Abstract
The DNA damage response (DDR) pathway regulates DNA repair and cell survival, and inactivating mutations in DDR genes can increase tumour mutational burden (TMB), a predictive biomarker of treatment benefit from anti-PD-1/PD-L1 immunotherapies. However, a better understanding of the relationship among specific DDR mutations, TMB and PD-L1 expression is needed to improve translational strategies. Here, we determined genomic alteration frequencies in selected DDR genes that are clinically actionable biomarkers and investigated their association with TMB and PD-L1 in bladder, colorectal, non-small cell lung, ovarian and prostate cancers using the FoundationInsights® web portal. Our results not only confirm known associations, such as mismatch repair and POLE gene mutations with high TMB, but also identify significant associations between mutations in the SWI/SNF chromatin remodelling genes ARID1A and SMARCA4 and high TMB in multiple tumour types. Mutations in the ATR gene were associated with high TMB in colorectal and prostate cancers; however, associations between individual DDR mutations and high PD-L1 expression were uncommon and tumour-type specific. Finally, we found that high TMB and high PD-L1 expression were poorly associated, emphasising their independence as predictive biomarkers for immune checkpoint inhibitor use.
Similar content being viewed by others
Introduction
Genomic instability has long been recognised as a cancer-enabling characteristic1,2. DNA damage response (DDR) pathways and repair mechanisms play a key role in preserving human genomic stability, and as they encompass multiple processes and components, their activation depends on the type of DNA damage and phase of the cell cycle3,4. Defective DNA repair due to aberrations in DDR components plays a critical role in tumour genomic instability, thereby resulting in cancer development and progression1,2,5.
Given their role in cancer, there is significant interest in DDR genes as prognostic markers, predictors of response to therapy and targets for therapy3,4. Genomic alterations (GA) in genes involved in the homologous recombination (HR) DNA damage repair pathway, such as the BRCA1/2 pathway, result in homologous repair deficiency and can be used to identify patients with tumours that may be sensitive to treatment with agents such as poly(ADP-ribose) polymerase (PARP) inhibitors in breast, ovarian (OC), prostate (PC) and pancreatic cancers. Additionally, these GAs may enhance tumour cell sensitivity to DNA-damaging chemotherapeutic drugs, such as platinum analogues and temozolomide6,7,8. GAs in HR genes, other than those in BRCA1/2, have been identified in many cancers, and preclinical studies suggest that specific HR GAs confer sensitivity to inhibitors of other DDR genes, such as ataxia-telangiectasia mutated (ATM), ataxia-telangiectasia and RAD3-related (ATR) and DNA-dependent protein kinase9. Furthermore, the tumour DNA repair landscape may play an important role in determining the response to immune checkpoint blockade10 because patients harbouring tumours with extensive DNA damage have been shown to respond better to immunotherapy11. These observations indicate the potential benefit of further exploring the role of DDR GAs and the use of recognised predictive biomarkers of immunotherapy to appropriately stratify patients based on tumour mutational burden (TMB), microsatellite instability (MSI) and PD-L1 expression12,13,14. Furthermore, available evidence suggests that combining DDR inhibitors with immunotherapy may have synergistic effects in tumours harbouring DDR GAs10,15.
Comprehensive genomic profiling (CGP) is being increasingly used for routine clinical management of patients with cancer16 and several studies have used CGP data to examine the potential of GAs and genomic instability signatures to inform targeted treatment in various cancers. For instance, in a study of approximately 3500 PC samples, 57% harboured GAs that are investigational biomarkers for targeted therapies17. In another study, 17% of ~17,500 gastrointestinal tumours had alterations in one or more of the 10 DDR genes that are known therapeutic targets18 and 21% of tumours with DDR GAs had high TMB, supporting the potential use of a combination of DDR inhibition with immunotherapy18. Furthermore, biallelic BRCA1/2 alterations were associated with increased genomic loss of heterozygosity (gLOH) in diverse solid tumours19, and biallelic HR GAs, other than those in BRCA1/2, were associated with increased genomic scar scores20. Thus, data from these and other studies point towards novel strategies for patient selection that will benefit from utilising genomics-based DDR mutational biomarkers for DDR inhibition and immunotherapy.
Although the consequences of DDR deficiency are becoming increasingly clear, significant gaps in knowledge remain, warranting further studies on the therapeutic impact of individual DDR GAs. As identifying associations between these DDR GAs and other biomarkers (e.g., TMB and PD-L1) and signatures of genomic instability (e.g., MSI and gLOH) is expected to have an increasingly important role in biomarker-driven precision oncology, we selected 35 DDR genes from the FoundationOne® gene panel in which somatic and germline mutations leading to genomic instability and increased mutation rates have been documented across a range of cancers1,21,22. These 35 genes can be grouped into eight DNA damage and response functions or pathways: base excision repair, damage sensor, Fanconi anaemia, HR, mismatch repair, nucleotide-excision repair, SWI/SNF chromatin remodelling and TP53 pathway (Table 1). Except for MDM2 and MDM4 from the TP53 pathway, which are oncogenes that can acquire copy number gains in certain cancers, the selected DDR genes are tumour suppressor genes that can undergo inactivating mutations or copy number loss in cancer. SWI/SNF chromatin remodelling genes play many roles, including regulation of chromatin assembly, transcription and DNA repair, and several of these genes are important tumour suppressor genes23. Here, we investigated the two most frequently mutated SWI/SNF genes, ARID1A and SMARCA4. Further, we focused our study on five tumour types (bladder cancer [BC], colorectal cancer [CRC], non-small cell lung cancer [NSCLC], OC and PC), to permit exploration of any association between DDR GAs and clinically relevant immunotherapy biomarkers (PD-L1, TMB) and identification of potential opportunities for combination therapy. Thus, this analysis was designed to provide comprehensive molecular profiling data on DDR deficiency in selected advanced solid tumours for which the immunotherapy is currently, (i) an established treatment option in most or all patients (BC, NSCLC), (ii) not generally indicated, but used in certain subgroups, such as MSI-high (MSI-H) or TMB-high (CRC, OC, PC).
