Purpose:

We sought to identify biomarkers that predict overall survival (OS) and response to immune checkpoint inhibitors (ICI) for patients with gastric cancer.

Experimental Design:

This was a retrospective study of multiple independent cohorts of patients with gastric cancer. The association between tumor ACTA2 expression and OS and ICI response were determined in patients whose tumors were analyzed with bulk mRNA sequencing. Single-cell RNA sequencing (scRNA-seq) and digital spatial profiling data were used to compare tumors from patients with gastric cancer who did and did not respond to ICI.

Results:

Increasing tumor ACTA2 expression was independently associated with worse OS in a 567-patient discovery cohort [HR, 1.28 per unit increase; 95% confidence interval (CI), 1.02–1.62]. This finding was validated in three independent cohorts (n = 974; HR, 1.52 per unit increase; 95% CI, 1.34–1.73). Of the 108 patients treated with ICI, 56% of patients with low ACTA2 expression responded to ICI versus 25% of patients with high ACTA2 expression (P = 0.004). In an analysis of a publicly available scRNA-seq dataset of 5 microsatellite instability-high patients treated with ICI, the patient who responded to ICI had lower tumor stromal ACTA2 expression than the 4 nonresponders. Digital spatial profiling of tumor samples from 4 ICI responders and 5 ICI nonresponders revealed that responders may have lower ACTA2 expression in α-SMA–positive cancer-associated fibroblasts (CAF) than nonresponders (median: 5.00 vs. 5.50).

Conclusions:

ACTA2 expression is associated with survival and ICI response in patients with gastric cancer. ACTA2 expression in CAFs, but not in other cellular compartments, appears to be associated with ICI response.

Translational Relevance

Biomarkers currently used to predict survival and therapy response in gastric cancer care are modest in efficacy. By analyzing data from multiple independent cohorts of patients with gastric cancer, we found that lower tumor ACTA2 expression was associated with improved overall survival and immune checkpoint inhibitor response. Single-cell RNA sequencing revealed that low ACTA2 expression in stromal cells, and not in tumor or immune cells, was associated with immune checkpoint inhibitor (ICI) response. Finally, analysis by digital spatial profiling suggested that ICI response was associated with low ACTA2 expression in α-SMA–positive fibroblasts. These findings show that ACTA2 expression is a promising biomarker to guide care for patients with gastric cancer.

Precision oncology care is based on biomarkers that predict survival and response. Currently, cancer patient survival is estimated using the tumor–node–metastasis staging system, which does not account for the molecular heterogeneity of cancer. In the case of gastric cancer, numerous groups have proposed genomic profiling schemes that are associated with survival (1, 2). We recently reported a 32-gene signature that is associated with overall survival (OS) and response to both chemotherapy and immune checkpoint inhibitors (ICI; ref. 3).

While prognostic biomarkers that estimate survival provide important information to set expectations for patients and physicians for the course of disease, these biomarkers often do not affect management decisions. In contrast, predictive biomarkers may be essential in deciding treatment strategy. For example, patients with gastric cancer with high HER2 expression benefit from the addition of trastuzumab to their therapy regimen (4).

ICIs have dramatically improved outcomes for patients with certain types of cancer (5, 6). However, recent clinical trials showed that the objective response rate is only 11% to 16% in patients with advanced gastric cancer (7–10). Thus, most patients with gastric cancer treated with ICI suffer treatment-related toxicities without any clinical benefit (11). Currently available biomarkers to predict which patients are likely to benefit from ICI therapy, such as the combined positivity score (CPS) that quantifies programmed death-ligand 1 (PD-L1) expression of tumor and immune cells, are limited in their utility (10). While gastric cancers that are microsatellite instability-high (MSI-H) or Epstein–Barr virus (EBV)-positive tend to respond well to ICI, these two subtypes comprise only a minority of cases (1, 12–17). In addition, while increased tumor mutational burden (TMB) is associated with response to ICI, its use as a predictive biomarker is minimal after adjusting for microsatellite stability status (18). Thus, novel predictive biomarkers are needed to improve the precision of ICI therapy in patients with gastric cancer.

In this study, we evaluated tumor ACTA2 expression as a prognostic biomarker for OS and determined its association with response to ICI. We also analyzed publicly available single-cell RNA sequencing (scRNA-seq) data and performed digital spatial profiling of tumors from patients with gastric cancer who did and did not respond to ICI.

Patient cohorts

The study was approved by the Institutional Review Board of the College of Medicine at Yonsei University and the Catholic University of Korea. We analyzed samples from 567 patients with gastric adenocarcinoma who underwent surgical resection at Severance Hospital, Yonsei University College of Medicine (Seoul, Korea) from 1999 to 2010. These gene expression profiles were generated using Illumina Human-6 V2 Expression BeadChips. Detailed information regarding data processing is available on the description page of each individual series of the Gene Expression Omnibus (GEO) databases. Raw microarray data were transformed to the log2 base scale and then were preprocessed by quantile normalization using the quantilenorm function in MATLAB R2018b (3). In addition, we analyzed data from an additional 17 patients from Severance Hospital, Yonsei University College of Medicine (Seoul, Korea) from 2014 to 2017, and 28 patients treated at Seoul St. Mary's Hospital (Seoul, Korea) from 2018 to 2020. We also examined data from cohorts previously published by The Cancer Genome Atlas (TCGA) project (1), Asian Cancer Research Group (ACRG; ref. 2), Sohn and colleagues (19), Kim and colleagues (17), and Chida and colleagues (20). For the scRNA-seq analysis, we examined data previously published by Kwon and colleagues (15). For the pooled data (n = 974), we used “ComBat” in “sva” package (R version 4.1.1) to remove possible batch effects in the expression values across the data sets.

