Introduction#
Gastric cancer, with gastric adenocarcinoma (GAC) as its main histological type, is one of the most common malignant tumors worldwide and a major cause of cancer-related deaths. A significant portion of gastric cancer patients are diagnosed at an advanced stage, which largely limits the effectiveness of treatment and the prognosis for patients. Although surgical resection remains a mandatory pillar of treatment, several studies, including JCOG9501 and JCOG9502 (series studies by the Japan Clinical Oncology Group), indicate that gastric cancer patients do not benefit from extended resection. In the past decade, neoadjuvant and perioperative therapies have brought new hope. The MAGIC trial showed that for resectable stage II/III gastric cancer patients, three cycles of preoperative and three cycles of postoperative ECF (epirubicin, cisplatin, and 5-fluorouracil) chemotherapy could increase the 5-year survival rate from 23% to 36% compared to surgery alone (MAGIC: the Medical Research Council Adjuvant Gastric Infusional Chemotherapy). The FLOT4-AIO trial further demonstrated that the FLOT (5-fluorouracil, folinic acid, oxaliplatin, and docetaxel) regimen led to better pathological response rates, R0 resection rates, and overall survival (OS) compared to ECF or ECX (epirubicin, 5-fluorouracil, and capecitabine). It is recognized that preoperative chemotherapy can increase the chances of radical resection, eliminate early microscopic spread, and allow for preoperative response assessment of adjuvant therapy. With the emergence of new drugs such as immune checkpoint inhibitors (ICIs), chemotherapy remains the most fundamental and accessible component of perioperative treatment for gastric cancer.
On the other hand, preoperative treatment for gastric cancer remains controversial, especially in East Asian countries. There is heterogeneity in the response to preoperative chemotherapy, and understanding its mechanisms is limited. Biomarkers that predict patient responses to preoperative chemotherapy are needed to optimally stratify patients for treatment. Emerging evidence suggests that immunity is involved in patients' responses to chemotherapy. Choi et al. reported that the expression of programmed cell death ligand 1 (PD-L1) in tumor specimens can predict the benefits of adjuvant chemotherapy after D2 gastrectomy in stage II/III gastric cancer. Kim et al. used paired preoperative and treatment-period gastric biopsy samples during standard first-line chemotherapy and found that chemotherapy induced the infiltration of natural killer (NK) cells, polarization of macrophages, and increased antigen presentation in treatment responders. However, existing studies in the field of gastric cancer immunology have mainly focused on local immune responses in the tumor microenvironment (TME), with little known about the relationship between systemic immunity and responses to gastric cancer chemotherapy.
Gastric cancer is a systemic disease. The immune response stimulated by tumor burden and antitumor treatment is coordinated across different tissues. Analyzing the systemic immune landscape or the immune macroenvironment described by Hiam-Galvez et al. in patients receiving preoperative chemotherapy is crucial for a comprehensive understanding of cancer immunity and treatment resistance mechanisms. Existing systemic immune-inflammation indices, such as the neutrophil-to-lymphocyte ratio (NLR), primarily rely on blood cell counts, which limits their dimensionality. Serum immunoproteomics, with its high content, would be an ideal reflection of systemic immunity. In this study, we collected serum samples from gastric adenocarcinoma patients who received preoperative chemotherapy before, during, and after surgery, and studied their immunoproteomics using an antibody-based proteomics platform (Olink Target 96 Inflammation panel). We also collected surgically resected tumor samples from these patients and assessed the tumor microenvironment using multiplex immunofluorescence (mIF), immunohistochemistry (IHC), and RNA sequencing (RNA-seq). We investigated the dynamic changes in serum immunoproteomics and their correlation with the tumor microenvironment, identifying biomarkers that predict tumor shrinkage, overall survival (OS), and progression-free survival (PFS) in patients receiving preoperative chemotherapy.
