Introduction

Zinc And Ring Finger 3 (ZNRF3) and Ring Finger Protein 43 (RNF43) are two closely related single-pass transmembrane E3 ligases with significant roles in embryonic development, tissue homeostasis and regeneration, and diseases (Basham et al, 2019; Hao et al, 2012; Koo et al, 2012; Lee et al, 2020; Planas-Paz et al, 2016; Sun et al, 2021; Szenker-Ravi et al, 2018). Extensive research has demonstrated their involvement in critical processes such as limb development (Szenker-Ravi et al., 2018), liver zonation (Planas-Paz et al., 2016), and mammalian sex determination (Harris et al, 2018). These two enzymes also function as tumor suppressors, as evidenced by the promotion of intestinal and adrenal hyperplasia (Basham et al., 2019; Koo et al., 2012) and hepatocellular carcinogenesis (Mastrogiovanni et al, 2020) upon tissue-specific inactivation of Znrf3/Rnf43. Similarly, Rnf43 deficiency leads to thickened mucosa, hyperplasia and cellular atypia in the stomach (Neumeyer et al, 2019), and enhanced tumor growth in a mouse model of inflammatory colorectal cancer (Eto et al, 2018). Conversely, overexpression of ZNRF3 and RNF43 suppresses cancer cell proliferation, migration and invasion, and drug resistance in multiple human cancer cell lines (Jiang et al, 2013; Pangestu et al, 2021; Qiu et al, 2016; Radaszkiewicz et al, 2021; Zhou et al, 2013). Importantly, in cancer patients, ZNRF3 and RNF43 are frequently inactivated by gene deletion or loss-of-function mutations (Hao et al, 2016; Yu et al, 2020). For instance, ZNRF3 is the most altered gene in adrenocortical carcinomas (ACCs), homozygously deleted in approximately 20% of ACCs (Assie et al, 2014). RNF43 is mutated in approximately 15% of endometrial cancer, 12% of stomach cancer, 11% of colorectal cancer, and 7% of pancreatic cancer (Tu et al, 2019). These findings highlight the clinical significance of ZNRF3/RNF43 and the importance of fully understanding the molecular mechanism(s) by which they affect development and cancer.

The primary known molecular function of ZNRF3/RNF43 is the regulation of WNT signaling. As transmembrane E3 ligases, they ubiquitinate the WNT signaling receptor Frizzled, targeting it for degradation thus dampening WNT signaling (Hao et al., 2012; Koo et al., 2012). Conversely, R- spondins (RSPO1-4) act as antagonistic peptide ligands, binding to ZNRF3/RNF43 and promoting their auto-ubiquitination and membrane clearance, thus maintaining WNT receptor levels and potentiating WNT signaling (Carmon et al, 2011; de Lau et al, 2011; Dubey et al, 2020; Glinka et al, 2011; Hao et al., 2012; Koo et al., 2012; Lebensohn & Rohatgi, 2018; Park et al, 2018; Park et al, 2020; Szenker-Ravi et al., 2018). Furthermore, activation of WNT signaling transactivates ZNRF3 and RNF43 expression via β-Catenin (Hao et al., 2012; Koo et al., 2012), creating a negative feedback loop that tightly regulates WNT signaling during normal development and tissue homeostasis. However, in cancer cells, alterations such as recurrent RSPO gene fusion or ZNRF3/RNF43 deletion or mutations disrupts this feedback loop, resulting in hyperactive WNT signaling and subsequent cancer development and progression (Hao et al., 2016; Seshagiri et al, 2012).

Although most molecular functions of RSPO-ZNRF3/RNF43 have been linked to their modulation of WNT signaling, emerging evidence suggests that ZNRF3 and RNF43 may possess WNT- independent functions. For instance, RSPO and WNT ligands exhibit non-equivalent roles in various processes, including mammary epithelial cell growth (Klauzinska et al, 2012), mammary side branches (Geng et al, 2020), intestinal stem cell self-renewal (Yan et al, 2017), and cochlea development (Mulvaney et al, 2013). Furthermore, loss of RNF43 function promotes mouse gastric epithelium proliferation and human gastric cancer cell xenograft growth, without detectable impact on WNT signaling activity (Neumeyer et al, 2020; Neumeyer et al., 2019). Additionally, the inhibitory effects of RSPO-ZNRF3 on BMP signaling are not abolished by β-Catenin knockdown (Lee et al., 2020). Moreover, approximately 40% of ZNRF3 or RNF43-mutant colon tumors also harbor alterations in either APC or CTNNB1, suggesting the involvement of these E3 ligases beyond the regulation of WNT signaling (Cerami et al, 2012; Gao et al, 2013). This is further supported by a recent report of a hotspot RNF43 G659 frameshift mutation that does not appear to affect WNT signaling (Fang et al, 2022). However, the extent to which ZNRF3 and RNF43 act on other substrates, apart from WNT receptors, to influence development and cancer has been unclear.

In this study, we report a new function of ZNRF3 and RNF43. Initially, we observed a strong negative correlation between ZNRF3/RNF43 mRNA levels and EGFR protein expression in datasets of human adrenocortical carcinomas and colorectal cancers. Subsequently, we investigated the impact of depleting ZNRF3 or RNF43 on EGFR clearance at the cell surface in mouse embryonic fibroblasts (MEFs) and human cancer cell lines including APC-mutated colorectal cancer cells. Notably, we discovered that ZNRF3/RNF43 interact with EGFR through their extracellular domains, leading to EGFR ubiquitination and degradation via the E3 ligase RING domain. Additionally, we substantiated our findings in multiple cell lines, human cancer cell xenograft models, and genetically engineered mouse models, wherein the loss of ZNRF3/RNF43 resulted in elevated EGFR levels and facilitated cancer progression.

Results

EGFR protein levels negatively correlate with ZNRF3/RNF43 mRNA expression in multiple human cancers

To elucidate novel signaling pathways regulated by ZNRF3/RNF43 in cancer, we conducted integrative proteogenomic analyses of human cancer datasets using LinkedOmics (Vasaikar et al, 2018). Given the critical role of ZNRF3 in adrenal hemostasis (Basham et al., 2019) and its frequent deep deletion in approximately 20% of adrenocortical carcinomas (ACCs) (Assie et al., 2014), our initial focus was to examine the ACC dataset from TCGA, comprising 92 samples, to identify proteins exhibiting correlations with ZNRF3 mRNA levels. Among the proteins evaluated by Reverse Phase Protein Array (RPPA), our analysis revealed EGFR protein levels to be most negatively correlated with ZNRF3 mRNA levels (r = -0.50 and p = 4.5e-4) (Fig. 1A and Fig. S1A). Notably, ACC tumors with deep deletion in ZNRF3 exhibited highest EGFR protein levels compared to ACC tumors with other alterations in the ZNRF3 locus (Fig. 1B). These findings suggest that disruption of the gene function of ZNRF3 may lead to upregulation of EGFR in ACCs.