Results
Clinical characteristics and genomic features of the patient population
To assess the prevalence of GAs in DDR genes, we analysed CGP results for tumours from 159,638 patients: 7803 with BC; 44,646 with CRC; 70,496 with NSCLC; 21,209 with OC; and 15,484 with PC (Table 2). The relative numbers and demographics were as expected for patients with solid tumours. TMB-high status was more prevalent in NSCLC (n = 24,510; 34.8%) and BC (n = 2611; 33.5%) than in CRC (n = 3876; 8.7%), PC (n = 748; 4.8%), or OC (n = 709; 3.3%; Table 2). Among patients with PD-L1 data (Table 2), the prevalence of PD-L1-high status was 52.8% in BC (n = 1344), 33.2% in OC (n = 2065), 32.4% in NSCLC (n = 9434), 13.2% in PC (n = 486) and 5.1% in CRC (n = 1656). The use of different cut-offs for PD-L1-high status in different tumour types precluded any comparison of these data.
DDR mutational landscape across tumour types
Analysis of the DDR mutational landscape showed that DDR GAs were common across tumour types but that they were non-uniformly distributed according to variant type and frequency (Fig. 1a–e, Table 3, Supplementary Fig. 1). As expected, TP53 (Fig. 1f, Table 3) was the most frequently altered DDR gene across all five tumour types (68.1%), followed by ARID1A (7.2%), ATM (4.6%), SMARCA4 (3.9%), MDM2 (3%), BRCA2 (3.3%), BRCA1 (2.4%), MUTYH (1.8%), CHEK2 (1.7%), ATR (1.2%), and MSH6 (1.1%). The TP53 pathway-related gene MDM2 was commonly amplified in BC (8.7%; copy number variant [CNV] 8.72%, short variant [SV] 0.14%) and NSCLC (4.40%; CNV 4.38%, SV 0.0014%). GAs in the other DDR genes were rare (≤1.0%).
In terms of GA based on DDR function, GAs in the following genes were relatively uncommon across tumour types: mismatch repair (<3.0%), Fanconi anaemia (<2%), and base– and nucleotide–excision repair (<2.5%). In contrast, HR genes were frequently altered in OC (BRCA1, 9.3%; BRCA2, 5.0%) and PC (BRCA2, 8.8%). Alterations in DNA damage sensor genes were also prevalent across tumour types, with ATM mutations found in ~5% of BC, NSCLC, CRC and PC specimens. Furthermore, GAs related to SWI/SNF chromatin remodelling were frequent in BC (ARID1A, 23.3%), NSCLC (ARID1A, 6.4%; SMARCA4, 7.0%), CRC (ARID1A, 7.2%) and OC (ARID1A, 8.3%) but were rare in PC ( < 2.5%).
SV mutations were the most common GA type across tumour types and genes (Fig. 1); specifically, SV mutations accounted for 47.1% of the TP53 GAs in BC, 69.7% in CRC, 60.3% in NSCLC, 73.5% in OC and 34.2% in PC (Fig. 1f). As expected, the TP53 pathway genes MDM2 and MDM4 almost exclusively harboured gene CNVs.
Distribution of germline and somatic status of SV mutations in DDR genes
SV mutations were assessed for germline– or somatic–only status. Across the tumour types analysed, SV mutations of somatic origin were predominant in TP53 (Fig. 2f; BC, 44.4%; CRC, 58.4%; NSCLC, 53.9%; OC, 62.2%; PC, 26.9%), the DNA damage sensor gene ATM (BC, 3.6%; CRC, 3.3%; NSCLC, 3.0%; OC, 1.1%; PC, 2.6%), SWI/SNF chromatin remodelling genes ARID1A (BC, 17.4%; CRC, 5.1%; NSCLC, 5.1%; OC, 5.1%; PC, 1.4%), SMARCA4 (BC, 2.0%; CRC, 0.8%; NSCLC, 4.9%; OC, 0.4%; PC, 0.2%) and in the mismatch repair gene MSH3 (BC, 0.3%; CRC, 1.6%; NSCLC, 0.4%; OC, 0.2%; PC, 0.6%; Supplementary Fig. 2). The proportion of germline SV mutations in BRCA1 and BRCA2 was markedly elevated in OC (BRCA1, 3.4%; BRCA2, 1.7%; Fig. 2d) and PC (BRCA2, 2.2%; Fig. 2e) compared to other cancers, as was the proportion of germline SV mutations in the HR genes RAD51C and RAD51D in OC (Supplementary Fig. 2). In addition, germline mutations were primarily identified (Fig. 2a–e) in MUTYH (BC, 1.3%; CRC, 1.5%; NSCLC, 1.3%; OC, 1.1%; PC, 1.2%) and CHEK2 (BC, 1.0%; CRC, 0.6%; NSCLC, 0.7%; OC, 0.6%; PC, 1.0%).
DDR GA type and relationship with TMB
We identified a sizeable proportion of tumours with multiple DDR GAs in the TMB-high tumour type BC and the highest proportion of tumours without DDR GAs in the TMB-low tumour type PC (Fig. 3 and Supplementary Fig. 3). This suggests that, in some cases, DDR GAs may be a consequence of a TMB-high phenotype. However, in some cases, such as the TMB-low tumour type OC, the highest proportion of tumours with one or more DDR GAs was observed, arguing against a simple relationship between TMB-high status and multiple GAs (Fig. 3 and Supplementary Fig. 3). However, for each tumour type, a weak but significant positive correlation was observed between the number of DDR GAs and TMB (BC: Spearman ρ = 0.308; CRC: Spearman ρ = 0.172; NSCLC: Spearman ρ = 0.358; OC: Spearman ρ = 0.261; PC: Spearman ρ = 0.206; all p < 2.2E–16).
Median TMB based on GA type (CNV, rearrangement and SV) was also assessed (Table 4). As noted above, SVs were the most common type of GA across tumour types, and this was also true for the tumours with available TMB data (range 50.7–84.1%). PC, which had the lowest median TMB overall (1.74) (Table 2), had the lowest proportion of SVs (50.7%) but a high proportion of rearrangements (4.8%) and CNVs (9.5%) compared with the other tumour types. However, median TMB was similar for PC with each of these GA types, as was the case for CRC. In contrast, OC, which also had a relatively low median TMB (2.5), had the highest proportion of SVs (84.1%) and CRC tumours with rearrangements had a higher median TMB than those with SV or CNVs. In BC and NSCLC, the median TMB for tumours with CNVs was lower than that for the other two GA types. Overall, no apparent relationship between the proportion of tumours with the different GA types and median TMB was observed.
