The Transcriptomic Landscape of Mismatch Repair-Deficient Intestinal Stem Cells

Abstract

Lynch Syndrome (LS) is the most common cause of hereditary colorectal cancer (CRC) and is secondary to germline alterations in one of four DNA mismatch repair (MMR) genes. Here we aimed to provide novel insights into the initiation of MMR deficient (MMRd) colorectal carcinogenesis by characterizing the expression profile of MMRd intestinal stem cells (ISC). A tissue-specific MMRd mouse model (Villin-Cre;Msh2LoxP/LoxP) was crossed with a reporter mouse (Lgr5-EGFP-IRES-creERT2) to trace and isolate ISCs (Lgr5+) using flow cytometry. Three different ISC genotypes (Msh2-KO, Msh2-HET, and Msh2-WT) were isolated and processed for mRNAseq and mass spectrometry followed by bioinformatic analyses to identify expression signatures of complete MMRd and haplo-insufficiency. These findings were validated using qRT-PCR, immunohistochemistry, and whole transcriptomic sequencing in mouse tissues, organoids, and a cohort of human samples, including normal colorectal mucosa, pre-malignant lesions, and early-stage CRC from LS patients and Familial Adenomatous Polyposis (FAP) patients as controls. Msh2-KO ISCs clustered together with differentiated intestinal epithelial cells from all genotypes. Gene set enrichment analysis indicated inhibition of replication, cell cycle progression, and the Wnt pathway and activation of epithelial signaling and immune reaction. An expression signature derived from MMRd ISCs successfully distinguished MMRd neoplastic lesions of LS patients from FAP controls. SPP1 was specifically upregulated in MMRd ISCs and colocalized with LGR5 in LS colorectal pre-malignant lesions and tumors. These results show that expression signatures of MMRd ISC recapitulate the initial steps of LS carcinogenesis and have the potential to unveil novel biomarkers of early cancer initiation.

Introduction

Lynch Syndrome (OMIM# 120435, LS) is a hereditary cancer syndrome predisposing patients to develop colorectal cancers (CRC) as well as tumors of the endometrium, ovary, stomach, and small intestine¹. LS has an estimated prevalence of 1 in 279, thus affecting over 1 million individuals in the US². LS is secondary to germline mutations in one of the DNA mismatch repair (MMR) genes (MLH1, MSH2, MSH6, and PMS2) that control post-replicative DNA proofreading, thus ensuring genomic integrity³. MMR deficiency (MMRd) accelerates the acquisition of secondary somatic mutations in oncogenes and tumor suppressor genes that regulate different pathways, including cell fate, transcription, growth factors, and other DNA repair mechanisms, thus promoting carcinogenesis⁴. LS has an autosomal dominant inheritance causing an estimated lifetime risk of CRC development of 20–50% depending on the germline MMR gene that carries the mutation, and also a young age of onset, typically in the fourth decade of life^5–7. Despite the recommendation of annual or biannual endoscopic surveillance starting at age 20–25, LS patients continue developing interval cancers and are counseled to consider risk-reducing surgeries^{8, 9}.

The epithelium of the small and large intestine contains niches of stem cells located at the bottom of specialized finger-like invaginations that are arranged in functional units called crypts, which are surrounded by connective tissue, and the underlying lamina propria. These fast-cycling stem cells, refered to as crypt base columnar (CBC) cells, generate daughter cells that exit the stem cell niche to integrate into the transit-amplifying compartment by migrating upwards to the lumen of the gut. This process takes 4–5 days and gives rise to several differentiated and specialized cell subtypes, including enterocytes (nutrient uptake), goblet (mucus production), enteroendocrine (hormone production), and paneth cells (growth factor production exclusively in the small intestine)^10–12.

Studies in animal models have demonstrated that pluripotent stem cells acquiring an initiating mutational event become the ‘cell of origin’ in several malignancies, including intestinal cancers^{13, 14}. In fact, the transcriptomic and proteomic profiles of intestinal stem cells (ISC) under physiologic and APC-inactivating conditions have been successfully characterized in genetically engineered animal models that allow for lineage tracing of cells expressing Lgr5, a Wnt target gene and stem cell marker^{15, 16}. Furthermore, inactivation of MMR function via Msh2 deletion in mouse embryonic stem cells generates a mutator phenotype causing genomic instability and accumulation of subsequent somatic mutations leading to cancer development¹⁷, thus demonstrating characteristics of a cancer stem cell¹⁸. Therefore, we hypothesized that MMRd ISCs display a unique transcriptomic and proteomic profile that is different from their daughter cells. It is essential to understand the molecular and cellular landscape of MMRd tissue-specific stem cells in order to unravel the mechanisms behind the earliest stages of cancer initiation before macroscopic lesions are detectable, thus identifying specific targets for the development of novel cancer interception strategies and biomarkers for early detection of cancer in LS¹⁹.

Here, we present, for the first time, the whole transcriptomic and proteomic landscape of the intestinal epithelium with haploinsufficiency and complete deficiency of MMR functioning, which mimics the biology of normal colorectal and neoplastic epithelium in LS patients, respectively. This gene expression signature of MMRd ISCs derived from a genetically-engineered mouse model uncovers activated molecular pathways involved in the initiation and early steps of MMRd colorectal carcinogenesis.

Materials and Methods

Mice.

C57BL/6J strain of conditional knockout mice for MMR gene Msh2 (Msh2^LoxP/LoxP) were crossed with Villin-Cre (VC) transgene expressing mice to obtain VC-Msh2^LoxP/LoxP mice²⁰. We further crossed VC-Msh2^LoxP/LoxP mice with a reporter Lgr5^{EGFP-IRES-creERT2} mice to track and isolate Lgr5^EGFP+ stem cells by flow cytometry¹⁵. We generated Lgr5^{EGFP-IRES-creERT2};VC-Msh2^LoxP/LoxP as Msh2 null mice in the intestine (herein referred as Msh2-KO), Lgr5^{EGFP-IRES-creERT2};VC-Msh2^LoxP/+ as Msh2 haplo-insufficient (Msh2-HET), and Lgr5^{EGFP-IRES-creERT2};VC-Msh2^+/+ as mice with wild-type Msh2 function (Msh2-WT). Also, we generated intestinal organoids from Msh2-KO and Msh2-WT for validation of the expression of key genes in ISC. All animal experiments were approved by the institutional animal care and use committee (IACUC) of The University of Texas MD Anderson Cancer Center and the care of the animals was in accordance with institutional guidelines (IACUC protocol # 00000469-RN02).

