The prognostic value of immune infiltration signature and its potential mechanism in gastric cancer

doi:10.21203/rs.3.rs-19939/v1

Download PDF

Research article

The prognostic value of immune infiltration signature and its potential mechanism in gastric cancer

https://doi.org/10.21203/rs.3.rs-19939/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

The development of gastric cancer (GC) is closely associated with tumor microenvironment. Tumor infiltrating immune cells (TIICs), as a significant component of the tumor microenvironment, not only dominates the immune response of tumors, but is also valuable prognostic indicators. In our study, GC patients from GEO database and TCGA database were enrolled. The proportion of 22 kinds of TIICs were accurately evaluated by a deconvolution algorithm named CIBERSORT. The result showed TIICs component were varied in cancer tissue and adjacent tissue, besides the TIICs component had an effect on overall survival (OS). After LASSO cox regression and stepwise filter optimization analyzed, tumor tissues which were riches in M2 macrophages and activated mast cells, lack of plasma cells or activated CD4 memory T cells and neutrophils may cause a poor prognosis. Based on this result, a novel nomogram was stablished to assess the survival rate of GC patients. A survival immune risk (SIR) score model was further developed to validate the relationship between immune infiltration and pathological traits or oncogenic pathways in GC. In conclusion, this study may be contributed to finding new immunotherapy targets and providing individualize treatment strategies for GC patients.

Cancer Biology

Oncology

Gastric cancer

CIBERSORT

Immune cell

Tumor microenvironment

Overall survival

Gastric cancer (GC), as the sixth commonly diagnosed cancer, the mortality rate of which ranks second, same as liver cancer in global [1]. GC was reported approximately 1 million new cases and induced 780,000 deaths around the world in 2018 [1]. Meanwhile, 27,000 estimated new cases and 11,000 estimated deaths among the American in 2019 [2]. In recent years, with the improvement of techniques in surgery, endoscopic, radiotherapy, chemotherapy and targeted therapy, gastric cancer comprehensive treatment has made great progress [3]. It has been shown that early gastric cancer 5-year mortality rate was significantly reduced [4]. However, it is worth noting that advanced GC patients’ 5-year survival rate was remaining lower than 30% [5]. Although the specific mechanism of gastric cancer development remains unclarified, it was believed that the onsets of GC is highly related to personal diet, infection of H. pylori and genetic factors [3–5]. It’s believed that low survival rate was related to cancer peritoneal dissemination, blood transfer and lymphatic metastasis [6].

Tumor infiltrating immune cell (TIIC), as a significant component of the tumor microenvironment, marks the immune status of tumors. In recent years, some studies show that the presence of various subsets of TIIC have different influence on the long-term prognosis of patients with GC. For example, after analyzing 107 Chinese gastric cancer patients, Wang B et.al claimed that the high infiltration of CD68⁺ macrophages and Foxp3⁺ cells in tumor tissue may be benefited in positive prognosis [7], whereas other studies showed that the high density of intratumor FOXP3⁺ cells was correlated with an adverse prognosis [8, 9]. In addition, some literature has reported that the rich quantity of CD8⁺ T cells in both of intra-or extra-gastric tumor collated with good prognosis [10–12]. However, a recent study revealed that the increased CD8⁺ T cells may promote the expression of programmed death ligand 1 (PD-L1), which in turn causes a poor prognosis [13]. These contradictory results may due to the fact that most researchers use immunohistochemistry (IHC) to assess the composition or proportion of TIICs. However, the limitations of this method may cause experimental errors, IHC differentiates cell type by specific marker, but multiple TIICs may express the same marker on the surface, which makes it hardly to accurately measure the TIIC density in gastric cancer tissues. Alternatively, although the proportion of TIICs can be accurately quantified by flow cytometry, it requires a lot of labor and is high cost.

At present, with the advantages of celerity and accuracy, deconvolution was considered to be one of the best algorithms for calculating the content and proportions of cells in complex tissue. In the past few years, some researchers attempted to estimate the immune cell content from the profile of gene expression in tissues by designing variety of deconvolution algorithms, such as “CSSAM” [14] and “SCLVM” [15]. In 2015, Newman et al. demonstrated a new algorithm called CIBERSORT [16]. As a variant of the Support Vector Machine (SVM) algorithm, its performance has been proved to be superior to other methods. CIBERSORT was used to predict the abundance of TIICs based on standardized gene expression data. Results obtained from CIBERSORT was well verified by flow cytometry [17]. In recent years, some studies have successfully evaluated the composition of TIICs in various tumors by CIBERSORT, such as lung cancer [18], liver cancer [19], prostate cancer [20] and bladder cancer [21].

In this study, CIBERSORT was used to estimate density differences of 22 TIICs between tumor and non-tumor tissues of gastric cancer by analyzing 547 key genes expression based on public database of Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA). Then, least absolute shrinkage and selection operator (LASSO) Cox regression analysis was utilized to screen several TIICs which have a better prognosis value. Furthermore, an immune-risk score was constructed for further evaluating the potential relationship between immune microenvironment, clinical feature and oncogenic pathway.

2.1 Data from GEO and TCGA database

The transcriptome expression data and gastric cancer related clinical information was obtained from both GEO and TCGA. A total of 300 gastric cancer patients were included in the series of GSE66229 based on GPL570 platform by Asian Cancer Research Group, which provides the gene expression data of 300 tumor samples and extra 100 paired adjacent normal samples [22]. Additionally, another data of stomach adenocarcinoma were obtained from TCGA (https://portal.gdc.cancer.gov/) up to June 30, 2019, which contains HTSeq-FPKM type gene expression profiles and clinical information of 343 tumor samples and 30 adjacent non-tumorous samples. For TCGA cohort, at first, the method of variance modeling at the observational level (also known as voom) was used for converting the type of mRNA expression data to a microarray-similar result. Subsequently, data was normalized by “limma”, a package of R. For GEO cohort, data was directly normalized by “limma” without converting. Insufficient follow-up information data recognized as invalid data was deleted in two cohorts.

