"name":"Owen J. Sansom","email":"data:image/png;base64,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","orcid":"https://orcid.org/0000-0001-9540-3010","institution":"Cancer Research UK Scotland Institute","correspondingAuthor":true,"prefix":"","firstName":"Owen","middleName":"J.","lastName":"Sansom","suffix":""},{"id":466550621,"identity":"37c5f10c-a9ef-4c00-85d6-70f450e1f758","order_by":27,"name":"Sabine Tejpar","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABFElEQVRIie2RMUsDMRTH3yGkSzTry9SvcHJQEAv3VU6EutRD6VKhSG45l4rrCeJn6FTcTHnQKde5juJahNKpg2KO6ySE3iiY35LwJz/e/xEAj+dPwkHXl+oYAhzt8mSfgrViAFgTBWrFEuQNlPZdOaOrIdyKoJxtbp4pZq1xB9cvkAqHEpo0ocIASrU4l+WUDhg3HfloYCCVQ4F+SIc5YKhNKLMpMYb9KLLJ2UQ7ij2srPINGGsTbbMn4qy9iqIvq7w6FFhWU5SdArZPpggZ8uOPoJri2mX5mRCfoyz0onei5hch473r9/scB+gsdkkbPuoKUZT0pkansWjRRG/zbioc6++w34Lmd7IXMW7wyOPxeP4lP1VrWfJkk42NAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0003-3281-8643","institution":"KU Leuven","correspondingAuthor":true,"prefix":"","firstName":"Sabine","middleName":"","lastName":"Tejpar","suffix":""}],"badges":[],"createdAt":"2025-06-03 14:35:53","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":true,"conflictsOfInterestStatement":true,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":true},"doi":"10.21203/rs.3.rs-6812365/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6812365/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":84044385,"identity":"7a93e44e-7c5b-4b50-89d0-b6ca6cc441bb","added_by":"auto","created_at":"2025-06-06 06:58:01","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":5665756,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDifferent colorectal cancer molecular subtypes reveal fetal gene modules with YAP activation.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea,\u003c/strong\u003e Schematic of study design including sample collection and single-cell analysis from colorectal cancer patients and mouse models with four different genotypes. Created with Biorender.com\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eb,\u003c/strong\u003e Heatmap showing z-scored autocorrelation of selected Top 2000 genes implemented in Hotspot. Autocorrelation values were calculated for each gene across 51,444 tumor cells from 78 patients. Genes are grouped into 21 modules and are shown as color strips on the left and top.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ec,\u003c/strong\u003e GSEA plots of the seven selected gene modules to compare 24,316 malignant cells from iCMS3 patients (n = 32) and 27,128 cells from iCMS2 patients (n = 46). \u003cem\u003eP\u003c/em\u003e-values were adjusted by Benjamini-Hochberg correction. NES, normalized enrichment score.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ed-e,\u003c/strong\u003e Spatial scatter plot of two representative sections in iCMS2 and iCMS3 patients (left, iCMS2; right, iCMS3). Each dot indicates major cell types (left), and expression of \u003cem\u003eANXA1 \u003c/em\u003e(right) at the spatial level.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-6812365/v1/8aa68e285c25a2b79322d1ba.png"},{"id":84043426,"identity":"f593b910-40ea-49e5-97dd-4d3304028f4a","added_by":"auto","created_at":"2025-06-06 06:42:01","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":5784846,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFetal-like cell phenotypes in each iCMS group show high activation of YAP with different canonical WNT levels.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea,\u003c/strong\u003e UMAP projections of 27,128 malignant cells from 46 iCMS2 patients (left) and 24,316 malignant cells from 32 iCMS3 patients (right), are shown for each cell type. Each dot is colored according to cell type.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eb,\u003c/strong\u003e UMAP visualization showing the scaled score of epithelial cell signatures in each iCMS group (top, iCMS2; bottom, iCMS3). Colors represent scaled score of each signature.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ec,\u003c/strong\u003e Heatmap showing Pearson’s Correlation Coefficient (PCC) between the score from the module (black) and the signature from iCMS2/iCMS3 cell types (iCMS2 cell type signature; purple, iCMS3 cell type signature; orange).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ed,\u003c/strong\u003e Network analysis to identify the similarity between each iCMS cell type signature (left, iCMS2 group; right, iCMS3 group) and gene modules. Only edge passing PCC \u0026gt; 0.4 and P-value \u0026lt; 0.05 were selected. Edge thickness represents PCC, and nodes indicate the type of gene signatures (black, gene module; purple, iCMS2 cell type signature; orange, iCMS3 cell type signature).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ee,\u003c/strong\u003e Percentage of cells with high levels of the indicated gene signatures (WNT, YAP, and Fetal) across cell types, color-coded by the iCMS group they belong to.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ef,\u003c/strong\u003e PCA plot on the pseudobulk data, annotated with their corresponding iCMS groups (top) or cell types (bottom). Each dot represents the pseudobulk values for each cell type for each patient (n = 235).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eg,\u003c/strong\u003e Association with three indicated signatures (WNT, YAP, and Fetal) and PC1 or PC2 components from PCA analysis using pseudobulk data. Heatmaps above show the scaled score of three signatures according to PC1 (left) or PC2 (right) order. Heatmaps show genes that are positively or negatively correlated with PC1 or PC2, respectively (bottom).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eh, \u003c/strong\u003eScatter plots indicate the correlation between each signature (WNT, YAP, and Fetal) and PC1 (top) or PC2 (middle and bottom). \u003cem\u003eP\u003c/em\u003e-values were calculated using the Pearson Correlation Coefficient.\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-6812365/v1/2bffb28b1fde3a2a3fa30a2f.png"},{"id":84042664,"identity":"ffacd210-d6f8-4e9e-9061-e58bc5b21041","added_by":"auto","created_at":"2025-06-06 06:34:01","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":1812130,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDifferent characteristics within the fetal-programmed populations from iCMS2 and iCMS3 groups.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea,\u003c/strong\u003e Scaled score of the WNT and fetal-associated modules for cell subtypes in each iCMS group. The rows indicate cell types and their corresponding iCMS groups (purple. iCMS2; orange, iCMS3).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eb,\u003c/strong\u003e GSEA plots comparing the WNT or fetal-associated modules in metaplastic cells from iCMS3 (n = 6,943) versus entero-like type 1 cells from iCMS2 (n = 4,793). Benjamini-Hochberg adjusted \u003cem\u003eP\u003c/em\u003e-values.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ec,\u003c/strong\u003e Volcano plot (left) and selected differentially expressed genes (right) between metaplasia versus entero-like type 1 populations. Top 20 significantly different genes between two populations are indicated by color. All adjusted \u003cem\u003eP\u003c/em\u003e-values in the violin plot are less than 0.05. \u003cem\u003eP\u003c/em\u003e-values were adjusted by Bonferroni correction. log2FC, log2 fold change.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ed,\u003c/strong\u003e GSEA analysis using hallmark genesets from two different populations (metaplasia versus. entero-like type 1). Pathways with positive NES represent those enriched in the metaplastic cell groups, whereas negative NES indicates pathways enriched in entero-like type 1.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ee,\u003c/strong\u003e GSEA enrichment of the two selected pathways that are significantly differentially expressed in metaplastic cells versus. entero-like type 1 (top) with clustered heatmap of core genes for each pathway (bottom). Benjamini-Hochberg adjusted \u003cem\u003eP\u003c/em\u003e-values.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ef,\u003c/strong\u003e The number of significant ligand-receptor interaction pairs between each tumor cell type in each iCMS groups (left, iCMS2 group; right, iCMS3 group). The color of the heatmap indicates the number of significant interactions between each tumor cell type, and the bar plots show the significant sender (right) or receiver (top).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eg-h, \u003c/strong\u003eSelected ligand receptor pairs that significantly interact within each iCMS group. Each interaction pair is represented by color and only significant interactions (aggregate_rank \u0026lt; 0.05) are shown. iCMS2 fetal programming cells significantly interact through \u003cem\u003eL1CAM\u003c/em\u003e ligand (g), while iCMS3 fetal programming cells significantly crosstalk through \u003cem\u003eTFF1\u003c/em\u003e or \u003cem\u003eTFF2\u003c/em\u003e ligands (h).\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-6812365/v1/3402e8153337dcc5c79e345b.png"},{"id":84043429,"identity":"ff602f49-0899-4cd5-85d1-abf7998d9c9d","added_by":"auto","created_at":"2025-06-06 06:42:02","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":3692704,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eConserved patterns between human and mouse according to various WNT status from different genotypes and the transition of fetal programmed cells according to genetic alteration.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea,\u003c/strong\u003e UMAP projection of 6,360 malignant cells from 28 samples derived from five different mouse models (A, n = 4; AK, n = 3; AKPT, n = 9; KPN, n = 9; BA, n = 3) (top) and relative proportion of malignant cell types (bottom). Each color in the UMAP and bar graph indicate different cell types.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eb,\u003c/strong\u003e Network analysis of each iCMS cell type signature versus the signature from mouse cell types. Edge thickness represents PCC regarding the scores between each cell type, and node indicates signature types (white, moue cell type signature; purple, iCMS2 cell type signature; orange, iCMS3 cell type signature). Network analysis is performed using pairs with PCC \u0026gt; 0.4 and P-value \u0026lt; 0.05.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ec,\u003c/strong\u003e Scaled score of WNT/fetal modules (top) and average expression of selected genes associated with WNT/fetal signature for each cell type (bottom).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ed,\u003c/strong\u003e UMAP visualization of the scaled score for the five indicated cell signatures. The colors represent the scaled score for each signature.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ee,\u003c/strong\u003e Volcano plot between enterocyte-like and metaplasia-like groups. Genes that were significantly differentially expressed between the two groups (absolute FC \u0026gt; 2 and adjusted \u003cem\u003eP\u003c/em\u003e-value \u0026lt; 0.05) were colored in the plot (red, enterocyte-like; purple, metaplasia-like). All adjusted \u003cem\u003eP\u003c/em\u003e-values in the violin plot are less than 0.05. \u003cem\u003eP\u003c/em\u003e-values were adjusted by Bonferroni correction. log2FC, log2 fold change.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ef,\u003c/strong\u003e GSEA plots of the three selected pathways that are significantly differentially expressed in metaplasia-like versus. enterocyte-like populations. Benjamini-Hochberg adjusted \u003cem\u003eP\u003c/em\u003e-values.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eg, \u003c/strong\u003eComparison of pseudotime score between two fetal programmed cell populations. Boxplots inside the violin plots depict the median and interquartile range (IQR). The \u003cem\u003eP\u003c/em\u003e-value was calculated by Wilcoxon rank-sum exact test (two-sided).