Extracellular Matrix Interactions Provide Tumor Cells With an Escape Mechanism for Commitment to Differentiation

Stem cells located at the apex of cellular hierarchies are responsible for maintaining the intestine and colon epithelium. Similarly, cells that express stem cell markers have been identified in colorectal cancer, where they have been proposed to support growth and relapse of the tumor.1Schepers A.G. Snippert H.J. et al.Science. 2012; 6095: 730-735Crossref Scopus (838) Google Scholar,2de Sousa e Melo F. et al.Nature. 2017; 7647: 676-680Crossref Scopus (442) Google Scholar However, recently, cells distinct from proposed cancer stem cells have been found to contribute to tumor growth and distant metastasis.3Lenos K.J. Miedema D.M. Lodestijn S.C. et al.Nat Cell Biol. 2018; 20: 1193-1202Crossref PubMed Scopus (100) Google Scholar,4Fumagalli A. Oost K.C. et al.Cell Stem Cell. 2020; 26: 569-578Abstract Full Text Full Text PDF PubMed Scopus (110) Google Scholar It is consequently pertinent to understand how specific cancer drivers, such as loss of adenomatous polyposis coli (APC), affect cell behavior and cellular responses to differentiation. Intestinal stem cells express several markers, including leucine-rich repeat-containing G-protein coupled receptor 5 (LGR5) and leucine-rich repeats and immunoglobulin-like domains protein 1 (LRIG1). Activation of the bone morphogenetic protein (BPM) pathway is one of the strongest regulators of terminal differentiation and is inversely correlated with Wnt pathway activity. Although patterns of pathway activation are well established in the homeostatic tissue, much less is known about their activation in tumors. We analyzed spontaneous colorectal tumors forming in the APCmin mouse model to assess stem cell marker expression and activation of the BMP pathway via SMAD5-phosphorylation (pSMAD5). Analysis of adenomas in LGR5-eGFPxAPCmin revealed that LGR5-eGFP–positive cells expressed LRIG1 and that LRIG1 expression is more widespread, including the early transit-amplifying compartment5Wong V.W. Stange D.E. Page M.E. et al.Nat Cell Biol. 2012; 14: 401-408Crossref PubMed Scopus (314) Google Scholar (Supplementary Figure 1A). Despite the morphologic changes observed in APCmin adenomas, the separation of LRIG1 expression and pSMAD5 was preserved (Figure 1A and Supplementary Figure 1B and C). This suggested that distinct compartments were maintained, even under conditions of constitutive Wnt activation. To further probe whether the patterns of normal cell behavior were perturbed, despite the observed segregation into distinct compartments, we isolated Lrig1high and Lrig1low cells from normal tissue and adenomas (Figure 1B). The proliferative potential of cells can be assayed in vitro using organoid models, where undifferentiated cells give rise to organoids.6Sato T. et al.Gastroenterology. 2011; 141: 1762-1772Abstract Full Text Full Text PDF PubMed Scopus (2089) Google Scholar In line with the specific expression of LRIG1 in the stem cell compartment, LRIG1high cells were able to form organoids when isolated from normal tissue, whereas LRIG1low could not (Figure 1B, Supplementary Figure 1D). Surprisingly, both LRIG1high and LRIG1low cells isolated from adenomas formed organoids and could be maintained long term (Figure 1B and C; at least 7 passages). Notably, LRIG1low cells engrafted and formed large LRIG1+ tumors after orthotopic transplantation into Rag2KO mouse colons (Figure 1D and E). Taken together, these experiments demonstrated that although the segregation of domains was maintained in adenomas, the cellular potential of distinct cellular compartments was altered in the APCmin tumors. These results suggested that tumor cells neither underwent terminal differentiation nor became post-mitotic like LRIG1low cells isolated from normal tissue. To test this hypothesis, we evaluated the response to BMP-induced terminal differentiation in vitro using the organoid model. Importantly, both wild-type (WT) and mutant organoids responded to BMP by means of up-regulation of immediate