Fig. 1

Besides frequent deletions in ACC, ZNRF3 and RNF43 are also important tumor suppressors in the more commonly occurring colorectal cancer (CRC) (Bond et al, 2016; Koo et al., 2012); thus, we next analyzed the TCGA colorectal adenocarcinoma (n=629) and CPTAC colon adenocarcinoma (n=210) datasets. By incorporating RPPA-based (TCGA CO/READ) and mass spectrometry-based (CPTAC COAD) proteomic data, we confirmed EGFR protein also to be the one most negatively correlated with ZNRF3 and RNF43 mRNA expression in CRC (Fig. 1C and Fig. S1B-E). Of note, we utilized mRNA abundance of ZNRF3/RNF43 mRNA abundances rather than protein levels due to the lack of protein measurements for these two low-abundance enzymes in the current proteomic datasets. Mutations in ZNRF3/RNF43 are found in approximately 15% of CRCs overall, with high frequencies observed in the microsatellite instability-high (MSI-H) subtype (60%-80%) (Bond et al., 2016; Giannakis et al, 2014; Tu et al., 2019; Vasaikar et al, 2019). In particular, the RNF43 G659Vfs*41 frameshift mutation accounts for 40-50% of all RNF43 mutations in multiple MSI-H cancer types (Tu et al., 2019). Interestingly, CRCs harboring the RNF43 G659Vfs*41 mutation exhibited significantly higher levels of EGFR protein compared to RNF43 wild-type (WT) tumors (Fig. 1D and Fig. S1F). A similar difference was observed in stomach cancer, which also displayed a high frequency of the RNF43 G659Vfs*41 mutation (Fig. S1G). Additionally, in the microsatellite stable (MSS) subtype of CRCs, predominantly expressing WT RNF43, we found a negative association between RNF43 mRNA and EGFR protein levels (r=-0.36 and p=0.0018) (Fig. 1E). Using cBioPortal (Cerami et al., 2012; Gao et al., 2013), we also detected a negative correlation between ZNRF3/RNF43 expression and EGFR protein in multiple other cancer types, including prostate cancer where ZNRF3 or RNF43 is deleted or mutated with a rate of 5% (Fig. 1F). Collectively, these results suggest that disruption of ZNRF3 and RNF43 may up-regulate EGFR protein levels in human cancers.

ZNRF3 and RNF43 downregulate the EGFR protein level

To test whether ZNRF3 and RNF43 regulate EGFR protein, we performed gain-of-function and loss-of-function experiments in several cell lines. In MDA-MB-231, a breast cancer cell line with intermediate levels of ZNRF3/RNF43, overexpression of either ZNRF3 or RNF43 substantially reduced EGFR protein levels (Fig. 2A). Remarkably, ZNRF3 and RNF43 decreased EGFR as robustly as CBL, the best-known E3 ligase of EGFR (Levkowitz et al, 1998; Waterman et al, 1999). Similarly, in 293T cells, ZNRF3/RNF43 overexpression reduced the levels of EGFR (Fig. S2A). As predicted, the phosphorylated form of EGFR declined as well (Fig. S2A). Notably, the reduction in both total EGFR and P-EGFR levels reached an extent comparable to that achieved by CBL (Fig. S2A). Consistent with their known function, ZNRF3/RNF43 overexpression reduced the level of their substrate FZD5 (Hao et al., 2012; Koo et al., 2012) (Fig. S2A). Additionally, these two E3 ligases exhibited selectivity in their regulation of growth factor receptors since overexpression of ZNRF3 or RNF43 did not decrease levels of TGFβ receptor-I (Fig. S2B) or FGFR1 (Fig. S2C).

Fig. 2

To test the impact of ZNRF3/RNF43 loss on EGFR levels, we compared WT and Znrf3 knockout (KO) MEFs because the WT MEFs have adequate Znrf3 expression but very minimal Rnf43 expression (Lienert et al, 2011). We found that Znrf3 KO enhanced both EGFR and P-EGFR levels, both in the absence and presence of recombinant EGF (Fig. 2B). To test the effect of RNF43 KO, we utilized HT29, a colorectal cancer cell line that expresses WT RNF43 at a high level but minimal ZNRF3. CRISPR-KO of RNF43 enhanced EGFR and P-EGFR levels regardless of EGF stimulation (Fig. 2C). We validated these results using three additional independent RNF43 gRNAs (Fig. S2D) and extended our analysis to HCC1954, a breast cancer cell line (Fig. S2E). Since this line expresses the EGFR family member HER2, we also assayed the effects of RNF43 KO on HER2. We found that knockout of either ZNRF3 or RNF43 enhanced the protein levels of HER2 (Fig. S2E), suggesting that ZNRF3 and RNF43 downregulate the levels of EGFR and its family members and that the loss of these E3 ligases unleashes this family of receptor tyrosine kinases.

RSPO1-4 are the antagonistic ligands of ZNRF3 and RNF43 (Hao et al., 2012; Koo et al., 2012). They bind to both ZNRF3/RNF43 and LGR4/5/6, resulting in ZNRF3/RNF43 auto-ubiquitination and degradation (Hao et al., 2012; Koo et al., 2012). Approximately 18% of CRCs harbor amplification or mutations of RSPO1-4 or LGR4/5/6, and 10% of CRCs show recurrent RSPO2/3 gene fusions (Seeber et al, 2019; Seshagiri et al., 2012). These genetic alterations potentially lead to the inhibition of ZNRF3/RNF43, thereby activating WNT and EGFR signaling. Therefore, we tested whether RSPO affected EGFR protein levels in CRC. Short-term treatment with recombinant RSPO2 increased EGFR levels in two CRC cell lines (Fig. 2D and E) as well as 293T cells (Fig. S2F). Further, this impact of RSPO2 depended on the presence of intact ZNRF3/RNF43, as RSPO2 failed to elevate EGFR levels in HT29 cells further knocked-out for RNF43, the predominant one compared to ZNRF3 (Fig. S2G). Conversely, the RSPO2 R65A/Q70A mutant, which cannot bind to ZNRF3/RNF43 (Xie et al, 2013), failed to elevate EGFR levels (Fig. 2F). On the other hand, the RSPO2 F105A/F109A mutant, which cannot bind to LGRs (Xie et al., 2013), still enhanced EGFR levels similar to RSPO2 WT (Fig. 2E). These data collectively suggest that RPSO2 regulation of EGFR does not rely on LGR-binding but requires its interaction with ZNRF3/RNF43.