2.2 Estimation of immune infiltration in samples

After cohorts were normalized, a new algorithm called CIBERSORT (http://cibersort.stanford.edu/) was applied to accurately assessing the proportion of invasive immune cell types. Based on “LM22”, a characteristic matrix file consists 547 genes for defining 22 subtypes of TIICs, CIBERSORT can estimate the proportion of naïve B cells, memory B cells, plasma cells, CD8⁺ T cells, naïve CD4⁺ T cells, resting memory CD4⁺ T cells, activated memory CD4⁺ T cells, follicular helper T cells, regulatory T cells (Treg cells), gamma delta T cells, resting NK cells, activated NK cells, monocytes, M0 macrophages, M1 macrophages, M2 macrophages, resting dendritic cells, activated dendritic cells, resting mast cells, activated mast cells, eosinophils, and neutrophils in each sample by a deconvolution algorithm at 1000 permutations. The p-value contained in result of CIBERSORT reflects the reliability of this program. Sample with a p-value over 0.05 was excluded in subsequent analysis.

2.3 Statistical analysis

The t-test and Kruskal-Wallis test were applied to comparing continuous variables, χ2 test and Fisher’s exact were used for categorical variables. Differential infiltration status of each TIICs in tumor and normal tissue was analyzed by Wilcoxon’s Sign Rank test. The correlation between any two TIICs was estimated by Pearson correlation coefficient. To explore the effect of each TIIC on prognosis respectively, patients were divided into two groups depend on density according to the median of the proportion, and the relationship between TIICs and overall survival (OS) time and was demonstrated by the Kaplan-Meier (K-M) curve by long-rank test. To identify the key TIICs affect tumor prognosis, we set GEO cohort (GSE66229) as the train group and TCGA cohort as the test group, and the R package named “glmnet” was applied to implement LASSO COX regression for screening out several TIICs with higher prognostic value. Then a nomogram for predicting the probability of OS were created by R package named “rms” based on results of the multivariate Cox analysis including age, gender and screened TIICs as the variables. Receiver operating characteristic (ROC) curve was performed by R package named “ROC”, and the area under curve (AUC) was utilized as an indicator for revealing the discriminative capabilities of the nomograms and the result of LASSO model, and the calibration curve was also established for examining the accuracy of nomogram. Then, “survival immune risk” (SIR) score was established by screened TIICs based on the Z-index of Cox regression, and patients was divided into “high-” and “low-” SIR group based on the cutoff value calculated by R package named “OptimalCutpoints”. Finally, through R package named “clusterProfiler” [23], the differentiations of immune-related pathways from KEGG in two groups were analyzed by GSEA. All statistical analyses were executed by R (version 3.6.0, https://www.r-project.org) and R studio (version 1.2.1335, https://www.rstudio.com).

3.1 The characteristics of patients in GEO and TCGA cohort

The overall process for our study was simplified, as shown in a flowchart in Figure.1. Specific clinical information for each patient in GEO or TCGA cohort was collected into Supplementary Tables 1 and 2. After calculation by CIBERSORT, sample with p-value > 0.05 were discarded. 100 sets of non-tumor samples and 299 sets of tumor samples in GEO cohort and 13 sets of non-tumor samples and 166 sets of tumor samples in TCGA cohort remained. Clinical characteristics including age, gender, pathologic stage and TNM stage of these patients were counted and settled in Table 1.

3.2 Heterogeneity of TIIC distribution in normal and tumor samples

The proportion of 22 TIICs in GEO or TCGA cohort was summarized in histogram (Fig. 2A, 2B). Obviously, although the variation of TIICs in each cohort was distinct, the content of T cells CD4 naive in both cohorts was almost negligible. Therefore, we further analyzed the differential expression of other 21 TIICs in tumor and normal tissues from these two cohorts. As shown in Table 2 and Table 3, four TIICs accounted for more than half of the population in the normal samples of GEO cohort, including T cells CD4 memory resting, macrophages M2, plasma cells and mast cells resting, while plasma cells, T cells CD4 memory resting and T cells CD8 occupied more than half proportion of TIICs in the normal samples of TCGA cohort. For tumor samples, Macrophages M2, T cells CD4 memory resting, Plasma cells, Macrophages M1 and Neutrophils accounted for more than half of TIICs in GEO cohort, while T cells CD4 memory resting, T cells CD8, Macrophages M0, Macrophages M2 and Macrophages M1 accounted for the more than half of TIICs in TCGA cohort. Figures 2C and 2D revealed the difference in infiltrating TIICs between normal and tumor samples by violin diagrams with Wilcoxon’s Sign Rank test, and p < 0.001 was considered to be significant. In GEO cohort, there were 15 TIICs showed significant diversity in normal and tumor samples, with plasma cells, Macrophages M0, Macrophages M1 and Macrophages M2 also significantly different in the TCGA cohort. In both cohorts, the content of macrophages M0 and Macrophages M1 in tumor samples was significantly higher than that in normal samples, while the plasma cells were obviously lower than that in normal samples. It is worth noting that the comparison of Macrophages M2 showed contradictory results in two cohorts, comparing with tumor samples, the density of Macrophages M2 was higher in normal samples of the GEO cohort (18.37% ± 5.94%), while in the TCGA cohort, the infiltration of M2 cells in normal samples was lower (6.02% ± 2.45%). Figures 2E and 2F revealed the infiltration of 21 TIICs in samples of GEO and TCGA cohort by heatmap. Obviously, the composition of TIICs varies widely regardless within or between groups of normal and tumor. Therefore, the heterogeneity of TIICs in GC tissues may regulate different pathological processes, leading to different clinical prognosis.