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eh, \u003c/strong\u003eThe PCA plot of 3,755 fetal programmed cells from samples from the five mouse models, colored by cell type. Arrows indicate direction of transition as determined by the CellRank method.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ei, \u003c/strong\u003eUMAP visualization depicting the distribution of each genotype in fetal programmed cells. Each color except gray represents the five mouse models respectively.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-6812365/v1/700ed05217d3d2547c141fa8.png"},{"id":84042663,"identity":"34c55bb2-74e6-4fa0-870e-23b97b4f71ac","added_by":"auto","created_at":"2025-06-06 06:34:01","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":2093236,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eThe crosstalk between fetal programmed tumor cells and the microenvironment.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea-b,\u003c/strong\u003e The number of significant ligand-receptor interaction pairs between tumor cell populations tumor microenvironment (TME) cells within each iCMS group (left; iCMS2 tumor subtypes, right; iCMS3 tumor subtypes). Two bar plots on the left indicate the significant interaction pairs of all ligand-receptor pairs (a) and other two plots represent the significant interaction pairs of ECM genes only (b). On the X-axis, the black color indicates the sender, and the gray color indicates the receiver.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ec-f,\u003c/strong\u003e The significant ligand-receptor interaction pairs and their ligand expressions associated with chemokine signaling (c-d), and immune checkpoint molecules (e-f). The circos plots show significantly interacting ligand-receptor pairs between tumor cells and TME cells within each iCMS group (c and e). Each interaction pair represents by color, only significant interactions (aggregate_rank \u0026lt; 0.05) are shown. The number of significant interactions is indicated by numbers in brackets. The scaled expression of selected ligands for cell subtypes in each iCMS group (d and f). The rows indicate cell types and their corresponding iCMS groups (purple, iCMS2; orange, iCMS3) and columns represent selected ligands.\u003c/p\u003e","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-6812365/v1/8f03aa0ac20d6a47fbd8e148.png"},{"id":84042668,"identity":"d0cebf88-d04a-4193-8f66-7899f6e73aa9","added_by":"auto","created_at":"2025-06-06 06:34:01","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":1107031,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eThe different abundance of subpopulations within each iCMS group and their association with poor relapse-free survival.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea-d,\u003c/strong\u003e Estimated proportion of each tumor subtype within each iCMS group. The hierarchical clustered heatmap shows the estimated proportion of each cell population from 305 iCMS2 tumor samples (a) or 252 iCMS3 tumor samples (c). GSEA analysis between Entero-like type 1\u003csup\u003ehigh\u003c/sup\u003e group (b) or Metaplasia\u003csup\u003ehigh\u003c/sup\u003e group (d) versus other groups using fetal, YAP genesets and gene modules. adjusted \u003cem\u003eP\u003c/em\u003e-values were from Benjamini-Hochberg.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ee,\u003c/strong\u003e Kaplan-Meier relapse-free survival curves of 557 patients grouped by estimated proportion. \u003cem\u003eP-\u003c/em\u003evalue was determined by log-rank test.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ef,\u003c/strong\u003e Schematic sketch of CRC tumor cell landscape and representative mouse models according to WNT and fetal/YAP axes. Created with Biorender.com.\u003c/p\u003e","description":"","filename":"Figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-6812365/v1/873596289d9412fff80d0cba.png"},{"id":84044573,"identity":"752aaf05-ee52-4178-8b99-64a41892fa5e","added_by":"auto","created_at":"2025-06-06 07:06:17","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":20505622,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6812365/v1/dba93e4b-2b98-44c2-8b96-d22dd88c9661.pdf"},{"id":84042671,"identity":"7d01b58e-34ab-4ad2-9698-6aa1dbca70cd","added_by":"auto","created_at":"2025-06-06 06:34:01","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":4014023,"visible":true,"origin":"","legend":"","description":"","filename":"ExtendedFigures.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6812365/v1/87f728e8437c91f504ddd624.pdf"}],"financialInterests":"The authors declare potential competing interests as follows: O.J.S. has received research funding from AstraZeneca, Boehringer Ingelheim, Novartis.\nThe other authors declare no competing interests.\n","formattedTitle":"\u003cp\u003eEpithelial states in colorectal cancer are co-determined by YAP associated fetal programming and WNT signaling\u003c/p\u003e","fulltext":[{"header":"Introduction","content":"\u003cp\u003eIn recent years, a novel paradigm in colorectal cancer (CRC) has emerged questioning the dominant role of Wnt signaling in tumor progression and metastasis\u003csup\u003e\u003cspan additionalcitationids=\"CR2 CR3\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. In the reigning paradigm, CRC can follow one of several molecular pathways in its development including WNT, chromosomal instability (CIN), and microsatellite instability (MSI)\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. But most commonly, CRC pathogenesis is driven by mutations in the adenomatous polyposis coli (\u003cem\u003eAPC\u003c/em\u003e) gene, encoding APC protein that plays a pivotal role in tumor initiation and is a crucial component in the WNT signaling pathway\u003csup\u003e\u003cspan additionalcitationids=\"CR6 CR7\" citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. Leucine-rich repeat-containing G protein-coupled receptor 5 (\u003cem\u003eLGR5\u003c/em\u003e) is an established marker of normal adult stem cells and is encoded by canonical WNT signaling\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e. The abundance of \u003cem\u003eLGR5\u003c/em\u003e during tumor initiation underscores the essential role played by WNT activation in tumor carcinogenesis\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. However, cancers with weak expression of \u003cem\u003eLGR5\u003c/em\u003e\u003csup\u003e3\u003c/sup\u003e or lacking mutations associated with WNT pathways\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e suggest that CRC progression is feasible even with weak or absent WNT activity. Furthermore, our previous work on consensus molecular subtypes (CMS) from bulk transcriptomic and intrinsic-CMS (iCMS) single-cell RNA sequencing analyses demonstrated varying levels of WNT signaling\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e,\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eFollowing colonic injury, low levels of WNT with \u003cem\u003eLGR5\u003c/em\u003e deficiency are observed, driving transition from a typical adult stem cell to those expressing fetal stem cell markers\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e. Damage-induced epithelial cells show high expression of fetal/revival stem cell markers as well as YAP (Yes-associated protein) target genes\u003csup\u003e\u003cspan additionalcitationids=\"CR15\" citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e. This phenomenon suggests that YAP activation is critical in maintaining fetal characteristics during injury. Fetal programmed tumor cells were found to display active YAP signaling\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e,\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e. In tumors, a similar decrease of \u003cem\u003eLGR5\u003c/em\u003e has been observed with the emergence of fetal-like cells expressing \u003cem\u003eANXA1\u003c/em\u003e, a widely used marker for identifying fetal programmed cells, accompanied by YAP signaling\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eRegulation of YAP signaling during the development of CRC is complex and poorly understood. One postulated mechanism suggests that during both adenoma formation and tumorigenesis, WNT signaling activates YAP through an \u003cem\u003eAPC\u003c/em\u003e mutation, either independent of its involvement in the β-catenin complex\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e, or through the β-catenin destruction complex\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e. And the activated YAP suppresses the WNT homeostasis gene expression program\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e. In addition, over-expression of YAP induces the expression of WNT transcription targets \u003cem\u003eLGR5\u003c/em\u003e, as well as cyclin D1 (\u003cem\u003eCCND1\u003c/em\u003e), via its interaction with β-catenin in the nucleus\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e. Thus, the specific mechanisms by which YAP impacts WNT signaling can vary dependent on the cellular context, since both feedforward and negative feedback loops have been described to date. YAP activation is not only associated with the WNT signaling pathway, but is also regulated by the extracellular matrix (ECM), injury repairing status, oncogenic mutations such as BRAF\u003csup\u003eV600E\u003c/sup\u003e, and atypical protein kinase C (aPKC) deficiency\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e,\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e,\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e,\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eTo elucidate these states, we set out a study of human CRC patient samples in their full complexity and combined this with analysis of mouse models of CRC that have different genetic alterations in which the epithelial status is controlled, achieved through genetic manipulation. We provide an unbiased view of the diverse single-cell epithelial states in human tumors regarding different WNT activation and other key features such as fetal programming and YAP signaling in both human and mouse models and observed a conserved pattern of fetal programmed cells with high YAP activation. In particular, \u003cem\u003eKras\u003c/em\u003e mutation is relevant for driving the transition status of fetal programmed cells from WNT-high to WNT-low states, whereas \u003cem\u003eApc\u003c/em\u003e or \u003cem\u003eBraf\u003c/em\u003e mutant models are not involved in the transition, and the characteristics of fetal programmed cells are conserved. We also recapitulated a conserved pattern of fetal programmed cells marked by high YAP activation. This study proposes a new paradigm for the cellular heterogeneity of CRC by exploiting functional status, which could provide new therapeutic targets for inhibiting tumor progression.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eDiverse cell states from colorectal cancer cells across different molecular subtypes\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo identify different cellular phenotypes based on status, we analyzed untreated primary CRC tissues from 81 patients, integrating previously published data from 63 patients with confirmed iCMS classifications\u003csup\u003e13\u003c/sup\u003e and 18 in-house patients (Figure 1a, Extended Figures 1a\u0026ndash;b, Supplementary Table 1\u0026ndash;2).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo classify malignant cells in the 18 in-house tumor tissue samples, gradient boosting was employed using malignant cells from the previously defined cohort (Extended Figures 1c\u0026ndash;d). The iCMS classification for the in-house cohort was applied only to the malignant cells, resulting in the identification of 10 patients as iCMS2 and seven patients as iCMS3, except for one patient who had an insufficient number of tumor cells (Extended Figures 2a\u0026ndash;b). We also excluded two patients with low number of tumor cells, resulting in 78 patients\u0026rsquo; tumor tissues samples being used for downstream analysis.