ZNRF3 and RNF43 induce EGFR ubiquitination and degradation

After demonstrating that ZNRF3 and RNF43 regulate EGFR protein levels, we next sought to elucidate the underlying mechanism. We first examined whether ZNRF3 and RNF43 regulate EGFR transcripts and found that ZNRF3/RNF43 did not impact EGFR mRNA levels in both MEFs and HT29, as determined by qPCR (Fig. 3A and B). Then we studied the protein stability of EGFR in WT and RNF43 KO HT29 cells. We treated cells with recombinant EGF to induce EGFR internalization/degradation and then collected cell lysates at different time points to measure EGFR protein levels by Western blotting. We found that knockout of RNF43 delayed EGFR degradation and sustained P-EGFR levels (Fig. 3C). Since EGFR clearance is initiated at the cell membrane and accelerated by EGF treatment, we also evaluated the impact of ZNRF3/RNF43 on EGFR levels at the cell surface. Flow cytometry revealed that ZNRF3 knockout in MEFs (Fig. 3D) or RNF43 knockout in HT29 (Fig. 3E) increased cell surface EGFR and inhibited EGF-induced internalization of surface EGFR (indicated by the comparison of the black bar vs. red bar in Fig. 3D and E). These data together suggest that ZNRF3 and RNF43 induce cell surface EGFR internalization and degradation.

Fig. 3

EGFR degradation is primarily mediated by ubiquitination (Galcheva-Gargova et al, 1995; Levkowitz et al., 1998). To examine the impact of ZNRF3 and RNF43 on EGFR ubiquitination, we performed EGFR immunoprecipitation (IP) followed by ubiquitin (Ub) immunoblotting (IB). Remarkably, RNF43 overexpression in MDA-MB-231 cells enhanced EGFR ubiquitination as potently as CBL overexpression (Fig. 4A). Conversely, in RNF43 KO HT29 cells and MDA-MB- 231 cells, anti-Ubiquitin IP brought down substantially less EGFR protein (Fig. 4B and Fig. S3A), and anti-EGFR IP produced much less ubiquitinated forms of EGFR (Fig. 4C), indicating that RNF43 knockout diminished EGFR ubiquitination. Next, we asked whether the E3 ligase activity of ZNRF3/RNF43 is needed for regulating EGFR ubiquitination. The RING domain is required for ZNRF3/RNF43 E3 ubiquitin ligase function (Pickart & Eddins, 2004). Therefore, we compared the impact of ZNRF3/RNF43 vs. their RING domain deletion mutants (ΔRING) on EGFR protein levels. We found that both ΔRING mutants failed to decrease EGFR levels (Fig. 4D and E) or HER2 levels (Fig. S3B). Furthermore, while WT ZNRF3 overexpression increased EGFR ubiquitination detectable even in the absence proteasome or lysosome inhibitors, the ΔRING mutant failed to induce detectable upregulation of EGFR ubiquitination even in the presence of both proteasome and lysosome inhibitors (MG132 and BAF, respectively) (Fig. 4F). Together, these results demonstrate that ZNRF3 and RNF43 regulate EGFR ubiquitination and degradation through their E3 ligase activity.

Fig. 4

ZNRF3 and RNF43 form a complex with EGFR through the extracellular domain

We next investigated whether ZNRF3 and RNF43 interact with EGFR to regulate its ubiquitination. In a co-IP experiment using lysates from MDA-MB-231 cells co-infected with lentivirus carrying EGFR and Myc-tagged ZNRF3/RNF43, anti-EGFR IP pulled down overexpressed ZNRF3/RNF43 (Fig. 5A), indicating a complex between EGFR and ZNRF3/RNF43. Mass spectrometry of anti-EGFR immunoprecipitates identified endogenous ZNRF3 protein (Fig. S4A), confirming ZNRF3 as an interacting partner of EGFR. Importantly, a proximity ligation assay (PLA) provided visual confirmation of the close interaction between co-transfected ZNRF3 and EGFR in situ (Fig. 5B). Next, we sought to determine the specific regions involved in the interaction between ZNRF3/RNF43 and EGFR. ZNRF3 and RNF43 are transmembrane proteins comprising a long extracellular domain (ECD), a single-span transmembrane domain (TM), and a catalytic intracellular domain (ICD) (Fig. 5C). We thus generated constructs expressing ZNRF3 ECD-TM and TM-ICD, with an N-terminal 3X Myc-tag inserted after the signal peptide for detection (Fig. 5C). Immunofluorescence (IF) confirmed the membrane localization of these two peptides (Fig. 5D), after which we performed co-IP using anti-Myc. EGFR was co- immunoprecipitated with Myc-tagged ZNRF3 full-length (FL) and ECD-TM, but not TM-ICD, (Fig. 5E). Conversely, anti-EGFR IP pulled down ZNRF3 FL and ECD-TM, but not TM-ICD (Fig. 5E). Similar results were obtained when performing anti-HA IP of HA-tagged RNF43 FL and ECD-TM, which also brought down EGFR (Fig. S4B and C). These data collectively indicate that ZNRF3 and RNF43 interact with EGFR via their extracellular domains, in accordance with evidence indicating the ZNRF3/RNF43 interaction with Frizzled via extracellular domains (Tsukiyama et al, 2015). while in contrast with reports detecting ZNRF3/RN43 interaction with Frizzled via their intracellular domains and other adaptor proteins (Jiang et al, 2015).

Fig. 5

The protease-associated (PA) domain is the only conservative part between the extracellular fragments of ZNRF3 and RNF43 (Tsukiyama et al, 2021). Therefore, we tested whether this domain is necessary for the interaction between ZNRF3/RNF43 with EGFR. We created PA domain-deletion mutants of ZNRF3/RNF43 (Fig. S5A) and found that these deletion mutants (ΔPA) also interacted with EGFR (Fig. S5B and C), indicating that the PA domain is dispensable for ZNRF3/RNF43 interaction with EGFR. In accordance with previous reports that the PA domain is dispensable for suppression of WNT signaling (Jiang et al., 2015; Radaszkiewicz & Bryja, 2020), we found that ΔPA ZNRF3 retained the ability to suppress the Top-Flash WNT reporter (Fig. S5D).

Loss of ZNRF3 and RNF43 unleashes EGFR-mediated cell growth in 2D culture and organoids