Cox proportional hazards model

A multivariable Cox proportional hazards model was used to determine the association between ACTA2 expression and OS and contained the following additional covariates: sex, age (> 60 or ≤ 60), and stage. We used “coxph” in “survival” package in R (version 4.1.1) for our survival analyses.

scRNA-seq analysis

The scRNA-seq data of patients with MSI-H gastric cancer are retrieved from the European Nucleotide Archive (ENA; accession number: PRJEB40416). The authors kindly provided us with the same read count matrices as in the original study as well as an annotation of cell types. A normalized matrix is obtained by Seurat::NormalizeData() where the raw read count matrix is divided by the total count of each cell, multiplied by 10,000, and then transformed in log1p. A total of 5 patient samples collected prior to ICI treatment were included (one responder EP-72 and 4 nonresponders, EP-75, 76, 77, and 78). ACTA2 expression was compared between responders and nonresponders using the Wilcoxon rank-sum test.

Digital spatial profiling

Samples from 9 patients with gastric cancer treated with ICI at Seoul St. Mary's Hospital and Yonsei University were analyzed using the NanoString GeoMx Digital Spatial Profiler. Four patients were ICI responders and 5 were nonresponders. Formalin-fixed, paraffin-embedded slides were processed on the basis of standard GeoMx Digital Spatial Profiler instructions (MAN-10087–04). The slides were baked at 60°C for at least 1 hour, and then deparaffinized through Leica Biosystems BOND RX. Proteinase K was added on the tissue and then washed with buffers. The slides were incubated with the Cancer Transcriptase Atlas probe mix overnight. The slides were washed with buffer and stained with anti–α-SMA (Invitrogen, 53–9760–82), anti-CD45 (BioLegend, 12130230), and anti–pan-cytokeratin (PanCK; Novus, NBP2–33200DL594) antibodies for 2 hours. A total of 41 regions of interest (ROI) were placed on 20X fluorescent images scanned by GeoMx DSP. Each PanCK+, SMA+, and CD45+ regions in the ROI were segmented and a total of 123 area of interests (AOI) were generated [i.e., 3 AOIs (PanCK+, SMA+, and CD45+ AOI) within the ROI]. Oligoes from these regions were collected by DSP separately and transferred to 96-well plates. The oligoes then were uniquely indexed using Illumina's i5 x i7 dual-indexing system. Library purification was done following GeoMx DSP slide prep user manual (MAN-10087–04). Fastq files were further processed by DND system. DSP counts were further analyzed through GeoMx DSP data analysis software. Q3 dataset was generated by normalizing across all the AOIs using their 75th percentile of gene expression. The ACTA2 expression in the Q3 dataset is compared between responders and nonresponders using the Wilcoxon rank-sum test.

Data availability

Gene expression profiles of patients treated at Yonsei University can be found here: [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE183136] and [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE84437]. RNA sequencing data for patients treated with ICIs is available in the European Genome-Phenome Archive under the Dataset ID EGAD00001008091: [https://ega-archive.org/studies/EGAS00001005588]. The ACRG data file is available here: [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE62254]. Data from the Sohn and colleagues cohort is available here: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE13861] and [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE26942]. Data from the Kim and colleagues cohort is available here: [https://www.ebi.ac.uk/ena/browser/view/PRJEB25780]. Data from the Chida and colleagues cohort is available here: [https://www.ddbj.nig.ac.jp/dra/index.html]. Data from the Kwon and colleagues cohort is available here: [https://www.ebi.ac.uk/ena/browser/view/PRJEB40416?show=reads]. Bulk RNA sequencing data that support the findings of this study is available at cBioPortal [www.cbioportal.org; Stomach Adenocarcinoma (TCGA, Firehose Legacy)].

Code availability

The R scripts used for this study are available here: [https://github.com/hwanglab/Stromal_ACTA2_ICI_Analysis].

Tumor ACTA2 expression was associated with OS in patients with gastric cancer

The workflow to identify, test, and validate ACTA2 as a prognostic and predictive biomarker for patients with gastric cancer is presented in Fig. 1. We previously used microarray-based mRNA expression profiles from pre-treatment tumor samples of 567 patients who underwent resection of their gastric cancer at Severance Hospital, Yonsei University College of Medicine (Korea) and generated a 32-gene signature which stratified patients into 4 groups that were prognostic for OS (3). We noted that patients in the group with the worst prognosis had a significantly higher expression of ACTA2 compared with the remaining 3 groups (P < 0.0001, Supplementary Fig. S1).

Figure 1.

Study workflow.

Figure 1.

Study workflow.

Close modal

To determine if tumor ACTA2 expression independently predicts OS, we performed a multivariable analysis of the Yonsei cohort that included age, sex, tumor stage, and ACTA2 expression as covariates. The demographic, clinical, and pathologic characteristics of the Yonsei cohort are presented in Table 1. In this discovery cohort, we found that increasing ACTA2 expression was independently associated with worse OS [HR, 1.28; 95% confidence interval (CI), 1.02–1.62, per unit increase in ACTA2 expression; P = 0.04; Fig. 2A; Supplementary Table S1], along with age greater than 60 years (HR, 1.83; 95% CI, 1.43–2.35) and advanced stage (stage 3 HR, 3.42; 95% CI, 1.26–9.25; stage 4 HR, 19.04; 95% CI, 6.37–56.89).

Table 1.