\u003c/p\u003e\n\u003cp\u003eTo understand the states of malignant cells across the iCMS2 and iCMS3 groups, we performed hotspot analysis\u003csup\u003e24\u003c/sup\u003e using 78 patients and identified 18 out of 21 modules to be associated with specific biological characteristics (Figure 1b, Supplementary Table 3). Most of the 18 gene modules (n=16) were significantly enriched in either the iCMS2 groups or the iCMS3 groups, respectively (Figure 1c, Extended Figure 3a). Similar to a previous study\u003csup\u003e13\u003c/sup\u003e, the gene modules associating with metaplasia (Module 3) and goblet (Module 7) were enriched in iCMS3, and those associated with canonical WNT (Module 19, 20) were enriched in iCMS2 (Figure 1c). Three modules (Module 8, 11, and 15) were identified as fetal-associating modules. Of these, Module 15 (fetal-III) was enriched in iCMS2, while Modules 8 and 11 (fetal-I and fetal-II, respectively) were enriched in iCMS3. The implication is that fetal-associated characteristics can manifest in both iCMS2 (hyperactivated WNT status tumors) and iCMS3 (hypoactivated WNT status tumors) settings. Not only is YAP activation a key player in maintaining fetal characteristics during injury repair, but it also plays a crucial role within tumor systems\u003csup\u003e3,17\u003c/sup\u003e. The pivotal role of YAP is further substantiated by its consistent presence in a human tumor scRNA-seq dataset (Extended Figure 3b). Furthermore, Human colon panel Xenium spatial transcriptome data from two CRC patients showed the diversity of tumor cell states, including fetal programmed cell status (Extended Figure 4). The patient that identified as iCMS2 showed predominant expression of \u003cem\u003eLGR5\u003c/em\u003e in their tumor cells, while a tumor sample from the iCMS3 patient predominantly expressed \u003cem\u003eANXA1\u003c/em\u003e, a fetal marker (Extended Figure 4). Interestingly, in line with the single cell data, we found also one cluster expressing \u003cem\u003eANXA1\u003c/em\u003e (Tumor subtype 11) in the iCMS2 patient, showing that some tumor cells in iCMS2 patients also undergo fetal programming (Figures 1d\u0026ndash;e and Extended Figure 4). \u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTwo fetal programmed cell phenotypes in both iCMS groups in the presence of differing WNT levels\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo understand the molecular characteristics of hyperactivated and hypoactivated WNT signaling groups (iCMS2 and iCMS3 tumors, respectively), subpopulations of iCMS2 and iCMS3 tumor cells from 78 patients were identified (Figure 2a). Interestingly, we determined that the transcriptional properties of malignant cells are shared with the characteristics of normal epithelial cells, similar to previous research\u003csup\u003e25,26\u0026nbsp;\u003c/sup\u003e(Figure 2b).\u003c/p\u003e\n\u003cp\u003eIn the iCMS2 group, there are six subtypes encompassing stem-like cells with high expression of \u003cem\u003eLGR5\u003c/em\u003e (LGR5\u003csup\u003ehigh\u003c/sup\u003e stem), highly proliferative cells, two populations expressing enterocyte signatures (enterocyte-like type 1 and 2), goblet-like cells with \u003cem\u003eMUC2\u003c/em\u003e expression, and a small subset expressing tuft cell markers (Figures 2a\u0026ndash;b and Extended Figure 5a). In the iCMS3 group, six subtypes were also identified, including an undifferentiated population expressing \u003cem\u003eOLFM4\u003c/em\u003e, highly proliferative cells, goblet-like cells with \u003cem\u003eMUC2\u0026nbsp;\u003c/em\u003eexpression, and metaplastic cells expressing metaplasia signatures (\u003cem\u003eANXA10, CTSE, MUC5AC,\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;TFF2\u003c/em\u003e), recognized as gastric-type cell markers in CRC and serrated polyps\u003csup\u003e27\u003c/sup\u003e (Extended Figures 5a-b). Moreover, metaplastic cells exhibit shared signatures of both the absorptive lineage and goblet cell lineage (Figure 2b). Additionally, they display a pronounced metaplasia signature, as previously shown in serrated colorectal polyps\u003csup\u003e28\u003c/sup\u003e. To understand the association between malignant cell states and cell types, we analyzed relationships between gene module expression and cell type signatures (Figures 2c\u0026ndash;d). Entero-like type 1 and 2 cell types in iCMS2 groups were positively correlated with fetal and WNT modules, while metaplastic cell types in iCMS3 groups were positively correlated with fetal, EMT, goblet, and metaplasia modules. The entero-like type 1 in the iCMS2 group, and the metaplasia population in the iCMS3 group both expressed high levels of fetal programming and YAP signatures, even in the presence of varying WNT levels (Figure 2e, Extended Figure 5c). This indicates that fetal-programmed tumor cells with YAP dependency give rise to two different phenotypes in different WNT levels: entero-like type 1 cells and metaplastic cells.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eBoth the module approach and cell type identification approach demonstrate the single presence or combination of WNT and YAP states that characterize the CRC states. Additionally, pseudobulk analysis based on cell types from each patient identified that WNT activation and YAP, fetal reprogramming signaling are the important factors to distinguish cell types (Figures 2f-h, Extended Figure 5d) in iCMS2 and iCMS3 groups. \u0026nbsp;In particular, principal component 1 (PC1) shows the positive correlation of WNT-association genes (\u003cem\u003eASCL2, AXIN2\u003c/em\u003e, and \u003cem\u003eNKD1)\u003c/em\u003e, and PC1 can clarify the group from iCMS2 patients and iCMS3 patient. While YAP target (\u003cem\u003eF3, GADD45A, MSLN,\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;MYOF\u003c/em\u003e) and fetal (\u003cem\u003eANXA1, PLAUR,\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;EMP1\u003c/em\u003e) genes are negatively correlated with PC2 and distinguish the entero-like type 1 population in iCMS2 and metaplasia in iCMS3 from other cell types (Figures 2f-g). \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eWithin the cell types harboring fetal programming features, WNT levels are linked to distinct characteristics and genotypes\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eApplying two WNT-associated modules and three fetal-associated modules to each cell type, we observed different activation profiles between WNT and fetal modules (Figures 3a\u0026ndash;b). Within the three fetal modules, the entero-like type 1 group was enriched in fetal-III, and the metaplastic cell groups were enriched in fetal-I and II. We performed differential gene expression analysis between the two fetal programmed cell types. Results revealed that the enterocyte-like type 1 group exhibits higher expression of chemokine ligand groups, including \u003cem\u003eCXCL1, CXCL2, CXCL3, and CCL20\u0026nbsp;\u003c/em\u003e(Figure 3c). In particular, the \u003cem\u003eCXCL1\u003c/em\u003e gene can recruit immune cell and stromal cells as well as promoting cancer progression and migration\u003csup\u003e29-31\u003c/sup\u003e. The metaplastic cell group showed higher expression of mucin genes (\u003cem\u003eMUC1, MUC2,\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;MUC5B\u003c/em\u003e), known to contribute to the formation of mucosal layers and exert control over their microenvironment\u003csup\u003e32,33\u003c/sup\u003e. The distinctions between these two populations extend beyond the gene level to the pathway level, indicating distinct characteristics. GSEA results also showed enriched TNF-\u0026alpha; signaling via NF-kB in the entero-like type 1 population, and glycolysis and fatty acid metabolism in metaplastic cells (Figures 3d\u0026ndash;e). Interestingly, these pathways are associated with YAP activation level and fetal programming. Fatty acid metabolism is enriched in the metastatic process with high YAP activity\u003csup\u003e34\u003c/sup\u003e. TNF can activate YAP by inducing the nuclear translocation process of YAP\u003csup\u003e35\u003c/sup\u003e, and inducing a fetal state\u003csup\u003e36,37\u003c/sup\u003e. Additionally, entero-like type 1 and metaplasia cell populations are associated with different genotypes, \u003cem\u003eKRAS\u003c/em\u003e and \u003cem\u003eBRAF\u003c/em\u003e mutation, respectively (Extended Figure 6a).\u003c/p\u003e\n\u003cp\u003eFurthermore, we performed a ligand-receptor analysis to understand whether fetal programmed cell populations can actively affect other tumor cells (Figure 3f)\u003csup\u003e38\u003c/sup\u003e. In both samples having hyper- and hypo-activated WNT, YAP-high/fetal programmed cell populations (entero-like type 1 and metaplasia) exhibited the highest number of significant interactions with other tumor groups, acting as both a sender and receiver. Using different significant ligand-receptor pairs from entero-like type 1 and metaplastic cells, we also identified specific interactions for each YAP-high/fetal programmed cell type (Figures 3f\u0026ndash;h, Extended Figures 6b-e). The entero-like type 1 population interacted with other cell types by expressing \u003cem\u003eL1CAM\u003c/em\u003e, indicating its potential ability to initiate metastasis in CRC\u003csup\u003e39\u003c/sup\u003e, while metaplastic cells interacted significantly with other cell types by expressing trefoil genes (\u003cem\u003eTFF1\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;TFF2\u003c/em\u003e) and mucin genes (\u003cem\u003eMUC5AC\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;MUC6\u003c/em\u003e). Both cell types showed significant interaction with laminin (\u003cem\u003eLAMB3\u003c/em\u003e and \u003cem\u003eLAMC2\u003c/em\u003e) and integrin, which is associated with the epithelial\u0026ndash;mesenchymal transition (EMT)\u003csup\u003e40,41\u003c/sup\u003e (Extended Figure 6b-d).\u003c/p\u003e\n\u003cp\u003eTaken together, we revealed two different types of fetal programmed cell populations in both the WNT-high and WNT-low groups. Fetal programmed cells in the WNT-high group exhibited higher expression of TNF-\u0026alpha; signaling including chemokines, while the WNT-low group showed elevated expression of metabolism related genes. Interestingly, both groups have higher expression of EMT genes, suggesting that these populations can play a critical role in tumor progression and metastasis.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMouse tumors recapitulate the diversity of cell states observed in human tumors and reveal the transition of fetal programmed cells according to genetic alteration\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo investigate the potential of pre-clinical mouse models in mimicking human CRC biology, we analyzed 28 mouse colonic tumor tissue samples from five mouse models having different genotypes and different WNT activation levels, VilCreER \u003cem\u003eApc\u003c/em\u003e\u003csup\u003efl/fl\u003c/sup\u003e (\u003cem\u003eApc\u003c/em\u003e\u003csup\u003efl/fl\u003c/sup\u003e) (A),\u0026nbsp;VilCreER \u003cem\u003eApc\u003c/em\u003e\u003csup\u003efl/fl\u003c/sup\u003e \u003cem\u003eKras\u003c/em\u003e\u003csup\u003eG12D/+\u003c/sup\u003e (AK), VilCreER \u003cem\u003eApc\u003c/em\u003e\u003csup\u003efl/fl\u003c/sup\u003e \u003cem\u003eKras\u003c/em\u003e\u003csup\u003eG12D/+\u003c/sup\u003e \u003cem\u003eTrp53\u003c/em\u003e\u003csup\u003efl/fl\u003c/sup\u003e \u003cem\u003eTgfbr1\u003c/em\u003e\u003csup\u003efl/fl\u003c/sup\u003e (AKPT), VilCreER \u003cem\u003eKras\u003c/em\u003e\u003csup\u003eG12D/+\u0026nbsp;\u003c/sup\u003e\u003cem\u003eTrp53\u003c/em\u003e\u003csup\u003efl/fl\u003c/sup\u003e \u003cem\u003eNotch1\u003c/em\u003e\u003csup\u003eTg\u003c/sup\u003e (KPN), and VilCreER \u003cem\u003eBraf\u003c/em\u003e\u003csup\u003eV600E/+\u003c/sup\u003e \u003cem\u003eAlk5\u003c/em\u003e\u003csup\u003efl/fl\u003c/sup\u003e (BA) (Extended Figure 7a, Supplementary Table 5). Interestingly, mouse tumor cell types shared similar characteristics with human tumor cell types and the abundance of cell types depends on the various genotypic alterations (Figures 4a-b, Extended Figures 7b-c). For instance, \u003cem\u003eLgr5\u003c/em\u003e\u003csup\u003ehigh\u003c/sup\u003e stem cells in the mouse model showed a positive correlation with human \u003cem\u003eLGR5\u003c/em\u003e\u003csup\u003ehigh\u003c/sup\u003e stem cells, and this was predominant in \u003cem\u003eApc\u003c/em\u003e\u003csup\u003efl/fl\u003c/sup\u003e models. The mouse enterocyte-like population, more abundant in the \u003cem\u003eApc\u003c/em\u003e-drived and the KPN models, was positively correlated with human entero-like type 1 and \u003cem\u003eLGR5\u003c/em\u003e\u003csup\u003ehigh\u003c/sup\u003e cell types, especially, entero-like type 1. The metaplasia-like population was significantly dominant in the BA model, and most closely resembled with human metaplasia cell type. Notably, the AKPT and KPN models had more cell types and more diverse contributions for each cell type compared to the other models. Unique populations, such as the Tuft-like cells and the EE-like cells, were enriched in AKPT and KPN models (Extended Figure 7d).