An investigation of the clinical impact and therapeutic relevance of a DNA damage immune response (DDIR) signature in patients with advanced gastroesophageal adenocarcinoma

Background An improved understanding of which gastroesophageal adenocarcinoma (GOA) patients respond to both chemotherapy and immune checkpoint inhibitors (ICI) is needed. We investigated the predictive role and underlying biology of a 44-gene DNA damage immune response (DDIR) signature in patients with advanced GOA. Materials and methods Transcriptional profiling was carried out on pretreatment tissue from 252 GOA patients treated with platinum-based chemotherapy (three dose levels) within the randomized phase III GO2 trial. Cross-validation was carried out in two independent GOA cohorts with transcriptional profiling, immune cell immunohistochemistry and epidermal growth factor receptor (EGFR) fluorescent in situ hybridization (FISH) (n = 430). Results In the GO2 trial, DDIR-positive tumours had a greater radiological response (51.7% versus 28.5%, P = 0.022) and improved overall survival in a dose-dependent manner (P = 0.028). DDIR positivity was associated with a pretreatment inflamed tumour microenvironment (TME) and increased expression of biomarkers associated with ICI response such as CD274 (programmed death-ligand 1, PD-L1) and a microsatellite instability RNA signature. Consensus pathway analysis identified EGFR as a potential key determinant of the DDIR signature. EGFR amplification was associated with DDIR negativity and an immune cold TME. Conclusions Our results indicate the importance of the GOA TME in chemotherapy response, its relationship to DNA damage repair and EGFR as a targetable driver of an immune cold TME. Chemotherapy-sensitive inflamed GOAs could benefit from ICI delivered in combination with standard chemotherapy. Combining EGFR inhibitors and ICIs warrants further investigation in patients with EGFR-amplified tumours.


INTRODUCTION
Gastroesophageal cancer accounts for w1.3 million annual deaths globally. 1The majority of patients have advanced disease at diagnosis 2 and median survival in unselected trial populations in this setting is less than a year. 3Although there are more biomarker-driven novel therapies being approved, including targeted therapies and immune checkpoint inhibitors (ICIs), cytotoxic chemotherapy remains an important part of clinical management and there is an ongoing need to identify biomarkers of treatment response.
The GO2 trial investigated chemotherapy dose deescalation in an older and/or frailer population with advanced gastroesophageal cancer (ISRCTN44687907).In GO2, reduced dose doublet oxaliplatin/capecitabine (OX) chemotherapy (60% of standard dose of oxaliplatin 130mg/m2 on day 1, capecitabine 625mg/m2 on days 1-21, on a 21 day cycle) had non-inferior progression-free survival (PFS) and overall survival (OS) with improved patient experience compared with standard dose. 4The use of different chemotherapy doses provided a unique translational opportunity to investigate dose impact and potential biomarkers of response.
The DNA-damage immune response (DDIR) signature is a 44-gene transcriptional signature based on the loss of the Fanconi anaemia/BRCA (FA/BRCA) DNA-damage response pathway.Developed in breast cancer, the DDIR signature identifies patients who respond well to DNA-damaging neoadjuvant chemotherapy. 5,6The signature can be expressed as a continuous score or dichotomized into positive or negative.DDIR-positive tumours (exhibiting defective DNA damage repair) are characterized by an inflammatory tumour microenvironment (TME), up-regulation of interferon signalling genes, high lymphocytic infiltration, 7,8 and enhanced signalling through the cGAS/STING pathway. 9n a subset of gastroesophageal adenocarcinoma (GOA) patients, chemotherapy promotes antitumour inflammation within the TME by reorganizing the T-cell compartment and inducing innate signalling pathways, including cGAS/STING in tumour cells, which is associated with chemotherapy response. 10Supporting this, DDIR-positive early-stage GOAs benefit more from neoadjuvant platinum-based chemotherapy with improved pathological response and survival. 7his has not been investigated in advanced-stage disease.
DNA-damaging chemotherapeutic agents, e.g.platinums, target vulnerabilities inherent in tumours with defective DNA damage repair machinery, leading to neoplastic cell death and improved outcomes for example in tumours with homologous recombination deficiency. 11Investigation in GOA indicates no association between homologous recombination deficiency and response to platinum-based chemotherapy, however, suggesting that defective DNA damage repair in tumours alone may have a limited impact on chemotherapy response in GOAs. 12iven these findings, and the emerging role of an inflamed TME in chemotherapy and ICI response, we hypothesized that the combination of defective DNA damage repair and an inflammatory TME, captured by the DDIR signature, could predict response and long-term survival to the DNA-damaging agent oxaliplatin within the GO2 trial population and provide further understanding of the biological basis of response to both chemotherapy and ICIs.