EGFR is an important tyrosine kinase that mediates many cellular activities, especially cell growth (Wieduwilt & Moasser, 2008; Yue et al, 2021). To study the biological consequence of ZNRF3/RNF43 down-regulating EGFR, we first compared cell growth of WT and ZNRF3/RNF43-depleted cells. Znrf3 knockout in MEFs led to the upregulation of EGFR (Fig. 2B) and an increase in cell growth (Fig. 6A), without affecting canonical WNT signaling activity based on qPCR for WNT target genes (Sox2, Axin2, Ccnd1) (Fig. 6B). EGF and EGFR signaling are also critical for the culture and maintenance of organoids derived from Apc-deficient mouse intestinal adenomas (Sato et al, 2011), which exhibits constitutively canonical WNT signaling that is no longer modulated by the RSPO-ZNRF3/RNF43 control at the membrane receptor level (Tsukiyama et al., 2021; Tsukiyama et al, 2020). Therefore, we established intestinal tumor organoids using the intestinal polyps from the Apc^min mice (Evans et al, 1992) and supplemented RSPO1 in the organoid culture medium to investigate the effect of ZNRF3/RNF43 deactivation on organoid growth and EGFR. Remarkably, the addition of RSPO1 increased the size of Apc^min mouse intestinal tumor organoids (Fig. 6C and D) in agreement with a previous report (Lahde et al, 2021) although it did not significantly affect the frequency of organoids detected (Fig. 6E). Immunoblotting analysis confirmed elevated EGFR levels after RSPO1 treatment (Fig. 6F) while qPCR showed no significant increase of Egfr nor WNT target genes including Axin2, Myc, Sox2, and Cd44 (Fig. 6G). Furthermore, we performed functional assays in HT29 cells, which bear APC mutations. Overexpression of ZNRF3 in HT29 substantially inhibited cell growth (Fig. 6H) and reduced EGFR levels (Fig. 6I). Conversely, knockout of RNF43 in HT29 activated multiple EGFR downstream effectors, as RPPA detected increased levels of P-AKT, P-ERK, P-GSK3, P-STAT1, and P-PRAS40 (Fig. S6). Importantly, after confirming abolishment of EGFR phosphorylation (Fig. 6J), the EGFR activity inhibitor erlotinib was found to block the growth gain caused by RNF43 knockout (Fig. 6K). Together, these data indicate that loss of ZNRF3 and RNF43 in cancer cells can promote tumor cell growth through EGFR signaling without significantly engaging canonical WNT signaling.

Fig. 6

ZNRF3/RNF43 inactivation enhances EGFR and promotes tumor growth in vivo

To demonstrate the function of the ZNRF3/RNF43-EGFR signaling pathway in vivo, we first xenografted luciferase-labeled HT29 cells overexpressing either GFP or ZNRF3 into NSG mice by flank injection and monitored tumor growth by bioluminescence imaging (Fig. 7A). We found that overexpression of ZNRF3 significantly inhibited tumor growth (Fig. 7A). These tumors exhibited decreased EGFR and P-EGFR levels compared to WT tumors (Fig. 7B). However, canonical WNT signaling, based on levels of active-β-Catenin (non-phosphorylated at Ser33/37/Thr41; Fig. 7B), remained unaffected, suggesting that ZNRF3 suppresses tumor growth through down-regulation of EGRF signaling without substantial involvement of canonical WNT signaling.

Fig. 7

Furthermore, we validated this finding in human prostate cancer, where ZNRF3 and RNF43 mRNA levels negatively correlated with EGFR protein levels (Fig. 1F). Specifically, we tested whether ZNRF3/RNF43 loss affects EGFR signaling and promotes prostate tumorigenesis in vivo. By breeding mice that are floxed at Znrf3 and Rnf43 (Koo et al., 2012) with mice transgenic for Probasin-Cre (Wu et al, 2001), we generated prostate-specific Znrf3/Rnf43 knockout mice. At one year of age, these mice exhibited multi-focal prostatic dysplasia and multifocal invasive tumors, as shown by the H&E images (Fig. 7C) and pathological scores (Fig. 7D and S7A-D). Immunohistochemical staining revealed elevated levels of both EGFR and P-EGFR in Znrf3/Rnf43 knockout prostatic dysplasia and tumors, compared to WT prostate tissues (Fig. 7E), indicating EGFR signaling activation due to Znrf3/Rnf43 knockout. In contrast, while β-Catenin staining was more intense in knockout tissues and tumors than in WT mice for unknown significance, the β- Catenin signals were mainly restricted to the cell membrane in both WT and Znrf3/Rnf43 knockout mouse prostate tissues and tumors (Fig. 7E), suggesting no substantial activation of canonical Wnt signaling. Together, these results suggest that loss of ZNRF/RNF43 elevates EGFR levels and signaling, promoting tumor development.

Discussion

The two homologous E3 ubiquitin ligases, ZNRF3 and RNF43, and their antagonistic ligand RSPO have been extensively studied as WNT signaling modulators during normal development and diseases (Basham et al., 2019; Hao et al., 2012; Koo et al., 2012; Lee et al., 2020; Planas-Paz et al., 2016; Sun et al., 2021; Szenker-Ravi et al., 2018). Their primary role has been attributed to negatively regulating WNT receptors Frizzled through ubiquitination and degradation (Hao et al., 2012; Koo et al., 2012), and the early mass spectrometry analysis of 293T cells following inducible expression of RNF43 only detected downregulation of Frizzled and LRP5 (Koo et al., 2012). However, a few later reports suggested that ZNRF3/RNF43 might also regulate other substrates and exert distinct functions in context-dependent manners. For instance, ZNRF3/RNF43 may induce degradation of BMP receptor BMPR1A (Lee et al., 2020), maternal dorsal determinant Huluwa (Zhu et al, 2021), non-canonical WNT component VANGL2 (Radaszkiewicz et al., 2021), and cell adhesion protein E-Cadherin (Zhang et al, 2019) in different physiological or pathological conditions. In the context of cancer, ZNRF3 and RNF43 are frequently deactivated by gene deletion, mutation, or other means (Hao et al., 2016; Seshagiri et al., 2012). Therefore, this study focused on elucidating ZNRF3/RNF43-regulated signaling pathways in cancer. We discovered that EGFR protein is the most negatively associated with ZNRF3/RNF43 expression using proteogenomic data of cancer patients and that ZNRF3/RNF43 interact with EGFR and down- regulate EGFR through ubiquitination and degradation. Using multiple cell lines and animal models, we demonstrated that loss-of-function changes of ZNRF3/RNF43 elevate EGFR signaling activity to promote cancer progression.

In colorectal cancer (CRC), ZNRF3 and RNF43 mutations occur frequently (Bond et al., 2016; Giannakis et al., 2014; Tu et al., 2019; Vasaikar et al., 2019) and have been mainly associated with the potentiation of WNT signaling (Hao et al., 2016; Hao et al., 2012; Koo et al., 2012). However, our findings suggest that ZNRF3/RNF43 deactivation may unleash other key oncogenic pathways in CRC, especially in the context of APC/CTNNB1 co-mutations that bypass the need for WNT receptor activation for pathway activation (Cerami et al., 2012; Gao et al., 2013). We provide evidence that in human colorectal cancer cells with a hyperactive WNT pathway due to APC mutations, knockout of RNF43 promotes cancer cell growth by activating EGFR signaling. Additionally, we found that the RNF43 G659Vfs*41 mutant, a hotspot frameshift mutation in human colorectal tumors that may not potentiate WNT signaling (Li et al, 2020; Tu et al., 2019), elevates EGFR protein levels, suggesting that this mutation may exert its oncogenic role in cancer through EGFR upregulation. Our data also have clinical implications, as germline polymorphisms and tumor gene expression levels of ZNRF3/RNF43 and RSPO may be related to tumor response to anti-EGFR monoclonal antibody cetuximab-based treatment in CRC patients (Battaglin et al, 2020), providing insights into understanding the interplay between RSPO-ZNRF3/RNF43 and EGFR protein levels and anti-EGFR activity in CRC treatment.