\u003c/p\u003e\n\u003cp\u003eAmong the 7 mouse tumor cell types, we identified two cell types that have a higher fetal module score than other cell types (Figure 4c). The enterocyte-like population and metaplasia-like population both exhibited high fetal module and YAP signaling scores, as well as a high iCMS3 score. However, we observed that the WNT activation level differed between the two fetal programmed cell groups (Figure 4c-d). Specifically, the enterocyte-like population exhibited high WNT and iCMS2 scores, whereas the metaplasia-like population had a high score only for the fetal modules and YAP signaling, but not for WNT activation. Furthermore, a comparative analysis between the enterocyte-like and metaplasia-like mouse cell type supports that both groups demonstrated a similar pattern to human-derived cell types. The metaplasia-like group, resembling human metaplasia cells, showed higher expression of the mucin genes, fetal and YAP-related genes (\u003cem\u003eMuc13\u003c/em\u003e, \u003cem\u003eCd55, Emp1,\u003c/em\u003e \u003cem\u003eNt5e, and Asap1\u003c/em\u003e). Conversely, the enterocyte-like population was associated with a higher expression of chemokine ligand group, WNT, fetal and YAP-related genes (\u003cem\u003eCxcl1, Cxcl2, Axin2, Sox4, Ly6a, F3, and Ctgf)\u0026nbsp;\u003c/em\u003e(Figure 4e). Both gene expression and pathway signaling were shown to have close similarities among human and mouse cell-types. TNF-\u0026alpha; and WNT-\u0026beta; catenin signaling were predominant in the enterocyte-like group, whereas fatty acid metabolism was enriched in the metaplasia-like group (Figure 4f).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe fetal programmed population is dynamically variable by genotypic alterations\u003csup\u003e42\u003c/sup\u003e. To identify the dynamics of fetal programmed cells across different genotypic alterations, we performed the CellRank trajectory analysis\u003csup\u003e43\u003c/sup\u003e using only the two fetal programmed tumor cell populations (Figures 4g-i). Trajectory analysis combining with CytoTRACE algorithm\u003csup\u003e44\u003c/sup\u003e revealed that enterocyte-like cells are less differentiated than metaplasia-like cells, and these cells can be converted into the metaplasia-like population (Figures 4g-h). Interestingly, we identified that this transition was facilitated by the \u003cem\u003eKras\u003c/em\u003e mutation (Figure 4i). While the \u003cem\u003eApc\u003csup\u003efl\u003c/sup\u003e\u003c/em\u003e\u003csup\u003e/fl\u003c/sup\u003e model is distinguished only by the enterocyte-like population, AK, AKPT, and KPN cells are more likely to become metaplasia-like population. In particular, AKPT and KPN fetal programmed cells show more heterogeneity than other mouse models, suggesting that AKPT and KPN samples exhibit higher plasticity than others. This suggests that while the \u003cem\u003eApc\u003c/em\u003e and \u003cem\u003eBraf\u003c/em\u003e mutations are only associated with the two distinct fetal programmed cells respectively, \u003cem\u003eKras\u003c/em\u003e mutation can facilitate the switch between these two cell types. In addition to the mouse model, fetal programmed cells from human tumor samples showed the different differentiated states between two fetal cell subgroups (Extended Figures 7e-g). Trajectory analysis using entero-like type 1 and metaplasia populations revealed that metaplasia cells are more differentiated than entero-like type 1 cells, which is consistent with the mouse data (Extended Figure 7e).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eOverall, these results demonstrate different fetal programmed cell populations in diverse mouse models, and their transition associated with \u003cem\u003eKras\u003c/em\u003e mutational status. \u003cem\u003eApc\u003c/em\u003e and \u003cem\u003eBraf\u003c/em\u003e mutations may initiate the acquisition of fetal characteristics, as shown in several studies\u003csup\u003e20,22\u003c/sup\u003e, but these mutations are not sufficient to promote the transition to other states. \u003cem\u003eKras\u003c/em\u003e mutation can facilitate the transition from the WNT-high to the WNT-low state in fetal programmed cells. \u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTumor-specific immune and stromal populations are actively linked to the fetal-programmed phenotype\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFetal programmed cells are induced by mutations, but non-genetic factors can also force cells into a fetal state. Especially, the extracellular matrix (ECM), which is one of the key factors in the transition of epithelial cells into the fetal programmed cell state\u003csup\u003e14,45\u003c/sup\u003e. To understand the interaction between fetal programmed tumor cells and the tumor microenvironment (TME), we performed cell-cell interaction analysis using a curated ligand and receptor database\u003csup\u003e38\u003c/sup\u003e. \u0026nbsp;Significant ligand-receptor interaction pairs among tumor and tumor microenvironment cells showed that entero-like type 1 cells and metaplasia cells had particularly strong interactions with TME cells, especially the stromal cell population (Figure 5a, Extended Figure 8a). \u0026nbsp;The interaction of tumor and TME cells, especially with regard to ECM genes, showed the two fetal programmed populations in iCMS2 and iCMS3 (entero-like type 1 and metaplasia) could influence and be influenced by the stromal cell population (Figure 5b). Interestingly, CXC chemokine interactions between tumor and stromal cells were enriched in the iCMS2 tumor fetal programmed group. We identified CXC chemokine expression was predominant in the iCMS2 fetal programmed cells (entero-like type 1) (Figure 3c-d). Furthermore, the number of interactions between CXC chemokines (\u003cem\u003eCXCL1\u003c/em\u003e, \u003cem\u003eCXCL2\u003c/em\u003e, \u003cem\u003eCXCL3\u003c/em\u003e, and \u003cem\u003eCXCL8\u003c/em\u003e) and atypical chemokine receptor 1 (\u003cem\u003eACKR1\u003c/em\u003e) was higher in entero-like type 1 than in other groups (Figure 5c-d). These findings suggests that the two subtypes of fetal programmed cells can interact with stromal cells in different ways.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn addition to identifying stromal cells, we also confirmed that fetal programmed cells actively interact with T/NK lymphocytes (Extended Figure 8a). Among the significant interactions between tumor and T/NK cells, we found that the number of interactions for the immune checkpoint receptors, including \u003cem\u003eHAVCR2\u003c/em\u003e (TIM-3), \u003cem\u003eLAG3\u003c/em\u003e, and \u003cem\u003eTIGIT,\u003c/em\u003e was higher in fetal programmed cells from each iCMS group than in other tumor cell populations (Figure 5e). Higher expression of immune checkpoint ligands was observed in the fetal-programmed population in each iCMS group, particularly metaplastic cells (Figure 5f). This suggests that fetal programmed cells could be a potential target for immunotherapy.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFetal programmed subtypes are prognostic biomarker for relapse in CRC\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAs above, using single-cell RNA-sequencing data, we identified two groups of fetal reprogrammed cells with high YAP signaling (entero-like type 1 and metaplasia) that are affected by genetic and non-genetic factors. To understand whether these populations are prognostic, we used bulk gene expression data and clinical information from a previously published cohort\u003csup\u003e46\u003c/sup\u003e (Figures 6a-e). We first typed them for iCMS and confirmed that WNT activation is only enriched in iCMS2 samples (Extended Figure 9a), as seen in the single-cell data. The subgroups within each iCMS group are identified by estimating the proportion of cell types using the CIBERSORTx deconvolution method\u003csup\u003e47\u003c/sup\u003e (Figures 6a and 6c, Extended Figures 9b-c). Groups with high proportions of entero-like type 1 or metaplasia cells displayed higher expression of the fetal/YAP geneset, including three fetal modules (Figures 6b and 6d). Notably, these two groups are correlated with poor relapse and progression than other groups (Figure 6e). This suggests that the fetal programmed cell population is one of the most aggressive components in CRC tumors, accelerating cancer progression, and relapse. Moreover, compared with the CMS classification, CMS1 and CMS4 samples are more abundant in Entero-like type 1\u003csup\u003ehigh\u003c/sup\u003e and Metaplasia\u003csup\u003ehigh\u003c/sup\u003e groups (Extended Figure 9d), suggesting that these populations engage in active crosstalk with the immune and stromal compartments.\u0026nbsp;\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eAn exciting new paradigm is entering the field of colorectal cancer. Consistently, fetal reprogramming in CRC, from primary to metastasis, has been identified from diverse research\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e,\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e,\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e,\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u003c/sup\u003e. These discoveries and definitions are fragmented with underlying mechanisms largely unknown. However, the consistency of the data forms an important contrast to the WNT-signaling dominated disease paradigm. Interestingly, fetal reprogramming feature can be affected by genetic alteration in mouse organoid\u003csup\u003e\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e or tissue samples\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e, as well as tumor microenvironment from immunocompetent mice\u003csup\u003e\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u003c/sup\u003e. Much of the current data comes from mouse models, making it crucial to have a clear picture of what is going on in human tumors. To contribute to these new paradigms, we set out to take an unbiased view of the epithelial cell states in CRC tumors, integrating all emerging pre-clinical concepts in our analysis and data interpretation. In the iCMS classification system we had previously unveiled two distinct tumor cell groups that convey diverse genetic and transcriptomic profiles to elucidate the unique characteristics inherent in each subtype\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e. We now embarked on looking for further subclasses within the iCMS groups, aiming to identify the distribution of WNT activation and the novel fetal programming with YAP states across a comprehensive cohort of CRC samples. Within iCMS2 patients, we found predominant populations of LGR5\u003csup\u003ehigh\u003c/sup\u003e stem cells that strongly express WNT signaling genes while iCMS3 patients lack in LGR5 gene expression and display predominance of a goblet-like cell population. In both iCMS2 and iCMS3 patients, we observed a fetal-programmed tumor cell population\u0026mdash;entero-like type 1 in the iCMS2 group and metaplasia in the iCMS3 group\u0026mdash;both displaying high levels of YAP signaling in different WNT activation status. These cells have been suggested to play an important role in CRC relapse. Additionally, we described the distinct characteristics of these aggressive cells with different WNT levels. The expression profile of the entero-like type 1 population is more enriched in the TNF-α signaling pathway, in contrast to metaplastic cells, which are more enriched in a metabolic pathway that includes fatty acid metabolism. This suggests that different WNT status (WNT-high vs. WNT-low) may influence their different properties. These findings are also supported by a variety of mouse models that have observed different WNT levels. In the mouse models, two fetal programmed cell populations exhibited similar characteristics to human fetal programmed cells. Specifically, the enterocyte-like group, had a high fetal module score as well as a high WNT score, whereas the metaplasia-like group only had a high fetal module score. In addition to the fetal and WNT scores, we confirmed that the two cell types had characteristics similar to human fetal programmed cell types. The enterocyte-like group had higher score of TNF-a signaling pathway enrichment, whereas the metaplasia-like group had a higher score of fatty acid metabolism enrichment.