Patient samples
Formalin-fixed paraffin-embedded (FFPE) pre-chemotherapy tumour samples from 395 patients recruited to the GO2 trial 4 were obtained.Samples were registered within NHS Tayside [Research Ethics Committee (REC) approval 17/ ES/0130] and Grampian (REC 16/NS/0055) biorepositories.Only those with histologically confirmed adenocarcinoma and in whom RNA sequencing was successful were included in DDIR analysis.Radiological response was graded according to RECIST v1.1. 14or independent in silico validation, RNA sequencing from 306 oesophageal adenocarcinoma tumours was obtained from the Oesophageal Cancer Clinical and Molecular Stratification (OCCAMS) consortium.In addition, 124 pretreatment samples from patients with adenocarcinoma treated in NHS Grampian underwent EGFR FISH and immunohistochemistry (IHC) for CD8, CD4, FOXP3 and programmed death-ligand 1 (PD-L1).

Gene expression profiling
Biopsies were reviewed for pathological subtype before marking for macrodissection and samples containing at least 10% adenocarcinoma tissue by area were taken forward.Where tumour material was limited, endoscopic biopsy fragments from the same patient were pooled.Methodology for RNA extraction and analysis was carried out as previously described. 7Further details can be found in the Supplementary Materials, available at https://doi.org/10.1016/j.esmoop.2024.103450.

Microenvironment cell population analysis
The 'MCPcounter' (version MCPcounter_1.1.0)R package was downloaded from GitHub (https://github.com/ebecht/MCPcounter) and was used to generate microenvironment cell population (MCP) estimation scores for 10 stromal and immune cell infiltrates from the transcriptomic data of the cohorts. 15Estimates were compared between DDIR-positive and DDIR-negative to determine their stromal/immune content and the differences in cellular composition between the cancer types.MCP estimation scores were also generated according to EGFR FISH status.

EGFR fluorescence in situ hybridisation
EGFR FISH was carried out and scored using an established protocol 16 in NHS Grampian.Further details can be found in the Supplementary Materials, available at https://doi.org/10.1016/j.esmoop.2024.103450.

