Somatic mutations in DCC are associated with genomic instability and favourable outcomes in melanoma patients treated with immune checkpoint inhibitors

Abstract

Background

Deleted in colorectal cancer (DCC) encodes a transmembrane dependence receptor and is frequently mutated in melanoma. The associations of DCC mutation with chromosomal instability and immunotherapeutic efficacy in melanoma are largely uncharacterised.

Methods

We performed an integrated study based on biological experiments and multi-dimensional data types, including genomic, transcriptomic and clinical immune checkpoint blockade (ICB)-treated melanoma cohorts from public databases.

Results

DCC mutation was significantly correlated with the tumour mutational burden (TMB) in The Cancer Genome Atlas (TCGA), International Cancer Genome Consortium (ICGC) and ICB-treated melanoma cohorts. DCC expression levels were correlated with DNA damage response and repair (DDR) pathways responsive to irradiation (IR) in the Malme-3M and SK-MEL-2 cell lines. In the TCGA cohort, DCC-mutated samples presented more neoantigens, higher proportions of infiltrating antitumour immunocytes and lower proportions of infiltrating pro-tumour immunocytes than DCC wild-type samples. DCC-mutated samples were significantly enriched in activated immune response and DDR pathways. Furthermore, patients harbouring mutated DCC treated with ICB showed remarkable clinical benefits in terms of the response rate and overall survival.

Conclusions

Somatic mutations in DCC are associated with improved clinical outcomes in ICB-treated melanoma patients. Once further validated, the DCC mutational status can improve patient selection for clinical practice and future study enrolment.

Background

Recent clinical trials with immune checkpoint blockade (ICB) therapies have led to unprecedented advancements in patients with advanced melanoma [1, 2]. Unfortunately, only a minority subset of patients respond to ICB [3, 4]. The factors defining which patients will be responsive or resistant are not fully characterised [5]. Recent studies have proposed distinct biomarkers for the response to ICB, including the tumour mutational burden (TMB) and neoantigen load [6–8], DNA damage and genomic instability [9], the intensity of CD8⁺ T-cell infiltrates [10] and intratumoural programmed death ligand 1 (PDL1) expression [11]. Furthermore, these factors are functionally interrelated and work co-ordinately in individual tumour specimens [12]. This led us to explore other biomarkers that may affect two or more of the above factors to provide stronger predictive value for immunotherapeutic outcomes in melanoma.

Deleted in Colorectal Carcinoma (DCC) was found to be one of the most important tumour suppressors in colorectal cancer in the early 1990s [13, 14]. As a gene frequently silenced through allelic loss of heterozygosity (LOH) on chromosome 18q, DCC exhibits markedly reduced expression in more than 70% of colorectal cancer cases, as well as in many other neoplasms [15]. As a transmembrane receptor, namely, one of the dependence receptors, DCC plays a dual role in cell signalling depending on its ligand netrin-1. In the presence of this ligand, DCC activates classic signalling pathways implicated in cell survival, migration and differentiation. In the absence of this ligand, DCC does not remain inactive but rather transduces a pro-apoptotic signal [16–18]. These findings led to the hypotheses that DCC expression is a constraint for tumour progression and that DCC functions as a tumour suppressor gene [18].

More importantly, recent genomic data have highlighted that DCC is the third most frequently mutated gene in sun-exposed melanoma [19]. The high load of mutations in DCC, including inactivating mutations in particular, is associated with a high burden of damaging mutations [19]. It was also recently reported that genetic silencing of DCC is associated with increased tumour progression in conditional mouse melanoma models [20]. Many intensive studies have reported that melanoma possesses the outstanding characteristics of genomic heterogeneity and the highest TMB among many cancer types [21, 22]. A high TMB has been demonstrated to be a biomarker of responsiveness to immunotherapy in several tumour types [6]. In addition, ~15% of colorectal cancer cases display microsatellite instability (MSI) owing to the presence of germline mutations in DNA mismatch repair (MMR) genes (MLH1, MSH2, MSH6 or PMS2) [23]. Studies have shown that colorectal cancers with MMR deficiency are sensitive to programmed death receptor 1 (PD1) blockade [24]. Notably, while DCC has been extensively described as “deleted” in colorectal cancers and “frequently mutated” in melanoma, the biological function of DCC in cancer progression is complicated but largely unknown in terms of genomic instability.

Based on these observations and the fact that melanoma still represents a significant clinical challenge, we sought to investigate whether there are relationships between DCC and genomic instability that can be exploited to influence the tumour immune microenvironment and immunotherapeutic efficacy.

Methods

Genomic and clinical data

Somatic mutation data for the 467 melanoma samples in The Cancer Genome Atlas (TCGA) were downloaded from Genome Data Commons (https://portal.gdc.cancer.gov). We termed the TCGA cohort the discovery cohort. For the validation cohort, clinical and somatic mutation data for 183 samples were obtained from the International Cancer Genome Consortium (ICGC) (https://dcc.icgc.org/). Normalised RNA-sequencing data for 465 participants were obtained from the TCGA dataset (https://portal.gdc.cancer.gov). The genetic profile of mutation calls for 77 human melanoma cell lines, under the TCGA code skin cutaneous melanoma (SKCM), were downloaded from the Cancer Cell Line Encyclopedia (CCLE) (https://portals.broadinstitute.org/ccle) [25].

An ICB-treated cohort was analysed as an integrative molecular and clinical model for further validation. Somatic and clinical mutation data were obtained from six previous studies of melanoma patients undergoing immunotherapy [7, 8, 26–29]. In this cohort, Oncotator was used to annotate genomic point mutations and short nucleotide insertions/deletions (indels) with variant- and gene-centric information relevant to cancer researchers (https://software.broadinstitute.org/cancer/cga/oncotator) [30]. The clinically annotated patients with melanoma (n = 336) were treated with ICB agents (anti-PD1, anti-CTLA4 or a combination of both). Tumour response in these patients was defined by the RECIST v1.1 criteria. Patients with a complete response or partial response were grouped as responders, and those with other statuses, such as progressive disease and stable disease, were grouped as non-responders. The flow diagram of this study is depicted as Supplementary Fig. S1.