\u003c/p\u003e \u003cp\u003eIn primary cancers, fetal-like tumor cells having high YAP activity are thought to be associated with tumorigenesis, poor prognosis, chemotherapy resistance, and metastatic colonization\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e,\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e,\u003cspan additionalcitationids=\"CR51\" citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u003c/sup\u003e, and interestingly these cells are plastic during progression and metastasis\u003csup\u003e\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e\u003c/sup\u003e. Using human and mouse scRNA-seq data, we revealed that the distinct characteristics of fetal programmed cells could be regulated by genetic and non-genetic factors. Both human and mouse model exhibited fetal programmed cells, which were abundant in \u003cem\u003eRAS/RAF\u003c/em\u003e mutation groups. Specifically, \u003cem\u003eKras\u003c/em\u003e mutation could facilitate the transition from the hyper- to the hypo-activated WNT state in fetal programmed cells, unlike \u003cem\u003eBraf\u003c/em\u003e mutation which could only facilitate to be the hypo-activated WNT state only. Among various mouse models, AKPT and KPN mouse models showed more heterogenous patterns in fetal programmed cell population than other models, again advocating \u003cem\u003eKras\u003c/em\u003e mutation as an important factor to promote plasticity.\u003c/p\u003e \u003cp\u003eIn addition to genetic factors, we also revealed the supportive evidence that non-genetic factors can influence and be influenced by the fetal programmed cells. Using ligand receptor interaction analysis, the number of significant ligand receptor interactions between the two fetal programmed subpopulations and TME cells are higher than for the other tumor cell types, suggesting these cells can actively interact with microenvironment cells, especially, CAFs and ECs.\u003c/p\u003e \u003cp\u003eCollectively, our study provides a comprehensive map of epithelial cell states at single-cell resolution in both human and mouse, highlighting the critical role of fetal programmed cells in colorectal tumors (Fig.\u0026nbsp;6f). Using scRNA-seq data, we demonstrated two different fetal programmed cell groups with different WNT activation states. Fetal programmed cells exhibit plasticity and can be modulated by genetic and non-genetic factors. Notably, we revealed that \u003cem\u003eKras\u003c/em\u003e mutation can facilitate the transition of fetal programmed cells from the WNT-high to WNT-low state. These conserved patterns of fetal characteristics in both human and mouse ensure the availability of appropriate pre-clinical models for each colorectal subtype.\u003c/p\u003e"},{"header":"Online Methods","content":"\u003cp\u003e\u003cstrong\u003eCollection of patient tissue samples and scRNA-seq sequencing for colorectal cancer patients\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was approved by the institutional review boards of Commissie Medische Ethiek UZ KU Leuven/Onderzoek (approval no. S66460 and S63391) for the new in-house cohort. This study was carried out in accordance with ethical guidelines and all patients provided written informed consent.\u003c/p\u003e\n\u003cp\u003eTumor and adjacent normal tissue samples were rinsed with PBS and minced into pieces measuring \u0026lt;1mm\u003csup\u003e3\u003c/sup\u003e. The minced pieces were transferred to a digestion solution containing Collagenase P (Roche), DNAse I (Roche), and DMEM (Thermofisher Scientific), and incubated for 15 minutes at 37\u0026deg;C, shaking 3 times every 5 minutes. After incubation, samples were vortexed for 10 seconds and pipetted up and down for 1 minute. Samples were filtered using a 40\u0026micro;m nylon mesh (ThermoFisher Scientific), then 30ml of PBS was added. After centrifuging for 5 minutes at 4\u0026deg;C, the supernatant was discarded. Cell pellets were resuspended with RBC lysis (Roche) and incubated for 5 minutes at room temperature. They were centrifuged again for 5 minutes at 4\u0026thinsp;\u0026deg;C, then supernatant was discarded. The cell pellet was resuspended with buffer containing BSA (Invitrogen) and PBS, then filtered using a Flowmi Tip Strainer 40\u0026micro;m (Fisher Scientific) to remove debris. The fresh single-cell suspensions were loaded into the Chromium Chip G, following the manufacturer\u0026rsquo;s instructions. The Chromium Single Cell 5\u0026rsquo; v1.1 kit was used for the generation of libraries, which were sequenced using NovaSeq6000, and processed using CellRanger v6.0.2.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePre-processing and quality control of scRNA-seq data from patient tissue samples\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe raw gene expression matrix was filtered using the following criteria. DropletUtils\u003c/p\u003e\n\u003cp\u003e(1.14.2)\u003csup\u003e53\u003c/sup\u003e was used to remove empty droplets, and low-quality cells were removed using Unique Molecular Identifier (UMI) and number of genes. We only selected cells with \u0026gt;200 genes, \u0026gt;400 UMIs, and \u0026lt;25% of mitochondrial genes. DoubletFinder\u003csup\u003e54\u003c/sup\u003e(v2.0.3) was used for estimating doublets using the principal component from 1 to 10, and optimal pK value. After filtering out low-quality cells and doublets, the expression matrix was normalized and processed using Seurat\u0026rsquo;s standard pipeline (v4.1)\u003csup\u003e55\u003c/sup\u003e, 176,120 cells remained. Patient specificity was corrected using reciprocal principal component analysis (RPCA), and dimensional reduction and clustering was performed using RPCA output. The major cell types were annotated by the known marker gene expression, which was confirmed by CellTypist\u003csup\u003e56\u003c/sup\u003e, using the \u0026lsquo;Cells_Intestinal_Tract\u0026rsquo; model with default parameters (Supplementary Table 2). As the proliferating cluster was mixed with multiple cell types having proliferative characteristics, we defined each cell type after extracting the proliferative cluster only.\u003c/p\u003e\n\u003cp\u003eAssigned cell types were subclustered again by the same process, but for tumor epithelial cells, we filtered out those with \u0026lt;1,000 genes, which is the recommended approach for detecting malignant cells\u003csup\u003e57\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClassifying malignant cells from epithelial cells: In-house cohort\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe performed a gradient boosting method to exclude non-malignant populations from tumor samples from the in-house cohort (n=18), using a classifier that identifies non-malignant and malignant cells. Tumor and normal epithelial cells\u003csup\u003e13\u003c/sup\u003e were used to train the classifier to distinguish malignant cells. Epithelial cells in tumor samples with a prediction score exceeding 0.5 from the xgboost model were identified as malignant cells. It was confirmed that the remaining cells had patient specificity using a Chi-squared test (Extended Figure 1c\u0026ndash;d).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIntrinsic CMS (iCMS) scoring in malignant cells and identification of iCMS for each patient\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe identified iCMS labels using pseudobulk data from tumor cells for each patient. To calculate patient-level iCMS, we constructed a pseudobulk dataset by combining the UMI count matrix of tumor cells from each patient, except 1 patient who had \u003cimg width=\"12\" height=\"19\" src=\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAwAAAATCAMAAACTKxybAAAAAXNSR0IArs4c6QAAAD9QTFRFAAAAAAAAAAA6ADpmADqQOgAAOjoAOmaQOma2kDoAkGY6kLbbtmY6ttv/tv//27Zm29v/2////9u2/9vb///bT63gcAAAAAF0Uk5TAEDm2GYAAAAJcEhZcwAADsQAAA7EAZUrDhsAAAAZdEVYdFNvZnR3YXJlAE1pY3Jvc29mdCBPZmZpY2V/7TVxAAAAQ0lEQVQYV2NgoBYQ4WcXhJolxMXEIQBmi/CzMXNDRIV5WNl5oQpEOJn5EPYiy6DqAatBmAY1Em4P3DRGKGCBuYCQpwD5ZgIp0dofMAAAAABJRU5ErkJggg==\" alt=\"image\"\u003e50 tumor cells. The count matrix of 17 patients was normalized using DESeq2 (v1.34.0)\u003csup\u003e58\u003c/sup\u003e, and PCA was performed using 715 iCMS signature genes with \u003cem\u003eprcomp\u003c/em\u003e function in R. iCMS for each patient was identified using hierarchical clustering of iCMS signature genes and PCA analysis (Extended Figure 1e).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIdentification of shared programs from each gene module\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe auto-correlated gene modules were identified using Hotspot packages\u003csup\u003e24\u003c/sup\u003e with negative binomial models. The Top2000 genes were selected based on the Z-score of auto-correlations with significance (FDR \u0026lt; 0.01), and pair-wise gene associations were calculated. The modules were identified from a gene-gene affinity matrix, using the create_modules function, with min_genes_threshold = 40, core_only=False, and fdr_threshold=0.01, to recruit enough modules.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMaking the reference map using previously published cohort and label transfer to in-house cohort\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe performed sub-clustering analysis of multiple cell compartments from a previously published cohort that already identified major cell types (syn26844071), and only used samples from tumor or normal tissues. Specifically in tumor epithelial cells, only cells that identified as iCMS2 or iCMS3 were collected to eliminate the non-malignant population in tumor tissues. After extracting each cell type, RPCA was used to correct patient specificity. Before performing RPCA, we excluded 2 patients that had \u003cimg width=\"12\" height=\"19\" src=\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAwAAAATCAMAAACTKxybAAAAAXNSR0IArs4c6QAAAD9QTFRFAAAAAAAAAAA6ADpmADqQOgAAOjoAOmaQOma2kDoAkGY6kLbbtmY6ttv/tv//27Zm29v/2////9u2/9vb///bT63gcAAAAAF0Uk5TAEDm2GYAAAAJcEhZcwAADsQAAA7EAZUrDhsAAAAZdEVYdFNvZnR3YXJlAE1pY3Jvc29mdCBPZmZpY2V/7TVxAAAAQ0lEQVQYV2NgoBYQ4WcXhJolxMXEIQBmi/CzMXNDRIV5WNl5oQpEOJn5EPYiy6DqAatBmAY1Em4P3DRGKGCBuYCQpwD5ZgIp0dofMAAAAABJRU5ErkJggg==\" alt=\"image\"\u003e50 cells. Using RPCA output, we performed non-linear dimensional reduction and clustering using Seurat functions. Clusters expressing genes that cannot possibly coexist were labelled as doublets and eliminated from the analysis. Data integration between the published cohort and in-house cohort was performed using label transfer from Seurat. Through \u003cem\u003eFindTransferAnchors\u003c/em\u003e, \u003cem\u003eTransferData\u003c/em\u003e, and \u003cem\u003eMapQuery\u003c/em\u003e function, cell type and UMAP axes were projected from reference to in-house cohort samples. In total 51,444 malignant cells from 78 patients were used for the downstream analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePseudobulk analysis in malignant human cells\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo reduce the variance between each patient\u003csup\u003e59\u003c/sup\u003e and find the strongest component affecting epithelial subpopulation in the iCMS2 and iCMS3 groups, we converted the UMI count matrix to pseudobulk RNA count by combining UMI counts per gene for each patient and for each major cell type with a count of \u0026gt;50 cells. The aggregated count matrix for 235 pseudobulk samples was transformed to DESeq2 format to perform normalization. Principal component analysis (PCA) was applied on scaled data using 2,000 variable genes using the \u003cem\u003eprcomp\u003c/em\u003e function in R. The significantly correlated genes with each PC were selected by absolute of Pearson correlation coefficient (PCC) \u0026gt; 0.4 and \u003cem\u003eP\u003c/em\u003e-value \u0026lt; 0.05.