By identifying ZNRF3/RNF43 as crucial regulators linking EGFR and WNT signaling at the membrane receptor level, our work reveals a novel mechanism of EGFR and WNT coactivation, which is commonly observed in human cancers (Hu & Li, 2010). Further work should explore the extent to which EGFR and WNT receptors (Frizzled and LRP5/6) each transmit the effect of ZNRF3/RNF43 deactivation at different cancer stages and investigate the efficacy of blocking both EGFR and WNT signaling in treatment of ZNRF3/RNF43-deactivated tumors. Our findings also have implications for organoid culture and normal development. For example, while RSPO and EGF are both required for organoid growth and maintenance, our data suggest that RSPO can both help sustain surface EGFR levels by deactivating ZNRF3/RNF43 and synergize with EGF to boost EGFR signaling activity. Comparative studies on ZNRF3/RNF43 regulation of EGFR versus WNT receptors in terms of substrate abundance, enzyme-substrate binding affinity, in vivo ligand availability, and temporal and spatial regulatory mechanisms can provide a comprehensive understanding of the RSPO-ZNRF3/RNF43-regulated signaling network. A recent manuscript reports that RSPO1 addition to Apc-deficient mouse intestinal organoids led to an increased size but a decreased number, and implicated abnormal recruitment of the TGF-β/SMAD pathway via LGR5/TGFBRII heterodimers as a mechanism (Lahde et al, 2021). We did not detect enhanced TGF-β signaling in our Apc^min mouse intestinal tumor culture treated by RSPO1 for two days or eight days (Fig. S8). Furthermore, we did not observe an impact of ZNRF3/RNF43 overexpression on TGFβ receptor I protein levels in our 293T culture studies (Fig. S2B), although our bioinformatic analysis of patient CRCs revealed that SMAD4, a crucial transcription factor downstream of TGF-β receptors, was among the proteins downregulated by ZRNF3/RNF43 (Fig. 1C). Therefore, whether TGFβ signaling is directly affected by the RSPO-ZNRF3/RNF43- regulated signaling network needs to be further investigated.

Structural studies have identified the key amino acids and motifs mediating ZNRF3/RNF43 binding with their peptide ligand RSPO (Chen et al, 2013; Peng et al, 2013; Zebisch et al, 2013) and also suggested that RSPO has the closest structural homology with EGFR (domain IV) and insulin receptor (domain II) (Chen et al., 2013). Our co-IP data indicate that ZNRF3/RNF43 interact with EGFR through their extracellular domains by co-IP. It would also be interesting to determine further the specific domains and amino acids that are responsible for ZNRF3/RNF43 and EGFR interaction and to study whether clinically relevant point mutations at the extracellular domains of ZNRF3/RNF43 and/or EGFR hinder their binding affinity, thereby enhancing EGFR signaling to promote cancer initiation and progression.

RSPO1-4 bind to both ZNRF3/RNF43 and LGR4/5/6, resulting in ZNRF3/RNF43 auto- ubiquitination and degradation (Hao et al., 2012; Koo et al., 2012). RSPO2/3 can also inhibit ZNRF3/RNF43 independently of LGR4/5/6 (Lebensohn & Rohatgi, 2018; Park et al., 2018; Szenker-Ravi et al., 2018) because RSPO2/3 can bind to the RNF43/ZNRF3 with relatively high affinity, especially in the presence of heparan sulfate proteoglycans (HSPGs) (Dubey et al., 2020). Approximately 18% of CRCs harbor amplification or mutations of RSPO1-4 or LGR4/5/6 while 10% of CRCs have recurrent RSPO2/3 gene fusions (Seshagiri et al., 2012). These genetic alterations potentially lead to enhanced inhibition of ZNRF3/RNF43, activating WNT and EGFR signaling (Seeber et al., 2019). While the involvement of each co-receptor (LGR4/5/6; HSPGs) in regulating RSPO-ZNRF3/RNF43 degradation of EGFR requires more investigation, LGR4/5/6 may be dispensable for RSPO2-ZNRF3/RNF43 regulation of EGFR because we found that RSPO2 mutant that cannot bind to LGR4/5/6 still enhanced EGFR protein levels. It is worth noting that LGR4/5/6 may have a parallel mechanism controlling EGFR levels as we have previously observed that LGR4 can enhance EGFR signaling independently of WNT activation (Yue et al., 2021).

Our work also uncovers a new mechanism of regulating EGFR levels in cancer. Hyperactivation of EGFR in cancer has been attributed to activating mutations, gene amplification, aberrant gene expression, and defective endocytosis/degradation (Nakai et al, 2016). But these mechanisms cannot account for all the 60-80% of CRC cases whose EGFR levels are upregulated (Cohen, 2003). ZNRF3/RNF43 loss-mediated EGFR stabilization represents a novel mechanism of EGFR upregulation, explaining the elevated EGFR protein levels observed in many human cancers without EGFR gene amplification or overexpression. In our study, ZNRF3/RNF43 can regulate EGFR levels under basal culture conditions as well as with EGF ligand stimulation, but the commonly known EGFR E3 ligase, CBL, mediates ligand-dependent EGFR ubiquitination and degradation (Batzer et al, 1994; Duan et al, 2011; Grovdal et al, 2004; Levkowitz et al, 1999; Levkowitz et al., 1998; Waterman et al, 2002; Waterman et al., 1999). Therefore, future studies should investigate whether ZNRF3/RNF43 and CBL regulate EGFR ubiquitination and degradation differently, whether ZNRF3/RNF43 regulate both ligand-independent and ligand- dependent EGFR ubiquitination/degradation and consequently affecting different downstream signaling pathways, and whether different EGFR ligands or ligand concentrations direct EGFR to ZNRF3/RNF43 versus CBL for ubiquitination and degradation (Harris et al, 2003; Roepstorff et al, 2009). In addition, a feedback regulation often contributes to tight controls of signaling activation (Chandarlapaty, 2012). In accordance, ZNRF3/RNF43 regulation of WNT signaling is feedback-controlled by ZNRF3/RNF43 as WNT target genes (Hao et al., 2012; Koo et al., 2012). Whether and how EGFR signaling may affect levels and activities of ZNRF3/RNF43 and their partners including RSPOs and LGRs as a feedback regulation remain to be understood.