Clinical and pathologic features of the discovery cohort.

CharacteristicsN (%)
N 567 
Age 
 ≤60 years 275 (48.5) 
 >60 years 292 (51.5) 
Sex 
 Male 386 (68.1) 
 Female 181 (31.9) 
Stage 
 I 21 (3.7) 
 II 147 (25.9) 
 III 379 (66.8) 
 IV 20 (3.5) 
Tumor location 
 Antrum 316 (55.7) 
 Body 182 (32.1) 
 Cardia 44 (7.8) 
 Whole 6 (1.1) 
 Missing 19 (3.4) 
Lauren type 
 Diffuse 198 (34.9) 
 Intestinal 194 (34.2) 
 Mixed 25 (4.4) 
 Other 149 (26.3) 
 Missing 1 (0.2) 
Lymphovascular invasion 
 Positive 268 (47.3) 
 Negative 294 (51.9) 
 Missing 5 (0.9) 
Perineural invasion 
 Positive 127 (22.4) 
 Negative 432 (76.2) 
 Missing 8 (1.4) 
EBV 
 Positive 19 (3.4) 
 Negative 210 (37.0) 
 Missing 338 (59.6) 
Microsatellite instability 
 Yes 21 (3.7) 
 No 156 (27.5) 
 Missing 390 (68.7) 
Chemotherapy receipt 
 Yes 453 (79.9) 
 No 110 (19.4) 
 Missing 4 (0.7) 
CharacteristicsN (%)
N 567 
Age 
 ≤60 years 275 (48.5) 
 >60 years 292 (51.5) 
Sex 
 Male 386 (68.1) 
 Female 181 (31.9) 
Stage 
 I 21 (3.7) 
 II 147 (25.9) 
 III 379 (66.8) 
 IV 20 (3.5) 
Tumor location 
 Antrum 316 (55.7) 
 Body 182 (32.1) 
 Cardia 44 (7.8) 
 Whole 6 (1.1) 
 Missing 19 (3.4) 
Lauren type 
 Diffuse 198 (34.9) 
 Intestinal 194 (34.2) 
 Mixed 25 (4.4) 
 Other 149 (26.3) 
 Missing 1 (0.2) 
Lymphovascular invasion 
 Positive 268 (47.3) 
 Negative 294 (51.9) 
 Missing 5 (0.9) 
Perineural invasion 
 Positive 127 (22.4) 
 Negative 432 (76.2) 
 Missing 8 (1.4) 
EBV 
 Positive 19 (3.4) 
 Negative 210 (37.0) 
 Missing 338 (59.6) 
Microsatellite instability 
 Yes 21 (3.7) 
 No 156 (27.5) 
 Missing 390 (68.7) 
Chemotherapy receipt 
 Yes 453 (79.9) 
 No 110 (19.4) 
 Missing 4 (0.7) 
Figure 2.

ACTA2 expression was independently associated with OS in patients with gastric cancer. A, Multivariable analysis of the Yonsei cohort. B, Multivariable analysis of the pooled cohort. C, Kaplan–Meier survival analysis of MSI-H patients from the pooled cohort stratified by ACTA2 expression. D, Kaplan–Meier survival analysis of MSS patients from the pooled cohort stratified by ACTA2 expression. ACTA2-Low was defined as patients with ACTA2 expression in the bottom quartile and ACTA2-High was defined as patients with ACTA2 expression in the top 3 quartiles. Kaplan–Meier curves were compared by the log-rank method.

Figure 2.

ACTA2 expression was independently associated with OS in patients with gastric cancer. A, Multivariable analysis of the Yonsei cohort. B, Multivariable analysis of the pooled cohort. C, Kaplan–Meier survival analysis of MSI-H patients from the pooled cohort stratified by ACTA2 expression. D, Kaplan–Meier survival analysis of MSS patients from the pooled cohort stratified by ACTA2 expression. ACTA2-Low was defined as patients with ACTA2 expression in the bottom quartile and ACTA2-High was defined as patients with ACTA2 expression in the top 3 quartiles. Kaplan–Meier curves were compared by the log-rank method.

Close modal

To validate ACTA2 as a prognostic biomarker, we pooled data from 3 large independent cohorts that were previously published. They include reports by the ACRG (n = 300, GEO: GSE62254; ref. 2), Sohn and colleagues (n = 267; GEO: GSE13861 and GSE26942; ref. 19), and TCGA (n = 407; ref. 1). We found that in this 974-patient cohort, every 1 unit increase of ACTA2 expression was associated with an increased HR for death of 1.52 (95% CI, 1.34–1.73; P < 0.0001; Fig. 2B; Supplementary Table S2–S3). Moreover, among all genes comprising our original 32-gene signature, ACTA2 expression was strongly associated with OS (Supplementary Table S4–S5).

Patients with MSI-H gastric cancers have better outcomes than patients with microsatellite stable (MSS) disease (21). We stratified patients in the pooled cohort by microsatellite stability status and found that MSS patients had higher ACTA2 expression (Supplementary Fig. S2). To determine if ACTA2 expression is prognostic for OS in both MSI-H and MSS gastric cancers, we repeated the multivariable analysis in our pooled cohort and found that ACTA2 expression was prognostic for OS after adjusting for microsatellite stability status (Supplementary Table S6). Then, we stratified pooled cohort patients by tumor microsatellite stability status and performed a Kaplan–Meier survival analysis. Amongst the MSI-H patients, we defined patients in the top 3 expressing quartiles as high expressors and patients in the bottom expressing quartile as low expressors. We found that MSI-H patients with low ACTA2 expression had improved OS compared with patients with high ACTA2 expression (median OS: not reached vs. 77.2 months, P = 0.013; Fig. 2C). We also divided MSS patients into quartiles based on ACTA2 expression and found that lowest quartile ACTA2 expression had improved OS compared with patients with high ACTA2 expression (median OS: not reached vs. 32.0 months, P < 0.0001; Fig. 2D). We then repeated our analysis using an additional survival endpoint, recurrence-free survival (RFS), and found that ACTA2 was independently associated with RFS in our pooled cohort (Supplementary Table S7). These data confirmed that tumor ACTA2 expression was a robust prognosticator of survival in patients with gastric cancer.

