If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Université de Paris, Institut Cochin, INSERM, Centre National Recherche Scientifique, Paris, FranceEquipe Labellisée Ligue Nationale Contre Le Cancer, Paris, France
Université de Paris, Institut Cochin, INSERM, Centre National Recherche Scientifique, Paris, FranceEquipe Labellisée Ligue Nationale Contre Le Cancer, Paris, France
Université de Paris, Institut Cochin, INSERM, Centre National Recherche Scientifique, Paris, FranceEquipe Labellisée Ligue Nationale Contre Le Cancer, Paris, France
Université de Paris, Institut Cochin, INSERM, Centre National Recherche Scientifique, Paris, FranceEquipe Labellisée Ligue Nationale Contre Le Cancer, Paris, France
Université de Paris, Institut Cochin, INSERM, Centre National Recherche Scientifique, Paris, FranceEquipe Labellisée Ligue Nationale Contre Le Cancer, Paris, France
Université de Paris, Institut Cochin, INSERM, Centre National Recherche Scientifique, Paris, FranceEquipe Labellisée Ligue Nationale Contre Le Cancer, Paris, France
Université de Paris, Institut Cochin, INSERM, Centre National Recherche Scientifique, Paris, FranceEquipe Labellisée Ligue Nationale Contre Le Cancer, Paris, France
Université de Paris, Institut Cochin, INSERM, Centre National Recherche Scientifique, Paris, FranceEquipe Labellisée Ligue Nationale Contre Le Cancer, Paris, FranceAssistance Publique Hôpitaux de Paris, Hôpitaux Universitaires Paris Centre, Pathology Department, Hôpital Cochin, Paris, France
Laboratory of Predictive Oncology, Centre de Recherche en Cancérologie de Marseille, Institut Paoli-Calmettes, INSERM Unité Mixte de Recherche 1068, Centre National Recherche Scientifique Unité Mixte de Recherche 725, Marseille, France
Laboratory of Predictive Oncology, Centre de Recherche en Cancérologie de Marseille, Institut Paoli-Calmettes, INSERM Unité Mixte de Recherche 1068, Centre National Recherche Scientifique Unité Mixte de Recherche 725, Marseille, France
Laboratory of Predictive Oncology, Centre de Recherche en Cancérologie de Marseille, Institut Paoli-Calmettes, INSERM Unité Mixte de Recherche 1068, Centre National Recherche Scientifique Unité Mixte de Recherche 725, Marseille, France
Correspondence Address correspondence to: Christine Perret, PhD or Béatrice Romagnolo, PhD, Inserm U1016, Institut Cochin, 24 Rue du Faubourg Saint-Jacques, 75014 Paris, France.
Université de Paris, Institut Cochin, INSERM, Centre National Recherche Scientifique, Paris, FranceEquipe Labellisée Ligue Nationale Contre Le Cancer, Paris, France
Correspondence Address correspondence to: Christine Perret, PhD or Béatrice Romagnolo, PhD, Inserm U1016, Institut Cochin, 24 Rue du Faubourg Saint-Jacques, 75014 Paris, France.
Université de Paris, Institut Cochin, INSERM, Centre National Recherche Scientifique, Paris, FranceEquipe Labellisée Ligue Nationale Contre Le Cancer, Paris, France
Axin1 is a negative regulator of wingless-type MMTV integration site family, member 1 (Wnt)/β-catenin signaling with tumor-suppressor function. The Wnt pathway has a critical role in the intestine, both during homeostasis and cancer, but the role of Axin1 remains elusive.
Methods
We assessed the role of Axin1 in normal intestinal homeostasis, with control, epithelial-specific, Axin1-knockout mice (Axin1ΔIEC) and Axin2-knockout mice. We evaluated the tumor-suppressor function of Axin1 during chemically induced colorectal tumorigenesis and dextran sulfate sodium–induced colitis, and performed comparative gene expression profiling by whole-genome RNA sequencing. The clinical relevance of the Axin1-dependent gene expression signature then was tested in a database of 2239 clinical colorectal cancer (CRC) samples.
Results
We found that Axin1 was dispensable for normal intestinal homeostasis and redundant with Axin2 for Wnt pathway down-regulation. Axin1 deficiency in intestinal epithelial cells rendered mice more susceptible to chemically induced colon carcinogenesis, but reduced dextran sulfate sodium–induced colitis by attenuating the induction of a proinflammatory program. RNA-seq analyses identified an interferon γ/T-helper1 immune program controlled by Axin1 that enhances the inflammatory response and protects against CRC. The Axin1-dependent gene expression signature was applied to human CRC samples and identified a group of patients with potential vulnerability to immune checkpoint blockade therapies.
Conclusions
Our study establishes, in vivo, that Axin1 has redundant function with Axin2 for Wnt down-regulation and infers a new role for Axin1. Physiologically, Axin1 stimulates gut inflammation via an interferon γ/Th1 program that prevents tumor growth. Linked to its T-cell–mediated effect, the colonic Axin1 signature offers therapeutic perspectives for CRC.
We show that Axin1, a Wnt/β-catenin partner, has redundant function with Axin2, for Wnt down-regulation in intestine, and identify an unrecognized role of Axin1 in the control of immune inflammation that accounts for its tumor-suppressor function in colorectal tumorigenesis.
Despite the remarkable clinical advances in cancer diagnosis and treatment achieved in recent decades, the mortality rate of colorectal cancer (CRC) remains high worldwide, with almost 600,000 deaths each year.
A precise understanding of the molecular mechanisms underlying CRC pathogenesis therefore is crucial for the improvement of therapeutic strategies. Novel targets are emerging in the form of the proteins of the wingless-type MMTV integration site family, member 1 (Wnt)/β-catenin pathway, which is activated in almost all CRCs, mostly through mutations of the APC tumor-suppressor gene.
Improvements in our understanding of the biology of the Wnt/β-catenin pathway therefore are required for the development of new therapeutic opportunities.
Mutations of APC are common in CRC, whereas mutations of AXIN1, a gene encoding another component of the Wnt pathway, are infrequent in these cancers.
The potential tumor-suppressor function of Axin1 in CRC remains unclear, but compounds stabilizing this protein have been reported to suppress the growth of Wnt-driven CRCs.
By contrast, a tumor-suppressor function for Axin1 has been shown clearly in hepatocellular carcinoma (HCC), for which AXIN1 mutations are identified in almost 10% of cases.
However, most human HCCs with AXIN1 mutations studied have not been found to be associated with higher levels of β-catenin activation, and the HCCs developing in mice with Axin1 inactivation in hepatocytes do not display aberrant activation of the Wnt pathway.
Axin, a negative regulator of the Wnt signaling pathway, directly interacts with adenomatous polyposis coli and regulates the stabilization of beta-catenin.
Several lines of evidence indicate that Axin1 and its homolog Axin2 play a key role in down-regulating the stability of β-catenin, encoded by Ctnnb1, which is crucial for the cellular response Wnt/β-catenin signaling.
In the absence of the Wnt signal, Axins act as scaffold proteins, together with the tumor-suppressor Adenomatous polyposis coli, in the destruction complex targeting β-catenin for degradation. However, Axin1 also is involved in transducing the Wnt signal, leading to an increase in cytoplasmic and nuclear β-catenin levels and the expression of β-catenin target genes.
Axin1 also acts as a scaffold protein for other non-Wnt pathways, including the c-Jun N-terminal kinase, transforming growth factor-β, p53, and AMP-activated protein kinase pathways.
Given the complex functions of Axin1, we decided to investigate its role in normal intestinal homeostasis and its tumor-suppressor function in CRC. Using mice with an epithelium-specific inactivation of Axin1, we showed that Axin1 was dispensable for normal intestinal homeostasis and redundant with Axin2 for down-regulation of the Wnt pathway. Mice that have combined deficiencies for Axin1 and Axin2 phenocopied Apc-deficient mice. We found that Axin1 deficiency in intestinal epithelial cells promoted the development of azoxymethane (AOM)-induced colon tumorigenesis, but induced an improvement of dextran sulfate sodium (DSS)-induced colitis. Axin1 therefore promotes an inflammatory program with tumor-suppressor function in intestinal epithelial cells that we identified as an interferon (IFN)γ/helper T cell 1 (Th1) immune program. We defined an Axin1-dependent gene expression signature (GES), which we then used to probe databases of human CRC transcriptomes. Patients with tumor profiles defined by the signature as Axin1-deficient had poor antitumor immune response and poor clinical outcome, whereas those with the Axin1-proficient profile had a better survival and an immune profile of potential vulnerability to immune checkpoint blockade therapies.
Results
Axin1 Is Not Required for Homeostasis in the Intestinal Epithelium, but Its Loss Sensitizes Progenitors to the Toxicity Resulting From Tamoxifen-inducible Cre Recombinase Activation
For assessment of the function of Axin1 in the intestine, we generated an Axin1-conditional knockout mouse strain under the control of the villin promoter (Vil CreERT2;Axin1fl/fl, referred to hereafter as Axin1ΔIntestinal Epithelial Cell[IEC]), in which Axin1 selectively was inactivated in all intestinal epithelial cells. Control (Axin1fl/fl or Vil CreERT2;Axin1+/+) and Axin1ΔIEC mice were subjected to 4 daily intraperitoneal (IP) injections of tamoxifen, and were analyzed on day 5 postinduction. Western blot confirmed the absence of Axin1 in the small intestine and colon of Axin1ΔIEC mice, and its presence in these tissues in control mice (Figures 1A and 2A). Histologic analyses were performed on the small intestine and colon of mutant and control mice. Mutant mice showed no major change in intestinal architecture or cell fate relative to controls (Figures 1B–E, 2B–E, and 3). However, H&E staining showed the presence of numerous apoptotic cells in the crypts of the small intestine of Axin1ΔIEC mice. This finding was confirmed by active caspase-3 and terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick-end labeling (TUNEL) staining, revealing significantly larger numbers of cleaved caspase-3-positive and TUNEL-positive cells in the crypt in Axin1ΔIEC mice than in controls (Figure 1C and D). A similar increase in the number of cleaved caspase-3-positive and TUNEL-positive cells relative to controls was observed in the colon of Axin1ΔIEC mice (Figure 2C and D). We analyzed proliferation status in the small intestine and colon of Axin1ΔIEC mice. We found that the number of proliferating cells in these compartments of the mutant mice was no higher than that in control mice (Figures 1E and 2E).