In conclusion, our study unveils a novel ZNRF3/RNF43-EGFR signaling axis in cancer and provides critical insights into RSPO-ZNRF3/RNF43 signaling during cancer progression. Understanding how ZNRF3/RNF43 regulates EGFR and the crosstalk between EGFR and WNT receptors may offer potential therapeutic targets for cancer treatment. Additionally, our findings have implications in organoid culture and normal development, highlighting the significance of RSPO-ZNRF3/RNF43 signaling in various physiological processes. Further research on ZNRF3/RNF43 regulation of EGFR versus WNT receptors in different cancer stages can contribute to a comprehensive understanding of RSPO-ZNRF3/RNF43 signaling in cancer progression and provide valuable knowledge for potential therapeutic interventions.

Materials and Methods

Animal Studies

All experiments using mice were performed utilizing procedures approved by the Institutional Animal Care and Use Committee at Baylor College of Medicine and the Van Andel Institute. The Rnf43-flox; Znrf3-flox were obtained from the laboratory of Hans Clevers (Koo et al., 2012). CMV-Cre animals were ordered from Jackson Laboratories (JAX stock #006054) (Schwenk et al, 1995). C57Bl/6J used for backcrossing were ordered from Jackson Laboratories (JAX stock #000664). Probasin-Cre animals were developed in the laboratory of Pradip Roy-Burman (Wu et al., 2001). NOD.Cg-Prkdc scid Il2rg tm1Wjl /SzJ (NSG) mice were purchased from Jackson Laboratories (JAX stock #005557) and bred in-house.

Cell lines

MDA-MB-231, MDA-MB-468, and HEK293T cells were purchased from ATCC. HT29 and LS180 cells were provided by Q. Liu; HCT116 cells were provided by N. Shroyer; MEF cells were generated from e12.5 embryos using standard methods (Herrera et al, 1996). MDA-MB-231, MDA-MB-468, HT29, HCT116, HEK293T, and MEF cells were cultured in DMEM (CORNING, 10-013-CV) with 10% FBS and 1% penicillin/streptomycin. LS180 cells were cultured in RPMI1640 (CORNING, 10-013-CV) with 10% FBS and 1% penicillin/streptomycin. All cell lines were routinely tested for mycoplasma contamination.

Constructs

Plasmids expressing Flag-FZD5, RNF43, HA-RNF43, or HA-RNF43 (CD8 sp) were provided by Q. Liu. Plasmids expressing Myc-RNF43, Myc-ZNRF3, or Myc-ZNRF3 ΔRING were provided by F. Cong. Plasmids for Myc-ZNRF3-ECD-TM, Myc-ZNRF3-TM-ICD, Myc-ZNRF3 ΔPA, HA- RNF43 ΔPA, HA-RNF43 ΔRING, HA-RNF43-ECD-TM were constructed by subcloning using In-Fusion (TaKaRa, 639642). Plasmids were further constructed for lentiviral expression by subcloning into pBobi vector by In-Fusion.

Generation of RNF43 or ZNRF3 knockout cancer cell lines

HT29 RNF43 knockout cells were provided by Q. Liu (guide RNA sequence, g0: 5’- GGCTGCTGATGGCTACCCTGC-3’). Other CRISPR guide sequences for knocking out RNF43 were designed by http://crispr.mit.edu and cloned into lentiCRISPR v2 (Addgene 52961). Sequences were as follows: g1: 5’-TGGACGCACAGGACTGGTAC-3’; g2: 5’- CAGAGTGATCCCCTTGAAAA-3’; g3: 5’-GGGCAGCCAGCTGCAGCTGG-3’. Lentiviral construct for knocking out ZNRF3 was provided by Q. Liu (guide RNA sequence, 5’- AGGACTTGTATGAATATGGC-3’) (Jiang et al., 2015; Tu et al., 2019). Cancer cells were infected with lentivirus-carrying CRISPR constructs or vector, and then selected in the culture medium containing 2 µg/ml puromycin (InvivoGen, ant-pr-1).

Generation of Znrf3 knockout MEFs

Global null alleles of Rnf43 and Znrf3 were generated in vivo by crossing Rnf43-flox; Znrf3-flox mice (Koo et al., 2012) with CMV-Cre (Schwenk et al., 1995). The resulting mice were backcrossed to C57Bl/6J animals to remove the Cre transgene and generate global deletion mouse colonies. Global deletions were confirmed using allele-specific PCR for both Rnf43 and Znrf3. Mouse embryonic fibroblasts (MEFs) were generated using standard protocols (Durkin et al, 2013). Briefly, Rnf43-KO/+; Znrf3-KO/+ mice were crossed, and embryos were collected at E12.5. Yolk sacs were collected for genotyping, and embryo heads and internal organs were removed. The remaining tissue was minced using a razor blade and 0.5mL 0.05% Trypsin-EDTA was added. The suspension was incubated at 37°C for 30 minutes then cultured in MEF media (DMEM, 10% FBS, 1x PenStrep) for subsequent experiments.

Apc^min mouse intestinal tumor organoid culture

Intestinal polyps from 6-month-old female Apc^min mice (Evans et al., 1992) were isolated and maintained using previously described protocols (Sato et al., 2011; Xue & Shah, 2013), embedded in Matrigel (Corning, 356231), and cultured in basal culture medium (advanced Dulbecco’s modified Eagle medium/F12 (Invitrogen, 12634-028), penicillin/streptomycin (Invitrogen, 15140- 122), 10 mmol/L HEPES (Invitrogen, 15630-080), 2mM Glutamax (Invitrogen, 35050-061), 1× N2 (Invitrogen, 17502-048), 1× B27 (Invitrogen, 17504-044), 1 mmol/L N-acetylcysteine (Sigma- Aldrich, A9165-5G)) supplemented with 50 ng/ml mouse recombinant EGF (Invitrogen, PMG8043). The organoids were trypsinized with 0.05% Trypsin-EDTA (Invitrogen, 25300054) to obtain single cells, counted and embedded into Matrigel in equal numbers, cultured in basal culture medium supplemented with EGF or basal culture medium supplemented with EGF and 10% RSPO1 conditioned medium for 8 days. The organoids were imaged using Leica DMi8 Inverted Microscope and then harvested for immunoblot analysis.

RNA extraction and quantitative PCR

Cells were lysed with TRIzol reagent (Invitrogen, 15596-026). Total RNA was extracted by chloroform and isopropanol precipitation. cDNA was obtained using the SuperScript III First- Strand Synthesis System (Invitrogen, 18080-051). qPCR analyses were performed with primers listed below using SsoAdvanced Universal SYBR Green Supermix (BIO-RAD, 1725270) in at least duplicate in three independent experiments. Plotted are data normalized to ACTB or Actb and relative to the control. The following primers were used: ACTB: 5’- ACTCTTCCAGCCTTCCTTCC-3’, 5’-CAGTGATCTCCTTCTGCATCC-3’; EGFR: 5’- TGCCCATGAGAAATTTACAGG-3’, 5’-ATGTTGCTGAGAAAGTCACTGC-3’; Actb: 5’- CATTGCTGACAGGATGCAGAAGG-3’, 5’-TGCTGGAAGGTGGACAGTGAGG-3’; Egfr: 5′- GGACTGTGTCTCCTGCCAGAAT-3′, 5′-GGCAGACATTCTGGATGGCACT-3′; Axin2: 5’- ATGGAGTCCCTCCTTACCGCAT-3’, 5’-GTTCCACAGGCGTCATCTCCTT-3’; Sox2: 5’- AACGGCAGCTACAGCATGATGC-3’, 5’-CGAGCTGGTCATGGAGTTGTAC-3’; Ccnd1: 5’- GCAGAAGGAGATTGTGCCATCC-3’, 5’-AGGAAGCGGTCCAGGTAGTTCA-3’. The primer sequences for Tgfbr1, Tgfbr2, Smad4, Ep300, Bach1 were from ORIGENE.

