SELENOP modifies sporadic colorectal carcinogenesis and WNT signaling activity through LRP5/6 interactions

Although selenium deficiency correlates with colorectal cancer (CRC) risk, the roles of the selenium-rich antioxidant selenoprotein P (SELENOP) in CRC remain unclear. In this study, we defined SELENOP’s contributions to sporadic CRC. In human single-cell cRNA-Seq (scRNA-Seq) data sets, we discovered that SELENOP expression rose as normal colon stem cells transformed into adenomas that progressed into carcinomas. We next examined the effects of Selenop KO in a mouse adenoma model that involved conditional, intestinal epithelium-specific deletion of the tumor suppressor adenomatous polyposis coli (Apc) and found that Selenop KO decreased colon tumor incidence and size. We mechanistically interrogated SELENOP-driven phenotypes in tumor organoids as well as in CRC and noncancer cell lines. Selenop-KO tumor organoids demonstrated defects in organoid formation and decreases in WNT target gene expression, which could be reversed by SELENOP restoration. Moreover, SELENOP increased canonical WNT signaling activity in noncancer and CRC cell lines. In defining the mechanism of action of SELENOP, we mapped protein-protein interactions between SELENOP and the WNT coreceptors low-density lipoprotein receptor–related proteins 5 and 6 (LRP5/6). Last, we confirmed that SELENOP-LRP5/6 interactions contributed to the effects of SELENOP on WNT activity. Overall, our results position SELENOP as a modulator of the WNT signaling pathway in sporadic CRC.


Introduction
Both human observational and animal preclinical studies support tumor-protective roles for the micronutrient selenium in the gastrointestinal tract; however, human clinical trials have yet to corroborate these findings (1)(2)(3)(4)(5)(6)(7). Selenium is thought to exert its biological functions through incorporation into selenocysteinecontaining proteins, or selenoproteins (8). Among the known selenoproteins, selenoprotein P (SELENOP) is unique in that it contains multiple selenocysteines: 1 selenocysteine in an N-terminal antioxidant domain and 9 selenocysteines in a C-terminal selenium transport domain. Although SELENOP is largely synthesized by the liver and secreted into the plasma, SELENOP is also expressed in tissues such as the testes, muscle, kidney, brain, small intestine, and colon (9,10). Cells internalize extracellular, secreted SELENOP via receptor-mediated endocytosis, once SELENOP binds low-density lipoprotein receptor-related proteins (LRPs) on the cell surface (8,11). LRP1 and LRP2 (also known as megalin) have been identified as the SELENOP receptors in muscle and kidney, respectively (12,13), whereas LRP8 (also known as ApoER2) has been identified as the SELENOP receptor in bone, brain, and testes (14)(15)(16). However, the SELENOP receptor(s) in the colon and small intestine, where LRP1, LRP2, and LRP8 are lowly expressed, remains unknown (17).
In sporadic colorectal cancer (CRC), genetic and epigenetic alterations influenced by lifestyle, environmental, and dietary factors drive carcinogenesis through activation of oncogenes and inactivation of tumor suppressor genes (18). Conventional CRCs, which comprise 60%-85% of sporadic CRCs, are characterized by initial inactivation of the tumor suppressor gene adenomatous polyposis coli (APC) and resultant hyperactivation of WNT signaling (19). In canonical WNT signaling, a destruction complex targets cytoplasmic β-catenin for proteasomal degradation. Binding of WNT ligands to their coreceptors low-density lipoprotein receptor-relat-Although selenium deficiency correlates with colorectal cancer (CRC) risk, the roles of the selenium-rich antioxidant selenoprotein P (SELENOP) in CRC remain unclear. In this study, we defined SELENOP's contributions to sporadic CRC. In human single-cell cRNA-Seq (scRNA-Seq) data sets, we discovered that SELENOP expression rose as normal colon stem cells transformed into adenomas that progressed into carcinomas. We next examined the effects of Selenop KO in a mouse adenoma model that involved conditional, intestinal epithelium-specific deletion of the tumor suppressor adenomatous polyposis coli (Apc) and found that Selenop KO decreased colon tumor incidence and size. We mechanistically interrogated SELENOP-driven phenotypes in tumor organoids as well as in CRC and noncancer cell lines. Selenop-KO tumor organoids demonstrated defects in organoid formation and decreases in WNT target gene expression, which could be reversed by SELENOP