Figure 1Acute deletion of Axin1 in the small intestine of mice. (A) Western blot analysis of Axin1 and Gapdh in whole–small intestine tissue lysates from WT (2 Axin1fl/fl and 2 Vil CreERT2;Axin1+/+) and KO (4 Axin1ΔIEC) mice. Gapdh served as a loading control. The mice were killed on day 5, after 4 days of tamoxifen injections. (B–E) Histologic analyses of WT (Axin1fl/fl and Vil CreERT2;Axin1+/+) and KO (Axin1ΔIEC) mice. The mice were killed on day 5, after 4 days of tamoxifen injections. H&E staining of representative intestinal sections from WT and KO mice with apoptotic cells in KO mice highlighted by black arrowheads. Scale barfor H&E: 25 μm. Cleaved caspase 3, TUNEL, and Ki-67 staining of representative small intestine epithelium sections from WT and KO mice. Hematoxylin and methyl green were used as a nuclear counterstain for cleaved caspase 3 (scale bar: 25 μm), TUNEL staining (scale bar: 50 μm), and Ki-67 staining (scale bar: 50 μm). Determination of the mean number of cleaved caspase 3–positive cells and of TUNEL-positive cells per crypt was performed with a combined average of more than 40 consecutive whole crypts from 8 WT (4 Axin1fl/fl, 4 Vil CreERT2; Axin1+/+) and 4 KO (Axin1ΔIEC) mice (N = 4 per group). (F) Transient apoptosis induced by Cre ERT2 activation. Experimental design. KO (Axin1ΔIEC) mice received 4 daily injections of tamoxifen and then were killed on day 5 or day 15. Representative TUNEL staining on small intestine sections from KO (Axin1ΔIEC) mice on days 5 and 15 after tamoxifen injection. Methyl green was used as a nuclear counterstain. Scale bar: 25 μm. Determination of the mean number of TUNEL-positive cells per crypt was performed with a combined average of more than 40 consecutive whole crypts from WT (Axin1fl/fl) and KO (Axin1ΔIEC) mice (N = 4 per group). (G) Western blot analysis of Axin1 and Gapdh on whole intestine tissue lysates from WT (2 Axin1fl/fl and 1 Vil CreERT2;Axin1+/+;Axin1+/+) and 4 KO (Axin1ΔIEC) mice killed 15 days after induction. Gapdh served as a loading control. Significant differences as follows: ∗∗∗P < .001, and ∗∗∗∗P < .0001.
Figure 2Acute deletion of Axin1 in the mouse colon. (A ) Western blot analysis of Axin1 and Gapdh on whole colon tissue lysates from WT (2 Axin1flfl and 1 Vil CreERT2;Axin1+/+) and 4 KO (Axin1ΔIEC) mice killed on day 5 postinduction. Gapdh served as a loading control. (B–E) Histologic analyses of the colon of WT (Axin1fl/fl and Vil CreERT2;Axin1+/+) and KO (Axin1ΔIEC) mice. The mice were killed on day 5, after 4 days of tamoxifen injections. H&E staining of representative colon sections from WT (Axin1flfl or Vil CreERT2;Axin1+/+) and KO (Axin1ΔIEC) mice. Scalebarfor H&E: 25 μm. Cleaved caspase 3 (Scale bar: 25 μm), TUNEL (Scale bar: 25 μm) , and Ki-67 staining of representative colonic epithelium sections from WT (Axin1fl/fl or VillinCre+;Axin1+/+) and KO (Axin1ΔIEC) mice. Hematoxylin and methyl green were used as a nuclear counterstain for cleaved caspase 3, Ki-67, and TUNEL staining, respectively. Determination of the combined mean number of cleaved caspase 3–positive cells, TUNEL-positive cells per crypt of more than 20 consecutive whole colonic crypts in 8 WT (4 Axin1fl/fl, 4 Vil CreERT2;Axin1+/+) and 4 KO (Axin1ΔIEC) mice. Black arrowhead are Cleaved Casp3 and TUNEL positive cells respectively. ∗∗P < .01, ∗∗∗P < .001, and ∗∗∗∗P < .0001.
Figure 3Axin1 deficiency does not impair intestinal differentiation in mouse small intestine and colon. (A) Villin, Alcian blue, lysozyme, and chromogranin B staining of representative intestinal sections from WT (Axin1fl/fl or Vil CreERT2;Axin1+/+) and KO (Axin1ΔIEC) mice. The mice were killed on day 5, after 4 days of tamoxifen injections. Hematoxylin and fast red were used as a nuclear counterstain for villin, lysozyme, and chromogranin B staining (scale bar: 50 μm), and for Alcian blue staining (scale bar: 100 μm), respectively. Scale bar: 100 μm. (B) Villin, Alcian blue, and chromogranin B staining of representative colon sections from WT (Axin1fl/fl or Vil CreERT2;Axin1+/+) and KO (Axin1ΔIEC) mice. The mice were killed on day 5, after 4 days of tamoxifen injections. Hematoxylin and fast red were used as a nuclear counterstain for villin and chromogranin B staining (scale bar: 50 μm), and for Alcian blue staining (scale bar: 100 μm), respectively. Black arrowheads are chromogranin-positive cells.
Surprisingly, we found that the apoptotic phenotype observed in the crypt of Axin1ΔIEC mice mostly concerned progenitor cells and was transient (Figure 4). It disappeared 15 days postinduction, although Western blot continued to confirm the absence of Axin1 in the intestine of mutant mice (Figure 1F and G). We investigated the possibility that apoptosis might result from a sensitivity of Axin1-deficient crypts to the toxicity induced by CreERT2 activation because apoptosis levels were significantly higher in the crypts of control (Vil CreERT2;Axin1+/+) mice than in those of control Axin1fl/fl mice, although still much lower than those observed for Axin1ΔIEC mice (Figure 1C and D). Such toxicity has been described elsewhere.
Mutant and control mice were treated with 2 daily doses of tamoxifen and were killed on days 3 and 8 postinduction. The apoptosis observed on day 3 in mutant mice was transient and had disappeared by day 8 (Figure 4B). We also tested the effect of a new IP tamoxifen treatment and confirmed that tamoxifen reinjection induced de novo apoptosis in the crypts of Axin1ΔIEC mice (Figure 4C). Consistent with Cre genotoxicity, γH2AX foci were observed in the crypts of mutant mice at day 5 postinduction, and were associated with an accumulation of nuclear p53 (Figure 4D). Consistent with these data, the crypts of mice with intestinal epithelial cells deficient for both Axin1 and p53 (Vil CreERT2;Axin1fl/fl;p53fl/fl) had significantly fewer apoptotic cells on day 5 postinduction than the crypts of mice lacking only Axin1 (Figure 4E).
Figure 4Sensitivity of Axin1-deficient crypts to the toxicity induced by Cre ERT2activation in the small intestine. (A) Apoptosis mostly affected progenitor cells in the crypt of the small intestine. Representative TUNEL staining of Axin1ΔIEC mice on day 5 after tamoxifen injection, showing that most of the TUNEL-positive cells are located in the progenitor compartment of the crypt rather than in the intestinal stem cell (ISC) compartment of the crypt. Determination of the combined mean number of TUNEL-positive cells of more than 30 whole crypts in WT and KO mice in the progenitor compartment of the transit-amplifying (TA) cells and in the ISC compartment (N = 3 per group). The ISC and the TA compartment were defined relative to the Paneth cell zone by considering the TA cells as the cells of the crypts located above the Paneth cells. (B) Transient apoptosis induced by Cre ERT2 activation. Experimental design. WT (Axin1flfl) and KO (Axin1ΔIEC) mice received 2 daily doses of tamoxifen by injection and were killed on day 3 or day 8. Western blot of Axin1 and Gapdh on whole-intestine tissue lysates from WT and KO mice on days 3 and 8. Gapdh served as a loading control. Representative TUNEL staining on small intestine sections from WT and KO mice on days 3 and 8. Methyl green was used as a nuclear counterstain. Determination of the combined mean number of TUNEL-positive cells per crypt of more than 40 consecutive whole crypts in WT and KO mice on days 3 and 8 (N= 3 per group). Scale bar: 25 μm. (C) Effect of a new tamoxifen treatment. Experimental design. The first group of mice received 4 daily doses of tamoxifen and were killed on day 5 or 15. The second group of mice underwent the same protocol and received 4 daily doses of tamoxifen, however, on day 15, they received 2 new doses of tamoxifen before being killed on day 18. Representative TUNEL staining on small intestine sections from WT (Axin1fl/fl) and KO (Axin1ΔIEC) mice on days 5, 15, and 18. Methyl green was used as a nuclear counterstain. Determination of the combined mean number of the TUNEL-positive cells per crypt of more than 40 consecutive whole crypts in WT and KO mice on days 5, 15, and 18 (N = 3 per group). Scale bar: 25 μm. (D) Representative p53 and H2ax staining on intestinal epithelium sections from WT (Axin1fl/fl and Vil CreERT2;Axin1+/+) and KO (Axin1ΔIEC) mice on day 5, after 4 daily doses of tamoxifen. Hematoxylin was used as a nuclear counterstain. Scale bar: 25 μm. (E) Absence of apoptosis in crypts with a combined deficiency of both Axin1 and p53. Representative TUNEL staining on intestinal sections from Axin1-deficient (Axin1ΔIEC) mice and mice with combined Axin1/p53 deficiency (Vil CreERT2;Axin1fl/fl;p53fl/fl) on day 5. Methyl green was used as a nuclear counterstain. Determination of the combined mean number of TUNEL-positive cells per crypt of more than 40 consecutive whole crypts in WT (Axin1fl/fl); Axin1ΔIEC(Vil CreERT2;Axin1fl/fl); Axin1, p53ΔIEC(Vil CreERT2;Axin1fl/fl;p53fl/fl) mice (N= 3 per group). Scale bar: 25 μm. Significant differences as follows: ∗P < .05, ∗∗P < .01, ∗∗∗P < .001, and ∗∗∗∗P < .0001.
Our results show that Axin1 is dispensable for normal homeostasis of the intestinal epithelium but that its loss sensitizes crypt progenitors to the genotoxicity induced by CreERT2 activation.