target (ID1), loss of stem cell marker (LRIG1), and induction of differentiation marker (FABP1) (Figure 1F). However, whereas treatment of WT organoids with BMP for 24 hours caused loss of stem cell potential, APC mutant lines (CRISPR knockout or APCmin tumor cells) retained their stem cell potential (Figure 1G). To uncover the molecular characteristics associated with unique properties of mutant Lrig1low cells characterized by pSMAD5 expression in vivo, we performed gene expression profiling. Aligned with the segregation of stem cell and differentiation markers, RNA sequencing of LRIG1high and LRIG1low cells isolated from normal tissue and APCmin tumors revealed a remarkable conservation of expression patterns (Supplementary Figure 1E). Aligned with the compartmentalized expression of pSMAD5 and LRIG1 in adenomas, genes down- and up-regulated upon BMP stimulation7Qi Z. Li Y. et al.Nat Commun. 2017; 813824Google Scholar were enriched in LRIG1high and LRIG1low tumor cells, respectively (P = 5 × 10–9; P = 10–11). Collectively, this illustrates that loss of APC—the most prominent mutation found in colorectal cancer—alone enables phenotypically differentiated cells to self-renew when placed in a permissive environment. To identify the molecular mechanisms in tumors making LRIG1low cells refractory to terminal differentiation and thus distinguish them from their counterparts in normal epithelium, we performed gene set enrichment analysis. As expected, the gene set enrichment analysis revealed enrichment for the Wnt pathway, integrin, and focal adhesion signaling (Supplementary Figure 1F). At the transcriptional level, this included abundant expression of fibronectin (FN1), which was also detected at the protein level in adenomas (Figure 1H and I). We speculated that signaling downstream of extracellular matrix (ECM) via FN1/integrins could be responsible for the observed phenotypic differences. Aligned with previous observations, WT cells exposed to high integrin signaling required WNT signaling for survival (Supplementary Figure 1G).8Yui S. Azzolin L. et al.Cell Stem Cell. 2018; 22: 35-49Abstract Full Text Full Text PDF PubMed Scopus (285) Google Scholar To determine experimentally the role of the ECM-directed signaling, we investigated pharmacologic disruption of the pathway on differentiation-promoting stimuli. SRC is a pivotal kinase downstream of ECM-integrin signaling, and its activity can be blocked using specific inhibitors. Interestingly, and in alignment with recent studies using collagen matrices,9Koppens M.A.J. Davis H. et al.Gastroenterology. 2021; 161: 239-254Abstract Full Text Full Text PDF PubMed Scopus (10) Google Scholar WT cells can escape BMP-induced differentiation when exposed to fibronectin in the matrix (Figure 1J). However, upon SRC inhibition, both WT and APC-mutant cells lost the ability to self-renew after BMP exposure (Figure 1J and Supplementary Figure 1H and I). Taken together, our findings show that, in a tumor context, certain cell populations escape irreversible commitment to terminal differentiation. Careful characterization of different cell states during both steady-state homeostasis and in adenomas enabled us to identify mis-regulation of integrin signaling as a mechanism used by tumor cells to become refractory to differentiation signals. Importantly, the dependence of ECM interactions on their ability to resist differentiation is a potential target for therapeutic interventions. This is significant, given that the vast majority of the currently used chemotherapy regimens target highly proliferative cells. Combining traditional chemotherapy regimens with approaches blocking ECM interactions could prevent seemingly differentiated tumor cells from reverting into a proliferative state and contributing to tumor regrowth. The authors thank Marie Kveiborg (University of Copenhagen) for sharing mouse models. Pawel J. Schweiger and Marie Le Bouteiller contributed equally to this work. Pawel Schweiger, PhD (Data curation: Equal; Formal analysis: Equal; Investigation: Equal; Validation: Equal; Visualization: Equal; Writing – original draft: Supporting; Writing – review & editing: Supporting). Marie Le Bouteiller, PhD (Data curation: Equal; Formal analysis: Equal; Funding acquisition: Supporting; Investigation: Equal; Methodology: Equal; Validation: Equal; Writing – original draft: Supporting). Shiro Yui, PhD, MD (Formal analysis: Supporting; Funding acquisition: Supporting; Investigation: Supporting; Methodology: Lead). Malte Thodberg, PhD (Formal analysis: Supporting; Visualization: Supporting). Ditte Clement, MSc (Investigation: Supporting). Kim B. Jensen, PhD (Conceptualization: Lead; Data curation: Lead; Funding acquisition: Lead; Investigation: Lead; Project administration: Lead; Resources: Lead; Supervision: Lead; Writing – original draft: Lead; Writing – review & editing: Lead). Lgr5-eGFP-IRES-CreERT2 (MGI ID: 3764660), Lrig1-eGFP-IRES-CreERT2 (MGI ID: 5520983), and APCMin (MGI ID: 1856318) mouse strains were used for breeding. Mice used for experiments were 1- to 5-month-old males and females. Animals were housed in individually ventilated cages in accordance with best animal husbandry guideline recommendations of the European Union Directive (2010/63/EU). Danish Animal Experiments Inspectorate reviewed and approved all animal experiments. Transplantation of organoids was performed as described previously.e1Watanabe S. et al.Nat Protoc. 2022; 17: 649-671Crossref PubMed Scopus (5) Google Scholar Transplanted mice were analyzed 5–6 weeks after transplantations. Distal colons were dissected and processed for immunostaining. Tissue sections were obtained from tissue fragments fixed overnight in 4% paraformaldehyde, followed by incubation overnight in 30% sucrose, and embedding in OCT. Tissue cryosections were cut at 8 μm. For immunostaining, cryosections were permeabilized with 0.1% Triton X100 for 5 minutes, blocked with 10% donkey serum, 0.5% fish skin gelatin, 3% bovine serum albumin (excluded for staining with primary antibodies from goat) for 1 hour and stained overnight with the following primary antibodies: anti-GFP (ab13970, 1:500; Abcam), anti-CDH1 (13-1900, 1:400; Invitrogen), anti-LYZ (A0099 1:1000; Dako), anti-LRIG1 (AF3688, 1:50; R&D), anti-PSMAD5 (ab92698, 1:200; Abcam), anti-KI67 (ab15580, 1:250; Abcam), anti-FN1 (F3648, 1:200; Sigma), and anti-LEF1 (sc-8591, 1:200; Santa Cruz Biotechnology). Staining with Alexa-488/555/647-conjugated secondary antibodies (1:400; ThermoFisher) was carried out for 1 hour at room temperature. Nuclei were counterstained with 1 μM 4′,6-diamidino-2-phenylindole and sections were mounted using Mowiol. Samples were imaged using Zeiss Axio Observer or Leica SP5/8 TCS. For confocal images, maximum projection images were generated from Z-stacks using software from the microscope manufacturer. For fluorescence-activated cell sorting (FACS) analysis, mouse tissue was washed extensively with phosphate-buffered saline and dissected into small 2- to 3-mm pieces. Tissue fragments were chelated in a 10-mM EDTA buffer, as described previously.6Sato T. et al.Gastroenterology. 2011; 141: 1762-1772Abstract Full Text Full Text PDF PubMed Scopus (2089) Google Scholar Separated epithelium was digested using TrypLE (1×) or Dispase II (1 U/mL; 04-942-078-001; Roche) for 10 minutes at 37°C. After filtration, cells were stained using the following antibodies: anti–EPCAM-APC (347200, 1:100; BD), anti–CD45-PE-Cy7 (552848, 1:100; BD), and anti–CD31-PE-Cy7 (25-0311-82, 1:100; eBioscience) for 30 minutes on ice, washed in phosphate-buffered saline, and sorted using FACS Aria I/III (BD). Lineage-positive cells (CD45+ CD31+) were excluded. Results were analyzed using FlowJo software. Single cells isolated using FACS were seeded in 35-μL drops of Matrigel mixed with recombinant Jagged-1 (AS-61298, 1 μM; Anaspec); 1000 or 25,000 cells were seeded into a single well of a 48-well plate (734-1607; Corning). Cells were cultured in advanced Dulbecco’s modified Eagle medium/F12 basal medium (12634010; Life Technologies) supplemented with penicillin/streptomycin (15140122; Life Technologies), HEPES (15630080, 10 mM; Gibco), GlutaMAX (35050061, 2 mM; Gibco), Y-27632 (Y0503, 10 μM; Sigma), murine Noggin (250-38, 100 ng/mL; Peprotech), human epidermal growth factor (AF-100-15, 50 ng/mL; Peprotech), and CHIR99021 (361559, 3 μM; Calbiochem). R-spondin necessary for organoid culture was supplemented in the recombinant form as murine Rspondin-1 (3474-RS, 500 ng/mL; R&D) or conditioned medium used at 5% final concentration. Cells were seeded in triplicate and maintained in a 37°C incubator with 5% CO2. Culture medium was changed every 2–3 days. After 7 days, epithelial organoids were counted manually based on images acquired with a bright-field microscope (Leica DMIL LED or Thermo EVOS FL Auto 2). After the first passage, CHIR99021 and Y-27632 were omitted. Organoids were split mechanically by means of vigorous manual pipetting every 5–9 days. For BMP treatment experiments, cells were initially grown in full medium, then deprived of Noggin for 24 hours and incubated with BMP4 (5020-BP, 100–400 ng/mL; R&D) for 48 hours. Cells were then passaged using the standard procedure. For reseeding experiments (Figure 1G and J and Supplementary Figure 1G–I) cells were analyzed 5–9 days after passaging. For SRC inhibition experiments, cells were cultured in medium containing dasatinib (CDS023389, 100 nM; Merck) and passaged using the standard procedure. When dasatinib was used together with BMP, medium was deprived of Noggin. APCmin and APCsgKO cells were cultured without R-spondin. Reseeded cells were cultured for 7–9 days and analyzed. CRISPR-Cas9–mediated deletion of APC was performed following the protocol published previously.e2Shwank G. Cell Stem Cell. 2013; 13: 653-658Abstract Full Text Full Text PDF PubMed Scopus (967) Google Scholar Deletions in the Apc gene in mutant clones were confirmed using Sanger sequencing (GAT, Denmark). Human fibronectin (F2006; Sigma) was mixed directly with Matrigel to obtain the desired concentration. The WNT was prepared as conditioned medium, as described previously.e2Shwank G. Cell Stem Cell. 2013; 13: 653-658Abstract Full Text Full Text PDF PubMed Scopus (967) Google Scholar RNA was extracted from organoids and FAC-sorted cells using the PicoPure RNA isolation kit (KIT0204; ThermoFisher). cDNA was prepared from 100–500 ng of RNA using the SuperScript III Reverse Transcriptase (18080093; ThermoFisher), according to the manufacturer’s protocol. For RNA sequencing, libraries were prepared with SMARTer Stranded Total RNA-Seq Pico Input Mammalian v2 Kit (634411; Takara) and sequenced on a NextSeq using 500 High Output v2 Kit (FC-404-2005; Illumina). Reads were mapped to the mm10/GRch38 genome assembly using RSubread.e3Liao Y. et al.Nucleic Acids Res. 2019; 47: e47Crossref PubMed Scopus (760) Google Scholar Differential expression analysis was carried out using edgeRe4Robinson M.D. et al.Bioinformatics. 2010; 26: 139-140Crossref PubMed Scopus (20727) Google Scholar and limma.e5Ritchie M.E. et al.Nucleic Acids Res. 2015; 43: e47Crossref PubMed Scopus (15243) Google Scholar Gene ontology enrichment analysis of differentially expressed genes and KEGG pathway analysis were carried out with gProfilere6Raudvere U. et al.Nucleic Acids Res. 2019; 47: W191-W198Crossref PubMed Scopus (1692) Google Scholar using default settings. The statistical significance for gene enrichment of genes affected by BMP stimulation was assessed with Fisher exact test using the number of genes detected as the background. For quantitative polymerase chain reaction, SYBR Green PCR (Invitrogen) was used with custom primers. Reactions were run on a QuantStudio 6 Flex (Applied Biosystems). Expression levels were normalized using TUBB5 and ALDA expression. Statistical analyses of data other than RNA sequencing were conducted using R and Microsoft Excel. Plots were produced using Prism (GraphPad). One-tailed, unpaired Student t test was used, with P < .05 considered statistically significant (∗P ≤ .05; ∗∗P ≤ .01). Error bars represent SDs. Relative organoid counts were normalized against WT organoids and/or or no-treatment controls.