restoration. Moreover, SELENOP increased canonical WNT signaling activity in noncancer and CRC cell lines. In defining the mechanism of action of SELENOP, we mapped protein-protein interactions between SELENOP and the WNT coreceptors low-density lipoprotein receptor-related proteins 5 and 6 (LRP5/6). Last, we confirmed that SELENOP-LRP5/6 interactions contributed to the effects of SELENOP on WNT activity. Overall, our results position SELENOP as a modulator of the WNT signaling pathway in sporadic CRC. SELENOP modifies sporadic colorectal carcinogenesis and WNT signaling activity through LRP5/6 interactions out enterocyte and colonocyte populations, as well as in subsets of proximal progenitor, Paneth, goblet, and enteroendocrine cells ( Figure 1D). To corroborate these observations, we subjected human small intestinal organoids ("enteroids") to established directed differentiation protocols (26) and then measured SELE-NOP protein levels by ELISA. Indeed, we found that SELENOP protein was highly expressed among enteroids differentiated toward enterocytes, goblet cells, or Paneth cells ( Figure 1E). We observed similar trends in SELENOP transcript expression in enteroids skewed toward the enterocyte, goblet cell, or Paneth cell lineages (Supplemental Figure 4).
SELENOP expression progressively increases throughout conventional colorectal carcinogenesis. We next evaluated SELENOP expression in colorectal polyps and cancers. For these analyses, we used a previously published scRNA-Seq data set of conventional adenomas (adenoma-specific cells [ASCs]), serrated polyps (serrated-specific cells [SSCs]), microsatellite stable (MSS) cancers, and microsatellite instability-high (MSI-H) cancers (27). Stem and absorptive cells are thought to represent the tumor-initiating cell types for conventional adenomas and serrated polyps, respectively, that can lead to MSS and MSI-H cancers (27). Here, we observed high SELENOP expression in subsets of ASCs, SSCs, and MSS cancer cells ( Figure 2A). Moreover, in ASCs and MSS cancer cells, SELENOP expression was weakly correlated (r = 0.44, P = 0.01) with inferred stemness, as derived from Cellular Trajectory Reconstruction Analysis Using Gene Counts and Expression (CytoTRACE) analysis that computationally predicts cellular differentiation states from scRNA-Seq data (28) ( Figure 2B).
When we integrated this data set with its corresponding patient-matched normal tissue data sets (Supplemental Figure  5A), we observed increases in SELENOP expression from normal crypt stem cells to ASCs to MSS cancer cells ( Figure 2C). Similarly, in a single-nucleus RNA-Seq (snRNA-Seq) data set generated from patients with familial adenomatous polyposis (FAP) and from non-FAP patients (29) (Supplemental Figure 5B), SELE-NOP expression was greater in adenocarcinomas than in polyps or unaffected stem cells (Supplemental Figure 5C). We also noted higher SELENOP expression in SSCs than in absorptive cells; however, SELENOP expression did not differ between absorptive cells and MSI-H cancer cells (Supplemental Figure 5D). Although SELENOP expression levels did not differ (P = 0.263) between MSS and MSI-H cancers in this particular data set (27) (Figure 2A and Figure 2D), SELENOP expression was greater in mismatch repair-proficient (MMR-proficient) than MMR-deficient cancers in another scRNA-Seq data set (30) (Figure 2D), and this correlated with the proportion of stem-like cells present in each cancer type. Overall, these results suggest that upregulation of SELENOP expression throughout conventional colorectal carcinogenesis occurs as a function of stemness.
Selenop KO decreases colon tumor incidence and size in Apcdependent tumorigenesis. Since SELENOP upregulation correlated with the conventional adenoma-carcinoma sequence, we hypothesized that SELENOP deficiency would reduce stem cell-driven colorectal tumorigenesis. To model this, we crossed Selenop -/mice (31) onto the Lrig1-CreERT2/ + Apc fl/+ genetic background (32). Importantly, these mice were maintained on a defined, selenium-supplemented diet (1.0 mg selenium/kg) to control for micro-ed proteins 5 and 6 (LRP5/6) and frizzled (FZD) inhibits destruction complex activity and triggers nuclear translocation of β-catenin. In the nucleus, β-catenin binds T cell factor/lymphoid enhancer factor (TCF/LEF) transcription factors to induce transcription of WNT target genes (20). Importantly, upstream WNT ligands continue to activate WNT signaling, even in the context of downstream WNT signaling hyperactivation (e.g., APC loss of function) (21,22).