Tumor ACTA2 expression was associated with response to ICIs

We previously found that the 4 molecular subtypes established by the 32-gene signature predicted response to ICI (3). Tumors in one of the ICI nonresponsive groups were enriched for tumors that had high ACTA2 expression. We next investigated whether tumor ACTA2 expression was associated with response to immune checkpoint blockade. We established a 108-patient cohort of patients with advanced gastric cancer who were treated with ICI, which included 45 patients published by Kim and colleagues (European Nucleotide Archive: PRJEB25780; ref. 17), 18 patients published by Chida and colleagues (Sequence Read Archive of DNA DataBank of Japan: DRA013565; ref. 20), and 45 patients treated at our institutions (3). RECIST analysis was used to categorize patients either as responders if they had a complete or partial radiographic response or nonresponders if they had stable or progressive disease. Tumors from these patients were analyzed with bulk RNA sequencing and the patients were stratified on the basis of ACTA2 expression. Again, we classified patients in the top 3 expressing quartiles (n = 81) as high expressors and patients in the bottom quartile (n = 27) as low expressors. We found that 15 of 27 (56%) low ACTA2 expressors responded to immune checkpoint blockade, while only 20 of 81 (25%) high expressors responded (P = 0.004; Fig. 3A; Supplementary Table S8). These findings showed that tumor ACTA2 level was associated with response to ICI. Of note, we tested multiple thresholds to stratify ACTA2-Low and ACTA2-High patients and found that currently used threshold resulted in the greatest separation between groups (Supplementary Fig. S3).

Figure 3.

ACTA2 expression is associated with response to ICIs. The association of ACTA2 with response to ICIs in (A) all patients, (B) MSI-H tumors, (C) MSS tumors, (D) EBV-positive tumors, and (E) EBV-negative tumors. ACTA2-Low was defined as ACTA2 expression values in the bottom quartile and ACTA2-High was defined as the top 3 quartiles. The Fisher exact test was used to compare groups. CR, complete response; PD, progressive disease; PR, partial response; SD, stable disease.

Figure 3.

ACTA2 expression is associated with response to ICIs. The association of ACTA2 with response to ICIs in (A) all patients, (B) MSI-H tumors, (C) MSS tumors, (D) EBV-positive tumors, and (E) EBV-negative tumors. ACTA2-Low was defined as ACTA2 expression values in the bottom quartile and ACTA2-High was defined as the top 3 quartiles. The Fisher exact test was used to compare groups. CR, complete response; PD, progressive disease; PR, partial response; SD, stable disease.

Close modal

MSI-H patients are more likely to respond to ICI than patients with MSS tumors (12, 22). Microsatellite stability status data was available for 94% (101 of 108) of patients. We stratified patients by microsatellite stability status and found that ACTA2 expression was higher in MSS patients (Supplementary Fig. S4). Next, we analyzed the MSI-H patients and classified patients in the top 3 expressing quartiles as high expressors and patients in the bottom expressing quartile as low expressors and found 88% (7 of 8) of ACTA2-Low patients and 71% (15 of 21) of ACTA2-High patients responded to ICI (P = 0.63; Fig. 3B). We performed a similar analysis using MSS patients and response was observed in 33% (6 of 18) of ACTA2-Low patients versus only 11% (6 of 54) of ACTA2-High patients (P = 0.06; Fig 3C). Finally, we analyzed patients by EBV status, which was available for 78 patients. We classified patients in the top 3 ACTA2 expressing quartiles as high expressors and patients in the bottom expressing quartile as low expressors. As expected, the majority of EBV-positive patients responded to ICI. Fifty percent (1 of 2) of ACTA2-Low patients responded and 83% (5 of 6) of ACTA2-High patients responded (P = 0.46; Fig. 3D). Among EBV-negative patients, a response was observed in 39% (7 of 18) of ACTA2-Low patients and only 17% (9 of 52) of ACTA2-High patients (P = 0.10; Fig. 3E).

Low ACTA2 expression in cancer-associated fibroblasts was associated with ICI response

While ACTA2 is a marker for fibroblasts, previous work in lung and breast cancer showed that tumor ACTA2 affects cancer cell autonomous functions (23, 24). We next sought to determine in which cell type ACTA2 expression was associated with ICI response. We analyzed scRNA-seq data of 5 patients with gastric cancer with MSI-H tumors treated with ICI published by Kwon and colleagues (European Nucleotide Archive: PRJEB40416; ref. 15). Even though MSI-H tumors respond more frequently to ICI than MSS disease, 50% of MSI-H tumors still do not respond (15). Of the 5 patients in this analysis, 1 patient's tumor responded to ICI and 4 did not. We determined ACTA2 expression levels in tumor, stromal, and immune cells, which were defined using the annotations reported by Kwon and colleagues (15). We found that ACTA2 expression was similar in the tumor cells and immune cells of the responder and nonresponders. However, ACTA2 expression in stromal cells in the responder patient was significantly lower (mean: 0.142) than in the stroma cells of each of the 4 nonresponders [mean: 1.790 (P < 0.0001), 1.233 (P < 0.0001), 1.029 (P < 0.0001), 0.662 (P < 0.01); Fig. 4A; Supplementary Table S9].