Statistical analysis
Statistical analyses were conducted according to prespecified statistical analysis plans that were agreed upon before the inspection of any DDIR-stratified outcome data.Further details can be found in the Supplementary Materials and Supplementary Table 2, available at https:// doi.org/10.1016/j.esmoop.2024.103450.
A comparison of baseline characteristics demonstrated that the DDIR-analyzed cohort was representative of the whole adenocarcinoma trial population (n ¼ 492) and a comparison between those patients with and without available RNA sequencing data revealed that there was no evidence of selection biases based on the GO2 stratification factors (Supplementary Table S3, available at https://doi.org/10.1016/j.esmoop.2024.103450)or difference in OS [hazard ratio (HR) 0.95, 95% confidence interval (CI) 0.76-1.17;P ¼ 0.6] (Supplementary Figure S3, available at https://doi.org/10.1016/j.esmoop.2024.103450).
A total of 31/252 (12.3%) patients were classified as DDIR-positive, and the proportion of patients across the dose levels A, B and C was 33.7%, 29.8% and 36.5%,respectively.The DDIR-positive population was significantly older, as previously reported in the neoadjuvant setting, and frailer (Table 1).
Outcomes according to DDIR status in the GO2 trial Patients were followed up for a mandated 12 months after the commencement of systemic therapy. 4Radiological response and survival were analyzed in the 243 of the 252 patients who received chemotherapy.
In the DDIR-negative population, there was no dose level relationship with survival and also no difference in quality of life between dose levels (Supplementary Figures S12 and  S13, available at https://doi.org/10.1016/j.esmoop.2024.103450), however, dose level A (100% OX) was associated with a poorer OTU (P < 0.001) (Supplementary Figure S14, available at https://doi.org/10.1016/j.esmoop.2024.103450).Given the observed higher response rate, longer OS in DDIR-positive patients receiving dose level A (100% OX) and the known biology of the DDIR signature in other tumour types, we proceeded to investigate the association between DDIR score and immune/stromal composition, using gene expression profiles and MCP analysis. 15e DDIR-positive TME reflects an immune-rich subtype Using MCP analysis, 15 we identified significant differences in several immune cell types according to DDIR status (Figure 2A).In addition, there were consistent correlations between DDIR scores and T-cell, B-cell, and monocytic immune lineages, confirming an increase in immune cell infiltration in DDIR-positive advanced GOA [Figure 2B.Pearson r; T cells ¼ 0.646 (P < 0.001), B lineage ¼ 0.4396 (P < 0.001), monocytic lineage ¼ 0.5657 (P < 0.001)].There was a strong correlation between the DDIR score and the cytotoxic T-lymphocyte score (Figure 2C).
Increasing DDIR score was also associated with an increase in CD274 (PD-L1) expression and microsatellite instability (MSI) signature score 20 (Figure 3A and B).The MSI signature was validated internally using IHC (Supplementary Figure S15, available at https://doi.org/10.1016/j.esmoop.2024.103450).DDIR-positive patients had significantly higher expression levels of both CD274 and the MSI signature than DDIR-negative patients.Together, this supports the hypothesis that DDIR-positive advanced GOAs may also be more sensitive to ICI.
Gene comparison analysis was carried out to identify shared genes between the significant predictive signatures (https:// bioinformatics.psb.ugent.be/webtools/Venn).No genes were shared between all five signatures (Supplementary Table S11, available at https://doi.org/10.1016/j.esmoop.2024.103450);however, two genes, CXCL10 and IDO1, were shared between three of the signatures.Both CXCL10 and IDO1 genes are significant contributors to the DDIR signature.CXCL9 and CXCL11, the other CXCR3-related chemokines and CCL5, the other chemokine associated with the cGAS-STING pathway, were shared by two signatures.
As anticipated, there was a very good correlation between the DDIR score and CXCL10 RNA expression (R ¼ 0.72, P < 0.001) (Figure 4A).High RNA expression of CXCL10 (defined as the top 25%) was associated with a survival advantage in the population as a whole (Figure 4B) and this benefit was maintained on Cox regression analysis; high expression of CXCL10 was associated with an OS benefit in those who received dose level A (HR 0.45; P ¼ 0.02) (Figure 4C).

Consensus pathway analysis and EGFR
We have observed a relationship between the DDIR signature, ORR and survival in advanced GOA patients treated with platinum-based chemotherapy.There is also an association of the DDIR signature with increased immune infiltrate and biomarkers of response to ICIs.This may indicate the potential importance of the DDIR signature in predicting clinical outcomes for patients treated with chemotherapy and ICIs which form the basis of first-line standard-of-care treatments.Therefore, we investigated the potential upstream targetable determinants of the DDIR signature using consensus pathway analysis (http://cpdb.molgen.mpg.de/).This analysis identified EGFR as a key hub (Supplementary Figure S21, available at https://doi.org/10.1016/j.esmoop.2024.103450).EGFR gene RNA expression had no association with ORR, PFS or OS in the GO2 cohort (Supplementary Figure S22, available at https://doi.org/10.1016/j.esmoop.2024.103450),however, it was negatively associated with DDIR score (Supplementary Figure S23, available at https:// doi.org/10.1016/j.esmoop.2024.103450).None of the tumours with EGFR expression above 6 fragments per kilobase of tanscript per million mapped reads were DDIRpositive.EGFR-amplified GOAs are known to have higher EGFR RNA expression levels and are also known to benefit from treatment with EGFR inhibitors. 26,27This may therefore suggest that the DDIR-negative tumours with high EGFR RNA expression are EGFR-driven and targetable.
Validation of EGFR status and immune cell phenotype by immunohistochemistry.To validate the association between EGFR amplification (by FISH) and immune cell infiltration (by IHC), 124 GOA FFPE pretreatment tumour specimens within a TMA were analyzed as previously published. 17,28The TMA was an external patient cohort obtained from NHS Grampian, Scotland.Two cores per patient, from the centre of the tumour, were taken.QuPath Image analysis 19 was carried out on sections from the same tumour blocks to investigate CD8, CD4 and FOXP3 infiltration (Supplementary Figure S1, available at https://doi.org/10.1016/j.esmoop.2024.103450).These markers were selected due to the relationship between T cells and response to ICI. 29,30 The sections were also scored manually for IHC PD-L1 combined positivity score (CPS).