In this study, we delineated tumor-promotive roles for SELE-NOP in sporadic CRC through amplification of canonical WNT signaling activity via specific interactions with LRP5/6. In human single-cell RNA-Seq (scRNA-Seq) data sets, we discovered progressive increases in SELENOP expression from stem to adenoma to carcinoma cells. To test our hypothesis that SELENOP promotes intestinal tumorigenesis, we defined the effects of Selenop KO in an Apc-dependent adenoma mouse model. Here, Selenop KO decreased colon tumor incidence and size. Additionally, Selenop-KO tumor organoids demonstrated reduced organoid formation and WNT target gene expression, which could be reversed by SELENOP overexpression. Moreover, SELENOP increased canonical WNT signaling activity in noncancer and colon cancer cell lines. In defining the mechanism, we identified a proteinprotein interaction between SELENOP and LRP5/6 and mapped the specific LRP5/6 interaction domain on SELENOP. Furthermore, we established that SELENOP's LRP5/6 interaction domain mediates its effects on canonical WNT signaling activity. Overall, our results position SELENOP as a modulator of canonical WNT signaling activity in sporadic CRC.

Results
SELENOP is predominantly expressed by differentiated epithelial cells in the normal colon and small intestine epithelium. We first profiled the selenotranscriptome in WT mouse small intestine and colon epithelial isolates by reverse transcription quantitative PCR (RT-qPCR). Selenop was the most abundant selenoprotein mRNA in the small intestine epithelium ( Figure 1A), in agreement with prior measurements of selenoprotein mRNA levels in whole small intestine tissue (23). Selenop was one of several highly expressed selenoprotein mRNAs, including selenoprotein F (Selenof), glutathione peroxidase 1 (Gpx1), and glutathione peroxidase 2 (Gpx2), in the small intestine and colon epithelium ( Figure 1A). Additionally, we confirmed GPX1 (Supplemental Figure 1A) and GPX2 (Supplemental Figure 1B; supplemental material available online with this article; https://doi.org/10.1172/JCI165988DS1) protein expression in these tissues (for Supplemental Figure 1, A and B, see complete unedited blots in the supplemental material). We observed similar selenotranscript expression patterns in the Gut Cell Atlas scRNA-Seq data set (24) generated from normal human colon and small intestine epithelium (Supplemental Figure 2).
When we performed RNA ISH on WT mouse tissues with a validated Selenop RNAscope probe (Supplemental Figure 3), we predominantly detected Selenop in differentiated epithelial cells of the villi and crypts, as well as in stromal cells ( Figure 1B). We observed a similar pattern of SELENOP expression in human colon tissues ( Figure 1C). Together, these findings complement previously described SELENOP expression patterns in mouse and human colon tissues (25). In the Gut Cell Atlas scRNA-Seq data set (24), SELENOP was moderately to highly expressed through-with Apc ΔIE/+ Selenop +/+ or Selenop +/mice, despite similar survival rates ( Figure 3D), numbers ( Figure 3E), and dysplasia severity (Figure 3, F and G). Similarly, in the small intestine, we observed decreased tumor areas (Supplemental Figure 6A) in Apc ΔIE/+ Selenop -/mice as compared with Apc ΔIE/+ Selenop +/+ or Selenop +/mice, despite similar incidence rates (Supplemental Figure 6B), numbers (Supplemental Figure 6C), and dysplasia severity (Supplemental Figure 6, D and E). Altogether, these results propound tumorpromotive roles for SELENOP in Apc-dependent tumorigenesis.