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLigand\u0026ndash;receptor interaction within malignant cells, and between malignant cells and stromal or immune cells\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCell\u0026ndash;cell communication analysis was run using LIANA (v0.1.12)\u003csup\u003e38\u003c/sup\u003e, providing the aggregated rank from various methods and multiple databases. The aggregated rank was obtained from 5 different methods (NATMI, LogFC mean, Connectome, SingleCellSignalR, and cellphonedb) with consensus resource. The significant crosstalk lists were filtered using the value of aggregated rank as it can be interpreted as \u003cem\u003eP-\u003c/em\u003evalue (aggregate_rank \u0026lt; 0.05).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIdentification of signature genes\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSpecifically expressed gene groups for each cell type were identified using MAST, implemented to \u003cem\u003eFindAllMarker\u003c/em\u003e function in Seurat package. Significant genes were selected from the Top100 genes after passing the following criteria: log2FC \u0026gt; 0.25 and Adjusted P-value \u0026lt; 0.05.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGene signature scoring for each subtype\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe scoring for each signature was performed by calculating the average expression level of each cluster then subtracting the aggregated expression of control gene score sets. The control gene score was calculated by the average expression of randomly selected 100 genes, then replicated 10 times. The signatures that were used in this paper are summarized in Supplementary Table 4.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eXenium sample preparation and analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eXenium assays were run using the 322-gene pre-designed 10x Genomics Human Colon panel (1000642). Tissue sections of 5 \u0026micro;m were placed on the 10x Genomics Xenium slides, followed by drying for 30 min at room temperature, 3h at 42\u0026deg;C, and overnight in a desiccator at room temperature, according to the demonstrated protocol of the Xenium In Situ for FFPE - Tissue Preparation Guide (CG000578 \u0026ndash; Rev C). Next, slides were processed within 2 days for deparaffinization and decrosslinking according to the demonstrated protocol of Xenium In Situ for FFPE \u0026ndash; Deparaffinization \u0026amp; Decrosslinking (CG000580 \u0026ndash; Rev C), followed by immediate proceeding to Xenium probe hybridization, ligation, amplification, and autofluorescence quenching according to the demonstrated protocol of Xenium In Situ for FFPE \u0026ndash; Probe Hybridization, Ligation \u0026amp; Amplification User Guide (CG000582). Subsequently, slides were placed inside the 10x Genomics Xenium Analyzer instrument (LISCO, KU Leuven Institute for Single Cell Omics) for automated readout cycles and fluorescent image acquisition, on-board image pre-processing, signal decoding and cell segmentation. The Xenium output files were used for downstream analysis off-instrument. After the Xenium runs, slides were additionally processed for post-Xenium H\u0026amp;E staining according to the demonstrated protocol of the Xenium In Situ Gene Expression \u0026ndash; Post-Xenium Analyzer H\u0026amp;E Staining (CG000613). Stained slides were imaged using a custom Nikon Ti2 8-stage system (LISCO).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAnalysis of spatial transcriptomics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eXenium data were filtered using scanpy (v1.9.3)\u003csup\u003e60\u003c/sup\u003e. Cells expressing less than 10 transcripts were filtered, and genes expressed in less than 5 cells were filtered. The counts for each cell were normalized using \u003cem\u003escanpy.pp.normalize_total\u003c/em\u003e with default parameters, and then logarithmized. We computed principal component analysis and applied Uniform Manifold Approximation and Projection (UMAP) to perform dimensionality reduction. Cells were clustered into subgroups using Leiden clustering, implemented in \u003cem\u003escanpy.tl.leiden\u003c/em\u003e.\u003c/p\u003e\n\u003cp\u003eTo improve the accuracy of cell type annotation, we combined cell annotation based leiden clustering and label transfer based on single-cell RNA sequencing data. Leiden clusters were annotated using marker genes (Extended Figure 4). Subsequently, Robust Cell Type Decomposition (RCTD) was performed to identify cell type in Xenium. Single-cell RNA sequencing data with major cell type annotation was used as reference. Cells with low confidence in the status of \u0026quot;reject\u0026quot; and \u0026quot;doublet_uncertain\u0026quot; were removed, and only cells with robustness based on manual annotation and RCTD label transfer were retained to recalculate the principal components, neighborhood graphs, and visualized using UMAP graphs.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMouse tumor models\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMouse experiments were carried out in accordance with the UK Home Office regulations, project licences PP3908577 and 70/9112, with the approval of the Animal Welfare and Ethical Review Board of the University of Glasgow. Mice were genotyped by Transnetyx (Tennessee, USA). Mice were housed in a specific pathogen-free facility in individual ventilated and conventional open top cages with a 12-hours light/dark cycle and free access to standard chow diet and drinking water.\u003c/p\u003e\n\u003cp\u003eFive different intestinal cancer models were used in this study. Supplementary Table 5 summarizes the number, sex age, and genotypes used in this study.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eVilCreER Apc\u003csup\u003efl/fl\u003c/sup\u003e\u003c/em\u003e (A, n=4)) and \u003cem\u003eVilCreER Apc\u003csup\u003efl/fl\u003c/sup\u003e Lsl-Kras\u003csup\u003eG12D/+\u003c/sup\u003e\u003c/em\u003e (AK, n=3) models were induced with a single injection of 70 \u0026micro;L of 100 \u0026micro;M 4-hydroxytamoxifen (Sigma, H7904), via colonoscope-guided injection into the sub-mucosa wall of the mid colon using a Karl Storz TELE PACK VET X LED endoscopic video unit \u003csup\u003e61-64\u003c/sup\u003e. A and AK mice were sampled when they had a colonic tumor burden. \u003cem\u003eVilCreER Braf\u003csup\u003eV600E/+\u003c/sup\u003e Alk5\u003csup\u003efl/fl\u003c/sup\u003e\u0026nbsp;\u003c/em\u003e(BA) were induced with a single IP injection of 2 mg tamoxifen (T5648)\u003csup\u003e65,66\u003c/sup\u003e. Then aged until showing clinical signs of intestinal tumorigenesis as previously described\u003csup\u003e22\u003c/sup\u003e. All mice were induced between the age of 7-20 weeks.\u003c/p\u003e\n\u003cp\u003eOrganoids used for the KPN transplant model (n=9) were derived from \u003cem\u003eVillinCreER Kras\u003csup\u003eG12D/+\u003c/sup\u003e Trp53\u003csup\u003efl/fl\u003c/sup\u003e R26\u003csup\u003eN1icd/+\u003c/sup\u003e\u003c/em\u003e mice\u003csup\u003e1\u003c/sup\u003e induced with 2 mg Tamoxifen intraperitoneal and aged to clinical end point. Tumors were taken for culture either from the small intestine (BVKPN RKAC13.1e, n=2) and GFVKPND GFSD106.2e, n=5) or from liver metastasis\u003csup\u003e1\u003c/sup\u003e (BVKPN RKAC3.2f, n=2). Organoids used for the AKPT transplant model (n=9) were derived from spontaneous small intestine tumors from \u003cem\u003eVillinCreER\u003c/em\u003e \u003cem\u003eApc\u003csup\u003efl/fl\u003c/sup\u003eKras\u003csup\u003eG12D/+\u003c/sup\u003eTrp53\u003csup\u003efl/fl\u003c/sup\u003eTgfbr1\u003csup\u003efl/fl\u0026nbsp;\u003c/sup\u003e\u003c/em\u003emice: VAKPT RJV42.1a (n=4), VAKPT RJV51.1c (n=2) and VAKPT RJV6.2a (n=3)\u003csup\u003e1\u003c/sup\u003e. These organoid lines were injected by colonoscope-guided injections into the sub-mucosa of the mid colon of immune competent C5BL/6J mice (Charles River). Tumors were confirmed by colonoscopy and harvested for single cell RNA sequencing.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTumor Organoid Culture\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOrganoids were grown in growth factor reduced Matrigel (Corning, Catalog No. 356231) in a 3D system. Advanced DMEM/F12 (Invitrogen, Catalog No. 12634\u0026ndash;028) was supplemented with 2 mmol/L glutamine, 10 mmol/L HEPES, and 100 U/mL penicillin/streptomycin (ThermoFisher Scientific, 15140122), B27 (Invitrogen, Catalog No. 12587\u0026ndash;010), and N2-supplement (Thermo Fisher Scientific, 17502001). For culturing organoids, 100\u0026thinsp;ng/mL Noggin (Peprotech, Catalog No. 250\u0026ndash;38), and 50\u0026thinsp;ng/mL EGF (Peprotech, Catalog No. AF-100\u0026ndash;15) were added. Organoids were cultured in six-well plates (BD Falcon) at 37\u0026deg;C in an atmosphere containing 5% CO2 and 95% air, with passaging every 2\u0026ndash;3 days. Prior to injection, organoids from a single confluent well of 6-well cell culture plates (Corning, Catalog No. 353046) were harvested and dissociated by fiercely pipetting and were washed twice with PBS. Organoids were resuspended in 70 \u0026mu;l PBS and injected into a mouse in a single injection.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMouse tissue processing for scRNA-seq\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMouse colonic tumors were excised into PBS and then chopped with a Mcllwan Tissue Chopper. Chopped paste was transferred to GentleMACS C tubes (Miltenyi Biotec, 130-093-237) containing enzyme mix (Miltenyi Biotec, 130-096-730; 2.35 mL of RPMI1640, 100 \u0026micro;L Enzyme D, 50 \u0026micro;L Enzyme R, and 12.5 \u0026micro;L Enzyme A). GentleMACS program (37C_m_TDK_1) was run on the GentleMACS Octo Dissociator with Heaters (Miltenyi Biotec, 130-096-427). The digested samples were spun and filtered through a 70 \u0026micro;m strainer with adding 10 mL of RPMI-10%FBS-2 mmol/L Ethylenediaminetetraacetic acid (EDTA). The supernatant from the suspension was discarded after centrifugation at 400 RPM for 3 minutes at 4\u0026deg;C. The cell pellet was resuspended with 0.5mL buffer containing 2% FBS, 25mM HEPES, 2mM EDTA in PBS, and transferred to a FACS collection tube on ice. Dissociated cells were sorted using a BD FACSAria (BD Biosciences) and DAPI (Thermo Fisher Scientific, 62248) to remove dead cells. Sorted cells were loaded into the Chromium Chip G, according to the manufacturer\u0026rsquo;s instructions. The reagents from 10x Chromium Single Cell 3\u0026rsquo; v3 kit (10x Genomics) were used for the generation of libraries. Libraries were sequenced on a NovaSeq6000.