Crypt isolation from small intestine and Fluorescence-activated cell sorting (FACS).

Crypts from the small intestine of 8-week-old Msh2-WT, Msh2-HET, and Msh2-KO mice were harvested, and single cells from each genotype were subjected to FACS to isolate Lgr5^EGFP+ stem cells and Lgr5^EGFP- daughter cells.

Transcriptome and mass spectrometry analysis of mouse specimens.

RNA and total cellular extracts (TCE) were isolated from ISCs (Lgr5^EGFP+) and daughter cells (Lgr5^EGFP-) from all three mouse genotypes. RNA was extracted using TRIzol (Invitrogen) and RNA isolation kit (Ambion) and then subjected to library preparation and sequenced with an Illumina HiSeq4000 instrument. TCE were analyzed using tandem mass spectrometry. Detailed steps in the analysis of RNA-seq and proteomics along with in-depth bioinformatics analysis for differential gene expression and gene set enrichment analysis (GSEA) followed standard protocols and pipelines previously described²¹.

Transcriptomic analysis of human samples and validation of MMR-deficient stem cell expression signatures.

Tissue samples were acquired through endoscopic biopsies during routine screening colonoscopies from a total of 17 Familial Adenomatous Polyposis (FAP) patients (17 paired adenoma and normal mucosa) and 27 LS patients (11 matched tumor/adenomas and normal mucosa, 3 unmatched tumor/adenomas, and 18 with unmatched normal mucosa, Table S1). All patients were followed at The University of Texas MD Anderson Cancer Center (UTMDACC) for routine surveillance care. Written informed consent was obtained from all study participants, and the UTMDACC Institutional Review Board (IRB) approved this study (IRB #PA12–0327). Total RNA was isolated from tissues that have been flash-frozen or preserved in RNALater and subjected to mRNA sequencing (mRNAseq). This human mRNAseq data set from colorectal normal mucosa, polyp, and tumor samples were used for validation of the gene signatures obtained in mice. FAP normal mucosa and polyps were selected as the human counterparts of Msh2-WT mice (MMR-proficient), LS normal mucosa as Msh2-HET (MMR-haploinsufficient), while LS polyp and tumor samples that were hypermutant (mutation rate≥10/Mb estimated by whole-exome sequencing) or displayed MMRd by microsatellite instability (MSI) via PCR or immunohistochemistry of MMR protein as ‘Msh2-KO’ (MMRd). In addition, transcriptomic data from organoids of normal mucosa from LS²² and sporadic CRC patients²³ were used to further validate the MMR-haploinsufficient expression signature.

Gene expression analysis and Chromatin Immunoprecipitation (ChIP) assays.

Total RNA from ISCs and organoids isolated from the different genotypes as well as both Lgr5^EGFP+ and Lgr5^EGFP- fractions were analyzed by qRT-PCR for validation of the expression of critical genes following previously described methods²². To assess if SPP1 gene expression was epigenetically regulated by histone H3 lysine 27 methylation (H3K27me3), ChIP assays were performed in chromatin extracts of the mismatch repair proficient (MMRp) CRC cell line SW620 and the MMRd endometrial cancer cell line HEC59, which harbors bi-allelic inactivating MSH2 mutations with details provided in Supplementary Material and Methods. Primer sequences are included in Table S2.

Immunofluorescence and imaging of mouse tissues.

Freshly extracted small intestine from 8-week-old mice from Msh2-WT, Msh2-HET, or Msh2-KO were used to stain Lgr5 (GFP⁺) cells as a marker to visualize ISC and the expression of SPP1 (Table S3).

Fluorescent multiplex immunohistochemistry (IHC) staining of human specimens and automated quantitative imaging.

Formalin-Fixed Paraffin-Embedded (FFPE) tissue specimens of uninvolved normal colorectal mucosa (N=6), tubular/tubulovillius adenoma (N=6), and invasive adenocarcinoma (N=3) from a total of 8 LS patients were used (Table S4). Unstained slides were processed as described above and stained with antibodies against human LGR5 and SPP1. Manual fluorescent multiplex IHC staining was performed following a validated protocol using antibodies and reagents listed in Table S3.

Statistical analysis.

Comparisons between two experimental groups were performed with GraphPad Prism using Student’s unpaired t-test, and among more than two experimental groups were analyzed using one-way ANOVA with Tukey’s post-hoc analysis for multiple comparisons. The data are expressed as means ±SD from three technical replicates and three independent experiments.

Experimental details and computational methods can be found in Supplementary Material and Methods.

Results

Isolation of MMRd mouse ISC.