SELENOP increases canonical WNT signaling activity in noncancer and colon cancer cell lines. As SELENOP under-and overexpression in tumoroids decreased and increased WNT target gene expression, respectively, we hypothesized that SELENOP might directly amplify WNT signaling activity. To investigate this, we used 293 Super TOPFlash (STF) cells, which stably express a luciferase reporter of β-catenin/TCF/LEF-mediated transcription that serves as a direct readout of canonical WNT signaling activity (37). In 293 STF cells, combinatorial treatment with SELENOP and WNT3A increased TOPFlash activity to a greater extent than did treatment with WNT3A alone ( Figure  6A). As 293 STF cells are a noncancer cell line, we subsequently generated RKO (human colon adenocarcinoma) STF cells to confirm this observation and contextualize these findings in CRC. Importantly, RKO cells possess both WT APC and β-catenin and, as such, display intact WNT signaling (38). Similarly, exogenous SELENOP amplified WNT3A-induced TOPFlash activity in RKO STF cells ( Figure 6B). locus ( Figure 5A). When we dissociated and plated Apc ΔIE/+ Selenop +/+ -dCas9-VP64-NONTARGET and SELENOP tumoroids as single cells, more SELENOP-overexpressing cells formed tumoroids after 5 days, as compared with control cells ( Figure 5, B and C). As we and others have reported that additional WNT stimulation increased tumoroid growth even after Apc loss of function (21,22), we also measured levels of WNT target transcripts by RT-qPCR. Here, SELENOP-overexpressing tumoroids displayed higher Axin2, Lgr5, and Sox9 transcript levels than did control tumoroids ( Figure 5, D-F). Altogether, these results demonstrate that SELENOP overexpression rescued the effects of Selenop deficiency on tumoroid-forming capacity and WNT target gene expression.
SELENOP increases WNT target gene expression in human tumoroids. Additionally, we tested the effects of SELENOP treatment on WNT target gene expression in human tumoroid lines established from patients with stage II/III CRC (Supplemental Table 6). Although WNT target transcript levels differed among tumoroid lines, treatment with purified human SELENOP increased SOX9 levels in lines 32385, 35349, and 40299; LGR5 levels in line 35349; and AXIN2 levels in line 40299 (Supple- As SELENOP is a secreted protein, we hypothesized that secreted SELENOP would increase WNT signaling by an autocrine and/or paracrine mechanism. Indeed, lentiviral SELENOP overexpression in 293 STF cells ( Figure 6C) promoted WNT3Ainduced TOPFlash activity ( Figure 6D). Similarly, CRISPR activation-mediated (CRISPRa-mediated) SELENOP overexpression in RKO cells ( Figure 6E) or MC38 (mouse colon adenocarcinoma) cells ( Figure 6G) augmented WNT3A-induced TOPFlash activity ( Figure 6F, Figure 6H). Overall, it appears that exogenous or endogenous SELENOP augmented canonical WNT signaling.
As SELENOP is widely thought to bind heparan sulfate proteoglycans (HSPGs) (42), and HSPGs deliver WNT modulators and ligands to LRP5/6 (43), we hypothesized that HSPGs facilitate SELENOP-LRP6 interactions. Surprisingly, inhibition of HSPG synthesis (via sodium chlorate [NaClO 3 ] treatment) markedly enhanced co-IP of SELENOP and FLAG-LRP6 in 293T-FLAG-LRP6 cells ( Figure 7B; see complete unedited blots in the supplemental material). Conversely, treatment with heparin prevented SELENOP and FLAG-LRP6 co-IP in these cells ( Figure 7C; see complete unedited blots in the supplemental material). Furthermore, we investigated whether SELENOP accelerates LRP5/6 recycling to potentiate WNT signaling. We tested this hypothesis through biotinylation and isolation of cell-surface proteins with and without SELENOP treatment. Indeed, we found that SELE-NOP decreased cell-surface LRP6 levels ( Figure 7D; see complete unedited blots in the supplemental material). Thus, SELENOP interacted with LRP6 (unless sequestered by HSPGs), promoted LRP6 internalization, and thus amplified WNT signaling.

Discussion
In this study, we defined the role of SELENOP in sporadic colorectal carcinogenesis, which is predominantly initiated by mutations that hyperactivate the WNT signaling pathway. We observed increases in SELENOP expression throughout conventional adenoma to carcinoma progression. To test the functional consequences of Selenop deficiency on intestinal tumorigenesis, we used a mouse model in which intestinal epithelium-specific deletion of the tumor suppressor Apc and concomitant WNT signaling hyperactivation drive adenoma formation. In this model, Selenop KO was tumor protective. Underlying these phenotypes, we discovered a mechanism in which SELENOP modulated canonical expression in colorectal tumors as compared with normal colon tissues (45)(46)(47)(48), these studies did not stratify SELENOP expression by epithelial cell type and thus failed to account for the SELE-NOP expression gradient from crypt base to top in the normal colon. Namely, in comparisons with bulk normal colon tissues, we believe strong SELENOP expression in stromal and differentiated epithelial cells obscures the detection of meaningful, albeit subtle, differences in SELENOP expression from tumor-initiating cells to polyps and cancers.