Cell culture

The human melanoma cell lines SK-MEL-2 and Malme-3M were obtained from the American Type Culture Collection (Manassas, VA, USA). SK-MEL-2 cells were cultured in minimum essential medium (MEM) supplemented with 20% foetal bovine serum (FBS), 1% MEM non-essential amino acid solution and 1 mM sodium pyruvate. Malme-3M cells were cultured in Iscove’s modified Dulbecco’s medium (IMDM) containing 20% FBS. Cells were cultured at 37 °C in a 5% CO₂ humidified incubator. Cells were seeded for at least 24 h before irradiation (IR) or transfection with siRNAs by Lipofectamine 2000 (Invitrogen, MA, USA). All the commercial reagents for cell culture were purchased from Gibco (Supplementary Table S1). The SK-MEL-2 and Malme-3M cell lines were authenticated by STR profiling and tested for bacterial, fungal and mycoplasma contamination.

siRNA transfection

Three different double-stranded DCC-targeting short interfering RNA (siRNA) oligonucleotides were processed to verify gene silencing efficiency by western blotting. A validated universal negative control (NC) was used as the control for transfection. Two siRNA oligonucleotides (DCC-siRNA-1 and DCC-siRNA-3) showed higher DCC protein knockdown efficiency in the malme-3M cell line and were used in experiments. To perform transfections, cells were seeded in a six-well cell culture plate at a density of 1–1.2 × 10⁵ cells (~50% confluence). The transfections were carried out using Lipofectamine 2000 transfection reagent (Invitrogen), Opti-MEM I reduced serum medium (Gibco, MA, USA) without antibiotics and siRNA at a final concentration of 50 or 60 nM. The incubation time for the oligonucleotide-Lipofectamine 2000 complexes was 20 min. Following transfection, the total incubation time of the transfected cells before IR treatment or harvest was 48 h. A human DCC targeting siRNA and NC-siRNA were purchased from RiboBio Company (Guangzhou, China). The target sequences are shown in Supplemental File 3.

Plasmid constructs and transfection

For the generation of cell lines with inducible high expression of netrin-1, the human NTN1 coding sequence was cloned into the pcDNA3.1-3xFlag-C plasmid. Plasmids were transiently transfected into the Malme-3M and SK-MEL-2 cell lines to force netrin-1 expression. The cell lines were further tested for netrin-1 expression by western blot analysis after transfection for 48 h. To perform transfections, cells were seeded in a six-well cell culture plate (Corning, NY, USA) at a density of ~4 × 10⁵ cells (~80% confluence). The transfections were carried out using Lipofectamine 3000 transfection reagent (Invitrogen), Opti-MEM I reduced serum medium without antibiotics and a total plasmid amount of 1 µg per well. The incubation time for the plasmid-Lipofectamine 3000 complexes was 20 min. Following transfection, the total incubation time of the transfected cells before IR treatment was 48 h. The target-NTN1 plasmid and NC plasmid were purchased from Youbio Company (Changsha, China). The target sequences are shown in Supplemental File 3.

Irradiation procedures

Irradiation (IR) of cell lines was performed using a linear accelerator of the Varian synergy platform system (Palo Alto, CA, USA; dose rate, 400 cGy/min). Cells were irradiated in cell culture plates or chambered slides (6 Gy or 10 Gy). At different time points after IR, the cells were subjected to western blot analysis, immunofluorescence staining and cell viability testing.

Western blot analysis

Whole-cell lysates were extracted using SDS lysis buffer supplemented with a complete protease inhibitor cocktail (Roche, Basel, Switzerland) and phosphatase inhibitor (Roche). Proteins were separated on 4–12% or 8–16% SurePAGE gels (GenScript, Nanjing, China) and transferred to polyvinylidene fluoride (PVDF) membranes (Millipore, MA, USA) at 250 mA for 1.5 or 2.5 h by wet transfer. Then, the PVDF membranes were blocked with 5% non-fat powdered milk in Tris-buffered solution (TBS) containing Tween-20 for over 1 h and then incubated with primary antibodies (Supplementary Table S2) overnight at 4 °C. The bound antibodies were detected using incubation with horseradish peroxidase (HRP)-conjugated secondary antibodies (Supplementary Table S2) in dilution buffer. Proteins were visualised by enhanced chemiluminescence according to the manufacturer’s instructions (Millipore) and imaged on a Tanon 5200 chemiluminescent imaging system.

Immunofluorescence staining

Single-cell suspensions of melanoma cells (4–6 × 10⁴/mL) were cultured in 4-well chambered slides (Thermo Scientific, MA, USA) for more than 24 h. At different time points after IR, non-adherent cells and the culture medium were removed by washing with Dulbecco’s phosphate-buffered saline (DPBS). The adherent cells on the slides were fixed with 4% paraformaldehyde for 30 min and permeabilized with 0.2% Triton X-100 for 15 min at room temperature. Non-specific binding was blocked for 1 h with 5% bovine serum albumin (MedChem Express, NJ, USA) in DPBS. Primary antibodies diluted with 5% BSA were added, and the chambered slides were incubated at 4 °C overnight. After washing three times with DPBS, the slides were incubated with an Alexa Fluor 594-conjugated secondary antibody at a 1:300 dilution in the dark. The slides were washed three times in the dark, mounted with Prolong Gold Antifade Mountant with DAPI reagent (Invitrogen), and viewed under a Zeiss Axio Imager and Z2 fluorescence microscope (Carl Zeiss, Oberkochen, Germany). The γH2AX and p-BRCA1 foci in every cell were counted by eye from the stored images.

Cell viability assay

Cell viability was typically assessed in a 96-well format with Cell Counting Kit-8 (CCK-8; MedChem Express). After adding 10 μL of CCK-8 solution to each well of a plate and incubating for 2 h, the absorbance at 450 nm was measured on a microplate reader (BioTek, VT, USA). The cell viability of different siRNA-transfected cell lines under IR conditions is reported as the percentage relative to the negative control.