Results#
Study Population#
This study included 90 gastric adenocarcinoma patients who received preoperative chemotherapy followed by gastrectomy (Figure 1A). Patients who received immune checkpoint inhibitors (ICIs) during the preoperative period were excluded. Eligible patients were divided into responders (residual tumor/tumor bed ≤50% chemotherapy effect, Becker TRG score 1–2) and non-responders (Becker TRG score 3). Among the 90 patients, 36 (40%) achieved a tumor shrinkage score of 1–2 and were considered responders. Patients with better tumor shrinkage had significantly longer overall survival compared to non-responders (Figure S1A). Progression-free survival showed a similar trend, although without statistical significance (Figure S1B). The basic clinical characteristics of the patients are summarized in Table S1. Nearly half of the patients received two-drug cytotoxic chemotherapy, most commonly the SOX (S-1 plus oxaliplatin) or XELOX (capecitabine plus oxaliplatin) regimen. The remaining patients received three-drug cytotoxic chemotherapy, primarily the DOS (docetaxel, oxaliplatin, and S-1) regimen. As of the analysis date of March 1, 2022, the median follow-up time was 55.8 months (range from 3.2 to 82.7 months). In the overall population, the median progression-free survival was 39.8 months (95% confidence interval [CI], 32.7 to not reached [NR]), while the median overall survival was 63.9 months (95% CI, 51.8 to 74.1), with 45 deaths (50%).
Dynamic Serum Immunoproteomics Associated with Preoperative Chemotherapy Response#
We collected 37 preoperative, 8 intraoperative, and 83 postoperative serum samples from patients receiving preoperative chemotherapy, with 30 paired preoperative and postoperative serum samples (Figure 1A). The levels of 92 marker proteins in key immune and inflammatory pathways were measured using the Olink Target 96 Inflammation panel's proximity extension assay (PEA). Comparing protein levels in preoperative and postoperative serum samples revealed dynamic changes in serum immunoproteomics after preoperative chemotherapy. Among the 92 proteins, 18 showed significant changes in both paired and unpaired tests (Figure 1B, Figure S1C, and Figure S1D), indicating that preoperative chemotherapy triggered a complex systemic immune response. Notably, serum levels of C-X-C motif chemokine ligand 1 (CXCL1) and CXCL5 significantly decreased after preoperative chemotherapy (Figure S1D). Interestingly, Zhou et al. reported that CXCL1 and CXCL5, as CXCR2 ligands, can significantly promote the migration of gastric cancer cells and drive gastric cancer metastasis. Chemotherapy may help prevent gastric cancer metastasis by lowering serum levels of CXCL5 and CXCL1. In fact, CXCL1/5 levels decreased during the early cycles of preoperative chemotherapy (Figure S1E).
We further compared the dynamic changes in serum immunoproteomics among patients with different treatment responses. We found that responders exhibited more dynamic changes in serum immunoproteomics after treatment (Figure 1C and 1D). We also compared the absolute changes in protein levels between responders and non-responders after chemotherapy, finding that immune protein levels changed more significantly overall in responders (Figure 1E). For example, the decrease in serum CXCL5 levels after treatment was much less pronounced in non-responders compared to responders (Figure 1C–1F). The proteomics during treatment also appeared to differ between responders and non-responders (Figure S1E). For instance, in responders, serum interleukin receptor subunit b (IL-10RB) and IL-18 levels trended upward during chemotherapy, while this trend was not observed in non-responders (Figure S1F and Figure S1G), although this conclusion may be limited by the sample size.
Overall, these results indicate a complex systemic immune response to preoperative chemotherapy in gastric adenocarcinoma patients. Responders often exhibit a more dynamic systemic immune response after preoperative chemotherapy.
Tumor Microenvironment (TME) Associated with Patient Response to Preoperative Chemotherapy#
First, we compared the transcriptomes of tumor samples from patients with different treatment responses to gain general knowledge about local tumor characteristics. Gene set enrichment analysis (GSEA) revealed altered hallmark pathways in good responders (Figure 2A). Changes in pathways such as DNA replication and the cell cycle may indicate inhibition of cancer cell proliferation and tumor regression. Additionally, nearly half of the altered pathways were immune-related, such as chemokine signaling pathways and cytokine-cytokine receptor interaction pathways (Figure 2B and Figure 2C), indicating the importance of immunity in chemotherapy.