To understand the biology of MMRd ISCs at the earliest stage of carcinogenesis, we crossed a mouse model of intestinal tissue-specific (Villin-Cre, VC) inactivation of the MMR function via deletion of the essential ATPase domain of Msh2 in exon 12 (Msh2^LoxP/LoxP, thus resulting in VC-Msh2^LoxP/LoxP) with another mouse line (Lgr5^{EGFP-IRES-CreERT2} ) that allowed the isolation and tracing of ISCs expressing the validated stem cell marker Lgr5. Then, we isolated ISCs from the entire small intestine of the following mouse genotypes to model different clinical scenarios: Lgr5^{EGFP-IRES-creERT2+};VC-Msh2^+/+ (Msh2-WT) as a counterpart of ISC from the sporadic normal colorectal mucosa; Lgr5^{EGFP-IRES-creERT2+};VC-Msh2^LoxP/+ (Msh2-HET) from normal colorectal mucosa of LS patients; and Lgr5^{EGFP-IRES-creERT2+};VC-Msh2^LoxP/LoxP (Msh2-KO) from early premalignant lesions of LS patients, as well as their corresponding daughter cells that were Lgr5^{EGFP-IRES-creERT2-} (Figure 1). FACS was optimized using crypt preparations from GFP negative mice to specifically isolate epithelial cells expressing high levels of GFP (GFP^hi) that were considered as Lgr5^EGFP+ ISCs and GFP^-/EpCAM⁺ cells (Lgr5^EGFP-) as daughter (non-stem) cells. Cell fractions excluded lymphocytes by labeling total crypt cells with CD45 antibody (Figure S1A). A cohort of 119 mice was used to isolate a sufficient number of stem cells that afforded extraction of RNA and protein for transcriptomics and proteomics analyses, respectively. From a total of 27 Msh2-WT mice, we obtained a mean of 224,388 Lgr5^EGFP+ cells per mice; 35 Msh2-HET rendered a mean of 47,873 Lgr5^EGFP+ cells; and 57 Msh2-KO a mean of 10,392 Lgr5^EGFP+ cells. We observed that the number of stem cells recovered for each genotype decreased exponentially with the deletion of each Msh2 allele (Figure S1B-D and Figure S2), which could possibly be due to premature differentiation of ISC upon deletion of Msh2 allele.

Figure 1.

Schematic outline of the experimental design. Lgr5^{EGFP-IRES-creERT2} mice were crossed with Villin-Cre;Msh2^LoxP/LoxP mice (VC-Msh2^LoxP/LoxP). After crypt isolation from Msh2-WT, Msh2-HET, or Msh2-KO mice, FACS was performed to isolate GFP labeled Lgr5^EGFP+ stem cells or Lgr5^EGFP- daughter cells. Sorted cell populations were used to extract RNA and protein for transcriptomics and proteomics profiling by mRNAseq and tandem Mass Spectrometry, respectively. Bioinformatic analyses were used to identify differentially expressed genes and proteins in MMRd and haploinsufficient ISC. Validation of gene expression signatures was performed using both mouse tissue specimens and organoids as well as human cell lines and tissues from LS patients.

Transcriptomic profile of MMR haploinsufficient and MMRd stem and non-stem cells.

We identified the transcriptomes of Lgr5⁺ ISCs (Lgr5^EGFP+) and their daughter cells (Lgr5^EGFP-) in all three mouse genotypes using next-generation whole transcriptomics followed by principal component analysis (PCA). The transcriptome of Lgr5^EGFP+ cells of Msh2-KO clustered with daughter cells (Lgr5^EGFP- fractions) of all three genotypes (Figure 2A). However, each genotype showed a profile that was distinct from the others when samples were separately analyzed based on stem and daughter cell phenotypes (Figure S3A). Then, we examined the transcriptional differences of MMR haploinsufficient and MMRd ISCs (Msh2-HET and Msh2-KO mice, respectively) to MMRp ISCs (which were represented by Msh2-WT mice and were the comparator for these analyses), and their corresponding daughter cells (differentiated non-stem cells). The comparison of the transcriptomes of Lgr5^EGFP+ Msh2-KO and Msh2-WT stem cells identified a total of 340 significantly dysregulated genes (284 upregulated and 56 downregulated, Figure 2B), thus defining an expression profile of MMRd ISC. In contrast, we observed only 39 genes significantly dysregulated in Lgr5^EGFP- non-stem cells (Figure S3B). Then, a comparison of Lgr5^EGFP+ Msh2-HET and Msh2-WT stem cells rendered a gene profile for MMR haploinsufficiency with a total of 60 genes differentially dysregulated (50 upregulated and 10 downregulated, Figure 2B) and the same comparison of Lgr5^EGFP- daughter cells observed 13 genes (Figure S3B). When we changed the reference and compared the transcriptomes of Msh2-KO and Msh2-HET, we observed a total of 182 genes differentially expressed (131 upregulated and 51 downregulated) in the stem cells (Figure 2B) and 13 genes in Lgr5^EGFP- non-stem cell fractions (Figure S3C). We observed that 20 genes were dysregulated in common between Lgr5^EGFP+ stem cells of Msh2-HET and Msh2-KO (Figure 2C and Table 1), whereas only one gene was commonly dysregulated within Lgr5^EGFP- daughter cells of Msh2-HET and Msh2-KO when compared to Msh2-WT. In order to summarize the number of genes expressed in stem and daughter cells as well as the intersection among the different genotypes (Msh2 status) and cell types (Lgr5^EGFP+ or Lgr5^EGFP-), we generated an UpSet plot matrix that provides a visualization at a glance of these numbers (Supplementary Figure S3C). Overall, the gene profile showing the largest differences was between Msh2-KO and -WT stem cells, and the one with the closest expression was between stem and differentiated cells of the Msh2-KO genotype (Figure S3C). These results indicate that the loss of one or both alleles of Msh2 induces a unique transcriptional profile that perturbs the biology of MMRd ISC. In addition, the transcriptome of Lgr5^EGFP+ stem cells of Msh2-KO clustered with daughter cells (Lgr5^EGFP- fractions) of all three genotypes in the PCA plot and also displayed the lowest amount of differentially expressed genes compared to their daughter cells, thus indicating that a complete loss of Msh2 may lead to premature differentiation of ISC.

Figure 2. Bioinformatics analysis of gene expression from Lgr5^EGFP+ stem and Lgr5^EGFP- daughter cells.

