Intrinsic activation of β-catenin signaling by CRISPR/Cas9-mediated exon skipping contributes to immune evasion in hepatocellular carcinoma

Comprehensive analysis of clinical samples has recently identified molecular and immunological classification of hepatocellular carcinoma (HCC), and the CTNNB1 (β-catenin)-mutated subtype exhibits distinctive characteristics of immunosuppressive tumor microenvironment. For clarifying the molecular mechanisms, we first established human and mouse HCC cells with exon 3 skipping of β-catenin, which promoted nuclear translocation and activated the Wnt/β-catenin signaling pathway, by using newly developed multiplex CRISPR/Cas9-based genome engineering system. Gene set enrichment analysis indicated downregulation of immune-associated gene sets in the HCC cells with activated β-catenin signaling. Comparative analysis of gene expression profiles between HCC cells harboring wild-type and exon 3 skipping β-catenin elucidated that the expression levels of four cytokines were commonly decreased in human and mouse β-catenin-mutated HCC cells. Public exome and transcriptome data of 373 human HCC samples showed significant downregulation of two candidate cytokine genes, CCL20 and CXCL2, in HCC tumors with β-catenin hotspot mutations. T cell killing assays and immunohistochemical analysis of grafted tumor tissues demonstrated that the mouse Ctnnb1Δex3 HCC cells evaded immunosurveillance. Taken together, this study discovered that cytokine controlled by β-catenin signaling activation could contribute to immune evasion, and provided novel insights into cancer immunotherapy for the β-catenin-mutated HCC subtype.


Downregulation of immune-related gene sets by exon 3 skipping of β-catenin in HCC.
We performed RNA-seq analysis of the HuH7-CTNNB1 Δex3 and 3H3-Ctnnb1 Δex3 cells, and identified LGR5, RNF43, AXIN2 and TMPRSS2 as commonly upregulated genes (log2 fold-change > 1.5 and P-value < 10 -10 ), which was consistent with the results of quantitative RT-PCR analysis. Gene set enrichment analysis (GSEA) of the HuH7-CTNNB1 Δex3 and 3H3-Ctnnb1 Δex3 cells revealed the close relationship between activation of the β-catenin signaling and downregulation of immune-associated gene sets (Fig. 3a). The HALLMARK TNFA SIGNALING VIA NFKB (M5890), GO HUMORAL IMMUNE RESPONSE (M13774) and GO REGULATION OF HUMORAL IMMUNE RESPONSE (M14968) gene sets were negatively enriched in both of the human and mouse HCC cells (Fig. 3b). These findings suggested that β-catenin signaling activation could contribute to immune evasion, and were consistent with previous studies of clinical specimens 7,13,16 .  . 4a and Supplementary  Table 2a,b), and CCL20, CSF1, CXCL2 and GDF15 were commonly suppressed at the mRNA level (Fig. 4b), which was confirmed by quantitative RT-PCR analysis (Fig. 4c). Knockdown of β-catenin recovered the expression levels of CCL20, CSF1 and GDF15 in the HuH7-CTNNB1 Δex3 cells, whereas recovering the expression level of Ccl20 in the 3H3-Ctnnb1 Δex3 cells ( Supplementary Fig. 3). We next compared the expression levels of the four cytokine genes between HCC samples with and without mutations in exon 3 of CTNNB1 by using the TCGA data sets (Fig. 5), and identified significant downregulation of CCL20 (fold-change: 0.164; P-value: 1.58 × 10 -9 ) and CXCL2 (fold-change: 0.467; P-value: 0.002).   Fig. 6a. Monocytes and T lymphocytes were obtained from C57BL6/J mice, and activated by conditioned media of each cell lines. By co-culture with immune cells, the number of the 3H3-Ctrl cells was notably decreased by more than 50%, while the number of the 3H3-Ctnnb1 Δex3 cells was not changed (Fig. 6b). Three dimensional co-culture system also showed the similar results (Fig. 6c). Knockdown of β-catenin in the 3H3-Ctnnb1 Δex3 cells strikingly restored the T cell killing effect (Fig. 6d). For examining the role of cytokines downregulated by β-catenin signaling activation, we generated the 3H3-Ctnnb1 Δex3 cells ectopically expressing mouse Ccl20 and Cxcl2. Both Ccl20 and Cxcl2 overexpression in the 3H3-Ctnnb1 Δex3 cells elicited 14.0% and 19.0% reduction of cancer cell viability compared with control cells, respectively, indicating that overexpression of the two candidate cytokines could show a rescue of T cell killing (Fig. 6e). We next evaluated tumorigenic activity of the 3H3-Ctnnb1 Δex3 and 3H3-Ctrl cells by subcutaneously injection into C57BL6/J mice. Glutamine synthetase expression, which is a surrogate marker for β-catenin signaling activation in HCC, was enhanced in the grafted tumors of the 3H3-Ctnnb1 Δex3 cells. There