Therefore, we assessed the geographical immune landscape in surgically resected tumor samples using multiplex immunofluorescence (mIF). We used CD4, CD8, and Foxp3 staining to identify different types of T cells. We used CD68 and CD163 staining to identify macrophages (Figure 2D). We compared immune infiltration between responders and non-responders. The cell density of CD68+ macrophages and CD68+/CD163+ M2 macrophages was significantly higher in non-responders (Figure 2E and Figure S2A). Accordingly, Xing et al. reported higher CD68+ macrophage infiltration in non-responders after neoadjuvant chemotherapy for gastric cancer. M2 macrophages have also been shown to be involved in chemotherapy resistance in various cancers.
Meanwhile, we collected 24 preoperative endoscopic biopsy samples from the cohort. We analyzed the preoperative TME using mIF (Figure S2B). Notably, most endoscopic biopsies only obtained superficial mucosa of the stomach, which largely limited their representativeness of the entire tumor and comparability with surgically resected tissues (Figure S2C). In fact, mIF showed no difference in immune cell infiltration in the preoperative TME between responders and non-responders (revised Figure S2D), which may be due to limited biopsy depth and significant intratumoral heterogeneity in gastric cancer.
Overall, these results suggest that the postoperative TME is associated with responses to preoperative chemotherapy.
Correlation Between Serum Immunoproteomics and TME#
Given that most existing cancer immunology studies focus on the tumor microenvironment (TME), we assessed the correlation between systemic immunity and TME. We also identified correlations between serum immunoproteomics and immune cell infiltration in the TME. Interestingly, the postoperative TME appeared to correlate more with preoperative rather than postoperative serum immunoproteomics. Even with a smaller sample size, the correlation between preoperative serum immunoproteomics and immune cell infiltration was generally stronger (Figure 3A and Figure 3B). For example, higher preoperative serum fibroblast growth factor 21 (FGF21) levels were associated with less infiltration of CD68+ macrophages, while higher preoperative serum transforming growth factor b1 (TGF-b1) levels were associated with more infiltration of CD4+ T cells (Figure 3C and Figure 3D). In fact, TGF-b has been reported to have pleiotropic effects in regulating effector and regulatory CD4-positive cell responses. The correlation between postoperative serum immunoproteomics and postoperative immune cell infiltration was also observed. For instance, higher preoperative serum C-C motif chemokine ligand 11 (CCL11) levels were associated with more infiltration of CD4+/FOXP3+ T cells (Figure 3E). Wang et al. reported that CCL11 increased the proportion of CD4+CD25+Foxp3+ regulatory T cells (Tregs) in breast cancer. Further studies are needed to explore whether CCL11 regulates CD4+Foxp3+ Treg cell function in gastric cancer.
We also assessed the correlation between postoperative serum protein levels and the tumor mRNA levels of 92 immune genes. Among them, five immune genes showed statistically significant correlations, with only two being positively correlated, as expected (Figure 3F and Figure S3A–S3E). The correlations of TNFSF12 and CCL4 were actually marginal (Figure S3A and S3B). Overall, the correlation between serum protein levels and tissue gene mRNA levels was relatively weak.
These results illustrate the intercommunication and interdependence between systemic immunity and the tumor microenvironment. Studies of the tumor microenvironment cannot adequately reveal how the immune system comprehensively responds to gastric cancer and antitumor treatment.