Flow cytometric analysis

Cells were trypsinized and resuspended in antibody dilution buffer (0.5% BSA in PBS). 5 X 10^5 cells were incubated with EGFR (D1D4J) XP Rabbit antibody (PE Conjugate) (Cell Signaling Technology, 48685) for 1 hr at 4 °C. Cells were then washed with antibody dilution buffer and resuspended in antibody dilution buffer. The cells were then subjected to flow cytometry using a BD FACSCanto II (BD, NJ, USA). The resulting data were analyzed using the BD FACSDiva™ software.

Immunofluorescence

HEK293T cells were seeded on poly-D-lysine coated chamber slides (CORNING, 354632) and transfected with empty vector, or constructs expressing wild-type or fragments of RNF43 or ZNRF3. 48 hr after transfection, cells were fixed in 3.2% paraformaldehyde in PBS at RT for 15 min, permeabilized in 3% BSA / 0.1% saponin (Sigma-Aldrich, 47036) in PBS at RT for 30 min, and then incubated in the secondary antibody anti-Rabbit Alex 488 (Invitrogen, A32731) at RT for 1 hr. Afterward, the cells were stained with DAPI (Thermo Fisher Scientific, 62248) and mounted.

Immunohistochemistry

Immunochemistry staining was performed using the Vectastain Elite ABC system (Vector Laboratories, Burlingame, CA) and developed using DAB as chromogen (Agilent Dako, Santa Clara, CA, USA). Primary antibodies used in the study included EGFR (D1P9C) rabbit antibody (Mouse Preferred, Cell Signaling Technology, 71655), P-EGFR (Y1068) XP rabbit antibody (Cell Signaling Technology, 2234), Non-phospho (Active) β-Catenin (Ser33/37/Thr41) (D13A1) rabbit antibody (Cell Signaling Technology, 8814).

Proximity ligation assay

HCT116 cells were seeded to poly-d-lysine–coated chamber slide (Corning) and co-transfected with Myc-ZNRF3ΔR and EGFR. Two days after transfection, the cells were processed following the standard protocol provided by Sigma-Aldrich (Duolink In Situ Red Starter Kit Mouse/Rabbit). The primary antibodies, EGFR (D38B1) XP rabbit antibody (Cell Signaling Technology, 4267) and Myc-tag (9B11) mouse antibody (Cell Signaling Technology, 2276), were used to detect ZNRF3 and EGFR interaction. For negative control, no primary antibody was used. DAPI was used to stain nuclei. The stained cells were imaged using Leica DMi8 Fluorescence Microscope.

Western blot

Cells were lysed on ice using RIPA lysis buffer supplemented with protease inhibitors (Sigma- Aldrich, P8340) and phosphatase inhibitors (Sigma-Aldrich, P5726 and P0044). Cell lysates were then centrifuged at 14,000 rpm for 15 min at 4 °C. Supernatant was collected and the concentration of protein lysate was quantified by BCA assay (Thermo Fisher Scientific, 23225). Cell lysates were mixed with Laemmli sample buffer (BIO-RAD, 1610747) and β-mercaptoethanol (Thermo Fisher Scientific, 21985023) before boiling for 5 min at 95 °C. Equal amounts of protein lysates were loaded and run in 8-12% SDS-PAGE gel. Gels were transferred onto nitrocellulose membrane (Thermo Fisher Scientific, 88018) at 100 V for 90 min at 4 °C. Membranes were then blocked with 5 % non-fat milk in TBST at RT for 1 hr, incubated with primary antibodies overnight at 4 °C and secondary antibodies at RT for 1 hr, and scanned using the Odyssey LI-COR imaging system. Primary antibodies used in the study included EGFR (D38B1) XP rabbit antibody (Cell Signaling Technology, 4267), EGFR (A-10) mouse antibody (Santa Cruz, sc-373746), EGFR (D1P9C) rabbit antibody (Mouse Preferred, Cell Signaling Technology, 71655), P-EGFR (Y1068) XP rabbit antibody (Cell Signaling Technology, 3777), HA-tag mouse antibody (BioLegend 901501), HA-tag (C29F4) rabbit antibody (Cell Signaling Technology, 3724), Myc-tag (9B11) mouse antibody (Cell Signaling Technology, 2276), Myc-tag (71D10) rabbit antibody (Cell Signaling Technology, 2278), Flag-tag M2 mouse antibody (Sigma-Aldrich, F1804), Ub (P4D1) mouse antibody (Santa Cruz, sc-8017), CBL mouse antibody (Santa Cruz, sc-1651), Non-phospho (Active) β-Catenin (Ser33/37/Thr41) (D13A1) rabbit antibody (Cell Signaling Technology, 8814), β-Catenin mouse Antibody (BD Biosciences, 610153), GAPDH rabbit antibody (Santa Cruz, sc- 25778), β-Actin (Sigma, A5441) and α-Tubulin mouse antibody (Sigma-Aldrich, T9026). Secondary antibodies used included anti-Mouse IRDye 680RD (LI-COR, 926-68070) and anti- Rabbit IRDye 800CW (LI-COR, 926-32211).

Immunoprecipitation assay

Cells were lysed on ice using NP-40 or RIPA lysis buffer supplemented with protease inhibitors (Sigma-Aldrich, P8340). Cell lysates were then centrifuged at 14,000 rpm for 15 min at 4 °C. Supernatant was collected and the concentration of protein lysate was quantified by BCA. Protein lysates were incubated with either EGFR (Ab-13) mouse antibody (Thermo Fisher Scientific, MS- 609-P1) or Ub (P4D1) mouse antibody (Santa Cruz, sc-8017), and Protein G Sepharose beads (GE Healthcare, 17061801) overnight at 4 °C. Beads were washed with NP-40 or RIPA lysis buffer. Whole-cell lysates and immunoprecipitates were analyzed by western blot analysis.