Figure 4.

CAF ACTA2 expression was associated with ICI response. A,ACTA2 expression in stromal cells from MSI-H patient tumors, as determined by scRNA-seq by Kwon and colleagues. The red diamond indicates the mean ACTA2 expression value. The Wilcoxon rank-sum test was used to compare ACTA2 expression between patients. B, H&E image of a tumor that was analyzed with digital spatial profiling. C, The tumor was stained with anti-PanCK (green), CD45 (red), and α-SMA antibodies (yellow). The yellow box indicates the magnified area in D. D, ROIs plotted. E,ACTA2 expression in different cell compartments of tumors from ICI responders and nonresponders. Each circle indicates the mean ACTA2 expression level in an individual patient (i.e., the average value of all ROIs). The bolded line in the center of the box plot indicates the median ACTA2 expression value. The outer edges of the box plot indicate the interquartile range. The vertical lines indicate the range. The Wilcoxon rank-sum test was used to test the statistical significance. **, P < 0.01; ****, P < 0.0001; α-SMA, smooth muscle α-2 actin.

Figure 4.

CAF ACTA2 expression was associated with ICI response. A,ACTA2 expression in stromal cells from MSI-H patient tumors, as determined by scRNA-seq by Kwon and colleagues. The red diamond indicates the mean ACTA2 expression value. The Wilcoxon rank-sum test was used to compare ACTA2 expression between patients. B, H&E image of a tumor that was analyzed with digital spatial profiling. C, The tumor was stained with anti-PanCK (green), CD45 (red), and α-SMA antibodies (yellow). The yellow box indicates the magnified area in D. D, ROIs plotted. E,ACTA2 expression in different cell compartments of tumors from ICI responders and nonresponders. Each circle indicates the mean ACTA2 expression level in an individual patient (i.e., the average value of all ROIs). The bolded line in the center of the box plot indicates the median ACTA2 expression value. The outer edges of the box plot indicate the interquartile range. The vertical lines indicate the range. The Wilcoxon rank-sum test was used to test the statistical significance. **, P < 0.01; ****, P < 0.0001; α-SMA, smooth muscle α-2 actin.

Close modal

On the basis of the scRNA-seq data, we hypothesized that ACTA2 expression in cancer-associated fibroblasts (CAF) was associated with ICI response. We performed digital spatial profiling of tumor samples from 9 patients treated with ICI. There were 4 responders and 5 nonresponders (Supplementary Table S10). We used anti-PanCK, anti-CD45, and anti–smooth muscle actin (α-SMA) antibodies to mark tumor epithelial cells, immune cells, and CAFs, respectively. We compared the fluorescent staining to the hematoxylin and eosin (H&E) staining of a consecutively cut slide to identify PanCK-positive areas that corresponded to tumor based on morphologic features (Fig. 4BC). We identified ROIs that encompassed tumor cells and the surrounding tumor microenvironment (Fig. 4D). We analyzed a total of 27 ROIs from responders (mean: 6.25, range 4–13) and 14 ROIs from nonresponders (mean: 2.8, range 1–5). We then compared the mean ACTA2 expression between responder and nonresponder patients. ACTA2 expression did not vary significantly between ICI responders and nonresponders in either tumor cells (median: 3.11 vs. 3.22; P = 0.56) or immune cells (median: 3.87 vs. 3.61, P = 0.81; Fig. 4E). We found higher ACTA2 expression levels in α-SMA CAFs and noted a trend towards lower ACTA2 expression in α-SMA CAFs in ICI responders than nonresponders (median: 5.00 vs. 5.50, P = 0.29; Fig. 4E). In sum, the scRNA-seq and digital spatial profiling results suggest that ACTA2 expression in the tumor fibroblast compartment was associated with ICI response.

Gastric cancer is a genomically heterogeneous disease, with subtypes having distinct molecular and clinical features that are associated with prognostic and predictive information (1–3, 17). In this study, we identified and validated ACTA2 as a prognostic biomarker for both MSI-H and MSS gastric cancer. Providing accurate estimation of expected survival is an essential component of cancer care because the information helps set patient expectations and guide treatment planning. Incorporating molecular features such as microsatellite status and tumor ACTA2 expression with traditional features such as Lauren classification and pathologic staging will improve the precision of risk stratification and survival estimate.

Only approximately 15% of patients with gastric cancer respond to ICI, which means that most patients undergo futile therapy that puts them at risk for treatment-related adverse events (7–10). Indeed, 17% of patients treated with pembrolizumab experienced a severe grade 3 to 5 adverse event including neutropenia, and autoimmune disorders such as colitis, hepatitis, myocarditis, endocrinopathies, xerostomia, and ocular disorders (7, 8). Furthermore, Patrinely and colleagues recently showed that the proportion of ICI-treated patients who suffer immune-related disorders is higher than previously estimated, including 43% of patients whose immune-related disorder persisted for greater than 12 weeks (11). Thus, novel biomarkers that predict ICI nonresponse may improve patient outcomes both in terms of OS and quality of life.