Clinical Value of Classic Systemic Immune Inflammation Indices#
Classic systemic immune inflammation indices are mostly based on blood cell ratios and have been shown to correlate with clinical outcomes in patients. We were curious about the relationship between serum immunoproteomics and classic systemic immune inflammation indices. Therefore, we assessed the correlation between postoperative serum immunoproteomics and classic immune inflammation indices, including the neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), monocyte-to-lymphocyte ratio (MLR), and platelet distribution width (PDW) as well as common blood cell counts. Although most correlations were relatively weak (Figure S3F), serum CXCL5 and CXCL1 levels were strongly correlated with platelet counts (Figure S3G and Figure S3H). Since CXCL1 and CXCL5 are generally involved in neutrophil homeostasis and function, more work is needed to understand this unexpected but interesting correlation. We also assessed the relationship between classic systemic immune inflammation indices and TME characteristics. No correlations were observed between postoperative classic immune inflammation indices and immune cell infiltration in the TME (Figure S3I).
We further explored the clinical value of classic systemic immune inflammation indices and assessed the predictive value of NLR, PLR, MLR, and PDW for treatment response. We plotted the receiver operating characteristic (ROC) curves for these four indices, with the highest area under the curve (AUC) being 0.602 (Figure S3J). Proportional hazards regression showed the prognostic value of these four indices. In univariate Cox regression, these indices did not show significant prognostic value for OS or PFS, while in multivariate Cox regression, a higher NLR was associated with shorter OS, with a hazard ratio of 1.172 (95% CI, 1.0066–1.3639) (Figure S3K and Figure S3L). Accordingly, previous reports have shown that NLR is a negative prognostic factor for gastroesophageal junction and gastric adenocarcinoma. Overall, the prognostic value of these four indices is limited.
Postoperative Tumor Stroma PD-L1 Levels and Preoperative Serum PD-L1 Levels Predict Preoperative Chemotherapy Response#
PD-L1 is a key immune regulatory molecule. When interacting with its receptor PD-1, PD-L1 inhibits the immune response of cytotoxic T cells, thereby participating in tumor immune escape. Choi et al. reported that stroma PD-L1 levels can predict the efficacy of adjuvant chemotherapy after D2 gastrectomy in stage II/III gastric cancer based on the CLASSIC trial cohort. Using a similar scoring system based on PD1/PDL1 immunohistochemical staining, we found a trend of higher stroma PD-L1 staining scores in non-responders in surgically resected tumor samples (Figure 4A and Figure 4B). Stroma PD-1 staining showed a similar trend, although this was not statistically significant (Figure S4A and Figure S4B). However, PD-L1 staining in tumor regions showed no correlation with treatment response (Figure 4A). These results suggest that the level of stroma PD-L1 in tumors can predict responses to preoperative chemotherapy and indicate that the PD-1/PD-L1 pathway may play a role in chemotherapy resistance in gastric cancer.
However, due to its delayed nature, the predictive value of postoperative stroma PD-L1 may be significantly limited. An ideal predictive factor should be preoperative. Stroma PD-L1 staining from preoperative endoscopic biopsies could not predict treatment response (Figure S4C and Figure S4D). Therefore, we further assessed the clinical significance of preoperative serum PD-L1 levels. Interestingly, preoperative serum PD-L1 levels showed differences among patients with different treatment responses (Figure 4C). Before treatment, responders had lower serum PD-L1 levels, and treatment seemed to diminish this difference, as no significant differences were observed in postoperative samples (Figure 4E). Using ROC curves, we evaluated the predictive value of preoperative and postoperative serum PD-L1 levels for treatment response. The AUC for preoperative serum PD-L1 levels was 0.737 (95% CI, 0.569–0.904), while the AUC for postoperative serum PD-L1 levels was approximately 0.5 (Figure 4D and Figure 4F), indicating that preoperative serum PD-L1 levels are a promising predictive factor for treatment response to preoperative chemotherapy. Patients with higher preoperative serum PD-L1 levels (>5.084 normalized protein expression [NPX]) tended to show poorer treatment responses to preoperative chemotherapy (Figure S4E).