DCC mutation and the tumour mutational burden

Genomic instability and a high TMB are both significantly correlated with variations in DNA damage response and repair (DDR) related genes [31]. Mutations in BRCA1/2 (OMIM 113705 and OMIM 600185, respectively), TP53 (OMIM 191170), POLE (OMIM 174762) and MMR deficiency increase mutation rates in the cancer genome [31]. In this project, we termed MMR genes [MLH1 (OMIM 120436), MSH2 (OMIM 609309), MSH6 (OMIM 600678) and PMS2 (OMIM 600259)] the “MMR signature”. The MMR signature was considered to be mutated significantly if the four genes contributed to more than one gene mutation. In addition to a univariate analysis of the association of DCC mutation with the TMB, a multivariate logistic regression model was used to analyse the association between DCC mutation and the TMB by including genomic instability-related genes and clinical variables as confounding factors. The TMB was defined as the log₂ transformation of mutation counts per megabase. A univariate clustering approach (Ckmeans.1d.dp algorithm) in the R package Ckmeans.1d.dp (version 4.2.2) [32] was applied to determine the optimal cut-off points for a high TMB vs. a low TMB followed by recently broadly used values, such as 17 mutation counts per megabase and the log₂ transformed level [4.087]).

Ultraviolet (UV) mutational signature extraction

We used the method proposed by Kim et al. [33] to extract a UV mutational signature from melanoma samples. In this method, Bayesian variant non-negative matrix factorisation (NMF) was applied to decompose mutation portrait matrix A with 96 base substitution categories into the 2 non-negative matrices W and H (i.e., A ≈ W × H), where W indicates the extracted mutational signatures and H represents the mutational activities of each mutational signature. All determined mutational signatures were then compared with the 30 well-annotated signatures in the COSMIC database (version 2) based on cosine similarity.

Neoantigens, tumour-infiltrating immune cells and immune checkpoint molecules

We have learned that neoantigens are recognised by the immune system and can be targeted to increase antitumour immunity [34]. For the TCGA melanoma cohort, tumour-specific neoantigen prediction data were available and obtained from Rooney et al. [35, 36]. The predicted neoantigen load was defined as the log₂ transformation of the total number of neoantigens per sample.

To illuminate the distinct immune infiltration levels of DCC wild-type vs. mutated subgroups, we evaluated the abundances of 28 tumour-infiltrated immune cell types, which were recently reported by Charoentong et al. [37]. The 28 immunocyte types were divided into three categories to analyse their distinct immune functions: antitumour, pro-tumour, and intermediate tumour immunocytes. The CIBERSORT algorithm with a leukocyte signature matrix (LM22) was also applied to deconvolve external datasets of variably purified leukocyte subsets [38]. This method contains 547 genes that distinguish 22 human haematopoietic cell types, including seven T-cell types, naive and memory B cells, plasma cells, natural killer (NK) cells and myeloid subsets [38].

The first generation of immune checkpoint inhibitors used in melanoma primarily include anti-CTLA4 and anti-PD1/PDL1 [39]. Other emerging immune targets with reported preclinical efficacy have progressed to active investigation in clinical trials. These targets include co-inhibitory and co-stimulatory markers of the innate and adaptive immune systems. For example, LAG3, TIM3, TIGIT and IDO1 are being tested in clinical trials and play crucial roles in immunotherapy [40]. Therefore, we further compared the differential expression of these genes according to the DCC mutational status.

Gene set enrichment analysis

Gene set enrichment analysis (GSEA) was implemented with fgsea package (version 1.10.0) [41]. Signalling pathways in the Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) terms were used as the background database [42]. The normalised enrichment score (NES) was the primary statistic for examining gene set enrichment results. The P value adjusted by the FDR estimated the statistical significance of the enrichment score. A gene set with an FDR ≤ 0.05 was considered to be significantly enriched.

Statistical analysis

R software (version 3.6.1), IBM SPSS (version 19.0) and GraphPad Prism (version 8) were used to perform relevant statistical analyses and draw figures. Mutational patterns were visualised via GenVisR package (version 1.16.0) [43]. To increase the robustness of statistical inferences, two independent datasets were utilised: the TCGA cohort and the ICGC melanoma cohort. Continuous data were analysed using the Wilcoxon rank-sum test (Mann–Whitney test) for group comparisons. The chi-squared test was used to determine whether there were significant differences between the expected frequencies and the observed frequencies for the TMB, responsive status and other more categorical risk factors. A multivariate logistic regression model was performed to adjust confounding factors.

For immunofluorescence staining, at each time point examined after IR treatment, two-way ANOVA was used to test whether the numbers of γH2AX and p-BRCA1 foci were different among the differently treated cell lines. For the cell viability assay, two-way ANOVA was also used to test the difference in cell proliferative ability among the Malme-3M cell lines transfected with different siRNAs.

Kaplan–Meier curves were generated according to the P values, hazard ratios (HRs) and 95% confidence intervals (CIs) determined by the log-rank test. Univariate analyses were performed first, including one factor at a time to examine prognostic effects. Multivariate analyses were performed using a Cox proportional hazards model to define the independent effects of prognostic variables. All P values less than 0.05 were considered to be statistically significant. *, ** and *** indicate P < 0.05, P < 0.01 and P < 0.001, respectively.