Relationship between TMB and alterations in individual DDR genes
Individual GAs were plotted for each tumour type using Fisher’s exact test to estimate the odds ratios (ORs) and p values (Fig. 4a–e). As this analysis revealed that certain genes displayed particularly significant associations with TMB, we focused on genes with a log10(p value) of >20 and a log10(OR) of >0.5 and found significant associations between mutations in multiple mismatch repair genes and TMB in all tumour types (CRC: MLH1, MSH2, MSH6 and PMS2; NSCLC: MLH1 and MSH2; OC: MSH2 and MSH6; PC: MLH1, MSH2 and MSH6), except BC. We also found significant associations between mutations in the nucleotide excision repair gene POLE and TMB in CRC and OC. The SWI/SNF chromatin remodelling gene ARIDIA was significantly associated with TMB in CRC, OC and PC and SMARCA4 mutations were significantly associated with TMB in CRC. Finally, significant associations of ATR mutations, a DNA damage sensor gene, with TMB were observed in CRC and PC (Fig. 4b, e). In CRC, significant associations of HR (BRCA1, BRCA2, BARD1, PALB2 and BRIP1) and Fanconi anaemia (FANCA, FANCC and FANCG) genes with TMB were identified. This suggests a CRC-specific link between these repair processes and TMB.
Relationship between PD-L1 and alterations in individual DDR genes
We undertook an exploratory analysis of DDR GAs and PD-L1 protein expression using immunohistochemistry (IHC) data for PD-L1 obtained using the Dako 22C3 assay. For PD-L1, a tumour proportion scoring (TPS) threshold of 1% was used for all tumour types, except BC, wherein a combined positive score (CPS) threshold of 10 was used, and NSCLC, wherein a TPS threshold of 50% was used. PD-L1 data were available for 32.6%, 24.7%, 42.4%, 29.3% and 23.8% of the BC, CRC, NSCLC, OC and PC specimens, respectively.
In contrast to the interesting associations observed between certain DDR GAs and TMB, significant associations between DDR GAs and PD-L1-high status were uncommon (Fig. 5a–e). TP53 mutations were associated with PD-L1-high status in BC (OR 2.48, p = 5.48E–26), NSCLC (OR = 1.56, p = 7.2E–58) and OC (OR 1.65, p = 4.66E–13) but not in CRC or PC. The only other strong associations between genes and PD-L1-high status that were observed in these five tumour types were for ARID1A (OR 2.06, p = 9.43E–15) and MSH6 (OR 2.81, p = 1.86E–11) in CRC and MSH2 in PC (OR 3.24, p = 1.16E–4). No significant associations between genes grouped according to DDR pathway and PD-L1-high status were observed.
Overlap of TMB-high and PD-L1-high status and correlation with DDR GAs
In general, the reported correlation between TMB-high and PD-L1-high statuses is low. In this study, the proportion of TMB-high tumours that were also PD-L1-high differed across tumour types (Fig. 6). Overlap was the highest in the TMB-high BC (59.8% of TMB-high tumours were also PD-L1-high) and OC (45.2%), followed by NSCLC (35.2%), CRC (32.1%) and PC (30.1%). In tumours that were both TMB-high and PD-L1-high, a significant enrichment of GAs in individual DDR genes was observed in CRC (ARID1A: OR 4.39, p = 4.56E–17; TP53: OR 2.67, p = 3.01E–17; ATR: OR 7.55, p = 8.0E–4; ATM: OR 3.69, p = 5.95E–07; BRCA2: OR 4.10, p = 5.42E–05) and PC (TP53: OR 3.81, p = 4.07E–06). In NSCLC, we observed a negative relationship between TP53 mutation and TMB-high/PD-L1-high tumours (OR 0.90, p = 1.3E–3; Fig. 7). No significant relationships between individual DDR genes and TMB-high/PD-L1-high status were found in BC or OC.
Discussion
We used the FoundationOne® and FoundationOne®CDx assays to characterise GAs in 35 DDR genes for which somatic and germline mutations have been documented21,22 and to examine the association of these GAs with TMB and PD-L1 status in 159,638 solid tumours. Our data indicate that most tumours harbour GAs in at least one DDR gene (~60%–85% depending on tumour type) and that GAs are often seen in more than one DDR gene (~15–40% depending on tumour type). GAs in DDR genes varied according to type and frequency across the tumour types investigated, with TP53 being the most frequently altered across all five tumour types. We have previously described the molecular profiles of 269,371 clinically advanced and diverse malignancies, and while that data set included samples from 14 types of malignancies24, this report utilised data from five tumour types and represents an extended and in-depth analysis of the same.
GAs were relatively common in SWI/SNF chromatin remodelling genes ARID1A and SMARCA4, which are important tumour suppressor genes23, the TP53 pathway-related gene MDM2, the DNA damage sensors ATM and ATR, and BRCA1 and BRCA2, whereas GAs in other DDR genes were rare. Furthermore, while the overall frequencies of GAs in the genes studied were comparable to those reported previously17,18,19,25,26, higher rates of mutations in the SWI/SNF chromatin remodelling genes have been reported27. Although GAs in mismatch repair, Fanconi anaemia, base excision repair and nucleotide excision repair genes were relatively uncommon, HR genes were frequently altered in OC and PC. Mutations in MUTYH, a base excision repair gene involved in correcting DNA errors resulting from guanine oxidation, were also relatively common. In a retrospective analysis of next-generation sequencing data, pathogenic mutations in the nucleotide excision repair gene POLE were associated with a clinical benefit from immune checkpoint inhibition therapy28. In our study, significant associations were observed between POLE mutations and TMB in CRC and OC. SV mutations were the most common type of GA across all tumour types and genes.