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePre-processing and quality control of scRNA sequencing data from diverse mouse models\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMouse sequences were aligned using Cellranger version 6.1.2 with reference genome mm10 version 3.0.0 using the developer\u0026rsquo;s standard workflow. Data were then analysed using the R package Seurat, and samples were filtered to include cells with \u0026gt; 100 genes, \u0026gt; 400 UMI counts, \u0026lt; 5% mitochondrial genes and \u0026lt; 10% haemoglobin genes. Samples were integrated using RPCA before performing data scaling, dimension reduction with PCA and clustering using the standard Seurat workflow. The resulting cells were annotated using CellTypist with the \u0026lsquo;Adult_Mouse_Gut\u0026rsquo; reference, with cell types representing epithelial subtypes labelled as epithelial. To obtain high-quality epithelial cells, we applied tumor cell-specific cutoff (\u0026gt; 500 genes), similar to the criteria used for human tumor cells and subclustering using RPCA was performed. Additionally, sample-specific clusters or clusters that did not express epithelial cell markers were discarded.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAnalysis of tumor epithelial cells from mouse models\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSubclustering using RPCA was performed on identified malignant tumor cells. To apply the human gene signature to mice, we converted human genes to mouse orthologs using the babelgene R package (v22.9). The scoring for each signature was identical with the human signature scoring method previously mentioned.\u003c/p\u003e\n\u003cp\u003eThe gene set enrichment analysis (GSEA) was performed using a list of the differentially expressed genes using MAST, performed using \u003cem\u003eFindMarkers\u003c/em\u003e function in Seurat with a threshold set to \u0026gt;10% of cells. The significance was determined by Benjamini-Hochberg adjusted \u003cem\u003eP\u003c/em\u003e-value.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAnalysis of public bulk gene expression data and deconvolution based on CIBERSORTx\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNormalized gene expression data for CRC patient samples was obtained from GSE39582\u003csup\u003e46\u003c/sup\u003e and excluded patients without survival information. In total, 557 patients from GSE39582 were used for the downstream analysis. We classified iCMS based on the iCMS signature score from a previous paper\u003csup\u003e13\u003c/sup\u003e. The cell abundance from bulk gene expression was estimated using CIBERSORTx\u003csup\u003e47\u003c/sup\u003e with default parameters. First, a signature matrix was created using an internal single-cell reference matrix. To improve profile specificity across cell types, we only used signature genes for each cell type. Next, we imputed cell fractions from bulk gene expression using the signature matrix. GSEA was conducted by pre-ranked genes that were identified by a ordered list of differentially expressed genes with specific groups and others.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe statistical analyses were performed using R (v4.1). To determine significance, results were compared using the two-sided Wilcoxon rank-sum exact test, the one-way ANOVA, and Pearson\u0026rsquo;s chi-squared test. For the analysis of differential gene expression for each group, P-values were adjusted using the Bonferroni correction. Detailed descriptions are in the Methods and Figure legends.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe IMMUcan project has received funding from the Innovative Medicines Initiative 2 Joint Undertaking under grant agreement No 821558. This Joint Undertaking receives support from the European Union\u0026rsquo;s Horizon 2020 research and innovation programme and EFPIA. IMI.europa.eu. S.T. is supported by the Belgian Foundation Against Cancer (FAF-C/2018/1301) and a BOF-Fundamental Clinical Research mandate (FKO) from KU Leuven. Y.H. is supported by a FWO Junior Postdoctoral fellowship (12D5823N). S.T. and S.V. are supported by the Research Foundation Flanders (G0C9620N, G067821N). S.T is supported by Stichting Tegan Kanker grant (2020-082). S.S., M.L.M., K.L.G., T.R.M.L., M.W., R.A.R., A.D.C., and O.J.S. were supported by Cancer Research UK core funding to the CRUK Scotland Institute (A31287) and a CRUK Core programme award to O.J.S (DRCQQR-May21\\100002). \u0026nbsp;M.L.M., K.L.G., T.R.M.L., A.D.C., and O.J.S. were supported by a Cancer Research UK Accelerator Award (A26825). S.S., M.L.M., K.L.G., A.D.C., and O.J.S. were supported by Mark Foundation ASPIRE I Award. \u0026nbsp;S.S. and O.J.S. were supported by a CRUK Programme Grant (DRCNPG-Jun22\\100002). M.W was supported by CRUK Clinical Academic Training Programme (A29706) and Chief Scientist Office (CSO PCL/24/01) lectureships. T.V and K.V are supported by KU Leuven (C14/22/125), the Research Foundation Flanders (I001818N, I009724N) and Vlaamse Veerkracht (VV023-06 \u0026ldquo;PRISMO\u0026rdquo;). K.V and N.V. are supported by the Research Foundation Flanders (G005923N). \u0026nbsp;The authors would like to thank the Research Services at the CRUK Scotland Institute (CRUK SI): The Biological Service Unit (BSU), Histology, Molecular Technologies and Central Services, funded by CRUK core funding to the CRUK Scotland Institute (A31287). The authors also thank Catherine Winchester (CRUK SI) for critical review of the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contribution\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eConceptualization: Y.H., S.S., O.J.S., S.T.\u003c/p\u003e\n\u003cp\u003eData Curation: Y.H., K.L.G\u003c/p\u003e\n\u003cp\u003eFormal analysis: Y.H., K.L.G., Z.H., F.D.V.B., A.O., S.V\u003c/p\u003e\n\u003cp\u003eFunding Acquisition: O.J.S., S.T.\u003c/p\u003e\n\u003cp\u003eInvestigation: S.S., M.L.M., T.R.M.L., M.W., R.A.R., B.V.B., Y.M., L.L., V.P., A.M.P., N.V.\u003c/p\u003e\n\u003cp\u003eMethodology: Y.H., S.S., M.L.M., K.L.G., Z.H.\u003c/p\u003e\n\u003cp\u003eResource: S.S., M.L.M., T.R.M.L., M.W., R.A.R., K.V., G.R., G.D.H., X.S., G.B., A.D.H., H.P., T.V.\u003c/p\u003e\n\u003cp\u003eSupervision: A.D.C., O.J.S., S.T.\u003c/p\u003e\n\u003cp\u003eVisualization: Y.H., Z.H., F.D.V.B., A.O.\u003c/p\u003e\n\u003cp\u003eWriting \u0026ndash; Original Draft: Y.H., S.S., O.J.S., S.T.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe publicly available scRNA-seq and bulk gene expression dataset is referred to in the Methods (syn26844071 and GSE39582). Mouse scRNA-seq data for five mouse model is available under Gene Expression Omnibus (GEO), with accession number GSE280631. All other materials are available from the corresponding authors upon reasonable request.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eO.J.S. has received research funding from AstraZeneca, Boehringer Ingelheim, Novartis.\u003c/p\u003e\n\u003cp\u003eThe other authors declare no competing interests.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eJackstadt, R.\u003cem\u003e et al.\u003c/em\u003e Epithelial NOTCH Signaling Rewires the Tumor Microenvironment of Colorectal Cancer to Drive Poor-Prognosis Subtypes and Metastasis. \u003cem\u003eCancer Cell\u003c/em\u003e \u003cstrong\u003e36\u003c/strong\u003e, 319-336 e7 (2019).\u003c/li\u003e\n\u003cli\u003eCanellas-Socias, A.\u003cem\u003e et al.\u003c/em\u003e Metastatic recurrence in colorectal cancer arises from residual EMP1(+) cells. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e611\u003c/strong\u003e, 603-613 (2022).\u003c/li\u003e\n\u003cli\u003eVasquez, E.G.\u003cem\u003e et al.\u003c/em\u003e Dynamic and adaptive cancer stem cell population admixture in colorectal neoplasia. \u003cem\u003eCell Stem Cell\u003c/em\u003e \u003cstrong\u003e29\u003c/strong\u003e, 1213-1228 e8 (2022).\u003c/li\u003e\n\u003cli\u003eChen, G.T.\u003cem\u003e et al.\u003c/em\u003e Disruption of beta-Catenin-Dependent Wnt Signaling in Colon Cancer Cells Remodels the Microenvironment to Promote Tumor Invasion. \u003cem\u003eMol Cancer Res\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 468-484 (2022).\u003c/li\u003e\n\u003cli\u003eNguyen, L.H., Goel, A. \u0026amp; Chung, D.C. Pathways of Colorectal Carcinogenesis. \u003cem\u003eGastroenterology\u003c/em\u003e \u003cstrong\u003e158\u003c/strong\u003e, 291-302 (2020).\u003c/li\u003e\n\u003cli\u003eCancer Genome Atlas, N. Comprehensive molecular characterization of human colon and rectal cancer. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e487\u003c/strong\u003e, 330-7 (2012).\u003c/li\u003e\n\u003cli\u003eSchmitt, M. \u0026amp; Greten, F.R. The inflammatory pathogenesis of colorectal cancer. \u003cem\u003eNat Rev Immunol\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 653-667 (2021).\u003c/li\u003e\n\u003cli\u003eDow, L.E.\u003cem\u003e et al.\u003c/em\u003e Apc Restoration Promotes Cellular Differentiation and Reestablishes Crypt Homeostasis in Colorectal Cancer. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e161\u003c/strong\u003e, 1539-1552 (2015).\u003c/li\u003e\n\u003cli\u003eSchepers, A.G.\u003cem\u003e et al.\u003c/em\u003e Lineage tracing reveals Lgr5+ stem cell activity in mouse intestinal adenomas. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e337\u003c/strong\u003e, 730-5 (2012).\u003c/li\u003e\n\u003cli\u003eBarker, N.\u003cem\u003e et al.\u003c/em\u003e Crypt stem cells as the cells-of-origin of intestinal cancer. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e457\u003c/strong\u003e, 608-11 (2009).\u003c/li\u003e\n\u003cli\u003eMartinez-Ordonez, A.\u003cem\u003e et al.\u003c/em\u003e Hyaluronan driven by epithelial aPKC deficiency remodels the microenvironment and creates a vulnerability in mesenchymal colorectal cancer. \u003cem\u003eCancer Cell\u003c/em\u003e \u003cstrong\u003e41\u003c/strong\u003e, 252-271 e9 (2023).\u003c/li\u003e\n\u003cli\u003eGuinney, J.\u003cem\u003e et al.\u003c/em\u003e The consensus molecular subtypes of colorectal cancer. \u003cem\u003eNat Med\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 1350-6 (2015).\u003c/li\u003e\n\u003cli\u003eJoanito, I.\u003cem\u003e et al.\u003c/em\u003e Single-cell and bulk transcriptome sequencing identifies two epithelial tumor cell states and refines the consensus molecular classification of colorectal cancer. \u003cem\u003eNat Genet\u003c/em\u003e \u003cstrong\u003e54\u003c/strong\u003e, 963-975 (2022).\u003c/li\u003e\n\u003cli\u003eYui, S.\u003cem\u003e et al.\u003c/em\u003e YAP/TAZ-Dependent Reprogramming of Colonic Epithelium Links ECM Remodeling to Tissue Regeneration. \u003cem\u003eCell Stem Cell\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 35-49 e7 (2018).\u003c/li\u003e\n\u003cli\u003eChen, L.\u003cem\u003e et al.\u003c/em\u003e TGFB1 induces fetal reprogramming and enhances intestinal regeneration. \u003cem\u003eCell Stem Cell\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 1520-1537 e8 (2023).\u003c/li\u003e\n\u003cli\u003eAyyaz, A.\u003cem\u003e et al.\u003c/em\u003e Single-cell transcriptomes of the regenerating intestine reveal a revival stem cell. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e569\u003c/strong\u003e, 121-125 (2019).\u003c/li\u003e\n\u003cli\u003eHan, T.\u003cem\u003e et al.\u003c/em\u003e Lineage Reversion Drives WNT Independence in Intestinal Cancer. \u003cem\u003eCancer Discov\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 1590-1609 (2020).\u003c/li\u003e\n\u003cli\u003eCai, J., Maitra, A., Anders, R.A., Taketo, M.M. \u0026amp; Pan, D. beta-Catenin destruction complex-independent regulation of Hippo-YAP signaling by APC in intestinal tumorigenesis. \u003cem\u003eGenes Dev\u003c/em\u003e \u003cstrong\u003e29\u003c/strong\u003e, 1493-506 (2015).\u003c/li\u003e\n\u003cli\u003eAzzolin, L.\u003cem\u003e et al.\u003c/em\u003e YAP/TAZ incorporation in the beta-catenin destruction complex orchestrates the Wnt response. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e158\u003c/strong\u003e, 157-70 (2014).\u003c/li\u003e\n\u003cli\u003eGregorieff, A., Liu, Y., Inanlou, M.R., Khomchuk, Y. \u0026amp; Wrana, J.L. Yap-dependent reprogramming of Lgr5(+) stem cells drives intestinal regeneration and cancer. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e526\u003c/strong\u003e, 715-8 (2015).\u003c/li\u003e\n\u003cli\u003eDeng, F.\u003cem\u003e et al.\u003c/em\u003e YAP triggers the Wnt/beta-catenin signalling pathway and promotes enterocyte self-renewal, regeneration and tumorigenesis after DSS-induced injury. \u003cem\u003eCell Death Dis\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 153 (2018).