Currently used biomarkers that predict ICI efficacy in gastric cancer are limited in their efficacy. In the CheckMate-649 study, patients with a CPS ≥ 5 treated with nivolumab plus chemotherapy had improved OS compared with patients treated with chemotherapy alone (25). However, in the KEYNOTE-062 study, pembrolizumab did not improve OS of patients with advanced gastric cancer with a PD-L1 CPS ≥ 1 (10, 26). While the subgroup of patients with a CPS ≥ 10 treated with pembrolizumab had improved OS compared with chemotherapy-treated patients, this difference was not tested in accordance with the study's pre-specified statistical plan. Moreover, patients with a CPS ≥ 10 constitute only a small fraction of patients with gastric cancer (27). Similarly, while MSI-H status is also used as a biomarker to identify patients who are likely to respond to ICI, the large majority of patients with gastric cancer have MSS tumors (22). We found that ACTA2 expression was associated with ICI response in MSS patients. Thus, our identification of the association of ACTA2 expression with ICI efficacy in gastric cancer has the potential to refine the precision of therapy for patients with gastric cancer.

In our cohort, 22 of 29 (76%) MSI-H patients responded to ICI. While the ACTA2-Low patients did have higher rate of response than ACTA2-High patients (88% vs. 71%), it was not statistically significant. This likely reflects that MSI-H status is a dominant factor driving ICI responsiveness and other components have relatively minor contributions. This concept was recently demonstrated by Lee and colleagues, who analyzed samples from KEYNOTE-062, which was a phase III clinical trial that compared pembrolizumab with pembrolizumab and chemotherapy in patients with locally advanced/unresectable or metastatic gastric and gastroesophageal junction cancers (18). While they found that TMB was associated with pembrolizumab clinical efficacy, the clinical utility of TMB was much lower when MSI-H tumors were excluded. However, it is notable that 50% of MSI-H gastric cancers do not respond to ICI (15). Thus, it is important to understand the contributions of additional factors to improve the precision of gastric cancer care. Our analysis of scRNA-seq data from the 5 MSI-H patients, which showed the only patient who responded to ICI had significantly lower ACTA2 expression than the 4 nonresponders, support the notion that ACTA2 expression is associated with ICI response in MSI-H patients and provide a molecular basis for future studies to determine why a significant portion of MSI-H tumors do not respond to ICI (15). Future investigations will need to include a larger cohort that comprises both MSI-H and MSS patients to further delineate the biological basis for ICI response as pertaining to microsatellite status.

We performed digital spatial profiling and confirmed that ACTA2 was primarily expressed in α-SMA cells that morphologically resemble fibroblasts. Taken in context with our bulk RNA-seq findings that showed high ACTA2 expression being associated with ICI nonresponse, the data suggest that CAFs may be the key stromal compartment where ACTA2 expression is associated with ICI response. ACTA2 encodes α-SMA which is a cytoskeletal protein found primarily in mesenchymal cells, including CAFs. CAFs are a heterogeneous population of cells that modulate the tumor immune microenvironment, angiogenesis, and extracellular matrix remodeling (28). CAFs have been implicated in mediating therapy resistance through multiple hypothesized mechanisms including facilitation of epithelial-to-mesenchymal transition, promotion of an immune suppressive tumor immune microenvironment, secretion of extracellular matrix, and modulation of the vascular supply to limit drug delivery (28). CAF subtypes may also induce an immunosuppressive environment and regulate ICI resistance in breast and pancreatic cancer (29, 30). While there is a paucity of data regarding the range of CAF function in the context of gastric cancer immunotherapy, increased CAF infiltration has been shown to be associated with an immunosuppressive gastric cancer tumor microenvironment (31). Finally, Kwon and colleagues found that MSI-H patients who did not respond to ICI had a higher baseline proportion of CAFs and an increase in CAF abundance following ICI treatment (15).

Our report demonstrates that ACTA2 expression can provide clinically impactful information to guide therapy. However, our study is limited by the retrospective nature of our analysis that may be confounded by selection bias. In addition, while we showed that ACTA2 expression was associated with increased radiographic response to ICIs, response may not accurately predict OS. Thus, the prognostic and predictive utility of ACTA2 expression should be validated prospectively using the most clinically relevant endpoints, such as OS and/or quality of life. Finally, the molecular mechanism underpinning the utility of ACTA2 expression in CAFs, and CAF subpopulations, is an important topic for future study.

S. Park reports a patent for Marker Composition for Predicting Prognosis of Cancer, Method for Prognosis of Cancer and Method for Providing Information for Determining Strategy of Cancer Treatment pending. J.R. Clemenceau reports grants from US Department of Defense during the conduct of the study. S.C. Wang reports a patent for PCT/KR2021/018966 pending. T.H. Hwang reports nonfinancial support from KURE.AI Therapeutics, as well as nonfinancial support from KURE.AI outside the submitted work; in addition, T.H. Hwang has a patent for WO2022/186455 issued. No disclosures were reported by the other authors.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

S. Park: Conceptualization, data curation, software, formal analysis, validation, investigation, methodology, writing–review and editing. J.D. Karalis: Conceptualization, data curation, investigation, visualization, writing–original draft, project administration, writing–review and editing. C. Hong: Conceptualization, data curation, software, formal analysis, investigation, writing–review and editing. J.R. Clemenceau: Data curation, software, formal analysis, investigation, writing–review and editing. M.R. Porembka: Conceptualization, investigation, writing–review and editing. I.-H. Kim: Conceptualization, resources, writing–review and editing. S.H. Lee: Investigation, writing–review and editing. S.C. Wang: Conceptualization, resources, supervision, validation, investigation, methodology, project administration, writing–review and editing. J.-H. Cheong: Conceptualization, resources, supervision, validation, investigation, project administration, writing–review and editing. T.H. Hwang: Conceptualization, resources, software, supervision, funding acquisition, validation, investigation, visualization, methodology, project administration, writing–review and editing.