The objective of this study was to identify potential and clinically relevant associations between DDR GAs and TMB or PD-L1 status. The incidence of TMB-high (defined as ≥10 mutations per megabase [mut/Mb]) and PD-L1-high (score cut-off = TPS 1 for CRC, OC and PC; CPS 10 for BC; and TPS 50 for NSCLC) tumours differed among cancer types. Although the cut-offs used to assess TMB and PD-L1 statuses vary in the literature, the observed incidences of TMB-high and PD-L1-high tumours were generally similar to those reported previously29,30,31,32,33,34,35,36,37. Our data showed that TMB-high and PD-L1-high statuses in each of the five tumour types were poorly correlated. Overlap was the highest in NSCLC and BC, and PD-L1-targeted therapies are approved for both38,39,40,41,42,43. The TMB cut-off of 10 mut/Mb was based on a previous study on NSCLC44; however, emerging evidence suggests that this cut-off maybe too broad and additional insights can be gained from examining different cut-offs for different tumour types45,46,47,48.
Immune checkpoint inhibitors, such as pembrolizumab, nivolumab, avelumab, atezolizumab and durvalumab, are not only approved for use in patients with tumours that express PD-L1 but are also indicated for certain solid tumour types without the need to assess PD-L1 expression38,39,40,41,42,43,49,50,51,52,53,54. Furthermore, while pembrolizumab and nivolumab are indicated for use in patients with MSI-H or mismatch repair-deficient solid tumours, both in the USA and in Europe39,40,49,50, pembrolizumab is indicated for use in patients with TMB-high solid tumours in the USA49. A pan-cancer study indicated that TMB-high status predicts response to anti-PD-1 therapy, independent of PD-L1 status, attesting to the independent predictive value of these biomarkers55,56 and suggesting that GAs associated with TMB-high status may have clinical relevance.
Our data indicate that while most cancers harbour GAs in at least one DDR gene (~60%–85% depending on tumour type), the frequency of cancers with DDR GAs does not correlate with median TMB (e.g., 84.8% of OCs had at least 1 DDR GA but a median TMB of 2.5 mut/Mb), indicating that the relationship between DDR GA and TMB is complex. For example, in the relationship between individual DDR GAs and TMB-high tumour status, CRC, OC and PC specimens were enriched with multiple DDR GAs, particularly in genes involved in nucleotide excision repair, mismatch repair, SWI/SNF chromatin remodelling and sensing DNA damage. We hypothesise that GAs in these genes are more likely to result in SVs and, eventually, in high TMB, whereas HR mutations result in larger rearrangements and CNVs, which are not captured by TMB. In contrast, BC and NSCLC showed high TMB but had either no enrichment of specific DDR GAs (BC) or only a few enriched DDR GAs (NSCLC). One explanation could be that TMB-high status in BC and NSCLC is the result of damage caused by exogenous carcinogens, such as tobacco smoke, a known aetiologic factor for both tumour types57,58, rather than being primarily driven by early clonal DDR defects.
Average CNV has previously been identified as an indicator of high TMB and immunotherapy efficacy in NSCLC, and we found a relatively high proportion of CNVs in NSCLC and BC, which are both tumour types with high median TMB. Notably, a similar association between TMB, CNVs and DNA methylation was observed in a study on Chinese patients, but it highlighted the need for further research to identify TMB thresholds for personalised lung cancer immunotherapy in different patient populations59. We also report a relatively high proportion of CNVs in PC, which has low median TMB and for which immunotherapy has shown limited efficacy. No association between CNVs and TMB was observed in CRC or OC, which are also tumour types for which immunotherapy has shown limited efficacy60,61,62. The significance of these findings and the role of TMB versus CNVs in determining sensitivity to immunotherapy warrants further investigation.
The DNA damage sensor genes ATM and ATR encode key kinases that help orchestrate DDR. Although both kinases are activated by DNA damage and replication stress, ATM acts as a regulator of cell cycle checkpoints and double-strand break repair, whereas ATR regulates cell cycle progression and promotes fork repair to overcome replication stress63. Of note, the high median TMB tumour types, NSCLC and BC, were particularly enriched with GAs in ATM. It has been shown that ATM-deficient cells have higher levels of type I interferon-stimulated genes (ISGs) and that ATR inhibition of these cells further increases ISG levels, resulting in dendritic cell activation64. Thus, tumours with ATM GAs may be sensitive to a synthetic lethality approach with ATR inhibitors, with or without the addition of immune checkpoint inhibitors.
We also found that ARID1A was commonly altered in BC, NSCLC, CRC and OC and that TMB-high tumours were enriched with ARID1A GAs. Further, as cancer cell models deficient in ARID1A are more sensitive to ATR inhibition through a synthetic lethality mechanism65, there is a rationale to combine immunotherapy with ATR inhibitors in TMB-high tumours harbouring ARID1A GAs, which is based on data linking ARID1A GAs and TMB-high status to ATR inhibitor and immune checkpoint inhibitor sensitivity, respectively.
The association of mutations in the SWI/SNF chromatin remodelling genes, ARID1A and SMARCA4, as well as ATR with high TMB represent a testable clinical hypothesis. The association of clinical responses to immune checkpoint inhibitors and SWI/SNF mutations may be further evaluated in other tumour types using larger clinico-genomic cohorts to determine whether these mutations could serve as additional biomarkers for immune checkpoint inhibitor therapy. The association between mutations in ATR with high TMB also imply that ATR mutations can lead to high TMB and a subsequent high neoantigen burden, warranting clinical investigation into the efficacy of combination treatment with ATR inhibitors and immune checkpoint inhibitors. However, it is possible that mutations in the SWI/SNF family, particularly ARID1A mutations that co-occur in high TMB tumours, are bystander mutations of the high TMB/MSI mutation phenotype, and thus, may not be causally related. It has been recently shown that MSI-H cancers often contain BRCA1/2 mutations; however, these mutations are predominantly monoallelic and are considered bystander alterations because they are not associated with gLOH and did not confer PARP sensitivity in several patients66. Arguing against the concept that SWI/SNF and ATR mutations are merely bystander mutations of MSI-H is the fact that all the mutations in this study were classified as functional or pathogenic mutations. Therefore, these DDR gene mutation associations warrant further investigations into their allelic status, functional relevance and potential clinical utility.