Results

DCC mutation in melanoma

We first determined the prevalence of somatic DCC mutations in melanoma samples in the TCGA, ICGC or CCLE dataset. Of the 467 patients in the TCGA cohort, 289 (61.9%) were male, and the median (IQR) age was 58 (24) years. DCC was one of the frequently mutated genes in melanoma, occurring in 130 of the 467 patients [27.8%, 84 of 289 (29.1%) males, 46 of 178 (25.8%) females] in the TCGA cohort, 59 of 183 patients (32.2%) in the ICGC cohort and 13 of 77 (16.9%) melanoma cell lines in the CCLE cohort. Most were missense mutations (83/130, 61.5%), whereas the remaining were truncating mutations (nonsense, splice site, or frameshift alterations). Melanoma samples with DCC mutations had a significantly higher TMB than samples with wild-type DCC (Fig. 1a). Among the melanoma samples with mutated DCC, 68 of 130 (52.3%) also harboured alterations in genes related to the maintenance of genomic integrity, DNA replication proofreading and MMR, such as BRCA1/2, POLE and MLH1 (Fig. 1b). The mutational associations between DCC and DNA damage repair-related genes in the ICGC cohort are shown in Supplementary Fig. S2. In addition, the difference in the TMB between the TCGA and ICGC melanoma cohorts was not statistically significant (median TMB: 3.78 vs. 4.27; P = 0.11; Fig. 2a).

Fig. 1. Mutation profile of DCC in relation to genes associated with genomic instability in the TCGA cohort containing 467 melanoma samples.

a Mutation rates across TCGA melanoma samples. Mutation rates per megabase for synonymous and non-synonymous mutations. b Gene mutation patterns. Frequencies of synonymous and non-synonymous mutations, gene mutation patterns across each sample, and distribution of tumour stage by patient sex.

Fig. 2. Association of DCC mutation with a higher tumour mutational burden.

a Mutation rates of melanoma samples in the TCGA and ICGC cohorts. b Mutation rates of melanoma samples stratified by the DCC mutational status. Data are presented as the medians with interquartile ranges; TCGA melanoma cohort, n = 467; two-tailed Mann–Whitney test. c Association of DCC mutation with a higher TMB by multivariate logistic regression of the TCGA cohort of 408 melanoma samples. Data were adjusted by age, gender, stage and mutations in the BRCA1/2, TP53, POLE, MMRs genes and UV signature. Square data markers indicate estimated odds ratios. Error bars represent 95% CIs. d Mutation rates of melanoma samples stratified by the DCC mutational status. Data are presented as the medians with interquartile ranges; ICGC melanoma cohort, n = 183, two-tailed Mann–Whitney test.

BRAF was one of the most frequently mutated genes in melanoma, occurring in 236 of the 467 patients [50.5%; 192 of 236 (39.0%) BRAF^V600E mutation] in the TCGA cohort. We further validated the association between the BRAF and DCC mutational statuses. However, no statistically significant association was observed between the BRAF mutational status and TMB (Supplementary Fig. S3a), or DCC mutational status (Supplementary Fig. S3b) in the TCGA cohort. This suggested that DCC mutation was independent of BRAF mutation.

Association between DCC mutation and the tumour mutational burden

In the TCGA cohort, melanoma patients with mutant DCC had a significantly higher TMB than those with wild-type DCC (median TMB: 4.84 vs. 3.33; P < 0.001; Fig. 2b). In addition, we found that BRCA1/2 (62 of 467 [13.3%]), TP53 (65 of 467 [13.9%]), POLE (45 of 467 [9.6%]), MMR genes (54 of 467 [11.6%] and the UV signature (417 of 467 [89.3%]) were frequently mutated (Supplementary Table S3). Mutations in these genes caused a significantly higher mutational burden (all P < 0.001; Supplementary Fig. S4a–e). To rule out the possibility that the higher TMB was generated by mutations in genome repair genes rather than the DCC mutation itself, we utilised a multivariate logistic regression model with mutations in DDR genes and clinical confounding factors taken into consideration. When the TMB data were suitably divided into high and low subgroups with a cut-off point of 4.22, the association of DCC mutation with a higher TMB was still statistically significant after adjusting for these confounding variables (OR: 5.239, 95% CI: 2.956–9.284; P < 0.001; Fig. 2c).

Of the 183 melanoma patients in the ICGC cohort, a significantly higher TMB was also observed in patients with mutant DCC (median TMB: 5.75 vs. 3.12; P < 0.001; Fig. 2d). MMR genes (total 22 of 160 [13.8%]), the UV signature (87 of 160 [54.4%]) and genomic integrity maintenance genes, including BRCA1/2 (27 of 160 [16.9%]), TP53 (24 of 160 [15%]), POLE (15 of 160 [9.4%]), were also frequently mutated in the ICGC cohort (Supplementary Table S4). Mutations in these genes were also associated with a significantly higher TMB (all P < 0.001; Supplementary Fig. S4f–j). A multivariate logistic regression model including these gene mutations and clinical variables was performed to control for confounders. The association between DCC mutation and a high TMB was still statistically significant in this model (OR: 28.736; 95% CI: 2.369–348.538; P = 0.008; Supplementary Fig. S5). In the cohort containing 77 melanoma cell lines, a significantly higher TMB was also observed in the cell lines with mutant DCC (median TMB: 4.70 vs. 4.03; P = 0.002; Supplementary Fig. S4k). The associations between mutations genomic instability genes mutation and the TMB were also figured out in Supplementary Fig. S4l–n. These results demonstrate that DCC mutation is significantly associated with high numbers of somatic mutations in melanoma.

DCC expression is associated with DNA damage repair signals

In the Pan-Cancer Atlas, melanoma exhibits the highest prevalence of somatic mutations, which are known to be generated by exposure to ultraviolet light or by abnormalities in DNA maintenance [22]. DDR genes play key roles in maintaining human genomic integrity [31]. Therefore, we detected whether the DCC protein could influence DDR pathways. First, we investigated the effects of IR treatment on several physiological mechanisms in two melanoma cell lines. SK-MEL-2 and Malme-3M cells expressed the wild-type DCC protein. The mutational data of these two cell lines have been previously reported in CCLE. Using γH2AX foci as a surrogate for DNA double-strand breaks (DSBs), we detected the rates of DSB repair in the two cell lines following IR. Cells were irradiated (6 Gy) and stained with antibodies against phosphorylated H2AX at three time points (1 h, 16 h, 24 h) over a 24-h period. Figure 3a shows the images of γH2AX foci in the two cell lines captured at different time points during the 24-h period following IR (6 Gy). The data in Fig. 3b indicate that SK-MEL-2 cell showed a slightly lower rate of DSB resolution and retained nearly 40.5% of γH2AX foci at 16 h following IR (Supplementary Fig. S6a, c, e). However, Malme-3M cells exhibited a strikingly higher rate of γH2AX foci resolution and eliminated nearly 100% of their foci by 16 h (Fig. 3a, b and Supplementary Fig. S6b, d, e). The data inspired us to detect the effect of irradiation on DDR biomarkers by western blot analysis, such as ATM, ATR, Chk1 and Chk2 in DNA damage response pathways and BRCA1, Rad50 and Rad51 in DNA damage repair pathways.