(A) PCA plot of expression profiles from Lgr5^EGFP+ intestinal stem cells (solid colors) isolated from Msh2-WT (dark yellow), Msh2-HET (dark blue) and Msh2-KO (dark red) and from daughter cells (Lgr5^EGFP- and EpCAM+, light shades of respective colors). The first and second principal components are plotted in the X and Y-axis, respectively. Individual samples within each group are connected by a centroid. A total of 21 mice for Msh2-WT, 25 for Msh2-HET and 40 for Msh2-KO were equally distributed among three biological replicates to obtain ~10,000 ISC for each replicate per genotype; (B) Volcano plots illustrate genes expressed in ISC of Msh2-KO and Msh2-HET compared to Msh2-WT and in ISC of Msh2-KO as compared to Msh2-HET. X-axis presents Log2Fold change and Y-axis presents log10 of adjusted P-value for multiple comparisons from DESeq2 differential analysis. The horizontal dashed line represents FDR=0.05, while left and right vertical dashed lines represent Log2FC of ±1, respectively. Significantly down-regulated genes are displayed in green, upregulated in red, and non-significant in black; (C) Venn diagrams showing numbers of significantly expressed genes in Lgr5^EGFP+ stem cells and Lgr5^EGFP- non-stem cells for each genotype compared with Msh2-WT; (D) Validation of the expression of key signature genes from the MMRd and MMR-haploinsufficient signatures as well as stem and differentiation markers analyzed using qRT-PCR in FACS sorted Lgr5^EGFP+ (stem, left panels) and Lgr5^EGFP- (non-stem, right panels) cells from Msh2-WT, Msh2-HET and Msh2-KO. Data is presented as fold changes and depicted as relative gene expression levels compared to expression levels in Lgr5^EGFP+ cells of Msh2-WT as reference. Expression levels of Gapdh were used as an internal housekeeping gene for normalization. Error bars display ±SD. One-way ANOVA with Tukey’s multiple comparison post-hoc test, *P -value< 0.05, **P-value < 0.01, ***P-value< 0.001, ****P-value<0.0001.

Table 1.

Common significant genes in stem (upper section) and daughter cells (lower section) isolated from Msh2-HET and Msh2-KO mice. Genes selected had FDR≤0.05 and |Log2FC|≥1.

Gene Symbol	Fold Change Lgr5EGFP+ Het vs WT	Fold Change Lgr5EGFP+ KO vs WT
Spp1	6.019	19.335
Nr1h5	4.862	8.549
Rian	3.727	4.717
Jchain	3.682	3.763
Meg3	3.520	4.305
Gm28230	3.515	5.396
Dgkh	3.045	3.907
Arhgap45	2.943	3.587
Trpv3	2.938	2.771
P2ry4	2.612	3.041
Ccn3	2.564	2.613
B230206H07Rik	2.518	4.196
Tchh	2.437	3.435
Lifr	2.350	2.069
Ahnak	2.290	3.747
Sardh	2.236	3.526
Cacna1h	2.204	2.373
Sned1	2.195	2.307
Nlrp9b	2.040	4.672
Gm23547	0.361	0.361
	Fold Change Lgr5EGFP- Het vs WT	Fold Change Lgr5EGFP- KO vs WT
Muc5ac	6.638	3.905

Validation of stem and non-stem cell specific genes in MMRd ISCs and daughter cells.

We assessed dysregulated genes that overlapped between Msh2-HET and Msh2-KO in Lgr5^EGFP+ stem cells and Lgr5^EGFP- non-stem cells using qRT-PCR (Table 1). These markers were selected based on their potential roles in tumorigenesis and the correlation between their level of dysregulation and Msh2 allele dosage. We evaluated a total of four markers of stem cells (Spp1, Nr1h5, Ahnak, and Nlrp9b) and one of daughter cells (Muc5ac). Overall, all of them were confirmed to be expressed in both Lgr5^EGFP+ and Lgr5^EGFP- cells of both Msh2-HET or Msh2-KO mice. Moreover, they were significantly upregulated in stem cells and their expression correlated with the Msh2 allele dosage (Figure 2D, upper left panel). Of note, we observed a high level of upregulation of the marker Spp1 in both Msh2-HET (10-fold) and Msh2-KO (15-fold) stem cells when compared to Msh2-WT. In addition, the relative expression of Spp1 was significantly lower in Lgr5^EGFP- daughter cells than in Lgr5^EGFP+ stem cells within each respective genotype. In contrast, the expression of Nr1h5 showed significant differences in daughter cells across genotypes, while Ahnak and Nlrp9b did not show differences across genotypes (Figure 2D, upper right panel). In regards to Muc5ac, there were significant differences among daughter cells across different genotypes that were also seen among stem cells, thus making this marker relatively non-specific of cell-of-origin. We attempted to validate the gene expression of the novel stem cell markers observed in the Lgr5^EGFP+ stem cells using Msh2-KO and Msh2-WT mouse organoids as an ex vivo model of the stem cell compartment of the intestinal crypt. Initially, we did not confirm the same trends observed in sorted cells (Figure S3D, left panel). Therefore, based on the function of these genes, we reasoned that their expression could be influenced by the immune environment and surrounding stem cell niche. We repeated the expression assessment after stimulating the organoids for 24 hours with colony-stimulating factor 1 (M-CSF1) to recreate cues received by stem cells. Under these conditions, we observed a significant upregulation of Spp1 and Nlrp9b expression in Msh2-KO organoids and no significant changes in the Msh2-WT counterparts (Figure S3D, center panel). These results are consistent with the expression data acquired in vivo and strongly suggest that Spp1 and Nlrp9b expression levels were significantly and specifically enhanced in MMRd ISCs upon interaction with the stem cell niche, including immune cells, thus highlighting its role as a potential biomarker in MMRd carcinogenesis.

Since the transcriptomes derived from Msh2-KO Lgr5^EGFP+ ISCs clustered together with daughter cells of all genotypes, we hypothesized that Msh2-KO ISCs lost their stem-ness and underwent premature differentiation. Based on this observation, we performed qRT-PCR analysis of differentiation-specific genes including Krt20 and Alpi (markers for enterocytes), and Muc2 (marker for Goblet cells) and stem cell markers, Ascl2 and Olfm4 in intestinal samples from mice of the three genotypes. We confirmed a significant decrease in expression of stem cell markers, Lgr5, Ascl2, and Olfm4 in stem cells from Msh2-KO mice compared to those in Msh2-WT mice. Subsequently, we observed a strong stimulation in the expression levels of enterocyte markers, Krt20 and Alpi, in the stem cells of Msh2-KO mice (Figure 2D lower left panel). These results were confirmed ex vivo in mouse organoids that showed downregulation of surface stem cell markers and upregulation of differentiation signals (Figure S3D, right panel). In line with these results, no significant differences were observed in the expression of these markers in Lgr5^EGFP- cells across all genotypes (Figure 2D lower right panel). Therefore, these results indicate that loss of MMR function influences ISC homeostasis and promotes premature differentiation of ISCs.