Our results also indicate that tumour type-dependent selection of the correct DDR gene for biomarker studies and the relevant DDR inhibitor and immune checkpoint inhibitor for combination trial strategies will be essential. For example, we show that GAs in genes such as ARID1A, SMARCA4 and ATR are associated with TMB-high status in a tumour type-specific manner and may thus represent additional biomarkers for immunotherapy in these tumour types. Furthermore, our findings suggest that the combination of immune checkpoint inhibitors and ATR inhibitors to which tumours with the aforementioned GAs are sensitive to in preclinical models and early clinical data, warrants further investigation.
To provide maximum sensitivity, a PD-L1 IHC score of TPS ≥ 1% was used as an exploratory cut-off for indications without a CDx claim in this study, namely, PC, OC and CRC. One of the limitations of using PD-L1 IHC TPS ≥ 1% as the cut-off in the tumour types without a CDx indication is that there is a lack of data on clinical outcomes in patients treated with immunotherapy using this PD-L1 cut-off; therefore, this analysis was considered exploratory67. In addition, TMB-high has been previously shown to be a biomarker of response in mismatch repair-deficient CRC, albeit at a higher TMB threshold. We acknowledge that the lack of mismatch repair status is a limitation of this study68.
In conclusion, these results, derived using clinically available and actionable assays, advance our understanding of the association between DDR GAs in cancer, and support the investigation of immune checkpoint inhibitors, not only in tumour type-specific and biomarker-defined subgroups (e.g., mutated ARID1A, SMARCA4, ATR) but also in combination with ATR inhibitors, in the clinical setting.
Methods
Comprehensive genomic profiling (CGP)
CGP of clinical solid tumour cases was performed in a Clinical Laboratory Improvement Amendments (CLIA)-certified and College of American Pathologists (CAP)-accredited laboratory using the FoundationOne® and FoundationOne®CDx assays (Foundation Medicine, Inc., Cambridge, MA, USA) as previously described69,70. The data were derived from de-identified and research-consented cases across five different solid tumour types, namely, BC, CRC, NSCLC, OC and PC, and were profiled between August 2014 and July 2021.
Formalin-fixed paraffin-embedded specimens containing at least 20% tumour cells and yielding a minimum of 50 ng of extracted DNA were subjected to hybridisation-captured, adapter ligation-based library preparation to identify GAs (SVs), CNVs and rearrangements in all coding exons (FoundationOne®CDx: n = 309 genes; FoundationOne®: n = 395 genes) and selected introns (FoundationOne®CDx: n = 36 genes; FoundationOne®: n = 31 genes) in cancer-associated genes. Further analysis was performed on cases with GAs in any of the 35 genes selected based on an adequate representation of the DDR pathways and coverage by the Foundation Medicine assay. Of these, 25 DDR-associated genes were baited across both FoundationOne®CDx and FoundationOne® assays, while 10 DDR-associated genes were baited only on the FoundationOne®CDx assay (Table 1). The predicted somatic or germline status of the GAs was determined using previously described methods19,71. TMB was calculated as the number of non-driver somatic coding mut/Mb of the genome sequenced; TMB-high was defined as ≥10 mut/Mb and TMB-low as <10 mut/Mb72. Briefly, for each sample, we created a genome-wide copy number profile with circular binary segmentation and a Gibbs sampling Markov Chain Monte Carlo algorithm, on the basis of log-ratios to a process-matched control and allele frequencies at over 3500 genome-wide single nucleotide polymorphisms70,71. Germline/somatic mutation calls were predicted without a matched normal. Germline/somatic calls for an alteration were estimated by modelling the alteration’s allele frequency, taking into account tumour content, tumour ploidy, and local copy number. In validation testing of 480 tumour-only sequencing calls against matched normal samples, the accuracy was 95% for somatic and 99% for germline calls.71
Only GAs (SVs, CNVs, or rearrangements) that are described as functional or pathogenic in literature and are listed in the Catalogue of Somatic Mutations in Cancer repository or had a likely functional status (e.g. frameshift or truncation events in tumour suppressor genes) were included in this study73. Variants of unknown significance were excluded.
PD-L1 IHC analysis was performed and interpreted by experienced board-certified pathologists according to the manufacturer’s instructions in a CLIA-certified and CAP-accredited laboratory (Foundation Medicine, Morrisville, North Carolina, USA) for a subset of specimens in this cohort using DAKO PD-L1 IHC 22C3 pharmDx74. A TPS (proportion of PD-L1-stained tumour cells) method was used to score all tumour types (score cut-off of 1 for CRC, OC and PC, and 50 for NSCLC), except for BC, in which the combined positive scoring (CPS; proportion of PD-L1-stained tumour, lymphocyte and macrophage cells) method was used (score cut-off of 10), as previously reported67.
The prevalence of detected GAs in the 35 DDR genes were categorised according to gene and disease, including base substitutions, small indels, CNVs and rearrangements. TMB scores and PD-L1 expression data were retrieved from the Foundation Medicine FoundationInsights® web platform. Data on tumour site and patient age at the time of CGP were extracted from accompanying pathology reports, clinical notes and test order requisition forms.
Statistical analyses
All statistical analyses were performed using the R (v4.0.3) and Python (v2.7.6) software packages. For each tumour type, Fisher’s exact test was used to estimate ORs and p values for the relationships between biomarkers (TMB or PD-L1) and DDR GAs. All p values were two-sided and multiple hypothesis testing correction was performed using the Benjamini–Hochberg procedure to calculate false discovery rate. Spearman’s correlation and univariate linear regression were used to determine the association between the number of DDR GAs (across 35 DDR genes, including multiple GAs per gene, if present) per specimen and the TMB score for each tumour type.
Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
Data availability
The authors declare that all relevant aggregate data supporting the findings of this study are available within the article and its supplementary information files. In accordance with the Health Insurance Portability and Accountability Act, we do not have IRB approval or patient consent to share individualised patient genomic data, which contains potentially identifying or sensitive patient information and cannot be reported in a public data repository. Foundation Medicine is committed to collaborative data analysis and has well-established and widely used mechanisms by which qualified researchers can query their core genomic database of >500,000 de-identified sequenced cancers. More information and mechanisms for data access can be obtained by contacting the corresponding author or the Foundation Medicine Data Governance Council at data.governance.council@foundationmedicine.com.