While SELENOP expression was still lower in MSS cancers than in differentiated epithelial cells, we hypothesize that SELE-NOP upregulation throughout progression to malignancy fortifies tumor-promotive WNT signaling activity. Unlike in conventional CRCs, SELENOP expression was increased in serrated polyps, but not MSI-H cancers, as compared with tumor-initiating absorptive cells. Moreover, MMR-deficient tumors demonstrated decreased SELENOP expression as compared with MMR-proficient tumors.
WNT signaling activity through specific interactions with the WNT coreceptors LRP5/6.
We identified Selenop as the most highly expressed selenotranscript in the normal mouse small intestine epithelium, consistent with a selenotranscriptomic profile of whole mouse small intestine (23). To the best of our knowledge, we are the first to characterize selenoprotein mRNA expression specifically in the mouse colon and small intestine epithelium. When we examined SELENOP localization in situ, we observed a gradient of epithelial SELENOP expression up the crypt axis, as well as stromal SELENOP expression, in both mouse and human tissues. This expression pattern confirms prior findings in rat, mouse, and human small intestine/ colon tissues and supports SELENOP's recently proposed role as a crypt axis marker (9,10,25).
Our analyses revealed increases in SELENOP expression from tumor-initiating stem cells to adenomatous polyps and MSS cancers. Although others have reported reductions in SELENOP demonstrated to regress spontaneously in several animal models (52)(53)(54). To the best of our knowledge, we are the first to investigate the effects of Selenop KO on adenoma, not ACF, development in a genetically, not chemically, induced CRC mouse model.
As in sporadic CRC models, current evidence suggests that different selenoproteins modify colitis-associated carcinoma (CAC) by distinct mechanisms. In the AOM/dextran sodium sulfate (DSS) experimental CAC model, Gpx2-or Gpx3-KO mice developed more tumors than did WT mice (55,56). In contrast, Selenof-KO mice developed similar numbers of tumors, yet fewer ACFs, as compared with WT mice after AOM/DSS treatment (57). Notably, Selenop-KO mice developed fewer, smaller tumors than did Selenop-WT mice after an AOM/DSS protocol (10), which partially parallels our findings in experimental CRC. Additionally, Selenop-KO tumors from this CAC model displayed dysregulated WNT signaling, including transcriptional upregulation of the known WNT antagonists secreted frizzled-related proteins (SFRPs) 4 and 5 (10). Similarly, our Apc ΔIE/+ Selenop -/tumoroids showed defects in organoid formation and decreases in WNT target gene expression that could be reversed by SELENOP restoration. Thus, SELENOP may play similar roles in CAC and sporadic CRC.
We discovered that SELENOP is a modulator of canonical WNT signaling activity through interactions with the WNT coreceptors LRP5/6. Although SELENOP's effects on WNT signaling activity were previously undescribed, the literature supports roles for selenium itself as both a positive and negative regulator of WNT signaling activity. For example, both sodium selenate and selenomethionine administration activated WNT signaling in hippocampus tissue and primary neurons from a mouse model of Alzheimer's disease (58,59). However, selenomethionine treatment inhibited WNT signaling in HT-29 human colorectal adenocarcinoma cells (60). Similarly, selenium deficiency upregulated the transcription of WNT pathway targets and components in the normal mouse colon (61). Thus, the effects of selenium on WNT signaling activity may depend on tissue and disease context. LRP1, LRP2, and LRP8 mediate SELENOP uptake in different tissues (12)(13)(14)(15)(16). Among these known SELENOP receptors, the interactions between SELENOP and LRP8 are well studied. SELENOP's LRP8 interaction domain was previously mapped to 3 specific residues (Cys 343 , Gln 344 , Cys 345 ) within the region between SELENOP's fifth and sixth selenocysteines (62). As we mapped SELENOP's LRP5/6 interaction domain to the 42 aa between SELENOP's third and fourth selenocysteines (Sec 258 -Sec 299 ), SELENOP binds LRP8 and LRP5/6 with distinct sites. In addition to LRP binding sites, SELENOP contains one well-defined (Leu 79 -Leu 84 ) and 2 putative, histidine-rich (Thr 178 -Lys 189 and His 194 -Gln 234 ) heparin binding sites (42). As such, SELENOP is widely thought to bind cell-surface HSPGs (11). However, pretreatment with heparin failed to disrupt LRP8-SELENOP interactions (62). In contrast, pretreatment with heparin prevented LRP6-SELE-NOP interactions, and inhibition of HSPG synthesis promoted LRP6-SELENOP interactions. Thus, HSPGs may sequester SELE-NOP from LRP5/6, as they do other WNT modulators and ligands to fine-tune WNT signaling activity (43).