Figure 2G reveals the correlation between each TIICs in tumor sample of GEO and TCGA cohort. The red represented positive correlation while the blue represented negative correlation. The correlation coefficient, also known as the R value, was shown in the box where every two TIICs intersect. The size of the bubble visualized the closeness of the correlation. As a result, two most positively correlated TIICs in the GEO cohort were activated CD4⁺ T cell and macrophages M1 with an R value of 0.65 (R = 0.42 in TCGA cohort), while in the TCGA cohort they were neutrophils and activated mast cells with an R value of 0.46 (R = 0.39 in GEO cohort). In contrast, the two TIICs with the most negative correlation in GEO cohort were resting CD4 memory T cells and activated CD4 memory T cells with an R value of -0.58 (R = -0.49 in TCGA cohort), while in the TCGA cohort they were CD8 T cells and resting CD4 memory T cells with an R value of -0.52 (R = -0.53 in GEO cohort).

3.3 Correlation of TIIC infiltration with clinical survival rate

We are following explored the effects of TIICs as a single factor on the survival rate of GC patients by analyzing clinical information provided in GEO and TCGA cohort, and results were summarized in Supplemental Figure S1. Among 21 TIICs, the activated CD4 memory T cells and the resting CD4 memory T cells showed a close correlation with survival rate in both two cohorts. As shown in Fig. 3A-3D, tumor with high-density activated CD4 memory T cells improved the long-term survival rate (p < 0.0001 in GEO, p = 0.029 in TCGA), while high-density resting CD4 memory T cells may significantly shorten patient’s survival time (p = 0.0096 in GEO, p = 0.034 in TCGA). In addition, high-density of macrophages M1 in the GEO cohort and CD8 T cells in the TCGA cohort were closely associated with a better prognosis, although not significant enough in each other’s cohort (Fig. 3E-3H).

3.4 Establishment of GC survival predictive nomogram

To further explorer the TIICs with effective prognosis value, we considered the GEO cohort as a training group and established an immune risk model. In addition, the TCGA cohort was set as the test group to verify the reliability. B cells naïve, Plasma cells, T cells CD4 memory activated, T cells gamma delta, Macrophages M2, Mast cells activated and Neutrophils were selected by the program of LASSO regression analysis and tenfold cross-validation (Fig. 4A, 4B), but B cells naïve and T cells gamma delta were excluded by stepwise filter optimization method. Finally, in the nomogram shown in Fig. 4C, age, gender and five kinds of TIICs were considered as indicators for predicting the prognosis. The “point” bar shown on the top reveals the independent risk of each indicators, and the “total point” bar on the bottom was the synthesis risk corresponding to the 1-, 3- and 5- survival rate. Obviously, the M2 macrophages and activated mast cells were strongly associated with poor prognosis, while plasma cells, activated CD4 memory T cells and neutrophils contributed to a better prognosis. In addition, ROC curves and calibration curves performed the predictive accuracy of this nomogram in the train group (GEO cohort) (Figure. 4D, 4F) and test group (TCGA cohort) (Figure. 4E, 4G), which demonstrated that this nomogram could precisely determine the prognosis survival rates.

3.5 The establishment and validation of survival immune risk (SIR) score model

To further explore the overall role of these five most useful prognostic TIICs in the course of disease, we constructed a "survival immune risk (SIR) score" consisting of the fractions of the above 5 TIICs (The formula based on the Z-index of Cox regression: SIRS = 2.724*Macrophages M2 + 3.030*Mast cells activated – 2.184*Plasma cells – 3.133*T cells CD4 memory activated – 2.643*Neutrophils). Then, SIR score was divided into “high-” and “low-” group based on the cutoff value of -0.03630041 (AUC = 0.687). As shown in Fig. 5A, patients with higher SIR in train group significantly impaired the survival prognosis, which was validated by test group (Figure. 5B). Two ROC curves perform the predictive accuracy of SIR for 1-, 3-, and 5-year survival rates of the train group (Figure. 5C) and test group (Figure. 5D).

3.6 The relationship between SIR score and clinical features

Then we analyzed the relationship between the SIR risk and clinical features including pathological stage, TMN stage and the pathological classification based on Laurén’s criteria in samples of GEO cohort. Obviously, the SIR score increased as the elevated pathological stage (Figure. 6A, p < 0.001), and patients in T3 and T4 stage has notability higher SIR score than T2 stage patients (Figure. 6B). In addition, N3 stage patients showed a significant higher SIR score than other patients (Figure. 6C). However, SIR scores are not different in different M stages (Figure. 6D). Laurén classification of gastric cancer was mainly divided into intestinal type carcinoma, diffuse gastric carcinoma and mixed type carcinoma according to cancer cell morphology and histochemistry. As shown in Figure. 6E, gastric cancer patients with diffuse type tumors possessed higher SIR scores than intestinal type. It is worth mentioning that in TCGA cohort, statistically significant differences were not exist in patients with different clinical traits (Supplemental Figure. S2)

3.7 Identification of SIR score associated oncogenic pathways

Patients in GEO cohort with high- and low- SIR scores were included in the GSEA analysis consist of KEGG (Supplemental Figure. S3). Obviously, the results showed that tumor tissues with high SIR scores significantly enhanced the capacity of oxidative phosphorylation, DNA repair, and glucose metabolism (Figure. 7A), while the pathways of immune cell response, cytokine reactivity, regulation of cytoskeleton and focal adhesion were enriched in low-SIR score tissues (Figure. 7C). However, similar pathways were not noticed in the TCGA cohort.