References
Hanahan, D. & Weinberg, R. A. Hallmarks of cancer: Prospective comprehensive genomic profiling of primary and metastatic prostate tumors. Cell 144, 646–674 (2011).
Hanahan, D. Hallmarks of cancer: new dimensions. Cancer Discov. 12, 31–46 (2022).
Curtin, N. J. DNA repair dysregulation from cancer driver to therapeutic target. Nat. Rev. Cancer 12, 801–817 (2012).
Pilié, P. G., Tang, C., Mills, G. B. & Yap, T. A. State-of-the-art strategies for targeting the DNA damage response in cancer. Nat. Rev. Clin. Oncol. 16, 81–104 (2019).
Hoppe, M. M., Sundar, R., Tan, D. S. P. & Jeyasekharan, A. D. Biomarkers for homologous recombination deficiency in cancer. J. Natl. Cancer Inst 110, 704–713 (2018).
Alnahhas, I. et al. Characterizing benefit from temozolomide in MGMT promoter unmethylated and methylated glioblastoma: a systematic review and meta-analysis. Neurooncol. Adv. 2, vdaa082 (2020).
LeVee, A. et al. Clinical utility of olaparib in the treatment of metastatic castration-resistant prostate cancer: a review of current evidence and patient selection. Onco. Targets Ther. 14, 4819–4832 (2021).
Wagener-Ryczek, S., Merkelbach-Bruse, S. & Siemanowski, J. Biomarkers for homologous recombination deficiency in cancer. J. Pers. Med 11, 612 (2021).
Trenner, A. & Sartori, A. Harnessing DNA double-strand break repair for cancer treatment. Front. Oncol. 9, 1388 (2019).
Mouw, K. W., Goldberg, M. S., Konstantinopoulos, P. A. & D’Andrea, A. D. DNA damage and repair biomarkers of immunotherapy response. Cancer Discov. 7, 675–693 (2017).
Wang, Z. et al. Comutations in DNA damage response pathways serve as potential biomarkers for immune checkpoint blockade. Cancer Res. 78, 6486–6496 (2018).
US Food & Drug Administration. APPROVAL FOR THE PD-L1 IHC 22C3 PHARMDX https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpma/pma.cfm?id=P150013 (2015).
Approval Order To Expand The Intended Use Of FoundationOne®CDx (F1CDx) To Include High Tumor Mutational Burden (TMB) At The Cut-Off Of Greater Than Or Equal To 10 Mutations Per Megabase (Mut/Mb) In Patients With Solid Tumors Who May Benefit From Treatment With KEYTRUDA® (Pembrolizumab). In: Administration UFD, ed. P170019 Supplement S0162020. (2020).
Approval Order To Expand The Intended Use Of FoundationOne®CDx (F1CDx) To Include A Companion Diagnostic (CDx) Indication For The Detection Of Microsatellite Instability High (MSI-H) Status In Patients With Solid Tumors Who May Benefit From Treatment With KEYTRUDA® (Pembrolizumab). In: Administration UFD, ed. P170019 Supplement S0292022. (2022).
Pilger, D., Seymour, L. W. & Jackson, S. P. Interfaces between cellular responses to DNA damage and cancer immunotherapy. Genes Dev. 35, 602–618 (2021).
Cobain, E. F. et al. Assessment of clinical benefit of integrative genomic profiling in advanced solid tumors. JAMA Oncol. 7, 525–533 (2021).
Chung, J. H. et al. Prospective comprehensive genomic profiling of primary and metastatic prostate tumors. JCO Precis. Oncol. 3, https://doi.org/10.1200/po.18.00283 (2019).
Parikh, A. R. et al. Analysis of DNA damage response gene alterations and tumor mutational burden across 17,486 tubular gastrointestinal carcinomas: implications for therapy. Oncologist 24, 1340–1347 (2019).
Sokol, E. S. et al. Pan-cancer analysis of BRCA1 and BRCA2 genomic alterations and their association with genomic instability as measured by genome-wide loss of heterozygosity. JCO Precis. Oncol. 4, 442–465 (2020).
Takamatsu, S. et al. “Utility of Homologous Recombination Deficiency Biomarkers Across Cancer Types.”. JCO Precis. Oncol. 6, e2200085 (2022).
Knijnenburg, T. A. et al. Genomic and molecular landscape of DNA damage repair deficiency across the cancer genome atlas. Cell Rep. 23, 239–254.e6 (2018).
Teo, M. Y. et al. Alterations in DNA damage response and repair genes as potential marker of clinical benefit from PD-1/PD-L1 blockade in advanced urothelial cancers. J. Clin. Oncol. 36, 1685–1694 (2018).
Ribeiro-Silva, C., Vermeulen, W. & Lans, H. SWI/SNF: complex complexes in genome stability and cancer. DNA Repair 77, 87–95 (2019).
Wang, D. et al. Abstract 2062: Comprehensive molecular profiling of DNA damage response (DDR) deficiencies in advanced solid tumors using a real-world genomic database. Cancer Res. 81, 2062 (2021).
Jiang, T., Chen, X., Su, C., Ren, S. & Zhou, C. Pan-cancer analysis of ARID1A Alterations as Biomarkers for Immunotherapy Outcomes. J. Cancer 11, 776–780 (2020).
Kurzrock, R. et al. Prevalence of ARID1A mutations in cell-free circulating tumor DNA in a cohort of 71,301 patients and association with driver co-alterations. Cancers (Basel) 14, 4281 (2022).
Wang, N., Qin, Y., Du, F., Wang, X. & Song, C. Prevalence of SWI/SNF genomic alterations in cancer and association with the response to immune checkpoint inhibitors: A systematic review and meta-analysis. Gene 834, 146638 (2022).
Garmezy, B. et al. Clinical and molecular characterization of POLE mutations as predictive biomarkers of response to immune checkpoint inhibitors in advanced cancers. JCO Precis. Oncol. 6, e2100267 (2022).