We also assessed the treatment-period serum PD-L1 levels among patients with different treatment responses. In responders, serum PD-L1 levels appeared to increase during treatment. The treatment-period serum PD-L1 levels in responders were significantly higher (Figure S4F and Figure S4G). One potential reason for this difference could be the destruction of tumor cells. More samples and further studies are needed to confirm this finding and reveal the underlying mechanisms. We further measured the pathological correlation between PD-L1/PD-1 levels and serum PD-L1 levels. Among different pairs, preoperative serum PD-L1 levels showed the strongest correlation with postoperative stroma PD-1 levels (Figure S4H). Preoperative serum PD-L1 levels may be related to the infiltration of PD-1+ immune cells in tumors after chemotherapy.
Overall, these results suggest that both postoperative tumor stroma PD-L1 levels and preoperative serum PD-L1 levels can predict responses to preoperative chemotherapy, with preoperative serum PD-L1 levels likely having greater clinical significance.
Preoperative Serum CCL20 Levels Predict Response to Preoperative Chemotherapy#
Inspired by the findings on PD-L1, we further compared preoperative serum immunoproteomics among patients with different treatment responses, revealing 10 proteins with p <0.05 differences. Among them, preoperative CCL20 levels showed the most significant difference. Notably, we also compared postoperative serum immunoproteomics among patients with different treatment responses, with differences being much weaker compared to preoperative samples (Figure S5A).
Recent studies have established CCL20 as an important mediator of chemotherapy resistance in various cancers. As summarized in Figure S5B, Chen et al. reported that chemotherapy induces CCL20 through a positive feedback loop between nuclear factor kB (NF-kB) and CCL20, mediating chemotherapy resistance by upregulating ATP-binding cassette sub-family B member 1 (ABCB1) expression in breast cancer. Wang et al. reported that chemotherapy upregulates CCL20 in colorectal cancer cells via the FOXO1/CEBPB/NF-kB signaling pathway, while secreted CCL20 recruits regulatory T cells, promoting chemotherapy resistance. Liu et al. reported that cisplatin-stimulated classical activated macrophages (CAMs) promote ovarian cancer cell migration by increasing CCL20 production. Overall, existing studies suggest that the upregulation of CCL20 is induced by chemotherapy and that increased CCL20 production promotes chemotherapy resistance.
However, our study found that the above model may not hold in gastric cancer. We found that in responders to preoperative chemotherapy, serum CCL20 levels were significantly lower before treatment began (Figure 5B). Preoperative serum CCL20 levels predicted treatment response, with an AUC of 0.769 (95% CI, 0.614–0.925) (Figure 5C), indicating that there are differences in serum CCL20 levels before treatment in gastric cancer patients. Consistent with existing findings, CCL20 mRNA levels were upregulated in non-responders' tumors (Figure 5D). However, there were no differences in serum CCL20 levels between responders and non-responders after treatment, indicating a decoupling of serum and tumor CCL20 levels (Figure 5E). Interestingly, referencing the serum and tissue proteomics of resectable gastric cancer reported by Shen et al., we found that serum CCL20 levels in gastric cancer patients were elevated compared to healthy individuals (Figure 5F). CCL20 protein levels in tumor samples were also higher than in normal gastric tissues (Figure S5C). However, resecting the tumor through gastrectomy did not restore serum CCL20 levels but rather further increased serum CCL20 levels (Figure 5F). These results suggest that serum CCL20 is not a systemic reflection of tumor CCL20 but rather an important component of systemic immunity to gastric cancer and chemotherapy.
We also validated the proposed signaling model of CCL20 upregulation in existing studies. Kim et al. collected pre-treatment and treatment-period gastric biopsy samples from patients receiving first-line standard chemotherapy but not PD-1 blockade. We analyzed their transcriptomic data and found that chemotherapy did not increase CCL20 mRNA levels in tumor samples. Instead, CCL20 mRNA levels decreased after chemotherapy (Figure 5G). This finding challenges the hypothesis that CCL20 is induced by chemotherapy in gastric cancer. Meanwhile, there were also no differences in ABCB1, CEBPB, and FOXO1 mRNA levels between tumors of different responses (Figure 5H) or between biopsy samples before and after chemotherapy (Figure S5D). Conversely, higher preoperative serum CCL20 levels were associated with less infiltration of CD4+ T cells in tumors (Figure S5E). CD4+ T cells mediate immune responses and are crucial for achieving regulation and effective immune responses against tumors. Meanwhile, higher preoperative serum CCL20 levels were associated with more infiltration of PD-1+ or PD-L1+ cells in the stroma (Figure 5I and Figure S5F), which should be key mediators of tumor immune escape. Overall, these results suggest that serum CCL20 induces a systemic immune suppressive environment against chemotherapy.