was no difference of Ki-67 staining between the tumor specimens of the 3H3-Ctrl and 3H3-Ctnnb1 Δex3 cells ( Supplementary Fig. 4). Remarkably, the tumor size of the 3H3-Ctnnb1 Δex3 cells was larger than that of the 3H3-Ctrl cells ( Supplementary  Fig. 5), and immunohistochemical analysis revealed decreased CD8+ T cell infiltration in the tumor tissue of -log 10 P-value (a) Volcano plots of differentially expressed cytokine genes between the HuH7 and 3H3 cells with and without β-catenin signaling activation. (b) Venn diagram of cytokine genes downregulated in the HuH7-CTNNB1 Δex3 and 3H3-Ctnnb1 Δex3 cells. Twenty and sixteen genes were extracted from 114 genes registered in the CYTOKINE ACTIVITY gene set (fold-change < 0.5 and P-value < 0.01). (c) Quantitative PCR analysis of four candidate cytokine genes downregulated by β-catenin signaling activation. Error bars are the mean ± SD. P-values were calculated by Welch's t test. *P < 0.05; **P < 0.01; ***P < 0.001. www.nature.com/scientificreports/ the 3H3-Ctnnb1 Δex3 cells (Fig. 7a). The tumors of the 3H3-Ctnnb1 Δex3 cells showed a dramatic decrease of Ccl20 and Cxcl2 expression (Fig. 7b).

Discussion
Although previous studies have examined the relationship between the Wnt/β-catenin signaling pathway and immune surveillance, they are artificial due to overexpression of mutated β-catenin, such as β-catenin S37F driven by SV40 promoter 14 , β-catenin S33A;S37A;T41A;S45A driven by tyrosinase promoter 15 , and β-catenin ΔN90 driven by EF1α promoter 17 . To overcome this limitation, we first tried to knock in a mutated sequence of CTNNB1 exon 3 to human HCC cells with the help of the CRISPR/Cas9 system by using single strand DNA or plasmid donor, but it is difficult to obtain CTNNB1-mutated subclones because homology-directed repair is less dominant than non-homologous end joining (NHEJ) during double-strand break repair. We then mimicked activation of the β-catenin signaling by skipping exon 3, that is, joining the ends of intron 2 and intron 3 simultaneously cleaved by the CRISPR/Cas9 system. For this purpose, we established the novel multiplex CRISPR/Cas9-mediated genome engineering system of lentiCas9-Blast and improved lentiGuide-Puro plasmids, although Kabadi et al. have already produced a lentivirus vector containing a Cas9 transcription cassette and multiple sgRNA transcription cassettes 20 . This is because a Golden Gate cloning method is complicated compared with a conventional cloning method, and because two-vector system is superior to all-in-one vector system in functional viral titer 18 . Thus, our vector system enabled exon skipping with ease and efficiency, and could expand to other models, such as EGFR Δex19 and ERBB2 Δex16 for activation of the EGFR signaling pathway, and POLD1 Δex10 and POLE Δex9 for attenuation of exonuclease activity. Immune checkpoint inhibitors (ICIs) including anti-PD-1 and anti-PD-L1 antibodies have provided a revolutionary approach to cancer therapy, and clinical trials of ICIs for various types of cancer are now ongoing and successful. In HCC, two anti-PD-1 antibodies nivolumab and pembrolizumab prolonged patient survival in phase II trials, however both monotherapies failed in phase III trials unfortunately. Harding et al. have revealed that CTNNB1-mutated HCC is more accumulated in the ICI-resistant group than in the ICI-sensitive group 21 , which is consistent with the important finding that CTNNB1 mutation is enriched in non-T cell-inflamed tumors insusceptible to ICI therapy 13 . Since the mutation rate of CTNNB1 gene is relatively higher in HCC than in other types of cancer ( Supplementary Fig. 6), clinical trials of ICIs should be conducted or subanalyzed for HCC with wild-type and mutated β-catenin separately. As described above, it is possible that the β-catenin signaling regulates not immune checkpoint molecules but cytokines for control of tumor immune microenvironment, such as upregulation of IL-10 14 and downregulation of CCL4 15 . Ruiz de Galarreta et al. demonstrated that antigen-expressing MYC;Trp53 −/− HCC evaded the immune system by decreasing CCL5 expression through activation of the β-catenin signaling pathway, and that CCL5 overexpression restored immunosurveillance in antigen-expressing MYC;CTNNB1 ΔN90 HCC 17 . In contrast, this study demonstrated that endogenous active form β-catenin downregulated immune-associated signaling pathways in both human and mouse HCC by bioinformatic analysis, and that tumor-intrinsic β-catenin activation suppressed T cell cytotoxicity through cytokine secretion by in vitro assays.