J.D. Karalis holds a Physician-Scientist Institutional Award from the Burroughs Wellcome Fund (award no. 1018897). M.R. Porembka is a Dedman Family Scholar in Clinical Care. S.C. Wang is a Disease Oriented Clinical Scholar at UT Southwestern and supported by the NCI/NIH (K08 CA222611). J.-H. Cheong is supported by a grant through KHIDI, funded by the Ministry of Health & Welfare, Republic of Korea (HI14C1324). T.H. Hwang is the Florida Department of Health Cancer Chair at Mayo Clinic Florida and supported by the DOD (CA190578).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.

Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).

1.
Cancer Genome Atlas Research Network
.
Comprehensive molecular characterization of gastric adenocarcinoma
.
Nature
2014
;
513
:
202
9
.
2.
Cristescu
R
,
Lee
J
,
Nebozhyn
M
,
Kim
K-M
,
Ting
JC
,
Wong
SS
, et al
.
Molecular analysis of gastric cancer identifies subtypes associated with distinct clinical outcomes
.
Nat Med
2015
;
21
:
449
56
.
3.
Cheong
J-H
,
Wang
SC
,
Park
S
,
Porembka
MR
,
Christie
AL
,
Kim
H
, et al
.
Development and validation of a prognostic and predictive 32-gene signature for gastric cancer
.
Nat Commun
2022
;
13
:
774
.
4.
Bang
Y-J
,
Van Cutsem
E
,
Feyereislova
A
,
Chung
HC
,
Shen
L
,
Sawaki
A
, et al
.
Trastuzumab in combination with chemotherapy versus chemotherapy alone for treatment of HER2-positive advanced gastric or gastro-esophageal junction cancer (ToGA): a phase III, open-label, randomized controlled trial
.
Lancet North Am Ed
2010
;
376
:
687
97
.
5.
Garon
EB
,
Rizvi
NA
,
Hui
R
,
Leighl
N
,
Balmanoukian
AS
,
Eder
JP
, et al
.
Pembrolizumab for the treatment of non-small-cell lung cancer
.
N Engl J Med
2015
;
372
:
2018
28
.
6.
Larkin
J
,
Chiarion-Sileni
V
,
Gonzalez
R
,
Grob
JJ
,
Cowey
CL
,
Lao
CD
, et al
.
Combined nivolumab and ipilimumab or monotherapy in untreated melanoma
.
N Engl J Med
2015
;
373
:
23
34
.
7.
Fuchs
CS
,
Doi
T
,
Jang
RW
,
Muro
K
,
Satoh
T
,
Machado
M
, et al
.
Safety and efficacy of pembrolizumab monotherapy in patients with previously treated advanced gastric and gastroesophageal junction cancer: phase II clinical KEYNOTE-059 trial
.
JAMA Oncol
2018
;
4
:
e180013
.
8.
Shitara
K
,
Özgüroğlu
M
,
Bang
Y-J
,
Di Bartolomeo
M
,
Mandalà
M
,
Ryu
M-H
, et al
.
Pembrolizumab versus paclitaxel for previously treated, advanced gastric or gastro-esophageal junction cancer (KEYNOTE-061): a randomized, open-label, controlled, phase III trial
.
Lancet North Am Ed
2018
;
392
:
123
33
.
9.
Kang
Y-K
,
Boku
N
,
Satoh
T
,
Ryu
M-H
,
Chao
Y
,
Kato
K
, et al
.
Nivolumab in patients with advanced gastric or gastro-esophageal junction cancer refractory to, or intolerant of, at least two previous chemotherapy regimens (ONO-4538–12, ATTRACTION-2): a randomized, double-blind, placebo-controlled, phase III trial
.
Lancet North Am Ed
2017
;
390
:
2461
71
.
10.
Shitara
K
,
Van Cutsem
E
,
Bang
Y-J
,
Fuchs
C
,
Wyrwicz
L
,
Lee
K-W
, et al
.
Efficacy and safety of pembrolizumab or pembrolizumab plus chemotherapy vs chemotherapy alone for patients with first-line, advanced gastric cancer
.
JAMA Oncol
2020
;
6
:
1571
.
11.
Patrinely
JR
,
Johnson
R
,
Lawless
AR
,
Bhave
P
,
Sawyers
A
,
Dimitrova
M
, et al
.
Chronic immune-related adverse events following adjuvant anti–PD-1 therapy for high-risk resected melanoma
.
JAMA Oncol
2021
;
7
:
744
8
.
12.
Pietrantonio
F
,
Randon
G
,
Di Bartolomeo
M
,
Luciani
A
,
Chao
J
,
Smyth
EC
, et al
.
Predictive role of microsatellite instability for PD-1 blockade in patients with advanced gastric cancer: a meta-analysis of randomized clinical trials
.
ESMO Open
2021
;
6
:
100036
.
13.
An
JY
,
Kim
H
,
Cheong
J-H
,
Hyung
WJ
,
Kim
H
,
Noh
SH
.
Microsatellite instability in sporadic gastric cancer: its prognostic role and guidance for 5-FU based chemotherapy after R0 resection
.
Int J Cancer
2012
;
131
:
505
11
.
14.
Zang
YS
,
Dai
C
,
Xu
X
,