Although the SELENOP receptor(s) in the gastrointestinal tract remain unidentified, LRP5 and LRP6 are expressed at much higher levels than LRP1, LRP2, or LRP8 in the small intestine and While beyond the scope of the current study, these intriguing results raise the possibility that SELENOP plays distinct roles in conventional versus serrated colorectal carcinogenesis.
In an Apc-dependent mouse adenoma model, Selenop KO reduced colon tumor size and incidence. Although SELENOP remains relatively understudied in sporadic CRC, the literature supports distinct roles for different selenoproteins in azoxymethane-induced (AOM-induced) experimental CRC. For example, transgenic mice with a mutation in the selenocysteine transfer RNA (tRNA) gene that inhibits selenocysteine synthesis, and thus reduces global selenoprotein production, developed fewer early neoplastic lesions called aberrant crypt foci (ACF) than did WT mice after AOM treatment (49). Similarly, Gpx2-or Selenof-KO mice developed fewer ACFs than WT mice after AOM treatment; in the case of Gpx2-KO mice, this corresponded with a decrease in tumor numbers (50,51). In contrast, Selenop-KO mice developed more ACFs than did Selenop-WT mice after AOM treatment, although ACF progression to adenomas was not reported in this study (10). Importantly, studies that use ACFs as a primary readout of experimental tumorigenesis warrant cautious interpretation, as ACFs, while widely considered CRC precursors, have been appropriate. cDNA was synthesized from 2 μg total RNA with qScript cDNA SuperMix (95048100, Quantabio). TaqMan RT-qPCR was performed in triplicate with the TaqMan probes listed in Supplemental scRNA-Seq data analysis and visualization. Gut Cell Atlas scRNA-Seq expression data (24) was explored at https://www.gutcellatlas. org. Human colorectal polyp/cancer scRNA-Seq data (27,29) (HTA10, HTA11) are publicly available through the Human Tumor Atlas Network (https://data.humantumoratlas.org). Human CRC scRNA-Seq data (30) (GSE178341) are publicly available through NCBI's Gene colon (24,63). Therefore, LRP5/6 may represent bona fide receptors for SELENOP uptake in the gut. Our finding that SELENOP decreased cell-surface LRP6 levels raises the intriguing possibility that LRP6 mediates SELENOP internalization directly. As SELE-NOP's expression pattern opposes the WNT3A gradient along the crypt/villus axis, perhaps LRP6 shuttles SELENOP into WNT hi , SELENOP lo crypt base cells to facilitate synthesis of other selenoproteins and further amplify WNT signaling activity.
Taken together, our results present a role for SELENOP in WNT signaling modulation in the intestine, and perhaps in other tissues as well. Thus, our findings add yet another layer of complexity to the multimodal mechanisms of WNT signaling regulation in the intestine. This justifies further research into SELE-NOP's contributions to sporadic colorectal carcinogenesis.

Methods
Additional details can be found in the Supplemental Methods.