By comparing the present and previous studies 17 , CCL20 was commonly downregulated in HCC with β-catenin signaling activation, although there was no difference of CCL5 expression between human and mouse HCC cells with and without exon 3 skipping of β-catenin in this study. CCL20, alternatively named liver and activation-regulated chemokine (LARC), was originally discovered in the liver and strongly expressed in mononuclear cells near necrosis in the chronically inflamed liver and HCC. CCR6 is the selective receptor for CCL20, www.nature.com/scientificreports/ of neutrophils towards tumor tissues. Similarly to tumor-associated macrophages, tumor-associated neutrophils can be polarized into either an antitumoral (N1) or a protumoral (N2) phenotype; the N1 phenotype is induced by TGF-β blockade, and expresses immunoactivating cytokines and chemokines for killing cancer cells 23 . Thus, the β-catenin signaling pathway might suppress immune response through decrease of cytokine levels. We hypothesized two molecular mechanisms underlying downregulation of cytokines by β-catenin signaling activation. One is that active form β-catenin might indirectly suppress cytokine expression by controlling other transcription factors. Spranger et al. have previously reported that the Wnt/β-catenin signaling induces expression of a transcription repressor ATF3, and that ATF3 inhibits CCL4 expression in melanoma cells 15 , but the expression levels of transcription factors including ATF3 were not altered in human and mouse HCC cells with β-catenin signaling activation. The other is that β-catenin might change the epigenetic status of cytokine genes. Large-scale methylome data of HCC provided from the Cancer Genome Atlas Research Network clarified that downstream genes of the Wnt/β-catenin signaling pathway, such as LGR5, RNF43 and AXIN2, were demethylated at the promoter regions in CTNNB1-mutated HCC samples (Supplementary Table 3). However, the methylation levels of the promoter regions of CCL20 or CXCL2 were not different between HCC samples with and without CTNNB1 mutations, implying that other epigenetic systems such as histone modification might be involved in downregulation of cytokine genes 25 . In addition, since cross-regulation between the Wnt/β-catenin and NFκB signaling pathways has recently been identified in various types of cancer 26 , the similar molecular mechanism might also function in HCC cells.
In conclusion, this study enabled intrinsic β-catenin signaling activation by developing the highly efficient CRISPR/Cas9-based exon skipping system, and showed that it could contribute to immune evasion by suppressing immunoactivating cytokines including CCL20 and CXCL2. The CTNNB1-mutated HCC subtype accounts for approximately 30% of all cases ( Supplementary Fig. 6), but is refractory to ICI therapy. Since clinical trials evaluating recombinant cytokines as immunostimulants in cancer patients have recently been launched 24 , transarterial infusion of the candidate immunoactivating cytokines could also be effective to the subtype.

Methods
Ethics statement. The study was carried out in compliance with the ARRIVE guidelines. All methods were performed in accordance with relevant guidelines and regulations. All experimental protocols were approved by Institutional Review Board (G2018-132C5, Medical Research Ethics Committee for Genetic Research of Tokyo Cell culture. Human HCC cell line HuH7 was purchased from the American Type Culture Collection (Manassas, VA). Mouse cell line 3H3 was derived from HCC tumor grown in a C57BL/6J MC4R-KO mouse fed with high fat diet 19 . We examined mutations of Ctnnb1, Trp53, Braf and Hras, which are frequently observed in human HCC and carcinogen-induced mouse HCC, and only detected the Hras Q61L mutation in the 3H3 cells. They were cultured in RPMI-1640 and DMEM (Wako, Osaka, Japan) medium containing 10% fetal bovine serum (FBS), and 1% penicillin, streptomycin and amphotericin B (Wako), maintained in a humidified incubator at 37 °C in 5% CO 2 , and harvested with 0.05% trypsin-0.03% EDTA (Wako).