Cai
X
,
Wang
G
,
Wei
J
, et al
.
Comprehensive analysis of potential immunotherapy genomic biomarkers in 1,000 Chinese patients with cancer
.
Cancer Med
2019
;
8
:
4699
708
.
15.
Kwon
M
,
An
M
,
Klempner
SJ
,
Lee
H
,
Kim
K-M
,
Sa
JK
, et al
.
Determinants of response and intrinsic resistance to PD-1 blockade in microsatellite instability-high gastric cancer
.
Cancer Discov
2021
;
11
:
2168
85
.
16.
Bai
Y
,
Xie
T
,
Wang
Z
,
Tong
S
,
Zhao
X
,
Zhao
F
, et al
.
Efficacy and predictive biomarkers of immunotherapy in Epstein–Barr virus—associated gastric cancer
.
J Immunother Cancer
2022
;
10
:
e004080
.
17.
Kim
ST
,
Cristescu
R
,
Bass
AJ
,
Kim
K-M
,
Odegaard
JI
,
Kim
K
, et al
.
Comprehensive molecular characterization of clinical responses to PD-1 inhibition in metastatic gastric cancer
.
Nat Med
2018
;
24
:
1449
58
.
18.
Lee
K-W
,
Van Cutsem
E
,
Bang
Y-J
,
Fuchs
CS
,
Kudaba
I
,
Garrido
M
, et al
.
Association of tumor mutational burden with efficacy of pembrolizumab {plus minus} chemotherapy as first-line therapy for gastric cancer in the Phase III KEYNOTE-062 study
.
Clin Cancer Res
2022
;
28
:
3489
98
.
19.
Sohn
BH
,
Hwang
J-E
,
Jang
H-J
,
Lee
H-S
,
Oh
SC
,
Shim
J-J
, et al
.
Clinical significance of four molecular subtypes of gastric cancer identified by The Cancer Genome Atlas project
.
Clin Cancer Res
2017
;
23
:
4441
9
.
20.
Chida
K
,
Kawazoe
A
,
Suzuki
T
,
Kawazu
M
,
Ueno
T
,
Takenouchi
K
, et al
.
Transcriptomic profiling of MSI-H/dMMR gastrointestinal tumors to identify determinants of responsiveness to anti–PD-1 therapy
.
Clin Cancer Res
2022
;
28
:
2110
7
.
21.
Smyth
EC
,
Nilsson
M
,
Grabsch
HI
,
van Grieken
NC
,
Lordick
F
.
Gastric cancer
.
Lancet North Am Ed
2020
;
396
:
635
48
.
22.
Pietrantonio
F
,
Miceli
R
,
Raimondi
A
,
Kim
YW
,
Kang
WK
,
Langley
RE
, et al
.
Individual patient data meta-analysis of the value of microsatellite instability as a biomarker in gastric cancer
.
J Clin Oncol
2019
;
37
:
3392
400
.
23.
Lee
HW
,
Park
YM
,
Lee
SJ
,
Cho
HJ
,
Kim
D-H
,
Lee
J-I
, et al
.
Alpha-smooth muscle actin (ACTA2) is required for metastatic potential of human lung adenocarcinoma
.
Clin Cancer Res
2013
;
19
:
5879
89
.
24.
Jeon
M
,
You
D
,
Bae
SY
,
Kim
SW
,
Nam
SJ
,
Kim
HH
, et al
.
Dimerization of EGFR and HER2 induces breast cancer cell motility through STAT1-dependent ACTA2 induction
.
Oncotarget
2017
;
8
:
50570
81
.
25.
Janjigian
YY
,
Shitara
K
,
Moehler
M
,
Garrido
M
,
Salman
P
,
Shen
L
, et al
.
First-line nivolumab plus chemotherapy versus chemotherapy alone for advanced gastric, gastro-esophageal junction, and esophageal adenocarcinoma (CheckMate 649): a randomized, open-label, phase III trial
.
Lancet North Am Ed
2021
;
398
:
27
40
.
26.
Keenan
TE
,
Burke
KP
,
Van Allen
EM
.
Genomic correlates of response to immune checkpoint blockade
.
Nat Med
2019
;
25
:
389
402
.
27.
Schoemig-Markiefka
B
,
Eschbach
J
,
Scheel
AH
,
Pamuk
A
,
Rueschoff
J
,
Zander
T
, et al
.
Optimized PD-L1 scoring of gastric cancer
.
Gastric Cancer
2021
;
24
:
1115
22
.
28.
Ganguly
D
,
Chandra
R
,
Karalis
J
,
Teke
M
,
Aguilera
T
,
Maddipati
R
, et al
.
Cancer-associated fibroblasts: versatile players in the tumor microenvironment
.
Cancers
2020
;
12
:
2652
.
29.
Costa
A
,
Kieffer
Y
,
Scholer-Dahirel
A
,
Pelon
F
,
Bourachot
B
,
Cardon
M
, et al
.
Fibroblast heterogeneity and immunosuppressive environment in human breast cancer
.
Cancer Cell
2018
;
33
:
463
79
.
30.
Feig
C
,
Jones
JO
,
Kraman
M
,
Wells
RJB
,
Deonarine
A
,
Chan
DS
, et al
.
Targeting CXCL12 from FAP-expressing carcinoma-associated fibroblasts synergizes with anti–PD-L1 immunotherapy in pancreatic cancer
.
Proc Natl Acad Sci
2013
;
110
:
20212
7
.
31.
Liu
X
,
Yao
L
,
Qu
J
,
Liu
L
,
Lu
N
,
Wang
J
, et al
.
Cancer-associated fibroblast infiltration in gastric cancer: the discrepancy in subtypes pathways and immunosuppression
.
J Transl Med
2021
;
1919
:
325
.

Supplementary data