The development of tumors is controlled by a series of complicated biological processes. The tumor microenvironment (TME), mainly composed with fibroblasts, myofibroblasts, neuroendocrine cells, adipocytes, immune and inflammatory cells, blood and lymphatic vascular networks, and extracellular matrices [24], not only acts as a soil for the breeding of malignant tumors, but also affects the characteristics of tumor invasion and metastasis. In turn, tumor also affects the TME, making it more suitable for its growth [25]. The diversity of components, activation and function of TIICs in TME plays a significant role in regulation of biological behavior of tumors by various cytokines or signaling pathways [26].

In this study, we first analyzed infiltrating TIICs with significant differences between tumors and normal tissues. In both two cohorts, the density of plasma cells was significantly higher in normal tissues. It was reported that the enrichment of CD138⁺ and IGKC⁺ plasma cells has a positive effect on the prognosis of patients with gastric cancer or esophageal cancer [27]. Therefore, same as cellular immunity, humoral immunity dominated by plasma cells may play an important role in inhibiting the development of digestive system tumors,but there are few studies focus on this area, the mechanism is worthy for further research. A long-term follow-up study confirmed that macrophage was a key factor in the progression of non-cancerous gastric adenomas to GC [28], and another animal experiment validated that the number of macrophages were positively correlated with gastric tumor growth [29]. We found that both Macrophage M0 and M1 were enriched in tumor tissues in both cohorts, but M2 macrophages proportion in two cohorts was inconsistent. Thus, the relationship between polarization state of macrophages and gastric tumor formation need to be further clarified.

Previous clinical studies have explored the relationship between CD4⁺ T cells and the prognosis of gastric cancer [30, 31]. CD4⁺ lymphocytes are composed by helper T cells, regulatory T cells, and memory T cells, these cells secrete a variety of cytokines. Therefore, the heterogeneity of CD4⁺ T cells complicates their function in tumor microenvironment [32]. In our research, we set each TIICs as a single factor and found the highest two infiltration CD4⁺ T memory cell subtypes (activated and resting) which indicated completely different prognosis. In addition, they showed a significant negative correlation in the amount of infiltration. We also found that the high density of M1 macrophages in GEO cohort and CD8⁺ T cells in TCGA cohort promoted the prognosis. M1 macrophages secrete a variety of cytokines consist of IL-1β, IL-6, IL-12 and TNF-α, while CD8⁺ T cell can secrete cytokines such as interferon gamma and can express PD-L1, both cells showed a tumoricidal ability [13, 33]. Interestingly, we found a positive correlation between these two cells and activated CD4⁺ T cells in GC immune infiltration, which suggested that activated CD4⁺ T cells have a profound effect on the establishment of tumoricidal microenvironment.

In order to find the most valuable prognostic indicators, we applied LASSO regression and stepwise filter optimization method to analyze TIICs in the GEO database. Finally, five TIICs were screened and were set as multiple factors for Cox analysis with age and gender. Obviously, in addition to the activated CD4 memory T cells and plasma cells we mentioned above, neutrophils cells contribute to a better prognosis. The antibacterial function of neutrophils is well known, but they also play an important role in cancer biology [34]. Previous studies have explored the relationship between the infiltration of neutrophils and GC prognosis, but produced variety of results. For example, Caruso et al found that neutrophils infiltration was a favorable factor for relapse-free survival (RFS) [35], while Zhao et al claimed that CD15⁺ neutrophil infiltration in gastric cancer is associated with poor prognosis in the next research [36]. This contradiction may be due to the inadequacy of the method used to identify the strength of neutrophils, and the selected characteristic markers are also not specific. Recently, some researchers adopted CD66b as a more specific activation marker to identify neutrophils, and found that GC patients with high expression of CD66b possessed a survival benefit [37]. Here, through CIBERSORT, we further confirmed that neutrophil was one of the multifactor prognostic indicators that marked a better outcome. Moreover, some retrospective studies have shown that submucosal mast cells adjacent to the tumor are involved in advanced disease and metastasis of human GC [38, 39]. In 2019, an article published on Nature Communication revealed that the inactivation of mast cells in gastric cancer tissue effectively inhibited the accumulation of tumor-associated macrophages, reducing tumor cell proliferation and angiogenesis [29]. Interestingly, we found that the M2 macrophages and activated mast cells were strongly associated with poor prognosis, suggested that these cells can be promising GC treatment targets.