Sha, D. et al. Tumor mutational burden as a predictive biomarker in solid tumors. Cancer Discov. 10, 1808–1825 (2020).
Voutsadakis, I. A. Urothelial bladder carcinomas with high tumor mutation burden have a better prognosis and targetable molecular defects beyond immunotherapies. Curr. Oncol. 29, 1390–1407 (2022).
Huang, L.-J. et al. Prognostic significance of programmed cell death ligand 1 expression in patients with ovarian carcinoma: A systematic review and meta-analysis. Medicine (Baltimore) 97, e12858 (2018).
Huang, R. S. P. et al. Clinicopathologic and genomic characterization of PD-L1 positive urothelial carcinomas. Oncologist 26, 375–382 (2021).
Lin, G. et al. Prognostic significance of PD-L1 expression and tumor infiltrating lymphocyte in surgically resectable non-small cell lung cancer. Oncotarget 8, 83986–83994 (2017).
Palicelli, A. et al. What do we have to know about PD-L1 expression in prostate cancer? A systematic literature review. Part 1: Focus on immunohistochemical results with discussion of pre-analytical and interpretation variables. Cells 10, 3166 (2021).
Rosenbaum, M. W., Bledsoe, J. R., Morales-Oyarvide, V., Huynh, T. G. & Mino-Kenudson, M. PD-L1 expression in colorectal cancer is associated with microsatellite instability, BRAF mutation, medullary morphology and cytotoxic tumor-infiltrating lymphocytes. Mod. Pathol. 29, 1104–1112 (2016).
Srivastava, P. et al. PD-L1 Expression in colorectal carcinoma and its correlation with clinicopathological parameters, microsatellite instability and BRAF mutation. Indian J. Pathol. Microbiol. 64, 490–496 (2021).
Yarchoan, M. et al. PD-L1 expression and tumor mutational burden are independent biomarkers in most cancers. JCI Insight 4, e126908 (2019).
European Medicines Compendium. Summary of Product Characteristics: Bavencio 20 mg/mL concentrate for solution for infusion (accessed 23 May, 2022), https://www.medicines.org.uk/emc/product/8453/smpc.
European Medicines Compendium. Summary of Product Characteristics: KEYTRUDA 25 mg/mL concentrate for solution for infusion (accessed 23 May, 2022) https://www.medicines.org.uk/emc/product/2498.
European Medicines Compendium. Summary of Product Characteristics: OPDIVO 10 mg/mL concentrate for solution for infusion (accessed 23 May, 2022) https://www.medicines.org.uk/emc/product/6888.
European Medicines Compendium. Summary of Product Characteristics: LIBTAYO 350 mg concentrate for solution for infusion (accessed 23 May, 2022) https://www.medicines.org.uk/emc/product/10438.
European Medicines Compendium. Summary of Product Characteristics: Tecentriq 1,200 mg concentrate for solution for infusion (accessed 23 May, 2022) https://www.medicines.org.uk/emc/product/8442/smpc.
European Medicines Compendium. Summary of Product Characteristics: Imfinzi 50 mg/mL concentrate for solution for infusion (accessed 23 May, 2022) https://www.medicines.org.uk/emc/product/9495.
Marcus, L. et al. FDA approval summary: pembrolizumab for the treatment of tumor mutational burden-high solid tumors. Clin. Cancer Res. 27, 4685–4689 (2021).
Huang, R. S. P. et al. Durable responders in advanced NSCLC with elevated TMB and treated with 1L immune checkpoint inhibitor: a real-world outcomes analysis. J. Immunother. Cancer 11, e005801 (2023).
Ricciuti, B. et al. Association of high tumor mutation burden in non–small cell lung cancers with increased immune infiltration and improved clinical outcomes of PD-L1 blockade across PD-L1 expression levels. JAMA Oncol 8, 1160–1168 (2022).
Rousseau, B. et al. The spectrum of benefit from checkpoint blockade in hypermutated tumors. N. Eng. J. Med. 384, 1168–1170 (2021).
Samstein, R. M. et al. Tumor mutational load predicts survival after immunotherapy across multiple cancer types. Nat. Genet. 51, 202–206 (2019).
KEYTRUDA (pembrolizumab) injection, for intravenous use: Highlights of prescribing information (accessed 7 November, 2022) https://www.accessdata.fda.gov/drugsatfda_docs/label/2022/125514s110lbl.pdf.
OPVIDO (nivolumab) injection, for intravenous use: Highlights of prescribing information (accessed 7 November, 2022) https://www.accessdata.fda.gov/drugsatfda_docs/label/2021/125554s097lbl.pdf.
TECENTRIQ (atezolizumab) injection, for intravenous use: Highlights of prescribing information (accessed 7 November, 2022) https://www.accessdata.fda.gov/drugsatfda_docs/label/2016/761041lbl.pdf.
BAVENCIO (avelumab) injection, for intravenous use: Highlights of prescribing information (accessed 7 November, 2022) https://www.accessdata.fda.gov/drugsatfda_docs/label/2017/761049s000lbl.pdf.
IMFINZI (durvalumab) injection, for intravenous use: Highlights of prescribing information (accessed 7 November, 2022). https://www.accessdata.fda.gov/drugsatfda_docs/label/2017/761069s000lbl.pdf.
LIBTAYO (cemiplimab) injection, for intravenous use: Highlights of prescribing information (accessed 7 November, 2022) https://www.accessdata.fda.gov/drugsatfda_docs/label/2021/761097s007lbl.pdf.
Carbone, D. P. et al. First-line nivolumab in Stage IV or recurrent non-small-cell lung cancer. N. Engl. J. Med. 376, 2415–2426 (2017).
Hellmann, M. D. et al. Genomic features of response to combination immunotherapy in patients with advanced non-small-cell lung cancer. Cancer Cell 33, 843–852.e4 (2018).
Shirai, T. Etiology of bladder cancer. Semin. Urol. 11, 113–126 (1993).
Dela Cruz, C. S., Tanoue, L. T. & Matthay, R. A. Lung cancer: epidemiology, etiology, and prevention. Clin. Chest Med. 32, 605–644 (2011).