Fig. 3. Analysis of DDR activities response to irradiation in human melanoma cell lines.

a Cell lines were irradiated with 6 Gy and then fixed at various time points (1 h, 16 h, 24 h and 0 h as the untreated control) and probed with an anti-phosphorylated histone H2AX antibody that was detected with a secondary antibody for visualisation by fluorescence microscopy. Representative images of γH2AX foci in the nucleus in two DCC wild-type cell lines (SK-MEL-2 and Malmle-3M) at ×400 magnification. b Graphical representation of γH2AX foci in the two cell lines (Malmle-3M, blue line; SK-MEL-2, green line). Points, mean values of γH2AX nuclear events per cell (at least 100 cells) derived from at least 10 fields; bars, SD. Experiments in (a, b) were repeated twice independently with similar results. c Western blot analysis of the expression of the DDR-related proteins γH2AX, ATM, p-ATM, p-ATR, p-Chk1, p-Chk2, BRCA1, p-BRCA1, Rad50, Rad51 and DCC in SK-MEL-2, Malmle-3M, and Malmle-3M cell lines transfected with NC-siRNA, DCC-siRNA-1 or DCC-siRNA-3. The cell lines were irradiated with 6 Gy, and cell lysates were extracted at various time points (1 h, 16 h, 24 h and 0 h as the untreated control) after IR. d Verification of the DCC silencing efficiency in the transfected Malmle-3M cell line by western blotting. Mock: transfection reagent did not contain any siRNA, used to exclude possible effects of transfection reagents on cells; NC: a validated negative control siRNA; 1, 2 and 3: three DCC-targeting siRNA oligonucleotides (DCC-siRNA-1, DCC-siRNA-2 and DCC-siRNA-3, respectively). e Cell viability assay performed with DCC knockdown (red line and orange line) and wild-type DCC expression (blue line) cell lines. Cells were irradiated with 10 Gy, and cell viability relative to that of the untreated samples was measured at various time points (24 h, 48 h, 72 h and 96 h) following IR. Experiments were repeated three times independently with similar results. Points, mean value of cell viability of six replicates from one independent experiment; bars, SD; h hour.

Consistent with the IF results, compared with that at the control time point after 6 Gy IR (0 h), the expressions of γH2AX showed a marked increase at 1 h in the two cell lines, it showed decreased at 16 h and a return to the control level at 24 h in SK-MEL-2 cells, and a timely decrease to the control level at 16 h in Malme-3M cells (Fig. 3c). In SK-MEL-2 and Malme-3M cells, activation of ATM/ATR resulted in upregulation of the DNA damage response proteins of p-Chk1 and p-Chk2 to arrest cells at the G2/M transition point of the cell cycle; at the same time, upregulation of p-BRCA1 and Rad51 expression promoted DNA damage repair. The activated expression of most proteins, such as p-ATM and p-BRCA1, returned to control or lower levels within 24 h (Fig. 3c). These results demonstrate that in response to IR, SK-MEL-2 and especially Malme-3M cells can repair DSBs in an accurate and timely manner.

More significantly, after knocking down the expression of the DCC protein in Malme-3M cells with two different siRNA sequences (Fig. 3d), the expression patterns of DSB proteins (γH2AX) and cell-cycle checkpoint molecules (p-Chk1 and p-Chk2) were not altered within 24 h compared to those in the NC-siRNA cell line (Fig. 3c). However, the constitutive phosphorylation of ATM/ATR and BRCA1 remained stable up to 24 h, which was distinctly different from the phosphorylation patterns in the NC-siRNA cell line. The expression of Rad50 and Rad51 presented different tendencies between the DCC-siRNA cell lines and the NC-siRNA cell line within 24 h after IR (Fig. 3c). Moreover, in cell viability assessments, the NC-siRNA cell line and the DCC-siRNA cell lines showed differences in long-term cell viability following IR (10 Gy). In contrast to the NC-siRNA cell line, the DCC-siRNA cell lines exhibited less dramatic decreases in viability and remained ~70% viable at the 4th day after IR (Fig. 3e), which was consistent with the results of the western blot assay assessing cell-cycle checkpoint molecules (p-Chk1 and p-Chk2) (Fig. 3c). These data indicate that knockdown of DCC expression is associated with prolonged DNA repair kinetics but does not produce a pattern of IR-induced arrest at the G2/M cell-cycle checkpoint.

When we evaluated the expression of DCC in cell lines treated with IR (6 Gy), it was interesting to find that DCC expression was time-dependent within 24 h in the SK-MEL-2 and Malme-3M cell lines. DCC expression increased within 24 h after IR in cells transfected with DCC-siRNA1/3 (Fig. 3c). These results further indicated that the DCC protein may play important roles in the DNA damage repair signalling pathway activated in responsive to IR.