DISCUSSION
Advanced GOA is associated with a very poor prognosis.There is a need to identify biomarkers of response and the underlying biology.In this study, we present a molecular analysis in samples from a completed randomized clinical trial and an investigation of underlying biology relevant to both chemotherapy and immunotherapy with our findings being validated in independent patient cohorts.
We demonstrate that the 44-gene DDIR signature, which captures a combination of defective DNA damage repair mechanisms and an inflammatory TME, is associated with a higher response rate to platinum-based chemotherapy and improved OS.While the increased response rate was observed across all dose levels from the GO2 trial, improved survival, potentially via stimulation of immune surveillance, appears to require the non-de-escalated chemotherapy dose (level A, 100% OX).A higher DDIR signature score was also associated with an inflamed TME and, like triplenegative breast and colorectal cancer, 8,31 increased expression of biomarkers of ICI response.In contrast, EGFR amplification was associated with the reduction of expression of the DDIR signature and an immune cold TME.
The DDIR signature has predictive value for response to DNA-damaging chemotherapy in breast cancer and oesophageal adenocarcinoma in the curative setting, 5,7 but not in advanced colorectal cancer. 8Our investigation is the first to determine the interaction of chemotherapy dose with DDIR and, in doing so, has provided novel insights for the application of DDIR as a predictive biomarker as well as the underlying therapeutically relevant tumour biology.
DDIR positivity was observed in 12.3% of the GO2 cohort.This was lower than the observed 24% in the curative setting. 7It was also lower than the rates observed in triplenegative breast cancer (62%), 6 ovarian cancer (30%) 32 and colorectal cancer (19%-35%). 8,33he lower rates of DDIR positivity seen in the advanced gastroesophageal setting compared with the neoadjuvant setting may be due to the older/frailer patient population in GO2 (i.e. a changing disease biology with age).It may also reflect the impact of a differing biology across stage, which would support recent data in oesophageal adenocarcinoma, suggesting differences in mutational signatures with stage. 34Interestingly, advanced-stage colorectal cancer patients also have lower rates of DDIR positivity than in the localized setting. 33here may also be a contribution of a greater benefit of DDIR-positive tumours following DNA-damaging systemic therapy in the neoadjuvant/adjuvant setting. 7Accordingly, the DDIR phenotype impact on the TME may produce an initial improved response to neoadjuvant and adjuvant chemotherapy, but also results in longer-term disease control and immune surveillance. 7Additionally, previous platinum chemotherapy may alter the biology of the tumour or select out subgroups, for example, DDIR-negative that are resistant to chemotherapy. 35Supporting this concept, the DEBIOC study in oesophageal adenocarcinoma found that the post-neoadjuvant therapy DDIR signature score was significantly reduced. 36ithin the GO2 cohort, DDIR-positive patients had a better response rate and a non-de-escalated dose of chemotherapy was associated with improved OS (HR, 0.23; 95% CI 0.06-0.85,P ¼ 0.028).This improved OS occurred despite these patients being older and frailer (assessed by the GO2 frailty score), which has clinical relevance as reduced-dose chemotherapy is now widely adopted in this population.Of note, older adults (aged >75) had improved survival, independent of frailty, which may again indicate a different tumour biology according to age.
Importantly, the higher chemotherapy dose did not have a negative impact on patient experience or quality of life in the DDIR-positive population, suggesting that, overall, it was tolerated as well as the lower dose.This may be explained by the increased response with the higher dose, resulting in reduced tumour burden and improved symptom control.Together, these would improve treatment tolerance in the population.In addition, as toxicity reporting is capturing disease-related symptoms as well as treatment-related toxicity, 37 the impact of the higher dose on experienced toxicity will be reduced.
The DNA damage-induced DDIR signature represents an inflamed baseline TME 5,9 associated with increased T-cell, B-cell and monocytic immune lineages in both breast and colorectal cancer; 8 this was tested in the GO2 population, which confirmed an increase in lymphocytic infiltration with an increasing DDIR score.This suggests that in this population, DNA-damaging chemotherapy induces a radiological response.For long-term disease control and thus improved survival, however, a higher dose of chemotherapy may be required to stimulate immune surveillance.This might explain the observed increased response rate to platinumbased chemotherapy in DDIR-positive GOAs across all dose levels from the GO2 trial, but improved survival only in those treated with non-de-escalated higher chemotherapy dose (level A, 100% OX).