As summarized in Figure 5J, existing studies propose that the upregulation of CCL20 in tumors is induced by chemotherapy, and increased CCL20 production promotes chemotherapy resistance. However, we found differences in serum CCL20 levels before the start of chemotherapy. Patients with higher preoperative serum CCL20 levels tended to have poorer treatment responses. The underlying mechanism may be that serum CCL20 induces a systemic immune suppressive environment. These findings suggest that in patients with higher preoperative serum CCL20 levels, the combination of immunotherapy with chemotherapy may be more effective. Significant efforts have been made to develop inhibitors of the CCR6-CCL20 axis (CCR6 is the receptor for CCL20). Interfering with the CCR6-CCL20 axis through antibodies or antagonists shows promise in cancer treatment. Preoperative serum CCL20 levels may help select patients who are likely to benefit from CCR6-CCL20 inhibitors. Furthermore, these findings indicate that the preoperative period is an irreplaceable time window for patient stratification through serum protein biomarkers. Therefore, we decided to further establish a preoperative serum protein scoring system for predicting responses to preoperative chemotherapy.
A Preoperative Serum Protein Scoring System for Predicting Response to Preoperative Chemotherapy#
By comparing preoperative serum protein levels among patients with different treatment responses (Figure 5A), we included 15 proteins with p<0.1 in the consistency clustering. Based on the consistency cumulative distribution function (CDF) plot, incremental area plot, and manual inspection of the consistency matrix, we identified four preoperative serum subtypes (Figure 6A, 6B, and Figure S6A–S6H). Among them, cluster 2 was associated with significantly better treatment responses in patients (Figure 6C). This unreviewed clustering was also associated with patients' clinical characteristics, such as the Lauren classification of tumors. Clusters 1 and 4 were associated with a higher proportion of adenocarcinoma tumor types (Figure 6D).
Considering clinical practicality, we further established a preoperative serum response prediction score (PSRscore) using the least absolute shrinkage and selection operator (LASSO) model to predict responses to preoperative chemotherapy (Figure S6I and S6J). In brief, LASSO regression is a type of linear regression that uses shrinkage for variable selection or parameter elimination. With an appropriate l value, the formula for PSRscore is limited to the serum levels of four proteins: CCL3, IL-15Ra, CXCL5, and CCL20 (Figure 6F and Figure S6K). The ROC curve for PSRscore showed an AUC of 0.907 (95% CI, 0.814–1.000), determining a cutoff value of -0.843 (Figure 6E). Patients were divided into high and low PSRscore groups (Figure 6F). A low PSRscore was associated with significantly poorer treatment responses (Figure 6G). Additionally, patients with low PSRscore had numerically more PD1+/PD-L1+ cell infiltration in the stroma and higher tumor PD-L1 staining in postoperative tumors (Figure 6H and Figure S6L), which typically leads to indications for anti-PD-1/PD-L1 therapy.
In addition to CCL20, PSRscore also includes preoperative serum levels of CCL3, IL-15Ra, and CXCL5. Higher serum levels of CCL3 and IL-15Ra and lower levels of CXCL5 were associated with poorer treatment responses (Figure S6K). Studies have shown that CCL3 is involved in immune escape and chemotherapy resistance in various cancers. High levels of CCL3 are associated with increased intratumoral infiltration of Tregs, tumor-associated macrophages (TAMs), and myeloid-derived suppressor cells (MDSCs). CCL3-driven TAM recruitment has been considered a driving event for metastatic niches. Neutralizing antibodies and inhibitors for CCL3 have been developed and show promise in anticancer therapy. Currently, the roles of IL-15Ra and CXCL5 in chemotherapy resistance are not well understood, and more research is needed to explore their functions in gastric cancer.