Axin2 Compensates for the Loss of Axin1 in the Control of Wnt/β-Catenin Signaling
The absence of hyperproliferation of the crypt cells in Axin1ΔIEC mice suggested that Axin1 inactivation in the intestinal epithelium did not lead to an aberrant activation of Wnt/β-catenin signaling. We tested this hypothesis by quantitative reverse-transcription polymerase chain reaction (qRT-PCR) analysis of the expression of 2 canonical β-catenin target genes: Axin2 and Lgr5. The levels of Axin2 and Lgr5 expression in the various parts of the small intestine and colon did not differ significantly between mutant mice and controls (Figure 5A). We hypothesized that functional redundancy with Axin2 might account for the lack of phenotype of Axin1ΔIEC mice. We crossed these mice with Axin2-null mice with lacZ knock-in alleles,
referred to here as Axin2 knockout (KO) mice. Histologic analyses showed no disturbance to the architecture and cell proliferation of the small intestine and colon in these mice (Figures 5C–E and 6A–C). By contrast, mice with invalidations of both Axin1 and Axin2 in the intestinal epithelium (Vil CreERT2;Axin1fl/fl;Axin2lacZ,lacZ), hereafter referred to as double-knockout (DKO) mice, had a severe intestinal phenotype mimicking that observed after Apc inactivation in the intestinal epithelium.
These Vil CreERT2;Apcfl/fl mice are referred to hereafter as ApcΔIEC mice. DKO mice, similar to ApcΔIEC mice, died within 7 days of tamoxifen injection (Figure 5B), whereas Axin1ΔIEC and Axin2KO mice remained alive for at least 1 year after injection (Figure 5B and data not shown). Likewise, H&E analyses of the small intestine showed an enlargement of the crypt in DKO mice resembling that observed in ApcΔIEC mice
(Figure 5C). This enlargement was associated with a large increase in the number of proliferating and apoptotic cells in the crypts of DKO mice that was not observed in the crypts of mice lacking only Axin1 or Axin2 or in controls, except for the small transient increase in apoptosis observed in Axin1ΔIEC mice (Figures 1B and 5D and E). An increase in the number of Olfm4-positive cells was observed, highlighting an increase in the intestinal stem cell compartment (Figure 5F). Furthermore, we observed a mislocalization of Paneth cells in the DKO mice. These cells were scattered throughout the enlarged crypts, rather than being confined to the bottom of the crypt, as observed for control mice and mice lacking either Axin1 or Axin2 (Figure 5G). All of these histologic features mimicked the characteristics of ApcΔIEC mice.
A phenocopy of ApcΔIEC mice also was observed in the colon of DKO mice, with an increase in the crypt length, apoptosis and cell proliferation, and de novo expression of the Paneth cell marker lysozyme
Figure 5Mice with a combined deficiency of Axin1 and Axin2 in the small intestine phenocopy Apc-deficient mice. (A) Relative messenger RNA (mRNA) levels for Axin1, Axin2, and Lgr5, assessed by qRT-PCR analyses on the duodenum (Duo), ileum (Ile), and colon (Col) of WT (Axin1fl/fl) and KO (Axin1ΔIEC) mice (N = 5 per group). (B) Kaplan–Meier survival curves for WT (Axin1fl/fl, n = 4, gray), KO (Axin1ΔIEC, n = 4, blue), Axin2KO (n = 4, red), ApcΔIEC (n = 3, green), and DKO (n = 4, purple) mice. (C–G) H&E staining of representative small intestine sections from WT (Axin1fl/fl), KO (Axin1ΔIEC), Axin2KO, DKO, and ApcΔIEC mice. Scale bar: 50 μm. Determination of mean crypt length (N = 4 per group). TUNEL, Ki-67, Olfm4, and lysozyme staining of representative small intestine epithelium sections from WT (Axin1fl/fl), KO (Axin1ΔIEC), Axin2KO, DKO, and ApcΔIEC mice. Hematoxylin and methyl green were used as a nuclear counterstain for Ki-67, Olfm4, lysozyme, and TUNEL staining. Scale bar: 50 μm. Determination of the mean number of TUNEL-positive cells per crypt of more than 40 consecutive whole crypts in WT Axin1fl/fl, KO (Axin1ΔIEC), Axin2KO, DKO, and ApcΔIEC mice (N = 3 per group). Mislocalization of lysozyme-positive cells in the enlarged crypts, highlighted by black arrowheads. The mice were killed on day 5, after 4 days of tamoxifen injections. Significant differences as follows: ∗P < .05, ∗∗P < .01, and ∗∗∗P < .001.
Figure 6Mice with combined deficiencies of Axin1 and Axin2 in the colon are phenocopies of Apc-deficient mice. (A) Representative H&E staining of colon sections from WT (Axin1fl/fl), KO (Axin1ΔIEC), Axin2KO, DKO and ApcΔIEC mice. Determination of mean crypt length (N = 4 per group). (B–D) Representative TUNEL, Ki-67, and lysozyme staining on colon sections from WT (Axin1fl/fl), KO (Axin1ΔIEC), Axin2KO, DKO, and ApcΔIEC mice. Hematoxylin and methyl green were used as a nuclear counterstain for Ki-67, lysozyme, and TUNEL staining, respectively. Scale bars: 50 μm. The mice were killed on day 5, after 4 days of tamoxifen injections.
We then performed qRT-PCR to analyze the expression of several β-catenin target genes in the small intestine and colon of DKO, control, and mice deficient for Axin1 or Axin2. Axin2 expression did not increase significantly in Axin1-deficient mice, and vice versa (Figure 7), excluding any compensation of expression between these 2 genes. More importantly, the results confirmed the aberrant activation of Wnt/β-catenin signaling in DKO mice, with strong induction observed for several β-catenin target genes (c-myc, Ascl2, Sp5, Sox9, Ephb3) in both the small intestine and colon, as reported for Apc-deficient mice.
The lack of induction of Sp5 and Sox9 in the colon is consistent with the regional-specific response of the Wnt/β-catenin signaling that showed activation of partially distinct sets of target genes in the intestine.
In agreement with our data, a constitutive Wnt activation in organoids of all parts of the intestine showed an increase of Sox9 expression in the small intestine and no induction in the colon.
No induction of β-catenin target genes was observed in control mice or in mice with single deficiencies of Axin1 or Axin2 (Figure 7).
Figure 7Activation of the Wnt/β-catenin pathway in the intestine and liver of mice with combined deficiencies of Axin1 and Axin2. (A) Relative messenger RNA (mRNA) levels for Axin1, Axin2, Ascl2, Sp5, Sox9, and Ephb3 assessed by qRT-PCR analysis on intestinal epithelium sections from the (A) duodenum, (B) ileum, and (C) colon of WT (Axin1fl/fl, n = 3), KO (Axin1ΔIEC, n = 4), Axin2KO (n = 3), DKO (n = 4), and ApcΔIEC (n = 4) mice. The mice were killed on day 5, after 4 days of tamoxifen injections. (B) Glul and Ki-67 staining in the livers of WT (Axin1fl/fl), Axin1KOLIV (Axin1fl/fl injected with Ad Cre), Axin2KO (Axin2LacZLacZ), DKOLIV (Axin1fl/fl;Axin2LacZLacZ injected with Ad Cre), and ApcKOLIV (Apcfl/fl injected with Ad Cre) mice. Scale bars: 200 μm (toprow) and 50 μm (bottom row). Mice were killed 2 weeks after recombined adenovirus injection.
Thus, Axin2 compensated for the loss of Axin1 in the down-regulation of Wnt/β-catenin signaling. We performed additional analyses on the liver, in which we also previously had observed a lack of activation of Wnt/β-catenin signaling in mice with a targeted Axin1 inactivation in hepatocytes.
This aspect was of particular interest because the domain in which Axin1 and Axin2 are co-expressed is smaller in the liver lobule than in the intestine, in which it covers the entire crypt, with a gradient from the bottom to the top. In the liver, Axin1 and Axin2 co-expression is restricted to a ring of 1 to 3 pericentral hepatocytes around the central vein, in which Wnt/β-catenin signaling is activated.
By injecting an adeno-associated virus encoding the Cre recombinase into Axin1fl/fl or (Axin1fl/fl;Axin2lacZ, lacZ) mice, resulting in the inactivation of Axin1, or of the combined deletion of Axin1 and Axin2, respectively, in all hepatocytes within the liver, we observed that inactivation of Axin1 in hepatocytes did not disturb metabolic zonation, as shown by the absence of change in expression of the pericentral marker Glul, a well-defined liver β-catenin target gene
Similar results were obtained for the inactivation of Axin2 in the hepatocytes of Axin2KO mice (Figure 7D). However, in the livers of mice with a combined inactivation of Axin1 and Axin2 in hepatocytes, we observed a complete disruption of metabolic zonation, with Glul expression throughout the entire liver lobule (Figure 7D). These results attest to the occurrence of Wnt/β-catenin activity throughout the livers of DKO mice, suggesting that there was sufficient Axin2 expression outside the perivenous area to compensate for the loss of Axin1, despite the lack of activation of Wnt/β-catenin signaling. As in the intestine, we found that the combined loss of Axin1 and Axin2 in hepatocytes phenocopied Apc loss in the liver, both for the metabolic zonation and for the control of cell proliferation with a strong increase in cell proliferation shown by Ki-67 staining (Figure 7D). This is consistent with our previous work.
Axin1ΔIEC mice and their wild-type (WT) Axin1fl/fl littermates were treated with the carcinogen AOM and then with 3 cycles of 2.5% DSS in drinking water, separated by periods of normal drinking water. This treatment resulted in chronic inflammation that supported the growth of the neoplastic epithelium induced by AOM (Figure 8A). We assessed the colitis phenotype by monitoring changes in body weight. We found that Axin1-deficient mice lost significantly less weight than control mice at each cycle of DSS treatment, suggesting that Axin1 loss in intestinal epithelial cells may protect against DSS-induced toxicity (Figure 8B). At the experimental end point, 100 days after the initial AOM injection, a macroscopic evaluation of the dissected and washed colon indicated that tumors had developed in both mutant and control mice, mostly in the distal colon (Figure 8C). Axin1ΔIEC mice developed significantly more tumors than WT mice (Figure 8D). Further histologic analyses of the size of the tumors showed that Axin1ΔIEC mice had significantly larger numbers of small tumors, suggesting that Axin1 influences tumor initiation (Figure 8E). All of the tumors were well-differentiated adenocarcinomas (Figure 8F). We also evaluated cell proliferation by Ki-67 staining of the tumors. There were larger numbers of Ki-67–positive cells in the tumors of Axin1ΔIEC mice than in those of WT mice (Figure 8G). AOM is carcinogenic in rodent colonic epithelial cells through the induction of changes to the Wnt/β-catenin pathway, mostly through Ctnnb1 mutations.