RNA isolation, cDNA synthesis, and RT-qPCR. Colon and small intestine epithelia were isolated as previously described (64). Cells and organoids were homogenized in TRIzol Reagent (15596018, Invitrogen, Thermo Fisher Scientific) prior to RNA isolation with the RNeasy Mini (74106, QIAGEN) or Micro (74004, QIAGEN) Kit, as Human enteroid culture. Human jejunal organoids were a gift from James Goldenring (Vanderbilt University, Nashville, Tennessee, USA). These enteroids were established from deidentified tissue collected at VUMC and provided by the Western Division of the Cooperative Human Tissue Network (CHTN) in accordance with the IRB of VUMC. Enteroids were refed with Intesticult Organoid Growth Medium (06010, STEMCELL Technologies) every 4 days. For ELISA experiments, enteroids were refed every 2-3 days with media described in Supplemental Table 3. Enteroids were split and replated every 7-10 days as described below.
Enteroids were collected by centrifugation at 200g for 5 minutes at 4°C, gently sheared approximately 20 times by pipetting, then centrifuged again as above. Enteroid fragments were resuspended in growth factor-reduced (GFR) Matrigel (354230, Corning), plated in 4 approximately 12 μL plugs per well, incubated at 37°C for 30 minutes, and fed with 500 μL Intesticult Organoid Growth Medium.
Cohorts of 8-to 10-week-old Lrig1-CreERT2/ + Apc fl/+ Selenop +/+ , Selenop +/-, and Selenop -/mice were administered 3 daily i.p. injections of 2 mg tamoxifen (T5648, MilliporeSigma) dissolved in corn oil (Mazola). Mice were colonoscopically monitored for tumors on days 50, 64, 78, and 92 after the initial tamoxifen injection and then euthanized on day 100 (35) by experimenters blinded to their genotype. Small intestine and colon tissue was macroscopically imaged and analyzed and then Swiss-rolled and formalin-fixed for unstained and H&E-stained slide preparation by the VUMC Translational Pathology Shared Resource (TPSR). Colon tumor volume was calculated from length (L) and width (W) measurements with the formula W 2 × L/2 (69). H&E-stained slides were examined for dysplasia severity by a gastrointestinal pathologist blinded to genotype.
Tumoroids were collected by centrifugation at 200g for 5 minutes at 4°C, gently sheared twice through a 25 gauge needle, and then centrifuged again as above. For subculturing and expansion, tumoroid fragments were resuspended in GFR Matrigel and plated in 50 μL plugs.
Polyp, normal, and cancer tissue data sets from (27) were integrated with the Single-Cell Regulatory Network Inference and Clustering (SCENIC) pipeline (65,66). From the SCENIC-derived, z score-standardized AUCell values, the "scanpy.tl.umap" function was used to compute UMAP coordinates, 50-principal component decompositions with no feature selection, and k-nearest-neighbor graphs, with k equal to the square root of the number of cells projected. The UMAP visualization for the data set from ref. 29 was produced by the same procedure but with normalized count values. Strip plots were generated from downsampled data of the corresponding bar plots, to keep cell number for all data set categories equal to the cell number of the smallest category. transduced for 4 hours in concentrated lentivirus with 8 μg/mL polybrene and 10 μM Y-27632. Forty-eight hours later, cells and tumoroids were selected with the following concentrations of puromycin (P8833, MilliporeSigma) or blasticidin (ant-bl-05, InvivoGen): 1 μg/mL puromycin (293 STF, MC38, and RKO cells), 3 μg/mL puromycin (tumoroids), 5 μg/mL puromycin (YAMC cells), 5 μg/mL blasticidin (tumoroids), or 10 μg/mL blasticidin (YAMC STF cells).
FLAG IPs. 293T cells were cultured to approximately 50% confluence in 10 cm plates and then cotransfected with 2 μg pcDNA6-N-For enzymatic dissociation experiments, tumoroids were resuspended in TrypLE Express (12604013, Gibco, Thermo Fisher Scientific) with 10 μM Y-27632 (1254, Tocris Bioscience) and 50 μg/mL DNase I (D5025, MilliporeSigma), incubated at 37°C for 3 minutes, and filtered through a 70 μm cell strainer. Enzymatic dissociation was halted by addition of PBS (without calcium or magnesium) and centrifugation as above. Tumoroid cells were then resuspended in GFR Matrigel and plated at a density of 5,000 live cells per 50 μL plug. Tumoroid fragments per cells were incubated at 37°C for 30 minutes, then fed with 500 μL basal media supplemented with 20% (v/v) R-spondin-conditioned media and 10% (v/v) Noggin-conditioned media.