Biologically, the DDIR effect in GOA, similar to breast cancer, appears to be driven by the chemokines CXCL10 and CCL5, the pro-inflammatory functions of which include T-cell recruitment and expansion. 38The relationship between CXCL10 and DDIR score is also observed in advanced colorectal cancer. 8The combined CXCL10/CCL5 signature performed as well as the DDIR signature in predicting TCGA T-cell signatures.There was also a good correlation between this combined signature and the DDIR score, and it was also prognostic (both of response rate and survival).Together this suggests CXCL10/CCL5 expression may warrant further investigation as a narrowed biomarker of DDIR status and chemosensitivity.A similar finding is observed in triple-negative breast cancer where CXCL10 expression is related to a favourable prognosis. 39Importantly, CXCL10 appears to be an important prognostic marker for response to ICIs, [40][41][42] including in advanced oesophageal adenocarcinoma. 43s mentioned already, CXCL10 expression is associated with improved response to ICI therapy.Other predictors of ICI response and improved outcome are PD-L1 and MSI, as well as the presence of tumour-infiltrating lymphocytes (TILs). 44In the GO2 population, both PD-L1 (CD274) and the MSI signature (which includes CXCL10) are expressed at significantly higher levels in DDIR-positive patients.CXCL10 had a good correlation to both PD-L1 (CD274) and MSI signature scores, however, they appear to represent distinct populations.Therefore, it could be inferred that DDIRpositive patients are most likely to benefit from ICI therapy.
Overexpression of EGFR and gene CNG detected by FISH was associated with a less inflamed and immunologically colder TME.This may be a result of the known correlation between immune cold CIN tumours and EGFR amplification or a direct impact of EGFR signaling. 45Our consensus pathway analysis demonstrated that EGFR was a key hub and driver of the DDIR signature supporting a direct role for EGFR signalling.
The potential immunosuppressive role of EGFR is supported by evidence in other tumour groups.In breast cancer, EGFR positivity has been associated with increased FOXP3þ regulatory T cells, 46 which are known to suppress antitumour immunity.In non-small-lung cancer, EGFR signalling, via interferon regulatory factor 1 (IRF1), reduces the expression of both CXCL10 and CCL5 while also increasing regulatory T-cell recruitment via CCL22. 47Therefore, the mechanism of an EGFR signalling-induced immune cold TME could be via alteration of the chemokine milieu.
8][49] This has not been investigated in advanced GOA and is the subject of ongoing research within our group.
The strengths of this study are that it is a clinical trial cohort and thus the clinical outcome data are reliable, and findings have been validated in independent cohorts.We present data from a large sample size which is unique in being from an older population which better represents the patients we see in clinical practice.
However, our study has several limitations.Firstly, there was an unexpectedly low prevalence of DDIR positivity, and therefore the survival findings should be interpreted with caution.In addition, the limited tumour tissue available on FFPE blocks resulted in reduced sample size for subsequent IHC analysis (e.g.HER2).The rates of PD-L1 positivity within the TMA were lower than expected; this may be a consequence of sampling during the creation of the TMA, with samples more likely to be taken from the centre of the tumour specimen or the result of age-related deglycosylation of the extracellular domain of PD-L1, 50 and indicates that caution needs to be taken in the interpretation of these particular results.
We must also acknowledge that despite the DDIR signature containing several features which are known to be prognostic, it had no impact on PFS or OS in the population as whole.Added to this, the clinical significance of improved response rate can be questioned as it is not a good surrogate for the main outcomes of PFS and OS.The improvement in response rate, however, does imply increased sensitivity to chemotherapy, and the lack of clear survival benefit findings may reflect the impact of different dose levels, an underpowered study or the impact of the treatment on an older frailer cohort of patients.As such the clinical utility of the DDIR signature may be in fitter patients and also in giving an insight into underlying biology.