We investigated whether the higher cell proliferation level and larger numbers of tumors in Axin1ΔIEC mice were associated with an increase in β-catenin activity resulting from Axin1 deficiency. Immunohistochemical staining for β-catenin on Axin1ΔIEC and WT tumors showed no increase in the overall cytosolic/nuclear β-catenin staining (Figure 8H). An analysis of the expression of several β-catenin target genes in tumors developed in Axin1ΔIEC mice and in their WT counterparts confirmed the absence of an increase in β-catenin activity between tumors developed in Axin1ΔIEC and WT mice (Figure 8I). As expected, from the presence of Ctnnb1 mutation in AOM/DSS-induced colon tumors,
expression of the β-catenin target genes were induced in all the tumors (WT and KO) compared with the nontumoral tissue (Figure 8I). Collectively, these results showed that tumors that developed in the Axin1ΔIEC mice did not display higher levels of β-catenin activity than those that developed in WT animals.
Figure 8Axin1 deficiency promotes the development of colon tumors. (A) Experimental protocol for AOM/DSS treatment. (B) Body weight changes during the AOM/DSS treatment of WT Axin1fl/fl (n = 13) and KO (Axin1ΔIEC, n = 10) mice. (C) Representative image of colons from WT and KO mice on day 100 after AOM treatment. (D and E) Tumor distribution in the colon of WT and KO mice on day 100 after AOM treatment. (F) H&E staining of representative sections of colon tumors from WT and KO mice on day 100 after AOM treatment, with a magnified inset in the lower row. Scale bar: 2.5 mm. (G) Ki-67 staining of representative sections of colon from WT and KO mice on day 100 after AOM treatment, with a magnified inset in the lower row. (H) Representative immunostaining for β-catenin of colon tumors from KO and WT mice, with a magnified inset in the lower row. Scale bar: 100 μm. (I) Heatmap indicating the expression of several β-catenin target genes in colon tumors from KO and WT mice and in the colon of untreated KO and WT mice. TAM, Tamoxifen.
For confirmation of protection against DSS toxicity, we examined the response of Axin1ΔIEC mice to acute treatment with DSS. Axin1ΔIEC mice and their control Axin1fl/fl littermates were treated for 7 days with 3% DSS in drinking water, and then were allowed to recover on normal drinking water for 6 days. From day 5 of DSS treatment, the Axin1ΔIEC mice lost significantly less weight than the controls. Furthermore, most of the KO mice presented weight recovery on day 13, whereas none of the WT mice had recovered their normal weight at this time point (Figure 9A). The WT mice had a shorter colon than the KO mice after DSS treatment (Figure 9B). These results confirm that epithelial Axin1 deficiency leads to less severe colitis than was observed in WT mice after DSS treatment. Consistent with this less severe colitis, the histologic score, reflecting inflammation and epithelial damage, was significantly lower in Axin1ΔIEC mice than in controls (Figure 9C). Surprisingly, as shown in Figure 9D, there was no significant difference between the intestinal permeability, as measured by the tetramethylrhodamine B isothiocyanate (TRITC)-dextran assay, in the colon of Axin1ΔIEC and WT mice that were challenged by DSS (Figure 9D).
Figure 9Axin1 deficiency attenuates DSS-induced colitis. (A) Body weight changes after treatment with 3% DSS or regular water (control) in WT (Axin1fl/fl) and KO (Axin1ΔIEC) mice. WT (n = 9 for water, n = 15 for DSS) and KO (n= 10 for water, n = 15 for DSS) mice. (B) H&E staining of representative sections of colon after DSS treatment for WT and KO mice on days 7 and 13. Colon lengths after treatment with DSS or regular water of WT (n = 9) and KO (n = 7) mice. (C) H&E-stained images and histologic scores for colon sections after DSS treatment for WT and KO mice on days 7 and day 13. Histologic scores for colitis obtained for the WT and KO groups after treatment with DSS. Scalebar: 1 mm (D) Intestinal permeability assessed by fluorescein isothiocyanate (FITC)-dextran assay on day 5 of DSS administration (N = 5 per group). (E) Relative messenger RNA (mRNA) levels for Axin1, Il6, Il1β, Tnfα, Ccl2, and Cxcl10 assessed by qRT-PCR analysis of the colon of DSS-treated mice killed on day 7 (WT, n = 5; KO, n = 9) or day 13 (WT, n = 6; KO, n = 6), and of mice given regular water (WT, n = 9; KO, n = 10). Significant differences as follows: ∗P < .05, ∗∗P < .01, ∗∗∗P < .001, and ∗∗∗∗P < .0001.
Consistent with the colitis scores, the increase in messenger RNA levels for proinflammatory cytokines/chemokines (Il6, Il1β, Tnfα, Cxcl10, Ccl2) after DSS treatment was smaller in mutant mice than in controls (Figure 9E). The residual expression of Axin1 in mutant mice may result from its expression in nonepithelial intestinal cells. Altogether, these results suggest that a protective mechanism operates in Axin1ΔIEC mice, limiting gut inflammation after DSS treatment.
Collectively, these results show that Axin1 has a tumor-suppressor function in the intestine that does not appear to be linked to an increase in activation of the Wnt-β-catenin pathway. Axin1ΔIEC mice were prone to the development of larger numbers of colorectal cancers upon AOM/DSS treatment, despite having a weaker colonic inflammatory response after acute DSS treatment.
Axin1 Controls an Immune IFNγ/Th1 Inflammatory Response
We investigated how Axin1 might be able to both stimulate an inflammatory response in the gut and to protect against tumorigenesis by comparing first the transcriptome profiling data for tumors developing in Axin1ΔIEC and WT mice by RNA sequencing (RNA-seq) and then the transcriptome profiles of colon tissues from Axin1ΔIEC and WT mice treated with DSS on day 7 and day 13 that was compared with those for untreated WT mice receiving drinking water.
RNA-seq analysis of the tumors from Axin1-deficient and WT mice identified 327 differentially expressed genes (DEGs) (P < .05; fold change, >±2), 232 of which were down-regulated in the Axin1-deficient mice (Supplementary Table 1). Interestingly, functional annotation analysis with IPA (Quiagen) and ClueGo software identified the IFN signaling and immune Th1 pathways as the highest-ranking enriched pathways inhibited in the tumors of Axin1ΔIEC mice (Figure 10A, and Table 3 in Figure 10). Consistent with these results, gene set enrichment analysis (GSEA) with the Hallmark gene sets identified IFNγ response, allograft rejection, inflammatory response, and tumor necrosis factor (TNF)α signaling via nuclear factor-κB (NF-κB) as significantly down-regulated in the tumors of Axin1ΔIEC mice relative to those in WT mice (Figure 10B). In addition, GSEA with the Biocarta profile identified the interleukin (IL)12 pathway, a major pathway for the development of Th1-type immune responses against cancer,
as significantly down-regulated in the tumors of KO mice relative to their WT counterparts (Figure 10C). Consistent with these results, a heatmap of the immune Th1 signature showed this signature to be less strongly expressed in tumors from Axin1-mutant mice than in WT mice, and confirmed the decreased expression of Il12 (Figure 10D). Upstream regulator analysis with IPA software identified IFNγ/signal transducer and activator of transcription (STAT)1 among the top-ranking cytokine upstream regulators for which significant inhibition was predicted (Table 3 in Figure 10). Stat1 plays a vital role in IFNγ signaling and is essential for Th1 cell differentiation.
Collectively, these results suggest a down-regulation of the IFNγ/STAT1/Th1 immune response in the tumors of Axin1ΔIEC mice. Interestingly, these pathways are known to play a major role in antitumor immunity,
To support a decrease in the Th1-mediated immunosurveillance in the tumors of Axin1ΔIEC mice, we quantified by immunohistochemistry the infiltration of the lymphocytes T CD8+ in the tumors. As expected, a significant decrease in the number of CD8+ T cells was observed in tumors of Axin1ΔIEC mice compared with tumors that developed in their WT counterparts (Figure 10F). We also identified 4 gene sets that were up-regulated significantly in the tumors of KO mice relative to those of WT mice. These gene sets included genes encoding adenoviral early region 2 binding factor targets, G2M checkpoint, and mitotic spindle proteins. These data are consistent with the increase in Ki-67 staining we described in Figure 8G.
Figure 10Functional gene expression profiling of AOM/DSS-induced colon tumors in Axin1ΔIECmice and controls. AOM/DSS-induced colon tumors from WT and KO mice were subjected to RNA-seq (n = 3 mice per group). (A) IPA analysis of DEGs down-regulated in the tumors of KO mice relative to the tumors of WT mice. The graph illustrates the most enriched pathways displaying down-regulation in the tumors of KO mice. Significance is expressed as the –log10 of P values. (B) GSEA results, by the Molecular Signatures Database (MSigDB) Hallmark gene set, showing negative enrichment in inflammatory and immune pathways in the tumors of KO mice. (C) GSEA, by the MSigDB Biocarta gene set, showing negative enrichment for the IL12 pathway in the tumors of KO mice. (D) Heatmap showing down-regulation of the IFNγ/Th1 immune response in the tumors of KO mice. (E) GSEA showing positive enrichment in cell-cycle pathways in the tumors from KO mice. (F) CD8 infiltration in tumors. Representative CD8+ immunostainings of WT and KO colon tumors. Scale bar: 50 μm. Quantification of the CD8+ T cells per mm2 tumor tissue in WT and KO mice (N = 3 per group). Significant differences as follows: ∗P < .05. IRF, Interferon Regulator Factor; KEGG, Kyoto Encyclopedia of Genes and Genomes; NES, normalized enrichment score; NOD, nucleotide-binding oligomerization domain; NOM, nominal; TUM, Tumor; WP, Wikipathway.