Murine tumoroid image quantification. Tumoroids were imaged after 5 days with an EVOS FL2 Auto Imaging System (Thermo Fisher Scientific). The tumoroid number was quantified in Image (NIH) (70) by an experimenter blinded to the genotype.
Cell lines and maintenance. 293T (CRL3216), Hep G2 (HB-8065), and RKO (CRL2577) cells were purchased from the American Type Culture Collection (ATCC), which confirms cell line identity by short tandem repeat analysis. 293 Super TOPFlash (293 STF) cells were a gift from Ethan Lee (Vanderbilt University, Nashville, Tennessee, USA) and Jeremy Nathans (Johns Hopkins University, Baltimore, Maryland, USA) (21,37). Although 293 STF cells were not authenticated in our laboratory, they demonstrate the expected G418 resistance and WNT-induced TOPFlash reporter activity. 293T-FLAG-LRP6 cells were a gift from Victoria Ng and Ethan Lee (both from Vanderbilt University, Nashville, Tennessee, USA). MC38 cells were a gift from Barbara Fingleton (Vanderbilt University, Nashville, Tennessee, USA). YAMC cells, generated and as described in ref. 71, were obtained from the VUMC Digestive Disease Research Center (DDRC) GI Organoid Subcore.
Study approval. All animal experiments were carried out in accordance with protocols approved by the IACUC of VUMC. All human tissues were provided by the Western Division of the CHTN in accordance with the VUMC IRB.
Data availability. Values for all data points found in graphs can be found in the supplemental supporting data values file.

Author contributions
JMP designed and performed experiments, analyzed data, and wrote the manuscript. REB, NJB, APO, and SPS performed experiments. ZC and KSL analyzed scRNA-Seq data. MKW performed histological analyses and provided pathological expertise. SAA quantified tumoroid images. SK, VHN, JJT, and JJ generated reagents. JAG, EL, YAC, KSL, SPS, and CSW provided intellectual contributions to the experimental design and analysis. All authors edited and approved the manuscript.
Cell-surface biotinylation and isolation experiments. 293T cells were cultured to approximately 80% confluence in 10 cm plates and then treated with 3 mL complete DMEM or SELENOP-conditioned media for 2 hours. Cells were biotinylated and lysed with a Cell Surface Biotinylation and Isolation Kit (A44390, Pierce, Thermo Fisher Scientific) per the manufacturer's protocol. Lysate concentrations were quantified with a BCA Protein Assay Kit. Equal amounts of total protein were used for pulldown with NeutrAvidin Agarose (29200, Pierce, Thermo Fisher Scientific), and bound proteins were eluted with DTT (A39255, Pierce, Thermo Fisher Scientific).
pLX304-V5-mSELENOP plasmids (full-length and Δ258-299) were generated by Gateway cloning (75) (Thermo Fisher Scientific) per the manufacturer's protocol. Briefly, V5-mSELENOP was flanked by attB sites via PCR amplification from pCMV6-V5-mSELENOP (full-length or Δ258-299) using the primers listed in Supplemental Table 5 and Q5 Hot Start High-Fidelity 2X Master Mix (M0494S, New England BioLabs). attB-flanked PCR products were purified with the QIAquick PCR Purification Kit (28104, QIAGEN) prior to BP reactions with Gateway pDONR221 (12536017, Invitrogen, Thermo Fisher Scientific) using Gateway BP Clonase II Enzyme mix (11789020, Invitrogen, Thermo Fisher Scientific). attL/attR recombination (LR) reactions were then performed with the attB/ attP recombination (BP) reactions and pLX304 (25890, Addgene) using Gateway LR Clonase II Enzyme mix (11791020, Invitrogen, Thermo Fisher Scientific). All pLX304-V5-mSELENOP constructs were sequence verified by Plasmidsaurus. ported by the Prince Bernhard Cultural Foundation (Cultural Foundation Grant) and the Royal Netherlands Academy of Arts and Sciences (Academy Ter Meulen Grant). SK was supported by the Japan Society for the Promotion of Science (KAKENHI grants JP22K11715 and JP19K11782). We would also like to thank the VUMC TPSR and Western Division of the Cooperative Human Tissue Network for aid with histology and tissue procurement, respectively. This manuscript is dedicated to the late Dr. Raymond Burk, an international expert on selenium in human health, a dedicated scientist, compassionate clinician, and mentor to many.