Next, we integrate above five representative prognostic factors into a formula based on the Z-index of Cox regression to improve prognostic ability and obtain accurate risk stratification. Obviously, the SIR score constructed by us accurately separated OS curves, patients with higher SIR significantly impaired the survival prognosis, which facilitated us to explore the relationship between overall risk of TIICs and clinical information. As shown in Figure.6, besides M stages, patients meeting with elevated pathological, T and N stage were provided with higher SIR score, showing that these five core TIICs account for a gradual deterioration of tumor microenvironment as the disease progresses. In addition to the TMN stages, the Laurén classification has become the one of the most commonly used classification of GC in all classification systems, including intestinal type, diffuse type and mixed type [40]. The intestinal type is associated with intestinal metaplasia of the gastric mucosa and composed of Papillary adenocarcinoma, Tubular adenocarcinoma and Mucinous adenocarcinoma, while diffuse type often defined the poorly cohesive carcinomas, such as Signet-ring cell carcinoma, a type of GC generally considered to be one of the most malignant form. Finally, GC with an uncommon histology is described as Mixed type. In traditional, GC with diffuse type is early metastasized by blood circulation or implantation due to the lack of adhesion, resulting in a poor prognosis [41, 42]. Interestingly, in our study, we found that there is a great relationship between SIR score and Laurén classification. Patients with diffuse type possessed significant higher SIR scores than intestinal type, indicating a higher risk of immunological characteristics, which may be another reason for their poor prognosis. Unfortunately, there is no related study have focused on the immune microenvironment of GC with diffuse type, which requires further evaluate to provide a crucial rationale for property immunotherapies.

Finally, we searched for SIR score associated oncogenic pathways to further clarify the mechanism for its influence. Through GSEA analysis, we were surprised to find that tumor with higher SIR score had stronger metabolic capabilities, specifically in oxidative phosphorylation and glucose metabolism. Oxidative phosphorylation, closely related to the progression of GC, is a crucial prerequisite for aerobic glycolysis of tumors, a process known as the Warburg effect [43]. Glycolysis further provides the energy required for tumor growth, increases its size, and ultimately strengthens the tumor malignancy [44, 45]. In addition, genes involved in DNA repair were also up-regulated, while those involved in the regulation of the cytoskeleton and focal adhesions are down-regulated, contributing to enhance radiotherapy resistance and metastasis capacity of GC. It is worth mentioning that some important immune-related pathways were widely enriched in tissues with low-SIR scores, including immune cell response, cytokine reactivity, T cell receptor signaling pathway and trans endothelial migration of leukocyte, suggesting that GC tissues with high SIR scores may build an immunosuppressive microenvironment to enhance tumor escape by regulating immune response.

Before we start this research, a previous study has analyzed the correlations of TIICs and the prognostic value of GC patients by designing an immune risk score derived from 11 kinds of immune cells [46], which is meaningful but encountered some issues. At first, this study has been questioned by Zhao et al. recently in sampling method and concepts presented [47]. According to their description, the authors integrated multiple GSE datasets and then randomly divided them into training and validation groups, but the variance error between was not given, creating the possibility of data reuse and leading to questionable results. In our study, datasets from two totally different platforms, GEO and TCGA, were respectively adopted as training and validation groups, making the results more reliable. In addition, in their formula provided in supplements, the naïve CD4⁺ T cells were regarded as an important element, with a weight of 0.713, which is higher than any other immune cells. However, we found the content of naïve CD4⁺ T cells was almost negligible in both of GEO and TCGA cohort, which was also found in multiple kinds of cancer [20, 48, 49] .

In summary, our study evaluated the TIIC microenvironment of GC tumors and adjacent tissues using the CIBERSORT deconvolution algorithm. It's worth noting that not only the content of plasma cells is significant difference between cancerous tissue and adjacent tissues, but also has an important prognosis value of patients, deserving further in-depth research and develop immunotherapeutic. Subsequently, five most prognostic immune cells were screened by a series of analysis and formed an SIR model. We found that high-SIR is tightly associated with more malignant pathological grades and traits. Finally, through KEGG enriched by GSEA, we found that tissues with high-SIR were more likely to upregulate oncogenic pathways related to promote the development of GC. However, in TCGA cohort, not only there is no statistically significant differences in SIR score between patients with different clinical traits, but also similar oncogenic pathways were not noticed by GSEA analysis, which may be due to some factors such as sample size. In addition, besides mentioned above, there are still some limitations in our study, whether in the TCGA or GEO dataset, the information of few patients is incomplete, including TMN stage, Laurén classification and age, etc. In addition, as a retrospective study, sample size is relatively small, and potential treatment heterogeneity cannot be ruled out. Thus, a multi-center, large-sample and well-designed prospective study is necessary for further validation of our results.

Gastric Cancer

TIIC

Tumor Infiltrating Immune Cell

Overall Survival

SIR

Survival Immune Risk

PD-L1

Programmed Death Ligand 1

IHC

Immunohistochemistry

SVM

Support Vector Machine

GEO

Gene Expression Omnibus

TCGA

The Cancer Genome Atlas

LASSO

Least Absolute Shrinkage and Selection Operator

ROC

Receiver Operating Characteristic

AUC

Area Under Curve

GSEA

Gene Set Enrichment Analysis

Acknowledgements

Not applicable

Funding

Not applicable

Availability of data and materials

Datasets generated or analyzed during the current study are available in the repository GSE66229 of GEO(https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE66229)

and TCGA repository (https://portal.gdc.cancer.gov/).

Authors’ contribution

J.Z and TY. Y conceived, designed the project and collected the data. X.L and HX. D analyzed the data. CH.W and SS.X provided administrative support. All authors contributed to the manuscript writing and have read and approved the manuscript.

Competing interests

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Consent to publish

Not applicable

Ethics approval and consent to participate

Data in this study are obtained from public databases and do not require any of ethics approval or consent to participate.

Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA, Jemal A. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. Ca Cancer J Clin. 2018;68:394–424.
Siegel RL, Miller KD, Jemal A. Cancer statistics, 2019. Ca Cancer J Clin. 2019;69:7–34.
Thrumurthy SG, Chaudry MA, Hochhauser D, Mughal M. The diagnosis and management of gastric cancer. Bmj Br Medical J. 2013;347 nov04 1:f6367.
Hamashima C, Shabana M, Okada K, Okamoto M, Osaki Y. Mortality reduction from gastric cancer by endoscopic and radiographic screening. Cancer Sci. 2015;106:1744–9.
Isobe Y, Nashimoto A, Akazawa K, Oda I, Hayashi K, Miyashiro I, et al. Gastric cancer treatment in Japan: 2008 annual report of the JGCA nationwide registry. Gastric Cancer. 2011;14:301–16.
Maleki SS, Röcken C. Chromosomal Instability in Gastric Cancer Biology. Neoplasia. 2017;19:412–20.
Wang B, Xu D, Yu X, Ding T, Rao H, Zhan Y, et al. Association of Intra-tumoral Infiltrating Macrophages and Regulatory T Cells Is an Independent Prognostic Factor in Gastric Cancer after Radical Resection. Ann Surg Oncol. 2011;18:2585–93.
Zhou S, Shen Z, Wang Y, Ma H, Xu S, Qin J, et al. CCR7 Expression and Intratumoral FOXP3 + Regulatory T Cells are Correlated with Overall Survival and Lymph Node Metastasis in Gastric Cancer. Plos One. 2013;8:e74430.
Perrone G, Ruffini PA, Catalano V, Spino C, Santini D, Muretto P, et al. Intratumoural FOXP3-positive regulatory T cells are associated with adverse prognosis in radically resected gastric cancer. Eur J Cancer. 2008;44:1875–82.
Kim K-J, Lee KS, Cho HJ, Kim YH, Yang HK, Kim WH, et al. Prognostic implications of tumor-infiltrating FoxP3 + regulatory T cells and CD8 + cytotoxic T cells in microsatellite-unstable gastric cancers. Hum Pathol. 2014;45:285–93.
Liu K, Yang K, Wu B, Chen H, Chen X, Chen X, et al. Tumor-Infiltrating Immune Cells Are Associated With Prognosis of Gastric Cancer. Medicine. 2015;94:e1631.
Wen T, Wang Z, Li Y, Li Z, Che X, Fan Y, et al. A Four-Factor Immunoscore System That Predicts Clinical Outcome for Stage II/III Gastric Cancer. Cancer Immunol Res. 2017;5:524–34.
Thompson ED, Zahurak M, Murphy A, Cornish T, Cuka N, Abdelfatah E, et al. Patterns of PD-L1 expression and CD8 T cell infiltration in gastric adenocarcinomas and associated immune stroma. Gut. 2017;66:794.
Shen-Orr SS, Tibshirani R, Khatri P, Bodian DL, Staedtler F, Perry NM, et al. Cell type–specific gene expression differences in complex tissues. Nat Methods. 2010;7:287.
Buettner F, Natarajan KN, Casale FP, Proserpio V, Scialdone A, Theis FJ, et al. Computational analysis of cell-to-cell heterogeneity in single-cell RNA-sequencing data reveals hidden subpopulations of cells. Nat Biotechnol. 2015;33:155–60.
Newman AM, Liu CL, Green MR, Gentles AJ, Feng W, Xu Y, et al. Robust enumeration of cell subsets from tissue expression profiles. Nat Methods. 2015;12:nmeth.3337.
Chen B, Khodadoust MS, Liu CL, Newman AM, Alizadeh AA. Cancer Systems Biology, Methods and Protocols. Methods Mol Biology Clifton N J. 2018;1711:243–59.
Liu X, Wu S, Yang Y, Zhao M, Zhu G, Hou Z. The prognostic landscape of tumor-infiltrating immune cell and immunomodulators in lung cancer. Biomed Pharmacother. 2017;95:55–61.
Rohr-Udilova N, Klinglmüller F, Schulte-Hermann R, Stift J, Herac M, Salzmann M, et al. Deviations of the immune cell landscape between healthy liver and hepatocellular carcinoma. Sci Rep-uk. 2018;8:6220.
Meng J, Liu Y, Guan S, Fan S, Zhou J, Zhang M, et al. The establishment of immune infiltration based novel recurrence predicting nomogram in prostate cancer. Cancer Med-us. 2019.
Tan TZ, Rouanne M, Tan KT, Huang RY-J, Thiery J-P. Molecular Subtypes of Urothelial Bladder Cancer: Results from a Meta-cohort Analysis of 2411 Tumors. Eur Urol. 2018;75 CA Cancer J Clin 65 2015:423–32.
Oh SC, Sohn BH, Cheong J-H, Kim S-B, Lee JE, Park KC, et al. Clinical and genomic landscape of gastric cancer with a mesenchymal phenotype. Nat Commun. 2018;9:1777.
Yu G, Wang L-G, Han Y, He Q-Y. clusterProfiler: an R Package for Comparing Biological Themes Among Gene Clusters. Omics J Integr Biology. 2012;16:284–7.
Chen F, Zhuang X, Lin L, Yu P, Wang Y, Shi Y, et al. New horizons in tumor microenvironment biology: challenges and opportunities. Bmc Med. 2015;13:45.
Langley RR, Fidler IJ. The seed and soil hypothesis revisited—The role of tumor-stroma interactions in metastasis to different organs. Int J Cancer. 2011;128:2527–35.