We then analyzed the colon transcriptional profile of KO and WT mice treated with DSS on day 7 and day 13 (Supplementary Table 2). As expected, GSEA identified major pathways of gut inflammation enriched in the colon of WT mice treated with DSS on day 7 and day 13, respectively, that were compared with untreated WT animals (Figure 11A). We then used GSEA analysis to compare the transcriptome profiles of Axin1ΔIEC mice treated with DSS on day 7 and day 13, respectively, with their WT counterparts. Strikingly, we observed opposite patterns of change in the expression of the gut inflammatory gene sets, for which the WT DSS-treated group showed positive enrichment relative to the untreated WT group, whereas the Axin1-mutant DSS-treated group showed negative enrichment relative to the WT DSS-treated group (Figure 11A). Therefore, there was a significant down-regulation of gut inflammatory pathways in Axin1-deficient mice after treatment with a single dose of DSS on both day 7 and day 13 (Figure 11A). Interestingly, it was the same gut inflammatory pathways that were down-regulated in the tumors of mutant mice, including the IFNγ response, allograft rejection, inflammatory response, and TNF signaling via NF-κB (Figure 10B). These data are consistent with the qRT-PCR data presented in Figure 9D that showed a significant decrease of several proinflammatory cytokines and chemokines in colon of DSS-treated Axin1ΔIEC mice vs their WT counterparts. Overall, these data indicate that Axin1 loss in the colon attenuates gut inflammatory pathways, including IFNγ and TNF signaling.
Figure 11Functional gene expression profiling of the colon after acute DSS treatment in Axin1ΔIECmice and controls. Colon tissues from WT and KO mice treated with DSS for 5 days and killed on day 7 (D7) or day 13 (D13), and colon tissue from WT untreated mice, were subjected to RNA-seq (n = 3 mice per group). (A) GSEA results, by Molecular Signatures Database Hallmark gene set, showing a positive enrichment in inflammatory and immune pathways in colonic tissues from WT mice treated with DSS on day 7 (D7) and D13 (D13) relative to untreated WT mice receiving water (top), and an inverse negative enrichment in these pathways in DSS-treated KO samples relative to DSS-treated WT samples (bottom). (B) Venn diagram showing the number of DEGs up-regulated in WT DSS-treated mice relative to untreated WT mice both on D7 and D13, together with the number of DEGs down-regulated in the KO DSS-treated mice relative to the WT DSS-treated mice both on D7 and D13. A total of 385 genes were identified as commonly up-regulated in WT DSS-treated mice relative to untreated WT and down-regulated in the KO DSS-treated mice relative to the WT DSS-treated mice. (C) IPA analysis of the 385 genes up-regulated in DSS-treated WT mice and down-regulated in DSS-treated KO mice. (D) fluorescence-activated cell sorting (FACS) analysis of of T-cell effector cell responses from colonic lamina propria of DSS-treated mice. T-cell populations in freshly isolated immune cells from the lamina propria of the colon of Axin-deficient (KO) mice and WT mice treated by DSS and killed on D13 were analyzed by FACS analyses. Representative flow cytometry plots of IFNγ+ cells within CD4+Tcrβ+ (Th1) cells, of Rorγt+Il17+ T cells within CD4+Tcrβ+ cells (Th17) and of Foxp3+Il10a+ within CD4+Tcrβ+ cells (regulatory T cell [Treg]) cells from WT mice and KO mice are shown. Box plots quantifying the percentage of IFNγ+ (Th1), Rorγt+Il17+ (Th17), and Foxp3+Il10a+ (Treg) cells within the CD4+Tcrβ+ cell population (means ± SEM; WT, n = 5; KO, n = 10). Significant differences as follows: ∗P < .05. FOXP3, forkhead box P3; FSC-H, Forward side scatter-Height; IFNG, interferon gamma; KEGG, Kyoto Encyclopedia of Genes and Genomes; NES, normalized enrichment score; NOM, nominal; PD-1, programmed cell death 1; PD-L1, CD274 molecule; R-MMU, Reactome pathway; RORγt,RAR-related orphan receptor gamma t; TREM1, Triggering receptor expressed on myeloid cells 1.
For further identification of the pathways inhibited in the colon of Axin1ΔIEC mice relative to WT controls after DSS treatment, we identified the DEGs commonly up-regulated in the WT DSS-treated group relative to the untreated WT group (P < .05; fold-change, ≥2) and down-regulated in the mutant DSS-treated group relative to the WT DSS-treated group (P < .05; fold change, ≥-2), on both day 7 and day 13 (Supplementary Table 3, Figure 11B). IPA and ClueGO analyses of these 385 DEGs identified inflammation and the immune Th1 pathway as the most significantly enriched pathways in KO DSS-treated mice (Figure 11C and Table 4 in Figure 11). Upstream regulator analysis identified IFNγ/STAT1 and NF-κB among the top-ranking upstream regulators showing significant down-regulation in KO DSS-treated mice (Table 5 in Figure 11). We then assessed the lymphocyte infiltration in Axin1ΔIEC mice and controls upon DSS challenge using flow cytometry analysis. Lymphocytes were isolated from the lamina propria of colon of DSS-treated Axin1ΔIEC and control mice and characterized for expression of different markers that define the Th1, Th17, and regulatory T-cell subsets. Consistent with the RNA-seq data, a significant decrease in the number of CD4+TCRβ+IFNγ+ (Th1) cells was observed in the colonic lamina propria of Axin1ΔIEC mice that was associated with a significant decrease in Th17 cells, while the regulatory T-cell population showed a similar number of cells in KO and WT mice (Figure 11D).
Altogether, these results identified epithelial Axin1 as a proinflammatory factor acting by promoting the development of an IFNγ/Th1 immune response in the colon. Interestingly, this inflammatory response is known to have antitumor immune activity and likely explain the tumor-suppressor function of Axin1 in the colonic mucosa.
The Gene Expression Signature From Axin1-Deficient Mice Identifies CRC Patients With Poor Antitumor Immune Responses and Clinical Outcomes
We then investigated whether the Axin1-dependent 327-GES defined by subjecting Axin1ΔIEC and WT mice to the AOM/DSS protocol (Supplementary Table 1) was clinically and biologically relevant in patients with CRC. Our hypothesis was that if the Axin1 pathway indeed promotes the development of an appropriate IFNγ/Th1 antitumor response, Axin1-proficient tumors should be enriched in CRC patients from the consensus molecular subtype (CMS)1 (immune molecular subtype). We therefore applied the 327-gene signature to a transcriptomic database including 2239 primary clinical samples of CRCs (Table 1). Samples were classified either in the Axin1-proficient-like group (N = 1053), for those enriched with genes up-regulated in the WT mice, or in the Axin1-deficient-like group (N = 1186), for those enriched with genes up-regulated in the Axin1ΔIEC mice. Clinical, pathologic, and molecular variables then were compared between the 2 groups.
Table 1Description of the 11 Human Colon Cancer Data Sets Included
Study
Data source
Technological platform
Probe sets/genes, n
Colon samples, n
Primary colon cancers, n
Jorissen et al, Clin Cancer Res 2009. PMID: 19996206
GEO database: GSE14333
Array U133 Plus 2.0; Affymetrix
54,000
251
251
Sheffer et al, Proc Natl Acad Sci U S A 2009. PMID: 19359472
GEO database: GSE41258
Array U133 A; Affymetrix
22,000
289
168
Staub et al, J Mol Med (Berl) 2009. PMID: 19399471
GEO database: GSE12945
Array U133 Plus 2.0; Affymetrix
54,000
29
29
Smith et al, Gastroenterology 2010. PMID: 19914252
GEO database: GSE17538
Array U133 Plus 2.0; Affymetrix
54,000
232
232
de Sousa et al, Cell Stem Cell 2011. PMID: 22056143
Relative to Axin1-proficient-like tumors, Axin1-deficient-like tumors were enriched (Fisher exact test) in distal colon tumors from male patients, frequently stage 4, mismatch repair-proficient (pMMR) and of CMS2 subtype that is associated with Wnt activation (Table 2). As hypothesized, the Axin1-proficient-like tumors were enriched in the CMS1 subtype relative to Axin1-deficient-like tumors (30% vs 9%) (Table 2), and, to a lesser extent, enriched in the CMS4 subtype (34% vs 27%) (Table 2).
Table 2Clinical and Pathologic Characteristics of the Colon Cancer Samples for the Whole Series and According to Axin1-Proficient-Like Vs Axin1-Deficient-Like Status
Disease-free survival (DFS) data were available for 1843 patients who had undergone surgery for primary CRC. The median follow-up period was 43 months (range, 1–212 mo); 120 patients (20%) had experienced an event, and the 5-year DFS was 66% (95% CI, 33–68) (Figure 12A). The 5-year DFS was lower in the Axin1-deficient-like group (63%; 95% CI, 60%–67%) than in the Axin1-proficient-like group (68%; 95% CI, 65%–72%; P = 1.69E-02, log-rank test) (Figure 12B). In univariate analysis (Figure 12C, Wald test), pathologic extension stage (P = 1.48E-14), MMR status (P = 2.67E-03), CMS (P = 1.08E-03), and the Axin1 GES signature-based group (P = 1.71E-02) were associated with DFS. In multivariate analysis (Figure 12C, Wald test), extension stage, MMR status, and the Axin1 GES signature-based group remained significant, suggesting that these variables were of independent prognostic value, whereas CMS tended toward significance.
Figure 12Correlations of Axin1-proficient-like and Axin1-deficient-like colon cancers with DFS and molecular variables. (A) Kaplan–Meier DFS curves for patients with surgically treated colon cancers. (B) As in panel A, but according to the Axin1-proficient-like and Axin1-deficient-like groups. The P value for the log-rank test is shown. (C) Forest plots of hazards ratios between the Axin1-proficient-like and Axin1-deficient-like groups in univariate and multivariate Cox analyses for DFS. The P values shown were obtained by the Wald test. (D) Forest plots of odds ratios between the Axin1-proficient-like and Axin1-deficient-like groups in univariate logistic regression analysis for molecular variables. The P values shown were obtained in logit link tests and the variables are ranked with P values increasing from top to bottom. aDC, activated DC; APM, Antigen presenting Machinery; DC, Dendritic cell; dMMR, deficient mismatch repair; ICR, Immunologic Constant of Rejection; iDC, immature DC; NK, Natural Killer; pDC, Plasmacytoid dendritic cell; Tem, T effector memory; TFH, T follicular helper; Tgd, _T gamma delta; TGF, transforming growth factor; TLS, _tertiary lymphoid structures; TReg, regulatory T cell.