Proteomic profile of MMR haploinsufficient and MMR deficient stem and non-stem cells.

To assess the proteomic profile, we isolated total cellular proteins from stem and non-stem fractions of Msh2-WT, Msh2-HET, and Msh2-KO mice and performed tandem mass spectrometry (MS/MS). We identified an average of 1238 gene counts (proteins) from total cell extracts of stem cells from Msh2-WT (Table S5). A total of 797 and 830 proteins were observed from equal amounts of total protein of stem cells from Msh2-HET and Msh2-KO mice, respectively. The number of proteins identified from non-stem cells was higher than stem cell fractions observed in each genotype. Individual analysis of fold change expression obtained directly from mean spectral counts of Lgr5^EGFP+ ISCs revealed high levels of differentiation markers such as Krt20 (5.5-fold) and Fabp1 (5.8-fold) in Msh2-HET Lgr5^EGFP+ cells (Table S6) and even higher in Msh2-KO Lgr5^EGFP+ (Krt20, 6.5-fold; Fabp1, 7.3-fold; Fabp2, 2.0-fold, Table S7) compared to Lgr5^EGFP+ Msh2-WT cells. We also observed enrichment for proteins that are involved in cancer progression and migration, such as carbonic anhydrase 1 (Car1, 10.89-fold in Msh2-HET, and 21-fold in Msh2-KO) and actin-binding protein Gelsolin (Gsn, 2.5-fold in Msh2-HET and 9.8-fold in Msh2-KO)^{24, 25}. Thus, the proteomic expression of MMRd stem cells showed enrichment for cancer-associated markers that are involved in their malignant transformation.

Generation of a molecular profile to define a gene signature of MMRd.

Using the transcriptomic and proteomic profiles of Lgr5^EGFP+ and Lgr5^EGFP- cells for each genotype, we generated a molecular profile that defines a gene signature for MMR haploinsufficiency and MMRd. First, we compared the expression data of Lgr5^EGFP+ ISCs obtained from Msh2-KO to Msh2-WT mice and excluded those genes expressed commonly in Lgr5^EGFP- fractions in order to generate a list of unique genes that specifically represent LS colorectal neoplasia (pre-cancers and tumors), which is characterized by a complete MMRd. We observed a total of 48 differentially expressed genes (Table 2). The same approach was applied to the analysis of the comparison of the transcriptome of Msh2-HET to Msh2-WT to generate a list of genes exclusively and differentially expressed in MMR haploinsufficient ISCs. This list reflected the expression patterns of LS normal colorectal mucosa and a total of 5 differentially expressed genes were detected (Table 2). GSEA highlighted several relevant pathways in MMRd ISC biology. Among Lgr5^EGFP+ cells from Msh2-KO, the top observed pathways were related to Integrin Signaling, Focal Adhesion, and Inflammatory Response. Interestingly, we also observed downregulation of the WNT pathway (Figure 3A). The top enriched pathways in the Msh2-HET ISCs were TGFβ signaling, mitogen-activated protein kinase (MAPK) signaling, and Endochondral Ossification (Figure 3B). We found that the pro-inflammatory gene sets, Prostaglandin-Leukotriene Metabolism and Eicosanoids Synthesis, were depleted in Msh2-HET. In addition, we observed that gene sets for Cytoplasmic Ribosomal Proteins were downregulated in both Msh2-HET and Msh2-KO ISCs (Figure 3C). Finally, we compared our gene signatures reflecting different stages of MMR carcinogenesis with the previously published profile of mouse APC-driven stem cells¹⁶. We observed a minimal degree of overlap with the Msh2-KO signature (only 6 genes) and increased numbers of genes shared with Msh2-HET and Msh2-WT, thus consistent with the expectation that APC-driven carcinogenesis overlaps minimally with MMRd during the earliest stages (Figure S3E).

Table 2.

List of genes defining a signature of MMR haploinsufficiency (Msh2-HET, upper section) and MMRd (Msh2-KO, lower section). Genes selected had FDR≤0.05 and |Log2FC|≥1.

Gene Symbol	Fold Change Lgr5EGFP+ Msh2-Het vs Msh2-WT	Fold Change Msh2-Het Lgr5EGFP+ vs Lgr5EGFP-
Spp1	6.019	8.984
Meg3	3.520	0.413
P2ry4	2.612	0.442
Defa5	0.486	0.318
Gm49320	0.297	0.410
	Fold Change Lgr5EGFP+ Msh2-KO vs Msh2-WT	FoldChange Msh2-KO Lgr5EGFP+ vs Lgr5EGFP-
Mup22	27.554	12.855
Spp1	19.335	19.562
Slc26a9	11.831	6.639
Muc6	10.515	11.907
Ugt8a	10.462	2.432
Cubn	8.143	2.473
Pgc	7.723	32.672
Aqp5	7.414	14.058
Cyp2c55	5.917	3.439
Abca12	5.825	3.805
Slc5a4a	5.580	3.023
Gm28230	5.396	2.352
Mal	5.256	3.366
Car1	4.989	2.706
G6pc	4.820	2.433
Sprr1a	4.709	2.328
Slc5a12	4.637	2.482
Scara3	4.513	4.933
Gif	4.484	13.338
Slc30a10	4.259	2.472
Lct	4.215	2.923
Sptssb	4.087	2.970
Sgk2	4.058	2.795
Clca4a	4.052	2.490
Cyp3a25	3.804	2.669
Gdpd2	3.690	2.173
Gkn1	3.677	3.141
Tmigd1	3.602	2.732
Slc2a2	3.575	3.053
Aldh1a1	3.476	2.309
Fa2h	3.466	2.378
Bst1	3.453	2.303
1810065E05Rik	3.436	2.567
Anxa10	3.435	2.881
Slc10a2	3.298	2.114
Clu	3.193	4.456
2010109I03Rik	3.181	2.060
Pdzk1	3.000	3.215
Cyp2b10	2.804	3.551
Mafb	2.766	2.497
Slc5a4b	2.594	2.618
Slc16a9	2.581	2.080
Ifit1	2.473	2.235
Oas3	2.346	2.001
Ak4	2.264	2.020
Sema6a	2.225	2.220
Slc28a1	2.168	2.053
Itga7	2.060	0.000

Figure 3. Pathways modulated in MMRd and MMR haploinsufficient ISCs.