Knockdown of DCC expression inhibits the DNA damage repair signalling pathway

To ensure DCC silencing efficiency, we increased the final siRNA concentration from 50 to 60 nM. The efficiency of DCC expression knockdown in the Malme-3M and SK-MEL-2 cell lines transfected with DCC-siRNA is shown in Fig. 4 and Supplementary Fig. S7. We further found that after largely knocking down the expression of DCC, the expression of ATM, p-ATM, p-ATR, p-BRCA1, Rad50 and Rad51 was significantly lower level than that in the NC-siRNA cell line within 24 h after IR, and these effects were time-dependent within 24 h. The expression levels of MLH1, MSH2 and MSH6 in the knockdown cell lines were also slightly lower than those in the control cell line (Fig. 4). Furthermore, the expression patterns of γH2AX, p-Chk1 and p-Chk2 were almost the same between the knockdown and NC-siRNA cell lines, consistent with the results in Fig. 3c. We also examined IR-induced repair foci of containing p-BRCA1 in the Malme-3M cell lines transfected with NC-siRNA or DCC-siRNA1/3. Consistent with the results of the western blot assay, the number of p-BRCA1 foci tended to be higher in the NC-siRNA cell line than in the DCC-siRNA1/3 cell lines within 24 h after IR (Supplementary Fig. S8).

Fig. 4. DCC expression is associated with the DNA damage repair signalling.

Western blot analysis of the expression of the DDR-related proteins γH2AX, ATM, p-ATM, p-ATR, p-Chk1, p-Chk2, p-BRCA1, Rad50, Rad51, MLH1, MSH2, MSH6 and DCC in Malmle-3M and SK-MEL-2 cell lines transfected with NC-siRNA or DCC-siRNA-1. The cell lines were irradiated with 6 Gy, and cell lysates were extracted at various time points (1 h, 16 h, 24 h and 0 h as the untreated control) after IR. h hour.

Because DCC plays a dual role in cell signalling depending on its ligand netrin-1, we confirmed the expression of netrin-1 in the Malmle-3M and SK-MEL-2 cell lines by western blot analysis. The results showed that SK-MEL-2 and Malmle-3M cells exhibited almost undetectable expression of netrin-1 (Supplementary Fig. S9a), which was consistent with the report by Boussouar and colleague [20]. We then established Malme-3M and SK-MEL-2 cell lines with forced netrin-1 expression via transfection of the target-NTN1 plasmid to determine the effect on the DDR pathway. The results showed that DDR signalling was not inhibited by netrin-1 overexpression (Supplementary Fig. S9b). More interestingly, in the wild-type DCC expressing cell line, DCC-knockdown cell lines and netrin-1 overexpressing cell lines, the expression tendencies of DCC and DDR proteins, including ATM, p-ATM, p-ATR and p-BRCA1, remained consistent within 24 h after IR (Figs. 3c and 4 and Supplementary Figs. S7 and S9b). Taken together, these results demonstrate that DCC plays important roles in the DDR signalling pathway, which are independent of its ligand netrin-1. Knockdown of DCC expression inhibits the DNA damage repair pathway through upstream molecules.

Association between DCC mutation and the immune-active microenvironment

Given the knowledge that DNA damage and genomic instability have been found to shape the antitumour immune response [9], we confirmed the association between DCC mutation and the immune-active microenvironment based on the TCGA database. First, in silico predictions of major histocompatibility complex class I-binding mutated peptides showed that DCC-mutated melanomas presented more putative neoantigens (Fig. 5a, b). Second, tumour-infiltrating immune cell phenotype analysis illustrated that infiltrating activated CD4⁺ T cell and central memory CD4⁺ T cells were enriched at significantly higher proportions in patients with mutated DCC (P < 0.05; Fig. 5c). Among the pro-tumour immunocytes, infiltrating macrophages and type 2 helper T cells were present at significantly lower proportions in patients with mutated DCC (P < 0.01; Fig. 5c). The CIBERSORT method revealed that infiltrating of regulatory T cells were significantly enriched in patients in the DCC wild-type group (P < 0.05; Supplementary Fig. S10a). Third, the mRNA expression of CTLA4 was significantly upregulated in patients with mutated DCC (P < 0.05). Other checkpoints molecules, including LAG3, TIGIT and IDO1, exhibited trends towards higher median mRNA levels in the DCC-mutated group (Supplementary Fig. S10b).

Fig. 5. DCC mutation correlated with more predicted neoantigens and a more activated immune infiltrate environment.

a Predicted neoantigen load of 98 melanoma cases from the TCGA segregated into groups defined by the DCC mutational status. b Number of predicted SNV-derived neoantigens in DCC wild-type vs. DCC-mutated human melanomas from the TCGA cohort; n = 98, two-tailed Mann–Whitney test. c Distinct infiltration levels of immunocytes in DCC wild-type vs. mutated subgroups according to Charoentong and colleagues’ method. *P < 0.05; **P < 0.01. d, e Gene set enrichment plots of the top ten upregulated signalling pathways identified by KEGG (d) and GO (e) enrichment analyses of DCC-mutated melanoma samples in the TCGA database. NES indicates the normalised enrichment score.

Associations between DCC mutation and signalling pathways determined by GSEA

We next performed GSEA to determine the significant pathways resulting from DCC mutation based on the TCGA database. KEGG enrichment analysis of the genes in DCC-mutated samples showed highly significant enrichment for numerous immune cell processes, including allograft rejection, graft versus host disease and antigen processing and presentation (NES range: 1.87–2.02; all FDR < 0.05; Fig. 5d). Gene ontology (GO) enrichment analysis also showed highly significant enrichment for immune cell processes including antigen processing and presentation, adaptive immune response and immune system process (NES range: 2.36–2.47; all FDR < 0.05; Fig. 5e). We further performed GSEA to determine the enrichment patterns in high TMB tumours without DCC mutations to rule out the confounding effect of the TMB. The results showed that the KEGG and GO enrichment analyses did not detect any immune-active pathways in wild-type DCC samples with a high TMB (Supplementary Fig. S11). These data suggest that significant immune-active pathways in the tumour microenvironment maybe associated with DCC mutation.