The PSRscore scoring system helps stratify gastric adenocarcinoma patients and screen those who may not benefit from preoperative chemotherapy alone. For this group of patients, our work strongly suggests that they may benefit from combination immunotherapy, such as immune checkpoint inhibitors (ICIs) or CCL3/20 neutralizing antibodies/inhibitors (Figure 6I). Prospective trials can be designed to validate this strategy, and a validation cohort is needed to verify the sensitivity and specificity of this scoring system.
Prognostic Value of TME and Serum Immunoproteomics#
We further assessed the prognostic value of TME and serum immunoproteomics. All basic clinical characteristics with predictive value in univariate Cox regression, as well as age and sex, were included in multivariate Cox regression (Tables S2 and S3). Immune cells shown as OS or PFS predictive factors are listed in the forest plot along with their hazard ratios (Figure S7A and S7B). Kaplan-Meier curves for representative survival predictive factors are plotted (Figure S7C–S7F). No immune cell type was an independent predictor of OS, while the infiltration of CD68+ macrophages was confirmed to predict shortened PFS through log-rank tests, univariate Cox regression, and multivariate Cox regression (Figure S7C). Although not independent, the infiltration of CD68+ macrophages also showed negative prognostic value for OS through log-rank tests (Figure S7D).
Preoperative and postoperative serum proteins shown as OS or PFS predictive factors are also listed in the forest plot along with their hazard ratios (Figure 7A, 7B, S7G, and S7H). Kaplan-Meier curves for representative survival predictive factors are plotted (Figure 7C, 7D, S7I, and S7J). Among them, high postoperative serum IL-10RB levels were associated with significantly shortened OS and PFS, confirmed through log-rank tests, univariate Cox regression, and multivariate Cox regression (Figure 7C and 7D). This indicates that postoperative serum IL-10RB levels are a strong negative survival predictive factor for patients receiving preoperative chemotherapy. Notably, postoperative IL-10RB levels significantly increased after preoperative chemotherapy, suggesting its potential involvement in the response to preoperative chemotherapy (Figure S1D). Research on the role of IL-10 signaling in gastric cancer is still limited. More work is needed to understand the role of IL-10RB in preoperative treatment of gastric cancer.
Discussion#
In the past decade, efforts have been made to reveal the role of immunity in cancer. Immunotherapy has made breakthroughs in the treatment of gastric cancer, with immune checkpoint inhibitors becoming first-line treatments for advanced gastric or esophageal adenocarcinoma. However, no treatment has successfully challenged the dominance of chemotherapy in perioperative treatment of gastric cancer. Immunity is believed to play a key role in patients benefiting from perioperative chemotherapy. Existing studies have focused on local immune responses in the tumor microenvironment, while a better understanding of gastric cancer immunity must particularly assess systemic immunity. We used serum immunoproteomics and classic systemic immune inflammation indices to describe systemic immunity and investigate its association with the tumor microenvironment and treatment responses. We found that perioperative treatment induced a complex systemic immune response, manifested as dynamic immunoproteomics. At the same time, patients with better treatment responses showed more dynamic changes in serum immunoproteomics after treatment. The tumor microenvironment also showed associations with responses to perioperative chemotherapy. However, predicting potential treatment responses before treatment begins would be more practical. Excitingly, we found that preoperative serum levels of PD-L1 and CCL20 are predictive factors for responses to perioperative chemotherapy, consistent with their known roles in immune suppression. We further established a preoperative serum proteomics panel for predicting responses, capable of accurately screening patients who may not respond to perioperative chemotherapy alone. For this subset of patients, we believe they will benefit from combination therapies of immunotherapy and chemotherapy. Meanwhile, postoperative serum levels of IL-10RB were also identified as a strong predictive factor for prognosis in gastric cancer patients.