Bubble chart plots display statistically significant pathways enriched in Msh2-KO ISCs (A), Msh2-HET ISCs (B), and both Msh2-HET and Msh2-KO ISCs (C), using BH-adjusted P-value=0.05 as a cutoff. Pathways bolded were relevant in terms of function to the molecular biology of MMRd ISC. The size of circles represents adjusted P-value (larger circles represent smaller P-value). The colors of bubbles were determined by the sign and amplitude of normalized enrichment score (NES) with positively enriched pathways in red and negatively enriched pathways in green.

Then, we generated a unique list of 78 signature proteins for Msh2-KO (MMRd) as ratio of protein expression from 130 proteins found between Lgr5^EGFP+ and Lgr5^EGFP- cells of Msh2-KO mice and also from the ratio of protein expression from 117 proteins found between Lgr5^EGFP+ cells of Msh2-KO and Msh2-WT (Table S8). From the proteomic analysis, 27 proteins were found to overlap with 197 signature genes from mRNAseq analysis for Msh2-KO (FDR≤0.05, Figure S4A). Of note, our mass spectrometry results indicated that Spp1 protein was found to be expressed only in Lgr5^EGFP+ cell fractions of Msh2-KO animals. In addition, we analyzed pathway enrichment using Ingenuity Pathway Analysis (IPA) observing a representative network of proteins related to cellular development, cellular growth, and proliferation as top affected pathways with expression of KRAS as a pivotal network (Figure S4B). Similarly, a unique list of 52 signature proteins for Msh2-HET was derived from comparing the ratio of protein expression from 235 proteins found between Lgr5^EGFP+ and Lgr5^EGFP- cells of Msh2-HET mice and also from the ratio of protein expression from 187 proteins found between only Lgr5^EGFP+ cells of Msh2-HET and Msh2-WT (Table S9). DNA replication, recombination and repair, as well as RNA post-transcriptional modification pathways stood out from the Msh2-HET protein profile, which was PARP1 centered (Figure S4C). Overall, the Msh2-HET and Msh2-KO protein signatures revealed potentially interesting candidates that merits consideration for its influence on carcinogenesis in MMR deficiency.

Validation of the MMRd stem cell signature in LS human specimens.

To assess the biological significance of the MMRd and MMR-haploinsufficient gene signatures derived from ISC in mice, we applied both signatures to whole transcriptomics of normal colorectal mucosa and neoplastic lesions (both adenomas and tumors) from a cohort of LS and FAP patients (as MMRp controls; Table S1). For the Msh2-KO signature, we observed that 41 genes out of 197 signature genes were still significantly dysregulated (FDR≤0.05) and had the same fold change direction in LS hyper-mutant adenomas and tumors. This confirmed that these genes represent and recapitulate the biology of cells derived from an MMRd progenitor. Together, these 41 dysregulated genes were able to clearly separate pre-cancers and early-stage tumors in LS and FAP patients into two distinct clusters, with only two LS samples misclassified (Figure 4A). For the MMR-haploinsufficient signature (Msh2-HET signature), we had to relax the criteria because only one gene out of 27 (AHNAK) was significantly dysregulated and had the same fold change direction. Therefore, we examined 14 genes that had the same fold change direction in mouse and human data sets. Unsupervised hierarchical clustering observed two groups of samples with one group integrated by mostly LS samples with the exception of 4 lesions (Figure S5A). Of note, SPP1 showed a trend towards being upregulated in LS normal mucosa. Overall, the MMRd gene signature (Msh2-KO) was able to correctly classify neoplasms that are MMRd and therefore contain biomarker information that recapitulated early stages of MMRd carcinogenesis in humans. Then, we investigated if the signature comparing Msh2-KO and Msh2-HET was able to distinguish LS pre-cancers and early-stage tumors from normal mucosa. The original signature from mouse ISC contained a total of 182 genes (Figure S3C) and 56 genes retained statistical significance in human and shared the same fold change direction. The resulting gene profile was able to separate the samples into two groups with only three neoplastic lesions misclassified, thus showing a strong performance (Figure 4B). Finally, we combined together with the MMRd and haploinsufficient signatures (Table 2) to validate its performance to differentiate the expression patterns of human organoids derived from the normal mucosa of both LS carriers²² and patients diagnosed with sporadic CRC (as normal mucosa controls, Supplementary Material and Methods)²³. We observed that an 11-gene set of significantly expressed genes with corresponding human orthologs were able to precisely differentiate and segregate colorectal normal mucosa organoids of LS from sporadic individuals (Figure S5B).

Figure 4. Expression of Msh2-HET and Msh2-KO signatures in FAP and LS patient samples.

(A) Unsupervised hierarchical clustering heatmaps showing the expression pattern of selected Msh2-KO versus Msh2-WT signature genes in FAP polyps and LS hypermutant/MSI pre-cancer/tumor samples (FDR<0.05 in human and same fold change direction in both mouse and human); (B) Expression patterns of selected Msh2-KO versus Msh2-HET signature genes in normal mucosa from LS patients using row-centered and batch corrected expression data (FDR<0.05 in human and same fold change direction in mouse and human). Dendrograms indicate sample-sample Pearson correlation distances. The significance of genes in human comparison are indicated in a row covariate bar. Log2FC in human and mice for each gene are shown as scatter plots. Sample type is color-coded as follows: blue represents normal tissue from LS patients; red represents hypermutant/MSI adenomas/tumor tissue from LS patients, respectively; FC, fold change; LS, Lynch Syndrome.