More importantly, KEGG enrichment analysis also revealed prominent enrichment of signatures related to DNA replication, MMR and nucleotide excision repair (NES range: 2.00–1.76; all FDR < 0.05; Fig. 5d) in the DCC-mutated group. Furthermore, GO enrichment analysis revealed prominent enrichment of signatures related to chromosome segregation and DNA replication (NES range: 2.21–2.32; all FDR < 0.05; Fig. 5e). These results were consistent with our experimental results for human melanoma cell lines showing that knockdown of DCC expression might predominantly lead to the acceleration of the cell cycle and DNA replication in addition to inhibition of the DNA damage repair signalling pathway. These results suggest that unrepaired DNA damage that does not kill the cell by blocking replication would tend to cause replication errors and thus potentially increase the mutation probability.

DCC mutation is associated with a higher ICB response rate and overall survival rate

To determine the association between DCC mutation and the antitumour immunotherapy response, we analysed the ICB-treated cohort data for clinical verification. We first found that the presence of mutations in DCC and DDR-related genes was significantly associated with the TMB in the ICB cohort (all P < 0.001; Supplementary Fig. S12a–f and Supplementary Table S5). Patients with mutated DCC had a significantly higher TMB than those with wild-type DCC (median TMB: 4.814 vs. 2.828; OR: 5.683; 95% CI: 2.649–12.190; P < 0.001; Supplementary Fig. S12a, g), which coincided with our results for the TCGA and ICGC melanoma cohorts.

We next employed categorical data analysis to determine the associations between DCC mutation and clinical outcomes in the ICB-treated cohort. Age was categorised based on a cut-off of 60 years. WHO grades were categorised into M0 or M1a and M1b, M1c or NA groups. The ICB agents for patients were categorised into anti-CTLA4, anti-CTLA4 plus anti-PD1 and anti-PD1 groups. The ICB response for patients was also categorised into CR or PR (response; yes) and PD or SD (nonresponse; no) groups (Supplementary Table S6). Correlation analysis showed that mutated DCC was significantly associated with a higher response rate than wild-type DCC in ICB-treated patients (response rate: 45.3% vs. 25.9%; P = 0.002; Fig. 6a; Supplementary Table S6). To determine the confounding effect of other categorical risk factors on the impact of DCC mutation, logistic regression analysis was conducted by combining the DCC mutational status with multivariable risk factors. After the DCC mutational status was adjusted by the other factors, the odds ratios remained less than 1, suggesting no confounding effects between the DCC mutational status and other risk factors (Supplementary Fig. S13a).

Fig. 6. Association of DCC mutation with better ICB therapeutic efficacy in the ICB-treated melanoma cohort.

a Comparison of the response rate after treatment with ICB between DCC wild-type and DCC-mutated human melanoma cases; n = 336, chi-squared test. b Overall survival of 336 patients with melanoma treated with immune checkpoint inhibitor therapy. The Kaplan–Meier survival curves for melanoma patients stratified according to the DCC mutational status were generated with the GraphPad Prism programme. c Multivariate Cox regression analysis. Data were adjusted by age, sex, stage, ICB type and TMB. Square data markers indicate estimated hazard ratios. Error bars represent 95% CIs.

In addition, log-rank analysis showed that the DCC mutational status was significantly associated with the occurrence of longer overall survival (OS) in the DCC-mutated group than in the DCC wild-type group (median OS: 35.30 vs. 17.97 months; P < 0.001; HR = 0.538; Fig. 6b; Supplementary Table S7). Other categorical risk factors, such as WHO grade (M1c vs. M0 vs. M1a vs. M1b vs. NA), ICB agent (both vs. anti-CTLA4 vs. anti-PD1) and TMB (low vs. high) showed prognostic significance for OS (Supplementary Table S7). Cox regression analysis was then conducted by combining the DCC mutational status with multivariable risk factors to identify the confounding effects. After the DCC mutational status was adjusted by the other factors, the HRs remained less than 1, suggesting no confounding effects between the DCC mutational status and other risk factors (P = 0.002; HR = 0.536; Fig. 6c). Furthermore, we analysed the association of the DCC mutational status with OS in the TCGA and ICGC cohorts, in which the patients had not been treated with ICB. The data showed that the DCC mutational status was not significantly associated with OS (Supplementary Fig. S13b, c). These results demonstrate that DCC mutation is associated with a higher response rate and longer OS in melanoma patients given ICB therapy. The combination of the DCC mutational status with immunotherapy is a potent predictor of the clinical outcomes of melanoma patients.

Discussion

We first examined the somatic mutation profile of the 467 melanoma samples in the TCGA cohort and the 183 melanoma samples in the ICGC cohort. We observed that DCC was frequently mutated in melanoma, and more than 50% of DCC-mutated samples contained alterations in DDR genes in the TCGA cohort. The samples with mutated DCC were also characterised by a higher total mutational burden. The association of DCC mutation with the TMB was independent of the significant presence of mutations in BRCA1/2 and MMRs genes.