Oluwadara O, Giacomelli L, Brant X, Christensen R, Avezova R, Kossan G, et al. The role of the microenvironment in tumor immune surveillance. Bioinformation. 2011;5:285–90.
Fristedt R, Borg D, Hedner C, Berntsson J, Nodin B, Eberhard J, et al. Prognostic impact of tumour-associated B cells and plasma cells in oesophageal and gastric adenocarcinoma. J Gastrointest Oncol. 2016;7:848–59.
Taniyama D, Taniyama K, Kuraoka K, Zaitsu J, Saito A, Nakatsuka H, et al. Long-term follow-up study of gastric adenoma; tumor-associated macrophages are associated to carcinoma development in gastric adenoma. Gastric Cancer. 2017;20:929–39.
Eissmann MF, Dijkstra C, Jarnicki A, Phesse T, Brunnberg J, Poh AR, et al. IL-33-mediated mast cell activation promotes gastric cancer through macrophage mobilization. Nat Commun. 2019;10:2735.
Zhang R, Li F, Li H, Yu J, Ren X. The clinical significance of memory T cells and its subsets in gastric cancer. Clin Transl Oncol. 2014;16:257–65.
WAKATSUKI K, SHO M, YAMATO I, TAKAYAMA T, MATSUMOTO S, TANAKA T, et al. Clinical impact of tumor-infiltrating CD45RO + memory T cells on human gastric cancer. Oncol Rep. 2013;29:1756–62.
Yu P-C, Long D, Liao C-C, Zhang S. Association between density of tumor-infiltrating lymphocytes and prognoses of patients with gastric cancer. Medicine. 2018;97:e11387.
Isidro RA, Appleyard CB. Colonic macrophage polarization in homeostasis, inflammation, and cancer. Am J Physiology Gastrointest Liver Physiology. 2016;311:G59–73.
Piccard H, Muschel RJ, Opdenakker G. On the dual roles and polarized phenotypes of neutrophils in tumor development and progression. Crit Rev Oncol Hemat. 2012;82:296–309.
Caruso RA, Bellocco R, Pagano M, Bertoli G, Rigoli L, Inferrera C. Prognostic Value of Intratumoral Neutrophils in Advanced Gastric Carcinoma in a High-Risk Area in Northern Italy. Modern Pathol. 2002;15:831–7.
Zhao J, Pan K, Wang W, Chen J, Wu Y, Lv L, et al. The Prognostic Value of Tumor-Infiltrating Neutrophils in Gastric Adenocarcinoma after Resection. Plos One. 2012;7:e33655.
Zhang H, Liu H, Shen Z, Lin C, Wang X, Qin J, et al. Tumor-infiltrating Neutrophils is Prognostic and Predictive for Postoperative Adjuvant Chemotherapy Benefit in Patients With Gastric Cancer. Ann Surg. 2018;267:311–8.
Yano H, Kinuta M, Tateishi H, Nakano Y, Matsui S, Monden T, et al. Mast cell infiltration around gastric cancer cells correlates with tumor angiogenesis and metastasis. Gastric Cancer. 1999;2:26–32.
ZHAO Y, WU K, CAI K, ZHAI R, TAO K, WANG G, et al. Increased numbers of gastric-infiltrating mast cells and regulatory T cells are associated with tumor stage in gastric adenocarcinoma patients. Oncol Lett. 2012;4:755–8.
Berlth F, Bollschweiler E, Drebber U, Hoelscher AH, Moenig S. Pathohistological classification systems in gastric cancer: diagnostic relevance and prognostic value. World J Gastroentero. 2014;20:5679–84.
Guglielmi A, Manzoni G de, Tomezzoli A, Ricci F, Pelosi G, Laterza E, et al. [Prognostic value of histologic classifications of advanced stomach cancer: comparative study of Lauren’s and Goseki’s classifications]. Chir Italiana. 1997;49:45–9.
Qiu M, Cai M, Zhang D, Wang Z, Wang D, Li Y, et al. Clinicopathological characteristics and prognostic analysis of Lauren classification in gastric adenocarcinoma in China. J Transl Med. 2013;11:58.
Feichtinger RG, Neureiter D, Skaria T, Wessler S, Cover TL, Mayr JA, et al. Oxidative Phosphorylation System in Gastric Carcinomas and Gastritis. Oxid Med Cell Longev. 2017;2017:1–14.
Wang H-M, Huang C-M, Zheng C-H, Li P, Xie J-W, Wang J-B, et al. Tumor size as a prognostic factor in patients with advanced gastric cancer in the lower third of the stomach. World J Gastroentero. 2012;18:5470–5.
Im WJ, Kim MG, Ha TK, Kwon SJ. Tumor Size as a Prognostic Factor in Gastric Cancer Patient. J Gastric Cancer. 2012;12:164.
Zeng D, Zhou R, Yu Y, Luo Y, Zhang J, Sun H, et al. Gene expression profiles for a prognostic immunoscore in gastric cancer. Brit J Surg. 2018;105:1338–48.
Zhao Z, Wang Y. Comment on: Gene expression profiles for a prognostic immunoscore in gastric cancer. Brit J Surg. 2019;106:1416–6.
Zhu G, Pei L, Yin H, Lin F, Li X, Zhu X, et al. Profiles of tumor-infiltrating immune cells in renal cell carcinoma and their clinical implications. Oncol Lett. 2019;18:5235–42.
Ge P, Wang W, Li L, Zhang G, Gao Z, Tang Z, et al. Profiles of immune cell infiltration and immune-related genes in the tumor microenvironment of colorectal cancer. Biomed Pharmacother. 2019;118:109228.