The role of intratumoral PD-L1 in immune suppression and chemotherapy resistance has been confirmed. However, research on soluble PD-L1 is limited. Our study found that there are differences in serum PD-L1 levels before the start of chemotherapy. Patients who respond to chemotherapy often have lower serum PD-L1 levels. Further studies are needed to explore whether soluble PD-L1 plays a role in chemotherapy resistance. Similar findings were observed with CCL20, a chemokine known to be involved in chemotherapy resistance in various cancers. Our study suggests that the CCL20-induced chemotherapy resistance model proposed in other cancer types may not hold in gastric cancer. Viewing changes in CCL20 as a result of chemotherapy deprives clinicians of the initiative to stratify and intervene in patients before treatment. In contrast, our findings show that patients with different responses to chemotherapy have differences in serum immunoproteomics before chemotherapy begins, advancing the time window for patient stratification and intervention. Inspired by PD-L1 and CCL20, we developed a preoperative serum proteomics panel for predicting responses to perioperative chemotherapy, termed PSRscore. By calculating the preoperative serum protein levels of four immune proteins, patients can be divided into two groups. Patients with low PSRscore tend to have poorer treatment responses and may benefit from combination immunotherapy. This scoring system has significant clinical application potential for patient stratification. Notably, the establishment of PSRscore is based on an Asian cohort receiving platinum-based chemotherapy. The performance of these immune markers in non-Asian patients receiving taxane-based regimens needs further validation.
We believe that serum protein biomarkers have special clinical significance in preoperative stratification of gastric cancer patients. Almost all existing molecular classifications of gastric cancer rely on tumor tissues obtained from surgical or endoscopic resections. The TCGA classification is the most famous example, where microsatellite instability (MSI) patients have been shown to benefit more from immunotherapy, while genomically stable (GS) patients respond poorly to chemotherapy. However, these molecular classifications are rarely used in clinical practice. One important reason is that most molecular classifications rely on complex molecular techniques, such as qPCR, in situ hybridization, and even omics technologies, which are not widely available in most clinical settings. Additionally, in gastric cancer, obtaining tumor samples preoperatively relies on endoscopic biopsies. Gastric cancer exhibits significant intratumoral heterogeneity, and the limited depth of biopsies largely affects the representativeness of biopsy samples. Therefore, determining the molecular classification of gastric cancer before gastrectomy has been very challenging. In contrast, serum proteomics encompasses both systemic and tumor-local characteristics, making it sensitive and informative. Serum samples can be easily obtained in clinical settings with limited harm to patients. Serum protein biomarkers, such as prostate-specific antigen (PSA) or alpha-fetoprotein (AFP), have been used for decades in cancer diagnosis and follow-up. Various hospitals widely provide equipment for measuring serum proteins and trained personnel. These factors endow gastric cancer serum proteomics research with significant clinical relevance. Future efforts should establish serum protein classifications for gastric cancer to guide perioperative treatment.
Limitations#
There are some limitations in the study that need to be noted. First, the number of serum samples during treatment is relatively small, which limits the statistical power to draw certain conclusions. Second, multiplex immunofluorescence (mIF) only measured key immune cells in the tumor microenvironment (TME). Single-cell sequencing could better depict the TME. Third, some conclusions and recommendations from this study should be further validated in prospective cohorts or even randomized controlled trials of patients receiving preoperative chemotherapy. These limitations should be considered when interpreting the data.
Overall, we described the systemic immune system and tumor microenvironment in gastric cancer patients and demonstrated their association with responses to preoperative chemotherapy. We identified serum biomarkers for predicting treatment responses and prognosis. This work emphasizes the fundamental yet largely underestimated role of systemic immunity in preoperative chemotherapy for gastric cancer, supports a patient stratification strategy based on preoperative serum immunoproteomics, and highlights the importance of comprehensively depicting immunity in future research.