We then compared several molecular variables, notably variables relating to immune landscape composition, between the Axin1-deficient-like and Axin1-proficient-like groups (Figure 12D). In silico analysis with the 24 immune cell types defined as the immunome
showed much lower levels of immune infiltrate (T cells, B cells, dendritic cells, macrophages, neutrophils, and so forth), and, more specifically, smaller cytotoxic cell subsets (CD8, Th1, cytotoxic, and γδ T cells) in the Axin1-deficient-like samples. Other immune functional signatures supported these observations: an impaired T-cell–inflamed signature,
for both innate and adaptive immune effector cells. This finding also is consistent with a profound defect of activation of the IFNα, IFNγ, and TNFα pathways,
Gene signature of antigen processing and presentation machinery predicts response to checkpoint blockade in non-small cell lung cancer (NSCLC) and melanoma.
These data showed that Axin1-deficient-like tumors from CRC patients have a profound defect of the Th1/IFNγ/cytotoxic antitumor response. This result obtained from the whole series also was confirmed in each molecular subtype (data not shown). Finally, the Axin1-deficient-like tumors also showed an enhanced activation of the transforming growth factor-β and lactic acidosis pathways, which are major inhibitors of the cytotoxic activity.
These data highlight the clinical pertinence of our data from Axin1-deficient mouse models. Patients bearing CRCs with weak expression of the Axin1-dependent signature had a worse prognosis, associated with a significant decrease in the level of the immune antitumor response. Axin1 therefore appears to be an interesting new modulable target related to the antitumor immune response and clinical outcome.
Discussion
We show that Axin1 is redundant with Axin2 for down-regulation of Wnt/β-catenin signaling in the intestinal epithelium. Axin1 is dispensable for intestinal homeostasis, deficiency of both Axin1 and Axin2 in the intestinal epithelium is required to phenocopy intestine-specific Apc knockout mice. This might explain why the Wnt/β-catenin pathway does not appear to be involved in the tumor-suppressor role of Axin1 in the intestine. We identify an unrecognized role for Axin1 in the control of the immune response in the intestine capable of accounting for its tumor-suppressor function. We found that Axin1 deficiency in intestinal epithelial cells rendered mice more susceptible to chemically induced colon carcinogenesis, but reduced DSS-induced colitis by attenuating the induction of a proinflammatory program. RNA-seq analyses identified an IFNγ/Th1 immune program controlled by Axin1 that enhances the inflammatory response and protects against CRC. These observations in mice were complemented by in silico analysis of 2239 human CRC tumors, which showed that patients with weak expression of the Axin1-dependent GES had a worse prognosis and suggested that those with strong expression of the Axin1-dependent GES might benefit from immune checkpoint blockade.
We investigated the role of Axin1 in the intestine because of the crucial role of the Wnt pathway in the control of homeostasis and tumorigenesis of the gastrointestinal tract.
We found that Axin1 was dispensable for normal intestinal homeostasis, its loss being compensated by Axin2. Mice with double Axin1, Axin2 deficiencies in the intestinal epithelium phenocopied ApcΔIEC mice and died rapidly. The reason for the lethality likely is owing to the failed differentiation of the intestinal lineages induced by the aberrant Wnt activation that should alter the function of the intestine.
We expected the compensation of Axin1 loss by Axin2 for Wnt pathway down-regulation to be observed only in tissues in which Wnt signaling was activated because Axin2 is a canonical β-catenin target gene.
We therefore analyzed the combined inactivation of Axin1 and Axin2 in the liver, in which the activation of Wnt signaling is restricted to a ring of a few hepatocytes around the central vein constituting the perivenous region of the liver lobule.
As reported here for intestinal cells, no particular phenotype was detected in the corresponding mice. Surprisingly, the combined deletion of Axin1 and Axin2 in hepatocytes led to an aberrant activation of β-catenin signaling throughout the liver lobule, even in the periportal area in which Axin2 expression was not detected.
These results therefore suggest that the compensation of Axin1 loss by Axin2 also occurs in tissues in which Wnt/β-catenin signaling is not activated and in which Axin2 expression levels are thought to be very low, such as periportal liver tissues. This surprising result supports the computational model of Lee et al,
who considered Axin1 to be rate-limiting for the destruction complex owing to its low levels.
We identified a new role for epithelial Axin1 in controlling gut inflammation that might be associated with its tumor-suppressor function. RNA-seq analyses of the colon of DSS-treated Axin1ΔIEC and WT mice identified an IFNγ/Th1 proinflammatory program that was down-regulated in Axin1ΔIEC mice. Flow cytometry analyses confirmed the reduction of the Th1 and Th17 cells in the lamina propria of DSS-treated Axin1ΔIEC mice that corroborated the resistance of these mice to DDS-induced colitis. Thus, epithelial Axin1 appears to promote a proinflammatory response mediated by activation of the IFNγ/Th1 pathway. Interestingly, the same program also was found to be down-regulated in tumors developing in Axin1ΔIEC mice after the AOM/DSS protocol. The IFNγ/Th1 pathway is known to have tumor-suppressor functions responsible for tumor clearance in several cancers, including CRC,
which could explain the increased AOM–DSS–driven colon tumorigenesis in Axin1ΔIEC mice. Consistent with a decrease in immune surveillance, we detected by immunohistochemistry a decrease in the infiltration of the cytotoxic lymphocytes CD8+ in tumors that developed in Axin1ΔIEC mice. This correlates with the significantly large numbers of small tumors in the Axin1ΔIEC mice and suggests that epithelial Axin1 suppresses the initiation of tumorigenesis. The tumor-suppressor function of Axin1 in the intestinal epithelial cells also involved a negative control of cell proliferation as shown by the strong increase in Ki-67 staining in tumors of Axin1ΔIEC mice. Interestingly, the increase in cell proliferation in the tumors of Axin1ΔIEC mice was not associated with an increase in Wnt/β-catenin signaling, likely owing to compensation by its paralog Axin2, induced in the tumors (Figure 8I).
The precise mechanism by which intestinal epithelial Axin1 exerts its tumor-suppressor function needs further clarification. Our data showed that Axin1ΔIEC mice were more susceptible to AOM–DSS–driven colon tumorigenesis, but were more resistant to DSS-induced colitis. These observations are inconsistent with the paradigm that linked inflammation and cancer.
However, inflammation has the role of a double-edged sword in cancer and may either induce antitumor immunity or promote tumor development, specifically in CRC.
Indeed, in human beings, an IFNγ/Th1 response is characteristic of cancer immune surveillance, associated with Th1CD4+ and CD8+T cells, which directly regulate tumor cell cytotoxicity and induce tumor suppression.
Our data strongly suggest that intestinal epithelial Axin1 exerts its tumor-suppressor function by inducing an antitumorigenic IFNγ/Th1 response. How epithelial Axin1 promotes a proinflammatory IFNγ/Th1 response remains to be investigated. We examined if it was via an epithelium-extrinsic and/or -intrinsic function. Analysis of the intestinal permeability showed no significant difference between the increase in the DSS-induced clearance of the TRITC-dextran in the colon of Axin1ΔIEC and WT mice, suggesting no specific alteration in the epithelial barrier integrity between the KO and WT mice upon the DSS challenge. This is in agreement with data showing that the systemic response to bacterial antigens can occur independently of epithelial barrier disruption.
A defect in epithelial barrier integrity is not required for a systemic response to bacterial antigens or intestinal injury in T cell receptor-alpha gene-deficient mice.
These results favor an epithelium-intrinsic mechanism by which Axin1 promotes a proinflammatory response in the colon. From the RNA-seq/in silico analyses of both the tumors (Figure 10A, and Tables 1 and 2 in Figure 10) and the DSS-treated colon (Figure 11C, and Tables 3 and 4 in Figure 11), we can speculate that Axin1 participates in the transduction pathways that allow, under the sensing of microbial stimuli, the synthesis of immunoregulatory signals in the intestinal epithelial cells. In agreement with this hypothesis, a role for epithelial Axin1 in host–pathogen interactions has been described. Salmonella typhimurium decreases Axin1 protein expression in intestinal epithelial cell lines at the post-transcriptional level, whereas its overexpression inhibits Salmonella invasion bacterial infections in vitro.
The RNA-seq data identified IFNγ/Stat1 and TNFα as the inflammatory pathways targeted by epithelial Axin1. This is consistent with the phenotype of mice with a genetic or a pharmacologic inhibition of Stat1 that showed a decrease in the Th1 response and a reduction in chemically induced colitis.
Hyaluronan-mediated leukocyte adhesion and dextran sulfate sodium-induced colitis are attenuated in the absence of signal transducer and activator of transcription 1.
Strikingly, the specific deletion of Stat1 in intestinal epithelial cells phenocopied Axin1ΔIEC mice and showed a tumor-suppressor function of Stat1 in CRC. Stat1ΔIEC mice were more resistant to DSS-induced colitis and showed an increased tumor load upon the AOM/DSS protocol.
The mechanism by which epithelial Axin1 is able to activate the inflammatory pathways IFNγ/Stat1 and TNFα is largely unknown at present. Axin1 is a scaffolding protein for which many interactors have been described (https://www.ncbi.nlm.nih.gov/gene/8312) and has not yet been connected to any inflammatory pathway. Among the already identified interactors, Card9 and Traf2 have been implicated in intestinal inflammatory responses and cancer,
Tumor necrosis factor receptor-associated factor (TRAF) 2 controls homeostasis of the colon to prevent spontaneous development of murine inflammatory bowel disease.
; notably, the features of the tumor microenvironment such as a lack of T-cell infiltration, and low levels of Th1 cell activity resulting in low levels of immune cytotoxicity, are predictive of adverse outcomes in patients with CRC; conversely, the presence of cytotoxic immune infiltration has been identified as a favorable prognostic marker.