Table.1 Clinical characteristics of GC patients in GEO and TCGA cohort.

	GEO Cohort (n=299)	TCGA Cohort (n=166)
Gender
Female	101 (33.8%)	64 (38.6%)
Male	198 (66.2%)	102 (61.4%)
Age (years)
Mean (SD)	61.9 (±11.4)	66.1 (±10.8)
Median [Min, Max]	64.0 [24.0, 86.0]	67.5 [39.0, 90.0]
TNM-T
T1	0 (0%)	4 (2.4%)
T2	185 (61.9%)	34 (20.5%)
T3	91 (30.4%)	78 (47.0%)
T4	21 (7.0%)	45 (27.1%)
Tx	0 (0%)	5 (3.0%)
NA	2 (0.7%)	0 (0%)
TNM-N
N0	38 (12.7%)	43 (25.9%)
N1	130 (43.5%)	53 (31.9%)
N2	80 (26.8%)	30 (18.1%)
N3	51 (17.1%)	32 (19.3%)
Nx	0 (0%)	6 (3.6%)
NA	0 (0%)	2 (1.2%)
TNM-M
M0	272 (91.0%)	147 (88.6%)
M1	27 (9.0%)	13 (7.8%)
Mx	0 (0%)	6 (3.6%)
Pathologic Stage
Stage I	30 (10.0%)	16 (9.6%)
Stage II	96 (32.1%)	54 (32.5%)
Stage III	96 (32.1%)	66 (39.8%)
Stage IV	77 (25.8%)	17 (10.2%)
NA	0 (0%)	13 (7.8%)

Table.2 The fraction of TIICs in normal and tumor tissues of GEO cohort

	Normal Tissue (%)		Tumor Tissue (%)	P.values
B cells naive	2.32 ± 2.60	1.00 ± 2.01		< 0.001
B cells memory	2.24 ± 3.26	3.32 ± 3.21		< 0.001
Plasma cells	14.92 ± 5.08	12.54 ± 4.77		< 0.001
T cells CD8	3.29 ± 3.15	3.67 ± 3.90		0.630
T cells CD4 memory resting	20.89 ± 5.06	13.57 ± 6.81		< 0.001
T cells CD4 memory activated	0.88 ± 1.33	3.33 ± 3.12		< 0.001
T cells follicular helper	3.57 ± 3.06	5.28 ± 3.58		< 0.001
T cells regulatory (Tregs)	0.03 ± 0.17	0.18 ± 0.59		0.012
T cells gamma delta	4.85 ± 2.57	3.13 ± 3.15		< 0.001
NK cells resting	0.20 ± 0.66	0.45 ± 1.07		0.005
NK cells activated	1.59 ± 1.52	2.93 ± 2.37		< 0.001
Monocytes	0.58 ± 1.21	0.24 ± 0.82		< 0.001
Macrophages M0	0.06 ± 0.30	4.85 ± 4.46		< 0.001
Macrophages M1	4.20 ± 1.58	7.57 ± 3.10		< 0.001
Macrophages M2	18.37 ± 5.94	15.58 ± 4.27		< 0.001
Dendritic cells resting	2.42 ± 1.79	2.12 ± 2.42		0.013
Dendritic cells activated	1.87 ± 1.41	4.13 ± 3.00		< 0.001
Mast cells resting	8.24 ± 4.11	2.86 ± 3.09		< 0.001
Mast cells activated	0.76 ± 1.53	2.37 ± 3.80		< 0.001
Eosinophils	4.51 ± 2.91	3.73 ± 3.15		0.011
Neutrophils	4.22 ± 3.20	7.08 ± 3.91		< 0.001

Table.3 The fraction of TIICs in normal and tumor tissues of TCGA cohort

	Normal Tissue (%)	Tumor Tissue (%)	P.value
B cells naive	3.56 ± 3.55	4.70 ± 5.05	0.381
B cells memory	3.02 ± 5.34	2.26 ± 5.83	0.216
Plasma cells	30.26 ± 8.21	4.49 ± 4.83	< 0.001
T cells CD8	15.15 ± 4.40	12.66 ± 9.56	0.095
T cells CD4 memory resting	17.66 ± 6.28	15.89 ± 8.16	0.421
T cells CD4 memory activated	0.97 ± 1.09	5.37 ± 5.77	0.003
T cells follicular helper	2.56 ± 2.48	2.40 ± 2.08	0.979
T cells regulatory (Tregs)	4.73 ± 2.10	6.04 ± 3.86	0.186
T cells gamma delta	0.21 ± 0.58	0.41 ± 1.19	0.705
NK cells resting	0.34 ± 0.60	1.25 ± 2.04	0.282
NK cells activated	1.61 ± 2.56	1.77 ± 2.20	0.879
Monocytes	0.98 ± 0.77	0.54 ± 0.89	0.007
Macrophages M0	1.74 ± 2.03	12.11 ± 11.34	< 0.001
Macrophages M1	2.22 ± 1.02	7.68 ± 4.46	< 0.001
Macrophages M2	6.02 ± 2.45	11.88 ± 6.34	< 0.001
Dendritic cells resting	1.71 ± 1.23	2.10 ± 2.78	0.501
Dendritic cells activated	0.82 ± 0.95	1.26 ± 2.58	0.558
Mast cells resting	4.16 ± 3.90	2.97 ± 3.66	0.112
Mast cells activated	0.69 ± 1.10	2.27 ± 4.64	0.992
Eosinophils	0.45 ± 0.70	0.43 ± 1.21	0.249
Neutrophils	1.14 ± 0.97	1.45 ± 3.28	0.054

Download PDF

Version 1

posted

You are reading this latest preprint version

The prognostic value of immune infiltration signature and its potential mechanism in gastric cancer

Status:

Version 1

Abstract

Figures

1. Background

2. Methods

3. Results

4. Discussion

5. Conclusion

Abbreviations

Declarations

References

Tables

Supplementary Files

Status:

Version 1