Immunoprecipitation and mass spectrometry

The IP-MS experiments were performed as previously described (Chen et al, 2019). BCG823 cells were starved overnight and then treated with 50 ng/ml EGF for 120 mins before sample collection. For each IP experiment, 1 mg protein lysate was incubated with 5 µg EGFR (Ab-13) antibody for 2 hr at 4 °C and cleared by ultra-centrifugation (100,000 g, 15min). The supernatant was then incubated with 30 µl 50% protein A-Sepharose slurry (GE Healthcare, 17-0780-01) for 1 hr at 4 °C. The bead-bound complexes were washed 4 times with NETN buffer (20 mM Tris pH7.5, 1 mM EDTA, 0.5% NP-40, 150 mM NaCl) and eluted with 20 µl 2X Laemmli buffer and heated at 95°C for 10 min. IP samples were resolved on a NuPAGE 10% Bis-Tris gel (Life Technologies, WG1201BX10) in 1x MOPS running buffer; the gel was cut into 3 molecular weight regions plus the IgG heavy and light chain bands. Each band was in-gel digested overnight with 100 ng of trypsin which cleaves peptide chains at the C-termini of lysine or arginine in 20 µl of 50 mM NH4HCO3 at 37 °C. Peptides were extracted with 350 µl of 100% acetonitrile and 20 µl of 2% formic acid, and dried in a Savant Speed-Vac. Digested peptides were dissolved in 10 µl of loading solution (5% methanol containing 0.1% formic acid) and subjected to LC-MS/MS assay as described previously (Dharmat et al, 2018).

Reverse phase protein array (RPPA)

RPPA assays were carried out as described previously with minor modifications (Creighton & Huang, 2015). Protein lysates were prepared with modified TPER buffer supplemented with a cocktail of protease and phosphatase inhibitors. Protein lysates at 0.5 mg/ml of total protein were denatured in SDS sample buffer containing 2.5% β-mercaptoethanol at 100 °C for 8 min. The Aushon 2470 Arrayer (Aushon BioSystems) with a 40 pin (185 μm) configuration was used to spot samples and control lysates onto nitrocellulose-coated slides (Grace Bio-labs) using an array format of 960 lysates/slide (2880 spots/slide). The slides were probed with a set of 216 antibodies against total and phospho-proteins using an automated slide stainer Autolink 48 (Dako). Each slide was incubated with one specific primary antibody, and a negative control slide was incubated with antibody diluent instead of primary antibody. Primary antibody binding was detected using a biotinylated secondary antibody followed by streptavidin-conjugated IRDye680 fluorophore (LI- COR Biosciences). Total protein content of each spotted lysate was assessed by fluorescent staining with Sypro Ruby Protein Blot Stain according to the manufacturer’s instructions (Molecular Probes). Fluorescence-labeled slides were scanned on a GenePix 4400 AL scanner, along with accompanying negative control slides, at an appropriate PMT to obtain optimal signal for this specific set of samples. The images were analyzed with GenePix Pro 7.0 (Molecular Devices). Total fluorescence signal intensities of each spot were obtained after subtracting the local background signal for each slide and were then normalized for variation in total protein, background and non-specific labeling using a group-based normalization method. For each spot on the array, the background-subtracted foreground signal intensity was subtracted by the corresponding signal intensity of the negative control slide (omission of primary antibody) and then normalized to the corresponding signal intensity of total protein for that spot. Each image and its normalized data were evaluated for quality through manual inspection and control samples. Antibody slides that failed the quality inspection were either repeated at the end of the staining runs or removed before data reporting. Samples (in four biological replicates) were extracted from the normalized data and then log2 transformed. The median value of the three technical replicates was used for statistical analysis. Welch’s t-tests were used for group comparisons. FDR-adjusted p-value < 0.05 was considered statistically significant.

Subcutaneous xenograft tumor model and IVIS imaging

5 X 10^5 luciferase-labeled HT29 cells were suspended in 0.1 ml serum-free DMEM and injected into the right back flank of 21 to 24-week-old NSG mice. Bioluminescence was measured once a week, by injection of 100 µl 15 mg/ml D-luciferin (LUCNA-1G, Goldbio) via the intra-orbital sinus. Mice were imaged using IVIS Lumina II (Advanced Molecular Vision). The acquired bioluminescence signals were normalized to the day 0 bioluminescence intensity.

Statistics

All statistical analyses in our study were conducted using GraphPad Prism or R. Two-tailed t-test without equal variation assumption was used for data comparison between two experimental groups. For data comparison with at least 3 groups, one-way ANOVA test was performed first to assess the overall difference among groups. If differences existed, uncorrected Fisher’s LSD test were performed to assess the significance of differences between two groups. Two-way repeated measures ANOVA and uncorrected Fisher’s LSD test were used to assess the difference between data sets with time series measurements, including growth curves of cell proliferation and in vivo BLI signals. p-values less than 0.05 were considered significant.

Author contributions

F.Y., X.L., and Y.L. developed the concepts, designed the experiments, interpreted data, and wrote the manuscript. F.Y., A.T.K., P.D.S., M.N.M, W.J., J.T., Z.S., and X.L. performed the experiments.

Y.W. performed mass spectrometry analysis. S.H. helped design and performed proteomics experiments. F.Y. and Y. D. performed statistical and bioinformatic analyses. G.H. evaluated slides and provided pathological reports. Q.L., B.O.W., B.Z., N.F.S., X.W, and X-H. F. provided advice, technical expertise, and critical reagents. All authors read and approved the final manuscript.

Competing interests

Authors declare that they have no competing interests.

Data and materials availability

All data are available in the main text or the supplementary materials.

Acknowledgements

We thank Dr. Feng Cong for the generous gifts of reagents. We also thank Dr. Xuan Wang from the BCM Antibody-based Proteomics Core/Shared Resource for her technical assistance in performing reverse phase protein array; Ms. Xi-Lei Zeng and Ms. Xiaomin Yu from the Texas Medical Digestive Diseases Center (DDC) GEMS core for assistance in organoid culture; Ms. Cassandra Diegel from the Van Andel Institute for assistance. This work was financially supported by NIH/NCI (R01 CA204926 and CA271498 to Y.L.), DOD-CDMRP (BC191649 and BC191646 to Y.L.), and NIH-T32 grant (#1T32CA203690-01A1 to A.K. under Dr. Suzanne Fuqua). This project was also supported by the Breast Center Pathology Core as part of the SPORE program (P50 CA186784), Cytometry and Cell Sorting Core with funding from CPRIT (RP180672) and NIH (S10 RR024574), Antibody-based Proteomics Core with funding from CPRIT (RP210227) and NIH (S10 OD028648 to S.H.), DDC GEMS core with funding from NIH/NIDDK (DK56338), and resources from the Dan L. Duncan Cancer Center with funding from NIH/NCI (P30 CA125123). The Van Andel Institute (VAI) provided additional support, including contributions from the VAI Vivarium (RRID:SCR_023211) and VAI Pathology and Biorepository Core (RRID:SCR_022912). P.D.S was supported by a post-doctoral fellowship from the American Cancer Society (PF-20-109-01) and M.N.M is supported by the NIH/NIDCR (K08DE3109).