RNA-seq and flow cytometry analyses of tumors from Axin1-deficient mice predicted a tumor microenvironment with low levels of IFNγ/Th1 cell activity. We therefore investigated whether the Axin1-dependent gene signature that we identified was of prognostic value in human CRC. In our 2239-sample series, the Axin1 signature-based classification was found to be an independent prognostic feature: patients from the Axin1-proficient-like group had a longer DFS than those from the Axin1-deficient-like group. The difference in 5-year DFS was small (5%) between the 2 groups, but was significant in multivariate analysis. These Axin1-proficient-like patients also had molecular traits suggestive of a significantly stronger Th1/IFNγ/cytotoxic antitumor response that would make them more likely to respond to immunotherapy than Axin1-deficient-like patients and opens up possibilities for treatment. First, patients with the Axin1-proficient-like signature were enriched for CMS1 class, which is associated with a microsatellite instability -high status and predicted to respond best to immunotherapy, and thus might be expected to benefit from immune checkpoint blockade. Second, the use of tankyrase inhibitors could be optimized; these drugs stabilize Axin proteins
although their clinical use is limited by their toxicity. The combined use of an immunotherapy approach together with tankyrase inhibitors might be expected to lower the effective concentration range for the use of these inhibitors, at least in the group of patients bearing the Axin1-proficient-like signature.
In conclusion, our study showed a new mechanism by which Axin1 has tumor-suppressive function in CRC, which might open new therapeutic perspectives in colorectal cancer.
Materials and Methods
Ethical Compliance Statement
All animal procedures were performed according to French legal regulations (Ministère de la Recherche, de l’Enseignement Supérieur et de l’Innovation), with the approval of the Ethics Committee of Paris Descartes University (Project autorisation pour projets utilisant animaux à des fins scientifiques [APAFIS] 8722 and 8612).
The human in silico study is based on public data from published studies in which the informed patients' consent to participate and the ethics and institutional review board already were obtained by authors. The study was approved by our Institutional Review Board from Institut Paoli-Calmettes (Marseille, France; Comité d’Orientation Stratégique).
Mice
Mice with an inducible specific deletion of Axin1 in intestinal epithelial cells (Axin1ΔIEC) were obtained by crossing Axin1fl/fl mice
were obtained from Anne-Amandine Chassot (Nice, France). For all experiments, approximately equal numbers of male and female mice were analyzed.
Mice were housed in colony cages in specific pathogen-free conditions, under a 12-hour light/dark cycle, in a controlled-temperature environment (21°C). They were fed ad libitum with a standard laboratory chow diet (65% carbohydrate, 11% lipids, and 24% proteins; SAFE 03, FRANCE).
Unless otherwise specified, we injected 4 doses of tamoxifen, 1 per day for 4 days (1 mg/mouse of tamoxifen, T5648; Sigma-Aldrich) in corn oil IP into 8- to 12-week-old mice.
For the inactivation of Axin1 in hepatocytes, Ad5-CMV-cre and Ad5-GFP (2 × 109 plaque forming unit in 150 mL) were injected into the retro-orbital vein of 8-week-old mice anesthetized by isoflurane inhalation as previously described.
Six- to 8-week-old WT (Axin1fl/fl) and KO (Axin1ΔIEC) mice received 3% DSS (weight/volume) in drinking water for 7 days and then were given regular drinking water until they were killed. Mice from each strain and gender were kept in separate cages. Mice were monitored daily for weight loss and clinical disease (soft feces).
A histologic score was obtained according to the criteria described by Erben et al
Briefly, 6- to 8-week-old WT (Axin1fl/fl) and KO (Axin1ΔIEC) mice received a single IP injection of AOM (10 mg/kg body weight; Sigma). Mice from different strains and sexes were kept in separate cages. Over the next few weeks, mice were exposed to 3 cycles of treatment with 2.5% DSS (weight/volume) in drinking water for 7 days, with a 14-day interval in which the mice received regular water between cycles, for recovery. Mice were killed on day 100 after AOM injection.
Immunoblot Analysis
Total protein extracts were obtained from 100 mg frozen mouse liver or from mouse primary hepatocytes homogenized in lysis buffer (50 mmol/L Tris-HCl, pH 7.5, 150 mmol/L NaCl, 5 mmol/L EDTA, 30 mmol/L Na4P2O7, 50 mmol/L Sodium fluoride [NaF], 1% Triton X-100 [Sigma-Aldrich], 1 mmol/L dithiothreitol) protease inhibitor cocktail (32953; Pierce, Thermo Fisher Scientific) supplemented with phosphatase inhibitor cocktail (8867; Pierce, Thermo Fisher Scientific) in a bead mill, with the Tissue Lyser disruption system (Qiagen, Hilden, Germany). Proteins were resolved by sodium dodecyl sulfate–polyacrylamide gel electrophoresis. The resulting bands were transferred to nitrocellulose membranes, which were blocked by incubation with 5% bovine serum albumin or 5% milk. Blots were incubated overnight at 4°C with specific primary antibodies, washed, incubated with the corresponding horseradish-peroxidase–conjugated secondary antibodies (Cell Signaling), and developed by an enhanced chemiluminescence technique (Thermo Fisher Scientific). Primary antibodies were obtained from Cell Signaling Technologies (Axin1, ref: C76H11, 1:500) and Santa Cruz Biotechnology (glyceraldehyde-3-phosphate dehydrogenase [Gapdh], ref: sc-25778, 1:1000).
Histology and Immunohistochemistry
Immediately after death, the entire gastrointestinal tract of the mouse was removed, splayed open along its length, and rolled up from the proximal to the distal end to form a Swiss roll. Tissues were fixed by incubation in 4% paraformaldehyde overnight at 4°C and embedded in paraffin wax. Sections (3-μm thick) were cut and stained with H&E. Immunohistochemistry was performed as previously described.
We treated 5-μm sections with 3% hydrogen peroxide for 15 minutes at room temperature. Antigens then were retrieved by boiling for 15 minutes in citrate buffer (10 mmol/L, pH 6) or for 40 minutes in Tris-EDTA buffer (100 mmol/L Tris, 12.6 mmol/L, pH 9) in a microwave pressure cooker (EZ Retriever; Biogenex). The sections were incubated in blocking solution (2% goat serum, 1% bovine serum albumin diluted in Tris Buffer Saline Tween) for 20 minutes at room temperature. They then were incubated overnight at 4°C with the primary antibodies against the following proteins, diluted in blocking solution: cleaved caspase-3 (9661S, 1:200; Cell Signaling), Ki-67 (12202S, 1:1500; Cell Signaling), lysozyme (EC.3.2.1.17, 1:200; Dako), Olfm4 (CS 39141S, 1:400; Cell Signaling), p53 (NCL-p53-CM5p, 1:200; Leica), γH2AX (05-636, 1:300; Millipore), chromogranin B (HPA012602, 1:1000; Atlas Antibodies), villin (ab130751, 1:500; Abcam), glutamine synthetase (610518, 1:400; BD Biosciences), and CD8 (14-0808-80, 1:100; eBiosciences). Specific binding was detected with a biotinylated secondary antibody and ABC reagent (Vector) for immunohistochemistry, and the signal was revealed with 3,3′-diaminobenzidine tetra hydrochloride solution (Vector). Hematoxylin was used as a nuclear counterstain. Apoptosis was assessed with a TUNEL assay kit, in accordance with the manufacturer's instructions (QIA33; Calbiochem). In the crypt, the 5–6 cells located above the +4 cells were considered transit-amplifying cells.
Quantification of the labeled positive cells was performed in almost 40 consecutive crypts in the small intestine and in 20 consecutive crypts in the colon. This provides an average of positive cells in the crypts per animal. Quantification was performed on 3 or 4 different mice in each group and statistics were performed on these combined averaged numbers.
Mouse livers were cut into 3-mm–thick sections and fixed by incubation in 4% paraformaldehyde for 12 hours. They then were embedded in paraffin. Immunohistochemistry was performed as previously described.
An antibody against glutamine synthetase was used (610518, 1:400; BD Biosciences).
Intestinal Permeability
Four hours (average time of intestinal transit in mice) before killing, overnight fasted mice were orally gavaged with a solution of TRITC dextran 4 kilodaltons at 50 mg/mL in phosphate-buffered saline (PBS) (TdB Labs, Sweden). Blood then was collected at the tail tip. Plasma was analyzed for the TRITC dextran 4 kilodalton concentration using an automatic fluorescence microplate reader at 544-nm excitation and 580-nm emission wavelengths to determine the intestinal epithelium permeability. TRITC concentrations were determined using standard curves prepared by diluting various amounts of TRITC-dextran in PBS.
RNA Extraction and Quantitative Real-Time PCR Analysis
Total RNA was extracted from mouse tissues with TRIzol Reagent (Life Technologies) in accordance with the manufacturer’s protocol. Reverse-transcription was performed with 1 μg total RNA and the Transcriptor First-Strand Complementary DNA Synthesis Kit (Roche Diagnostics), with random hexamer primers. Quantitative PCR was performed with the Light Cycler 480 Sybr Green I Master kit (Roche) and specific primers (Eurogentec) on a Light Cycler 480 thermocycler (Roche). RNA levels were calculated by the 2(-Delta Ct) method, with 18S as the internal control, relative to RNA levels in control (Axin1flfl) mice.
RNA-Seq Analysis
RNA extracted from colon tissue (distal colon) was treated with DNAse1 according to the manufacturer’s protocol (RNA Clean and Concentrator-5, R1013; Zymo Research). Three different mice per genotype were used for RNA-seq analysis. Library preparation and sequencing were performed at the GenomIC’ sequencing facility using 3 RNA samples per genotype (https://www.institutcochin.fr/core_facilities/genomesequencing-studies). The libraries were prepared following the TruSeq Stranded messenger RNA protocol from Illumina, starting from 800 ng high-quality total RNA (RNA Integrity Number, >8). Paired-end (2 × 75 bp) sequencing was performed on an Illumina Nextseq 500 platform, 20–40 millions paired end reads were used for the analysis. After sequencing, a primary analysis was performed with AOZAN software (ENS, Paris) for demultiplexing and quality control analysis on the raw data (based on FastQC modules, version 0.11.5).