Expression of Spp1 in MMR-deficient mouse ISCs and LS patient specimens.

Our combined transcriptomics and proteomics analysis indicated that Spp1 expression is upregulated in ISCs of both MMRd mice and LS patients. To gain mechanistic insights in epigenetic regulation of SPP1, ChIP assays were performed in MMRp and MMRd cell line models. Our results indicated that levels of H3K27me3 epigenetically regulate the expression of SPP1 as a function of the MMR status (Figure S5C). Then, to further confirm the expression of SPP1 in stem cells (Lgr5^EGFP+ cells), we performed IHC in intestinal tissue sections of Msh2-WT, Msh2-HET, Msh2-KO mice using antibodies against Spp1 and GFP. Since GFP expression is under control of the Lgr5 promoter in mice, the level of GFP staining correlates to the level of Lgr5 cells. We observed enhanced staining of Spp1 in crypts of Msh2-HET and Msh2-KO mice that co-localized with Lgr5 cells (GFP⁺) located at the base of the crypts, thus confirming the upregulation of Spp1 in ISCs of MMR haplo-insufficient and MMRd tissues (Figure 5A). Finally, we examined the levels of SPP1 and LGR5 in a series of LS samples (Table S4) representing sequential steps of colorectal carcinogenesis: normal mucosa (n=6), pre-cancers (n=4, tubular adenomas), and tumors (n=3, adenocarcinomas). Quantitative imaging and analysis of single- and double-positive cells for SPP1 and LGR5 (Figure S6) showed the highest proportion of co-expressing stem cells (double-positive cells) in cancers compared to adjacent normal colorectal mucosa and pre-cancers (non-statistically significant trend, P-value=0.15 cancer vs. normal, Figure 5B and 5C). Despite the limited number of samples available for analysis, these results are consistent with the data from the LS mouse model, thus suggesting a potential role of SPP1 in the progression of colorectal MMRd carcinogenesis.

Figure 5. Expression and localization of Spp1 within crypts of MMR mouse models and LS patient specimens.

(A) Small intestine from 8 week-old Msh2-WT, Msh2-HET, and Msh2-KO mice was stained with antibodies against GFP to detect Lgr5⁺ cells (green) and Spp1 (red) by immunofluorescence. Panels show representative images for Lgr5 and Spp1 expression and location within crypts. Yellow arrows in merged images demarcate the co-localization of Lgr5 and Spp1 in crypts of tissues from Msh2-HET and Msh2-KO. Nuclei were counterstained with DAPI (blue). Scale bar is equivalent to 50 μm; (B) Representative images from FFPE tissue sections of H&E, nuclear counterstaining with DAPI (blue), immunostaining with anti-human LGR5 (TSA with Opal-520, green), and anti-SPP1 (TSA with Opal-570, red) antibodies, and composite images acquired using fluorescent multiplex immunohistochemistry. Regions of interest for digital image analysis include normal colon epithelium (top panel), adenomas (middle panel), and adenocarcinoma (lower panel) displaying single positive cells for LGR5 and SPP1, and double-positive cells (MERGE); Scale bar represents 100 μm and scale bar in insets are equivalent to 10 μm. (C) The number of positive cells for each marker and double positives reported was quantified as cell density and expressed by the number of cells per mm² using inForm advance image analysis software (considering that the total number of nucleated cells is 100%). The percentage of stem cells co-expressing SPP1 and LGR5 (double-positive cells) was higher in pre-cancers and cancers compared to normal adjacent tissue showing a non-statistically significant trend. Graph displays mean±SEM.

Discussion

In this study, we have used a tissue-specific mouse model of LS to identify, for the first time, the transcriptomic and proteomic profiles of ISCs displaying MMRd. We observed a significant loss of Lgr5⁺ stem cells upon deletion of each Msh2 allele, which posed additional technical challenges as we required a large number of animals to obtain sufficient numbers of cells to perform mRNAseq and mass spec analyses. We validated this observation in MMRd ISCs and intestinal organoids, which revealed a loss of stemness reflected by downregulation of Lgr5, Ascl2, and Olfm4, and upregulation of the differentiation-specific markers Krt20, Alpi, and Muc2. Therefore, our results suggest that stem cells from Msh2-KO prematurely exhibit a differentiated phenotype, as evident by the fact that MMRd stem cells clustered together with differentiated cells of all genotypes regardless of their MMR status. It is plausible that stem cells trigger a natural epigenetic response towards differentiation to avoid malignant transformation. In fact, this mechanism has been therapeutically exploited in the treatment of leukemia, where ‘differentiation therapies’ are used to treat Acute Promyelocytic Leukemia²⁶. More relevant for our disease context was the observation that inhibition of R-Spondin in CRC, a ligand for Lgr5 receptor, led to in-vivo differentiation and loss of stem-cell function^{27, 28}.

Enrichment analysis has pointed towards the dysregulation of other cellular pathways associated with the loss of stemness in MMRd ISC that could drive their transformation into a cancer stem cell phenotype¹⁸. We observed the downregulation of ribosomal proteins such as RPS7, which has been reported to act as a tumor suppressor gene inhibiting proliferation by decreasing hypoxia-inducible factor-mediated glycolysis in CRC²⁹, and RPS14, which has been reported to play a significant role in cell proliferation by negatively regulating the transcriptional activity of c-Myc, a key oncogene involved in colorectal carcinogenesis³⁰. Thus, we posit that the MMR system may have a direct role on stem cell maintenance and renewal. As MMRd prompts premature differentiation of stem cells in LS, a relatively small percentage of cells persist that sustain and acquire a cancer stem cell phenotype via dysregulation of key genes from cancer promoting pathways such as those that we have observed in our pathway enrichment analysis. The prematurely differentiated MMRd stem cells that have lost their stem-ness will become de-differentiated under specific conditions, which may yield pluripotency that subsequently drives the onset of carcinogenesis³¹. Therefore, induction of stem cell differentiation in LS could become a potential avenue for cancer interception that warrants further investigation. Another notable finding is the downregulation of Wnt signaling in MMRd ISC. This observation confirms that the key initiating step in LS carcinogenesis is inactivation of the MMR system within aberrant crypt foci, then leading to flat pre-malignant lesions upon acquisition of additional hits in other key oncogenic drivers, with activation of Wnt signaling at later stages and only in a fraction of the pre-malignant lesions. This agrees with previous models that were based on anecdotal observations and that now can be better substantiated in our results³².