Conclusions
In summary, our study shows that in advanced GOA, the DDIR signature can predict an improved response to oxaliplatin treatment.The OS benefit may require the standard, non-dose de-escalated chemotherapy regime.We have identified that the underlying biology of the DDIR signalling in GOA, similar to breast cancer, is associated with constitutive gene up-regulation of the chemokines CXCL10 and CCL5 and an inflamed TME.EGFR copy number gain and in particular amplification may have an inhibitory effect on this signalling; however, this needs further investigation.
This work also underscores the importance of the connection between DNA damage repair components and inflammation in the TME in determining GOA patient outcomes.Our data may provide rationale for the mechanistic investigation of the combination of ICI with EGFR inhibition in tumours with EGFR CNG as a means to enhance anticancer immune responses and improve the efficacy of immunotherapies.

Figure 1 .
Figure 1.KaplaneMeier curves stratified by DDIR status for (A) progression-free and (B) overall survival for 243 patients with advanced gastroesophageal adenocarcinoma treated with platinum-based chemotherapy in the GO2 trial and (C) progression-free and (D) overall survival for the subgroup of 82 patients treated with full dose platinum-based chemotherapy.CI, confidence interval; DDIR, DNA damage immune response; HR, hazard ratio; OS, overall survival; PFS, progression-free survival.

4 BFibroblastsFigure 2 .
Figure 2. (A) Microenvironment cell population (MCP) scores of the individual immune cell types according to DDIR status.(B) Increased immune infiltrates correlated highly with DDIR positivity.MCP scores of three immune infiltratesdT-cells (blue), B lineage (red), and monocytic lineage (green)dcorrelated with DDIR scores with a line of best fit for each immune infiltrate.The black dashed horizontal line denotes the DDIR positivity cut-off.(C) Cytotoxic lymphocyte MCP scores correlated with DDIR scores in the GO2 population.The black horizontal line denotes the cut-off for DDIR positivity.Boxplot denotes the distribution of values for DDIR-positive and -negative status in the cohort.DDIR, DNA damage immune response; NEG, DDIR-negative; NK, natural killer; POS, DDIR-positive.

Figure 3 .Volume 9 -Figure 4 .
Figure 3. (A) Correlation of CD274 RNA expression and DDIR Score in the GO2 GOA population.The black line denotes DDIR positivity.Boxplot denotes distribution of values for DDIR-positive and negative in the cohort.(B) Correlation of MSI Signature RNA expression score and DDIR Score in the GO2 GOA population.The black line denotes DDIR positivity.Boxplot denotes distribution of values for DDIR-positive and negative in the cohort.DDIR, DNA damage immune response; MSI, microsatellite instability.
and QuPath Version 0.3.2.Whole slide images (WSI) of immunostained TMA slides for CD4, CD8, FOXP3 and PD-L1 were imported.Further details can be found in the Supplementary Materials, available at https://doi.org/10.1016/j.esmoop.2024.103450.PD-L1 was scored manually by two independent observers, one of whom was an experienced gastrointestinal pathologist.Further details can be found in the Supplementary Materials, available at https://doi.org/10.

Table 1 .
Demographics of the GO2 adenocarcinoma population according to DDIR status.