The STAR algorithm (version 2.7.6a; Genomic'IC) then was used to align the Fastq files with the Ensembl Mus musculus GRCm38 reference, release 101 (Genomic'IC). Reads then were counted with RSEM (v1.3.1; Genomic'IC) and statistical analyses were performed on the read counts with R (version 3.6.3; Genomic'IC) and the DESeq2 package (DESeq2_1.26.0; Genomic'IC) to determine the proportion of genes differentially expressed between 2 conditions.
We used the standard DESeq2 normalization method (DESeq2 median of ratios with the DESeq function), with a prefilter of reads and genes (reads uniquely mapped on the genome, or up to 10 different loci with a count adjustment, and genes with at least 10 reads in at least 3 different samples).
In accordance with the recommendations for this package, we used the Wald test with the contrast function and the Benjamini–Hochberg false discovery rates control procedure to identify DEGs.
Significantly over-represented terms were identified with a right-tailed Fisher exact test, which calculated an overlap P value determining the probability of each term being associated with our lists of differential transcripts by chance alone.
The IPA z-score is a statistical measurement of correlation between the direction of the relationship and experimental levels of protein. It was calculated to assess the activation (positive z-score) or repression (negative score) of each term. The z-score was considered significant if its absolute value was greater than 2. GSEA was performed with Broad Institute GSEA software v4.1.0 (http://software.broadinstitute.org/gesa/index.jsp).
GSEA determined whether gene sets from the Molecular Signatures Database Hallmark collection (v7.4) were distributed randomly throughout our ranked transcript lists or whether they were located toward the top or bottom of these lists.
Briefly, small intestines were rinsed in cold PBS, minced into small pieces, incubated with EDTA (2 mmol/L), and digested with collagenase (1.5 mg/mL) for 30 minutes at 37°C. A discontinuous Percoll separation method (40% and 80%) was used to purify immune cells. Cell suspensions were centrifuged and the pellet was resuspended in 40% Percoll layered by 80% Percoll. The cells concentrated at the interface were collected and washed in cold PBS. Cells were suspended in RPMI1640 containing 5% fetal bovine serum, counted, incubated with human FcgammaR-binding inhibitor Receptor Binding Inhibitor Polyclonal Antibody (14-9161-73; eBioscience) and stained with specific antibodies.
Cells were subjected to fluorescence-activated cell sorting (FACS) analysis with the following antibodies: anti-CD4 (RM4-5 clone, 558107; BD Biosciences), anti-CD19 (1D3 clone, 550992; BD Biosciences) anti-IFNγ (XMG1.2 clone, 554412; BD Biosciences), or anti-transcription factor forkhead box P3 (FOXP3) (FJK-16s clone, 12-5773; eBioscience), anti-IL10 (JES5-16E3 clone, 554467; BD Biosciences), anti-IL17A (TC11-18H10 clone, 560220; BD Biosciences), and anti-RAR-related orphan receptor gamma (RORγ) (t) (REA278 clone, 130-123-248; Miltenyi Biotec). Dead cells were removed using the LIVE/DEAD Fixable Aqua Dead Cell Stain Kit (L34957; Invitrogen).
For intracellular staining, cells were pretreated with phorbol myristate acetate/ionomycin for 4 hours at 37°C 5% CO2 and brefeldin A was added for the last 3 hours. Cells then were stained with specific extracellular antibodies, permeabilized with BD cytoFix/cytoPerm solution (555028; BD Biosciences) or with transcription factor staining buffer set (00-5523-00; eBiosicence), and, finally, stained intracellularly for IFNγ, IL10, RORγ(t), IL17, or Foxp3. FACS analysis was performed with the BD LSR Fortessa flow cytometer (BD Biosciences) with FlowJo software (BD Bioscience).
Gene Expression Analysis on Colon Cancer Clinical Samples
We assessed the biological and clinical relevance of our 327-gene signature (Axin1-deficient mice vs Axin1-proficient mice) in human CRC samples by analyzing our database including clinical, pathologic, and gene expression data for primary CRC samples from 10 publicly available data sets and our own data set generated at the Institut Paoli-Calmettes (Marseille, France). Data were collected from the National Center for Biotechnology Information/GenBank Gene Expression Ommnibus, ArrayExpress, and The Cancer Genome Atlas (TCGA) databases (Table 1). Samples were profiled with DNA microarrays (Affymetrix) or by RNA sequencing (Illumina). The data set analyzed contained data for 2239 primary CRC samples. The characteristics of these samples are summarized in Table 2. Briefly, 88% of patients were older than age 50 years, and 53% were male. The proximal colon was the site of the CRC in 51% of cases. There were more stage 2 cancers (41%) than stage 3 cancers (34%). MMR status was classified as pMMR in 81% of cases. CMS determination identified 20% of samples as CMS1 (microsatellite instability immune), 32% as CMS2 (canonical), 17% as CMS3 (metabolic), and 31% as CMS4 (mesenchymal). DFS data were available for 1843 nonmetastatic patients with CMS subtyping data who had undergone surgery.
Data analysis required a preanalytical processing step, as previously described.
We then applied several multigene molecular signatures separately to each tumor in each data set. We first used the 327 DEGs between the tumors of Axin1-deficient mice and tumors from WT mice to define an Axin1-deficient vs Axin1-proficient expression profile. We then calculated the Pearson correlation distance between each sample and this profile: the samples with negative correlations were classified as Axin1-proficient-like (N = 1053) and those with positive correlations were classified as Axin1-deficient-like (N = 1186). The other gene expression signatures applied included the CMS classification,
Gene signature of antigen processing and presentation machinery predicts response to checkpoint blockade in non-small cell lung cancer (NSCLC) and melanoma.
In statistical analysis, the correlations between the Axin1-proficient-like and Axin1-deficient-like groups and the clinical and pathologic variables were calculated with the Fisher exact test. The DFS was calculated from the date of diagnosis until the date of metastatic relapse or death from CRC. Follow-up evaluation was measured from the date of diagnosis to the date of last contact for event-free patients. Survival was calculated by the Kaplan–Meier method, and curves were compared in log-rank tests. Univariate and multivariate prognostic analyses for DFS were performed by Cox regression analysis (Wald test). The variables tested in univariate analysis included Axin1-proficient-like vs Axin1-deficient-like status, patient age (>50 y vs ≤50 y), sex (male vs female), tumor site (proximal vs distal), pathologic stage (2 vs 1, 3 vs 1), MMR status (pMMR vs deficient MMR), and CMS classification (CMS2 vs CMS1, CMS3 vs CMS1, and CMS4 vs CMS1). Multivariate analysis incorporated all variables with a P value less than 5% in univariate analysis. The correlations of molecular variables with Axin1-proficient-like and Axin1-deficient-like status were assessed by logistic regression analysis with the lm function (R statistical package; significance estimated by specifying a binomial family for models with a logit link). All statistical tests were 2-tailed and were performed with an α risk of 5%. Statistical analysis was performed with the survival package (version 2.43) in R software (version 3.5.2)
Statistical Analysis
Statistical analyses were performed with GraphPad Prism (9.04). Graphic data are represented as means ± SD unless specified. Data were checked for normality and unpaired 2-tailed Student t tests were used to assess differences between groups. P values less than .05 were considered significant.
Acknowledgments
The authors thank S. Robine (Institut Curie, Paris, France), A. Berns (Netherlands Cancer Institute, Amsterdam, The Netherlands), and Anne-Amandine Chassot (University of Nice Sophia Antipolis, Nice, France) for supplying mutant mice. The authors also thank Trinath Jamma (BITS Pilani-Hyderabad Campus, Telangana, India) for help and support for the cell sort at the Flow Core Facility, the staff of the Cochin animal facility, Isabelle Lagoutte from the Imagerie du Vivant facility of Cochin Institute, and Lucie Adoux from the Genomic facility of Cochin Institute.
RNA sequencing data sets are available through the public, open-access repository National Institutes of Health Gene Expression Omnibus platform: =https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?accGSE196220. GEO accession number: GSE196220.
Axin, a negative regulator of the Wnt signaling pathway, directly interacts with adenomatous polyposis coli and regulates the stabilization of beta-catenin.
Gene signature of antigen processing and presentation machinery predicts response to checkpoint blockade in non-small cell lung cancer (NSCLC) and melanoma.
A defect in epithelial barrier integrity is not required for a systemic response to bacterial antigens or intestinal injury in T cell receptor-alpha gene-deficient mice.
Hyaluronan-mediated leukocyte adhesion and dextran sulfate sodium-induced colitis are attenuated in the absence of signal transducer and activator of transcription 1.
Tumor necrosis factor receptor-associated factor (TRAF) 2 controls homeostasis of the colon to prevent spontaneous development of murine inflammatory bowel disease.
Conflicts of interest The authors disclose no conflicts.
Funding Supported by the National Institute of Health and Medical Research (France), Ligue Contre le Cancer (Equipe Labellisée LNCC 2020), Labex Who Am I? (France; R.S. and C.L.P.), Cancer Research Personalized Medicine, and a Ministère de la Recherche et de l’Enseignement Supérieur fellowship (R.S. and D.C.).
Epithelial cells lining the intestine continually renew, with nearly complete replacement occurring at least weekly.1 The intestine can absorb nutrients and water because of precisely regulated cellular proliferation, differentiation, and death, all of which maintain the proper balance of stem and differentiated cells. Several signaling pathways are involved in intestinal homeostasis and Wnt signaling plays a key role. In cells not encountering a Wnt ligand, tumor-suppressor protein adenomatous polyposis coli (APC) and Axin form a scaffold that brings β-catenin in proximity with other proteins that mark β-catenin for proteasome-mediated destruction.
00231-4&id=gr1.jpg){kind=link}
00231-4&id=gr2.jpg){kind=link}
00231-4&id=gr3.jpg){kind=link}
00231-4&id=gr4.jpg){kind=link}
00231-4&id=gr5.jpg){kind=link}
00231-4&id=gr6.jpg){kind=link}
00231-4&id=gr7.jpg){kind=link}
00231-4&id=gr8.jpg){kind=link}
00231-4&id=gr9.jpg){kind=link}
00231-4&id=gr10.jpg){kind=link}
00231-4&id=gr11.jpg){kind=link}
00231-4&id=gr12.jpg){kind=link}