Cai, L. et al. Epigenetic alterations are associated with tumor mutation burden in non-small cell lung cancer. J. Immunother. Cancer 7, 198 (2019).
Santoni, M. et al. PD-L1 inhibitors for the treatment of prostate cancer. Curr. Drug Targets 21, 1558–1565 (2020).
Maiorano, B. A., Maiorano, M. F. P., Lorusso, D. & Maiello, E. Ovarian cancer in the era of immune checkpoint inhibitors: state of the art and future perspectives. Cancers (Basel) 13, 4438 (2021).
Gorzo, A. et al. Landscape of immunotherapy options for colorectal cancer: current knowledge and future perspectives beyond immune checkpoint blockade. Life (Basel) 12, 229 (2022).
Blackford, A. N. & Jackson, S. P. ATM, ATR, and DNA-PK: The trinity at the heart of the DNA damage response. Mol. Cell 66, 801–817 (2017).
Lopez-Pelaez, M. et al. Targeting DNA damage response components induces enhanced STING-dependent type-I IFN response in ATM deficient cancer cells and drives dendritic cell activation. Oncoimmunology 11, 2117321 (2022).
Williamson, C. T. et al. ATR inhibitors as a synthetic lethal therapy for tumours deficient in ARID1A. Nat. Commun. 7, 13837 (2016).
Sokol, E. S. et al. PARP inhibitor insensitivity to BRCA1/2 monoallelic mutations in microsatellite instability-high cancers. JCO Precis. Oncol. 6, e2100531 (2022).
Huang, R. S. P. et al. A pan-cancer analysis of PD-L1 immunohistochemistry and gene amplification, tumor mutation burden and microsatellite instability in 48,782 cases. Mod Pathol 34, 252–263 (2021).
Schrock, A. B. et al. Tumor mutational burden is predictive of response to immune checkpoint inhibitors in MSI-high metastatic colorectal cancer. Ann Oncol 30, 1096–1103 (2019).
Food and Drug Administration. FoundationOne®CDx: Summary of safety and effectiveness data (SSED). Food and Drug Administration. (accessed 7 November, 2022), https://www.accessdata.fda.gov/cdrh_docs/pdf17/P170019S006B.pdf.
Frampton, G. M. et al. Development and validation of a clinical cancer genomic profiling test based on massively parallel DNA sequencing. Nat Biotechnol 31, 1023–1031 (2013).
Sun, J. X. et al. A computational approach to distinguish somatic vs. germline origin of genomic alterations from deep sequencing of cancer specimens without a matched normal. PLOS Comp. Biol. 14, e1005965 (2018).
Chalmers, Z. R. et al. Analysis of 100,000 human cancer genomes reveals the landscape of tumor mutational burden. Genome Med. 9, 34 (2017).
Food and Drug Administration. PD-L1 IHC 22C3 pharmDx. Rx Only. SK006: 50 tests for use with Autostainer Link 48 (pdf). Food and Drug Administration, 2019 (cited 2022. FDA document on testing assats from FoundationOne) https://www.accessdata.fda.gov/cdrh_docs/pdf15/P150013S014C.pdf (2019)
Forbes, S. A. et al. COSMIC: somatic cancer genetics at high-resolution. Nucleic Acids Res. 45, D777–D783 (2017).
Acknowledgements
This study (including acquisition of data; analysis and interpretation of data; study supervision, conception and design; development of methodology; administrative, technical or material support) was sponsored by EMD Serono Research & Development Institute, Inc., Billerica, MA, USA, an affiliate of Merck KGaA (CrossRef Funder ID: 10.13039/100004755). Medical writing assistance was provided by Andy Noble of Bioscript Stirling Ltd, Macclesfield, UK and Anupama Singh of Merck Specialities Pvt. Ltd., Bengaluru, India, an affiliate of Merck KGaA. Editorial support was provided by Vasuprada Iyengar, also of Merck Specialities Pvt. Ltd. Writing and editorial support was funded by Merck (CrossRef Funder ID: 10.13039/100009945).
Author information
Authors and Affiliations
Contributions
D.W. conceptualised, designed and administrated the project, performed analysis and wrote the manuscript; B.E. provided advise during study planning, was involved in data interpretation and substantially contributed to composing the manuscript; K.M. contributed to data analysis, interpretation and manuscript writing; K.S. and M.M. contributed to data interpretation; I.G. interpreted the results and wrote the manuscript; J.S. provided scientific expertise and reviewed manuscript; G.L. conceived and designed the study and wrote the manuscript; Z.F. supervised the project, contributed to data interpretation and wrote the manuscript. All authors take accountability for the accuracy and integrity of the data and analyses reported here, and have read and approved the final manuscript.
Corresponding author
Ethics declarations
Competing interests
D.W., B.E., and Z.F. are employees of EMD Serono Research & Development Institute, Inc., Billerica, MA, USA, an affiliate of Merck KGaA and own stock in Merck. Ioannis Gounaris is an employee of Merck Serono Ltd., Feltham, UK, an affiliate of Merck KGaA and owns stocks in Merck. J.S. and G.L. are employees of Merck Healthcare KGaA, Darmstadt, Germany. K.M., K.S., and M.M. are employees of Foundation Medicine, Cambridge, MA, USA and stockholders of Roche Pharma AG.
Ethics
For Foundation Medicine (FMI) genomic analysis, approval for this study, including a waiver of informed consent and a HIPAA waiver of authorization, was obtained from the WCG Institutional Review Board (IRB; Protocol No. 20152817). The IRB granted a waiver of informed consent under 45 CFR § 46.116 based on review and determination that this research meets the following requirements: (i) the research involves no more than minimal risk to the subjects; (ii) the research could not practicably be carried out without the requested waiver; (iii) the waiver will not adversely affect the rights and welfare of the subjects.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Wang, D., Elenbaas, B., Murugesan, K. et al. Relationship among DDR gene mutations, TMB and PD-L1 in solid tumour genomes identified using clinically actionable biomarker assays. npj Precis. Onc. 7, 103 (2023). https://doi.org/10.1038/s41698-023-00442-4
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41698-023-00442-4