Our gene signature of MMR deficiency in stem cells is a frontier discovery that includes a unique set of genes with the potential of being a biomarker of early cancer progression. In fact, this signature is able to differentiate between MMRd and MMR-proficient neoplasia as well as organoids derived from normal tissues in the same contexts. Several individual genes integrated within this profile have been previously shown to be involved in different aspects of colorectal cancer progression. For example, Aldh1a1 expression, which is normally found in the bottom of the crypt, was shown to be enhanced during progression from normal epithelium to adenoma with increasing expression levels as cells move upwards in the colonic crypt³³. Another gene, Spp1, which encodes for the bone sialoprotein Osteopontin was found to be significantly and exclusively expressed in MMR haplo-insufficient and -deficient stem cells. Furthermore, our proteomic profiling recorded spectral counts only in Lgr5^EGFP+ stem cells in Msh2-KO mice. These observations were validated both in mouse and human samples where we observed enhanced staining of SPP1 in small intestine crypts of Msh2-HET and Msh2-KO mice co-localizing with Lgr5, and in LS adenoma and tumor samples, respectively. SPP1 has been reported to have a role in invasion and metastasis of several cancer types, including colon, ovarian, and breast, thus acting both in an autocrine and paracrine manner^34–36, and specifically promoting stem-cell-like properties in CRC^{37, 38}. In addition, overexpression of SPP1 in tumor cells has been associated with infiltration by tumor-associated macrophages³⁹, thus fueling tumor growth and angiogenesis⁴⁰. This fact was also thought to weaken CD8⁺ T cell-mediated immune response leading to immune tolerance by dampening host tumor immune surveillance^{41, 42}. In fact, we were only able to confirm the upregulation of Spp1 in mouse organoids when we culture them in supplemented media with M-CSF1, which has been shown to be involved in the maintenance of the intestinal stem cell niche through different cellular and molecular mechanisms involving Paneth cells⁴³ and VEGF⁴⁴, as well as mediating the effects of the immune-environment surrounding the stem cell niche, in particular tumor-associated macrophages. Thus, SPP1 functions via adhesive interactions at the tumor/host interface in several malignancies and can be a potential biomarker for MMR deficiency, hence warranting further investigation in LS carcinogenesis.

Our work has several limitations. First, our proteomics data was only able to partially validate the gene expression profiling of stem cells. We believe that this weak correlation between mRNA and protein profiles was due to the limited protein yielded from FACS-sorted stem cells in Msh2-KO mice (only ~100k cells per replicate). Therefore, the lack of comprehensive proteomics coverage due to low-abundance of proteins limits the detection of most proteins with required quantitative precision. These challenges were observed previously in other biological systems, especially in embryonic stem cells where a large number of animals have been required to gather a sufficient number of cells for the appropriate biological replicates to provide adequate statistical power in the analysis^{45, 46}. Second, it would have been ideal to utilize single-cell genomics to investigate and validate MMRd specific gene expression patterns in mouse and human ISC. However, our experiments were designed prior to recent developments that would have helped improve the feasibility and cost-efficiency of single-cell transcriptomics. Third, we have attempted to perform a validation of our newly-discerned MMRd stem cell signature in an internal and limited LS cohort. Itdeally, the validation would have been performed using an independent set of samples with a larger number of tissue specimens. The ultimate value of the gene signature as a whole or individual markers such as SPP1 remains to be established in future studies.

In conclusion, we have identified a gene signature of MMRd ISCs using the transcriptomic and proteomic profiles of the stem and non-stem cells from a MMRd mouse model. We have observed that the MMRd stem cell signature is able to correctly distinguish early MMRd from MMRp early neoplasia from samples of LS and FAP patients. Using systems biology approaches, molecular, and cellular studies in both mouse and human samples, we identified SPP1, which qualifies as a bona fide marker of MMR deficiency in LS patients. In summary, data presented in this study advance our understanding of ISC biology in LS patients that serves as the starting point to develop novel markers of early detection of progression of LS carcinogenesis and potential targets for cancer interception strategies in this patient population.

Significance statement:

The transcriptomic and proteomic profile of MMR-deficient intestinal stem cells display a unique set of genes with potential roles as biomarkers of cancer initiation and early progression.

Abbreviations:

APC

adenomatous polyposis coli

Ascl2

achaete-scute family bHLH transcription factor 2

CBC

crypt base columnar

CMMRD

constitutional mismatch repair deficiency

CRC

colorectal cancer

DEGs

differentially expressed genes

EGFP

enhanced green fluorescent protein

FACS

florescence-activated cell sorting

FAP

familial adenomatous polyposis

GSEA

gene set enrichment analysis

H&E

hematoxylin and eosin

Het

heterozygous

IHC

immunohistochemistry

IRB

Institutional Review Board

ISCs

intestinal stem cells

KO

Knockout

Krt20

keratin 20

Lgr5

leucine rich repeat containing G protein-coupled receptor 5

LS

Lynch Syndrome

MMRd

mismatch repair-deficient

MMRp

MMR-proficient

MS

mass spectrometry

Msh2

mutS homolog 2

NES

normalized enrichment score

qRT-PCR

quantitative reverse transcriptase-polymerase chain reaction

mRNAseq

RNA sequencing; Spp1, secreted phosphoprotein 1

UTMDACC

The University of Texas MD Anderson Cancer Center

WT

wildtype