Exon 3 skipping of β-catenin by multiplex CRISPR/Cas9-based genome engineering system. To generate the backbone plasmid for the CRISPR/Cas9 system, the lentiGuide-Puro (Addgene #52963) was modified by inserting a KpnI site in front of the U6 promoter and replacing the HindIII site behind the sgRNA scaffold with an EcoRI site, named as LG-U6. The LG-H1 plasmid was also produced by replacing the U6 promoter with the H1 promoter in the LG-U6 plasmid. The LG-U6 and LG-H1 plasmid for expressing sgRNAs targeting intron 2 and intron 3 of β-catenin (sgRNA-in2 and sgRNA-in3) were constructed following the manufacture's manual (Supplementary Table 4). The H1-sgRNA-in3 sequence was tandemly cloned into the EcoRI site of the LG-U6-sgRNA-in2 plasmid (Fig. 1a). The HuH7 and 3H3 cells were sequentially infected with the lentiviral vectors for constitutively expressing SpCas9 (lentiCas9-Blast; Addgene #52962) and simultaneously expressing sgRNA-in2 and sgRNA-in3, and then treated with 10 μg/mL blasticidin and 10 μg/mL puromycin, respectively. The subclones with β-catenin alleles lacking exon 3 were isolated by limiting dilution. Western blotting. After whole cell lysates were collected by using ice-cold RIPA buffer (Thermo Fisher Scientific), 30 μg of protein from each sample was subjected to electrophoresis through 10% sodium dodecyl sulfate-polyacrylamide gels and transferred onto Immobilon polyvinyldifluoride membranes (Millipore, Bedford, MA). The membrane was blocked with 5% skimmed milk or bovine serum albumin for an hour at room temperature, adequately cut, and then incubated overnight at 4 °C with primary antibodies as follows;β-catenin (

Isolation of T cells.
Eight-week-old male C57BL/6J mice were euthanized, and spleens were resected and disrupted with a flat plunger tip of a 5 mL syringe. After hemolysis, whole splenocytes were incubated in a nylon wool fiber column to remove B lymphocytes for an hour at 37 °C. T lymphocytes were collected and cultured in RPMI-1640 medium supplemented with 10% FBS, 1% ITS supplement (Thermo Fisher Scientific), 100 U/mL murine IL-2 (Peprotech, Cranbury, NJ) and 10 ng/mL murine IL-7 (Peprotech).
Immune-cell preparation. Isolation of mouse bone marrow and differentiation of DCs was performed as previously described 27 . Briefly, eight-week-old male C57BL/6J mice were euthanized, and bone marrow was flushed out from femur and tibia by using a 1 mL syringe and a 27G needle. Bone marrow-derived monocytes (BMDMs) were washed, and then cultured in DC differentiation medium as follows; RPMI-1640, 10% FBS, 1% penicillin-streptomycin-amphotericin B, 20 ng/mL murine GM-CSF (Peprotech) and 5 ng/mL murine IL-4 (Peprotech). Six days after preculture, differentiated bone marrow-derived dendritic cells (BMDCs) were further cultured in conditioned medium collected from the 3H3-Ctrl cells or 3H3-CTNNB1 Δex3 cells for 24 h to stimulate with cancer antigens.
T cell killing assay. A day after T lymphocytes were co-cultured with BMDCs for priming, cells were plated at 5 × 10 3 cells per well in a 24-well tissue culture plate (for two-dimensional culture) or ultra-low attachment plate (for sphere culture) with primed immune cells for 48 h. Advanced DMEM/F12 (Thermo Fisher Scientific) with 0.5% B-27 supplement (Thermo Fisher Scientific), 20 ng/mL human EGF (Peprotech) and 1 μg/mL human FGF-basic (Peprotech) was used for sphere formation. To evaluate cytotoxic activity of immune cells in twodimensional culture, cell viability was estimated by using CellTiter-Glo 2.0 reagent (Promega) with FLUOstar OPTIMA-6 microplate reader (BMG Labtech) according to the manufacturer's instructions. For sphere culture, cancer cell area was measured by using ImageJ software.
Overexpression of cytokine genes. Entire coding sequence of mouse Ccl20 and Cxcl2 were amplified by using PrimeSTAR MAX DNA Polymerase (TaKaRa Bio), and cloned into the XhoI and NotI sites of the CSII-EF-MCS-IRES-Hygro plasmid 28 . Cells were infected with the lentiviral vectors, and then treated with 500 μg/ mL hygromycin.
Bioinformatic analysis. Gene set enrichment analysis was performed with the MSigDB gene sets. Public genome, methylome and transcriptome data of 373 HCC samples were provided from the Cancer Genome Atlas Research Network, and downloaded from the cBioPortal site. Genome data divided them into 77 and 296 tumors with and without CTNNB1 hotspot mutations.