\u003c/li\u003e\n\u003cli\u003eLeach, J.D.G.\u003cem\u003e et al.\u003c/em\u003e Oncogenic BRAF, unrestrained by TGFbeta-receptor signalling, drives right-sided colonic tumorigenesis. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 3464 (2021).\u003c/li\u003e\n\u003cli\u003eOgasawara, N.\u003cem\u003e et al.\u003c/em\u003e Discovery of non-genomic drivers of YAP signaling modulating the cell plasticity in CRC tumor lines. \u003cem\u003eiScience\u003c/em\u003e \u003cstrong\u003e27\u003c/strong\u003e, 109247 (2024).\u003c/li\u003e\n\u003cli\u003eDeTomaso, D. \u0026amp; Yosef, N. Hotspot identifies informative gene modules across modalities of single-cell genomics. \u003cem\u003eCell Syst\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 446-456 e9 (2021).\u003c/li\u003e\n\u003cli\u003eLi, H.\u003cem\u003e et al.\u003c/em\u003e Reference component analysis of single-cell transcriptomes elucidates cellular heterogeneity in human colorectal tumors. \u003cem\u003eNat Genet\u003c/em\u003e \u003cstrong\u003e49\u003c/strong\u003e, 708-718 (2017).\u003c/li\u003e\n\u003cli\u003ePelka, K.\u003cem\u003e et al.\u003c/em\u003e Spatially organized multicellular immune hubs in human colorectal cancer. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e184\u003c/strong\u003e, 4734-4752 e20 (2021).\u003c/li\u003e\n\u003cli\u003eKim, J.H.\u003cem\u003e et al.\u003c/em\u003e Gastric-type expression signature in serrated pathway-associated colorectal tumors. \u003cem\u003eHum Pathol\u003c/em\u003e \u003cstrong\u003e46\u003c/strong\u003e, 643-56 (2015).\u003c/li\u003e\n\u003cli\u003eChen, B.\u003cem\u003e et al.\u003c/em\u003e Differential pre-malignant programs and microenvironment chart distinct paths to malignancy in human colorectal polyps. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e184\u003c/strong\u003e, 6262-6280 e26 (2021).\u003c/li\u003e\n\u003cli\u003eWang, D.\u003cem\u003e et al.\u003c/em\u003e CXCL1 induced by prostaglandin E2 promotes angiogenesis in colorectal cancer. \u003cem\u003eJ Exp Med\u003c/em\u003e \u003cstrong\u003e203\u003c/strong\u003e, 941-51 (2006).\u003c/li\u003e\n\u003cli\u003eZhuo, C., Ruan, Q., Zhao, X., Shen, Y. \u0026amp; Lin, R. CXCL1 promotes colon cancer progression through activation of NF-kappaB/P300 signaling pathway. \u003cem\u003eBiol Direct\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, 34 (2022).\u003c/li\u003e\n\u003cli\u003eYang, X.\u003cem\u003e et al.\u003c/em\u003e Collagen 1-mediated CXCL1 secretion in tumor cells activates fibroblasts to promote radioresistance of esophageal cancer. \u003cem\u003eCell Rep\u003c/em\u003e \u003cstrong\u003e42\u003c/strong\u003e, 113270 (2023).\u003c/li\u003e\n\u003cli\u003eHollingsworth, M.A. \u0026amp; Swanson, B.J. Mucins in cancer: protection and control of the cell surface. \u003cem\u003eNat Rev Cancer\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, 45-60 (2004).\u003c/li\u003e\n\u003cli\u003eSheng, Y.H.\u003cem\u003e et al.\u003c/em\u003e MUC1-mediated Macrophage Activation Promotes Colitis-associated Colorectal Cancer via Activating the Interleukin-6/ Signal Transducer and Activator of Transcription 3 Axis. \u003cem\u003eCell Mol Gastroenterol Hepatol\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 789-811 (2022).\u003c/li\u003e\n\u003cli\u003eLee, C.K.\u003cem\u003e et al.\u003c/em\u003e Tumor metastasis to lymph nodes requires YAP-dependent metabolic adaptation. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e363\u003c/strong\u003e, 644-649 (2019).\u003c/li\u003e\n\u003cli\u003eCaire, R.\u003cem\u003e et al.\u003c/em\u003e YAP Transcriptional Activity Dictates Cell Response to TNF In Vitro. \u003cem\u003eFront Immunol\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 856247 (2022).\u003c/li\u003e\n\u003cli\u003eShimomura, K.\u003cem\u003e et al.\u003c/em\u003e Sleeping Beauty transposon mutagenesis identified genes and pathways involved in inflammation-associated colon tumor development. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 6514 (2023).\u003c/li\u003e\n\u003cli\u003eFey, S.K., Vaquero-Siguero, N. \u0026amp; Jackstadt, R. Dark force rising: Reawakening and targeting of fetal-like stem cells in colorectal cancer. \u003cem\u003eCell Rep\u003c/em\u003e \u003cstrong\u003e43\u003c/strong\u003e, 114270 (2024).\u003c/li\u003e\n\u003cli\u003eDimitrov, D.\u003cem\u003e et al.\u003c/em\u003e Comparison of methods and resources for cell-cell communication inference from single-cell RNA-Seq data. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 3224 (2022).\u003c/li\u003e\n\u003cli\u003eGanesh, K.\u003cem\u003e et al.\u003c/em\u003e L1CAM defines the regenerative origin of metastasis-initiating cells in colorectal cancer. \u003cem\u003eNat Cancer\u003c/em\u003e \u003cstrong\u003e1\u003c/strong\u003e, 28-45 (2020).\u003c/li\u003e\n\u003cli\u003ePuram, S.V.\u003cem\u003e et al.\u003c/em\u003e Single-Cell Transcriptomic Analysis of Primary and Metastatic Tumor Ecosystems in Head and Neck Cancer. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e171\u003c/strong\u003e, 1611-1624 e24 (2017).\u003c/li\u003e\n\u003cli\u003eMalagoli Tagliazucchi, G., Wiecek, A.J., Withnell, E. \u0026amp; Secrier, M. Genomic and microenvironmental heterogeneity shaping epithelial-to-mesenchymal trajectories in cancer. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 789 (2023).\u003c/li\u003e\n\u003cli\u003eMzoughi, S.\u003cem\u003e et al.\u003c/em\u003e A Mutation-driven oncofetal regression fuels phenotypic plasticity in colorectal cancer. \u003cem\u003ebioRxiv\u003c/em\u003e, 2023.12.10.570854 (2023).\u003c/li\u003e\n\u003cli\u003eLange, M.\u003cem\u003e et al.\u003c/em\u003e CellRank for directed single-cell fate mapping. \u003cem\u003eNat Methods\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 159-170 (2022).\u003c/li\u003e\n\u003cli\u003eGulati, G.S.\u003cem\u003e et al.\u003c/em\u003e Single-cell transcriptional diversity is a hallmark of developmental potential. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e367\u003c/strong\u003e, 405-411 (2020).\u003c/li\u003e\n\u003cli\u003eQin, X.\u003cem\u003e et al.\u003c/em\u003e An oncogenic phenoscape of colonic stem cell polarization. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e186\u003c/strong\u003e, 5554-5568 e18 (2023).\u003c/li\u003e\n\u003cli\u003eMarisa, L.\u003cem\u003e et al.\u003c/em\u003e Gene expression classification of colon cancer into molecular subtypes: characterization, validation, and prognostic value. \u003cem\u003ePLoS Med\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, e1001453 (2013).\u003c/li\u003e\n\u003cli\u003eNewman, A.M.\u003cem\u003e et al.\u003c/em\u003e Determining cell type abundance and expression from bulk tissues with digital cytometry. \u003cem\u003eNat Biotechnol\u003c/em\u003e \u003cstrong\u003e37\u003c/strong\u003e, 773-782 (2019).\u003c/li\u003e\n\u003cli\u003eMoorman, A.\u003cem\u003e et al.\u003c/em\u003e Progressive plasticity during colorectal cancer metastasis. \u003cem\u003ebioRxiv\u003c/em\u003e, 2023.08.18.553925 (2023).\u003c/li\u003e\n\u003cli\u003eGoto, N.\u003cem\u003e et al.\u003c/em\u003e SOX17 enables immune evasion of early colorectal adenomas and cancers. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e627\u003c/strong\u003e, 636-645 (2024).\u003c/li\u003e\n\u003cli\u003eTouil, Y.\u003cem\u003e et al.\u003c/em\u003e Colon cancer cells escape 5FU chemotherapy-induced cell death by entering stemness and quiescence associated with the c-Yes/YAP axis. \u003cem\u003eClin Cancer Res\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 837-46 (2014).\u003c/li\u003e\n\u003cli\u003eHeinz, M.C.\u003cem\u003e et al.\u003c/em\u003e Liver Colonization by Colorectal Cancer Metastases Requires YAP-Controlled Plasticity at the Micrometastatic Stage. \u003cem\u003eCancer Res\u003c/em\u003e \u003cstrong\u003e82\u003c/strong\u003e, 1953-1968 (2022).\u003c/li\u003e\n\u003cli\u003eRoulis, M.\u003cem\u003e et al.\u003c/em\u003e Paracrine orchestration of intestinal tumorigenesis by a mesenchymal niche. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e580\u003c/strong\u003e, 524-529 (2020).\u003c/li\u003e\n\u003cli\u003eLun, A.T.L.\u003cem\u003e et al.\u003c/em\u003e EmptyDrops: distinguishing cells from empty droplets in droplet-based single-cell RNA sequencing data. \u003cem\u003eGenome Biol\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 63 (2019).\u003c/li\u003e\n\u003cli\u003eMcGinnis, C.S., Murrow, L.M. \u0026amp; Gartner, Z.J. DoubletFinder: Doublet Detection in Single-Cell RNA Sequencing Data Using Artificial Nearest Neighbors. \u003cem\u003eCell Syst\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 329-337 e4 (2019).\u003c/li\u003e\n\u003cli\u003eHao, Y.\u003cem\u003e et al.\u003c/em\u003e Integrated analysis of multimodal single-cell data. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e184\u003c/strong\u003e, 3573-3587 e29 (2021).\u003c/li\u003e\n\u003cli\u003eDominguez Conde, C.\u003cem\u003e et al.\u003c/em\u003e Cross-tissue immune cell analysis reveals tissue-specific features in humans. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e376\u003c/strong\u003e, eabl5197 (2022).\u003c/li\u003e\n\u003cli\u003eGavish, A.\u003cem\u003e et al.\u003c/em\u003e Hallmarks of transcriptional intratumour heterogeneity across a thousand tumours. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e618\u003c/strong\u003e, 598-606 (2023).\u003c/li\u003e\n\u003cli\u003eLove, M.I., Huber, W. \u0026amp; Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. \u003cem\u003eGenome Biol\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 550 (2014).\u003c/li\u003e\n\u003cli\u003eZhang, M.\u003cem\u003e et al.\u003c/em\u003e IDEAS: individual level differential expression analysis for single-cell RNA-seq data. \u003cem\u003eGenome Biol\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 33 (2022).\u003c/li\u003e\n\u003cli\u003eWolf, F.A., Angerer, P. \u0026amp; Theis, F.J. SCANPY: large-scale single-cell gene expression data analysis. \u003cem\u003eGenome Biol\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 15 (2018).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"KU Leuven","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":"","lastPublishedDoi":"10.21203/rs.3.rs-6812365/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6812365/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eIn the colonic epithelium, fetal programming is induced by Yes-associated protein (YAP) signaling, which is associated with tumorigenesis and progression in colorectal cancer (CRC). While CRC was long considered a WNT driven disease, here we show that fetal programmed cells with activated YAP signaling can arise independently of WNT activation. Fetal programmed cells, in the presence of hyper- or hypo-activated WNT, have different characteristics and are associated with poor relapse. Furthermore, using various mouse models, we demonstrated the conserved characteristics of two different fetal programmed cell states with different WNT activation, which can be switched from a hyperactivated to hypoactivated WNT state based on genetic alteration, particularly \u003cem\u003eKras\u003c/em\u003e mutation. Taken together, these data redefine the key determinants of epithelial cell states in colorectal cancer and integrate emerging preclinical biology into a robust landscape of states observed in human and mouse.\u003c/p\u003e","manuscriptTitle":"Epithelial states in colorectal cancer are co-determined by YAP associated fetal programming and WNT signaling","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-06-06 06:33:56","doi":"10.21203/rs.3.rs-6812365/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":"6a96d372-090f-4804-94bf-248eedf12847","owner":[],"postedDate":"June 6th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-06-06T06:33:57+00:00","versionOfRecord":[],"versionCreatedAt":"2025-06-06 06:33:56","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6812365","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6812365","identity":"rs-6812365","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}