To further investigate the underlying interaction between DCC mutation and genomic instability, we focused on building an inactive mutated DCC cell line by silencing the expression of the DCC protein. We did not find any hot-spot mutation regions in the full-length sequence of DCC in the Catalogue of Somatic Mutations in Cancer (COSMIC) database (https://cancer.sanger.ac.uk/cosmic). A high mutational load in DCC was associated with high a burden of damaging mutations in melanoma samples, as previously reported [19]. Therefore, we knocked down the expression of the DCC protein to confirm its functions in DDR signalling. More interestingly, we provided evidence that knocking down DCC expression prolonged or inhibited DDR signals in the Malme-3M and SK-MEL-2 cell lines within 24 h after IR. First, the phosphorylation of ATM at site Ser1981 was significantly prolonged or inhibited within 24 h after IR in the DCC-knockdown cell lines, which is central to controlling the DNA damage response by phosphorylating hundreds of substrates in response to DNA damage. Similarly, the expression of Rad50 exhibited the same tendency as that of p-ATM. This finding suggests that Rad50 is regulated by p-ATM in DNA damage repair pathways. This may also be explained by ATM being recruited to chromatin in response to DSBs in a process that requires the MRE11-RAD50-NBS1 (MRN) complex [44]. This complex both recruits ATM to DNA lesions and stimulates ATM/ATR kinase activity at DNA lesions [44, 45]. Second, the phosphorylation of BRCA1 by ATM is critical for proper responses to IR-induced DSBs and plays a major role in DSB repair by homologous recombination (HR) [46, 47]. Our data showed that the phosphorylation of BRCA1 at site Ser1524 was significantly prolonged or inhibited, similar to the patterns of p-ATM. Therefore, we proposed that DCC may involve in the ATM-related pathway to affect DNA damage repair mediated the HR pathway. Notably, the kinase Chk2 activated by ATM and the kinase Chk1 activated by ATR did not exquisitely resist IR, indicating that the knockdown of DCC expression did not inhibit G2/M cell-cycle checkpoint control. Furthermore, in the cell viability assay, we observed that the DCC-siRNA cell lines retained higher cell viability than the DCC wild-type cell line. Cell viability assessments performed on the 4th day following IR reflected the combined effects of cell death and cell proliferation in the entire population. One historical discovery reported that the expression of DCC activated caspase-3 and programmed cell death, or induced G2/M cell-cycle arrest by inhibiting cdk1 activity in tumour cells [48]. Another possible reason could be that DCC plays a dual role depending on its ligand. In the absence of netrin-1, an active signalling pathway results in apoptotic cell death in the NC cell line [17, 49]. Thus, we reasonably inferred that inactivation of DCC may lead to the inhibition of the DNA damage repair signalling pathway but not DNA replication, which eventually results in many unrepaired DNA errors and thus a mutational burden.

By analysis of immune cell subsets in the tumour immune microenvironment, we found that compared with DCC wild-type samples, melanoma samples with mutated DCC were characterised by correlations with more predicted neoantigens, a higher proportion of infiltrating antitumour immunocytes and a lower proportion of infiltrating pro-tumour immunocytes. Gene set enrichment analyses indicated that the adaptive immune response was significantly related to DCC mutation in melanoma samples. These findings demonstrated that tumour samples with mutated DCC might present a co-stimulated immune microenvironment that influences genomic instability. It would be of considerable clinical interest to investigate whether the co-stimulatory factors have coordinate with the immune response in the context of treatment with ICB agents. Therefore, we analysed 336 melanoma patients treated with ICB for clinical validation. The results showed that DCC mutation was associated with a better prognosis and potentially enhanced response to immunotherapy.

There are several limitations to our study. First, DCC encodes a 1447-amino-acid cell membrane-spanning protein with four immunoglobulin-like and six fibronectin type III-like extracellular domains and a 325-amino-acid cytoplasmic domain [50]. Netrin-1 signalling via the DCC extracellular domain has been implicated in promoting axon guidance and cell migration in the developing nervous systems [51]. In this study, we report that DCC plays an important role in ATM/ATR-related DDR pathways independent of its ligand netrin-1. This finding raised some questions. How does DCC interact with the DDR signalling pathways? Historical studies have reported that the intracellular domain of the Frazzled/DCC receptor functions as a transcriptional activator between the cytoplasm and the nucleus to regulate midline axon guidance [52, 53]. Therefore, we wondered whether the DCC intracellular domain can function as a transcriptional activator of DDR molecules to affect genomic instability in the nucleus. Our observations have paved the way for further research on the biological functional impact of DCC on genomic instability, which may represent a very interesting mechanism to investigate in the future. Second, the tumour immune microenvironment assay was performed based on a public database. Significant infiltrated immune cells in the DCC-mutated samples were not largely determined based on the limited data available. This suggests that the underlying anti-cancer immune-related mechanisms in DCC-mutated samples remain unclear. Future experimental studies are required to validate the differences in immune phenotypes between the two genotypes. Third, to validate the association of DCC mutation with the response to immunotherapy, the somatic mutational data and clinical data of patients in the ICB therapy cohort were aggregated from six previous studies. Although we proceeded with normalised annotation, samples with a greater diversity of in race, ethnicity and tumour histology or future clinical trials are needed for validation. Moreover, efforts to develop drugs that inhibit the interaction of netrin-1 with its receptors are ongoing. Several preclinical studies have shown that candidate drugs interfering with netrin-1-receptor interactions, used either alone or in combination with conventional chemotherapies, markedly inhibit tumour growth and metastasis development [20, 54–56]. An anti-netrin-1 antibody is currently being evaluated in a Phase I clinical trial of all solid tumours (https://clinicaltrials.gov/ct2/show/NCT02977195). Future work should also investigate whether combining netrin-1 interference with immune checkpoint inhibitors deserves to be assessed in clinical trials of combination therapies.

In summary, our research first provided new insight into the mechanistic consequences of DCC in genomic instability; and evaluated the prognostic and predictive contributions of DCC mutation in ICB-treated patients. These results collectively reinforce the view that DCC mutation may be exploited for the development of more precise ICB strategies in melanoma.

Author contributions

YL had full access to perform experiments and statistical analysis in this study. LZ conceived and designed this study. All authors performed acquisition, analyses, or interpretation of data. YL drafted of the manuscript under close supervision and critical revision of LZ. All authors contributed paper writing and proofreading.

Funding

This work was supported by the National Natural Science Foundation of China (81372429 to Lujun Zhao).

Data availability

The datasets used in this study are available from the TCGA, ICGC and CCLE repositories. The ICB-treated melanoma datasets generated during the current study are available in six previous studies and their supplementary data files as described in the ‘Genomic and clinical data’ subsection. All data supporting the conclusions of this study have been included within the article and the Supplemental Data.

Code availability

The computer codes used to generate results that support the paper’s conclusions are available from the corresponding author upon reasonable request.

Materials availability

All materials supporting the conclusions of this study have been included in the article and the Supplemental Data.

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

Not applicable.

Consent to publish

Not applicable.

Supplementary information

The online version contains supplementary material available at 10.1038/s41416-022-01921-4.