The Cancer Genome Atlas (TCGA) database (https://portal.gdc.cancer.gov/) and the GenotypeTissue Expression (GTEx) database were utilized to obtain the RNA-seq data and relevant clinical data across 33 tumor types and normal tissues of 15,776 samples. We also collected 375 STAD patients accompanied with 32 normal tissues from the TCGA database. Then, we transferred RNAseq data in FPKM format to TPM format and further analyzed expression pattern for LSM family members. GSE14210 dataset was obtained from the Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo/) for validation.

cBioPortal (www.cbioportal.org) is a web resource that allows for the exploration, visualization, and analysis of multidimensional cancer genomics data (11). We analyzed the genomic profiles of 13 members of the LSM family to visualize the full details of each type of mutation in each individual sample.

GEPIA2 (http://gepia2.cancer-pku.cn/#index) is a multidimensional cancer genome dataset that integrates large amounts of data from The Cancer Genome Atlas (TCGA) and the Genotype-Tissue Expression Project (12). Here, the correlation between LSM and clinical stage was evaluated, and the module “Similar Gene Detection” was used to identify the most similar 100 genes for LSM5, 7, 8.

Methsurv (https://biit.cs.ut.ee/methsurv/) was used to evaluate the methylation level (13). Here we analyzed the CpG sites of the LSM family members in the STAD samples.

Based on the median expression levels of LSM5, LSM7 and LSM8, the expression data (HTseqCounts) were divided into high and low expression groups and further analyzed by unpaired Student’ s t-test within the DESeq2 R package (3.6.3). Adjusted p <0.05 and |log2-fold change (FC)|>1.5 were considered as thresholds for the DEGs.

TIMER (cistrome.shinyapps.io/timer) is a public tool to comprehensively investigate tumor-immune interactions (14). In this work, “Gene module” and scatterplots were obtained to analyze the correlation between the expression level of LSM members and the infiltrating of diverse immune cells in STAD.

We determined the correlations between LSMs mRNA expression levels and the survival of GC patients using the KM Plotter (http://www.kmplot.com) (accessed on 18 December 2021) databases (15). This online public database is a robust platform for visualizing patients’ survival across several cancer types. Overall survival (OS) in GC based on 5-FU chemotherapy for the LSM gene family was set as the default in the KM-plot database.

The online STRING database (https://string-db.org/, V11.0) (accessed on 18 December 2021) was used to analyze all publicly available sources of information and predict protein-protein interactions in the organism (16). The STRING analysis data were imported using Cytoscape software (version 3.8.1). In this work, a network compose of LSM family and their 50 similar neighboring genes was constructed using a protein-protein interaction module.

ROC analysis of LSMs were realized by the pROC package. The calculated area under the curve (AUC) value ranges, which were from 0.5 to 1.0, indicated the discrimination ability of 50%–100%.

With the selection of 100 similar expression genes for LSM5 and LSM8 through the GEPIA2 dataset, the “org.Hs.eg.db” (v3.10.0) R package was used to convert entrez ID to the gene symbol. The “ClusterProfiler” (v3.14.3) R package was used for the functional annotation. GO analysis includes cellular component (CC), molecular function (MF), and biological process (BP).

Thirteen GC patients with neoadjuvant chemotherapy treated in the Third Affiliated Hospital of Sun Yat-Sun University between January 2013 and December 2021 were enrolled in this study. Fresh frozen normal stomach and tumor tissues were obtained. All patients enrolled in this study had provided the written informed consent. Experiments related to human or human samples were approved by the Ethics Committee of Third Affiliated Hospital of Sun Yat-Sun University ([2022]02-169-01). The effects of patients with 5-FU chemotherapy were evaluated according to RECIST(Response Evaluation Criteria In Solid Tumours)1.1 (17), including CR (complete response), PR (partial response), SD (stable disease) and PD (progressive disease). We then divided the patients into two groups, sensitive (CR, PR) or non-sensitive (SD, PD).

Total RNA was extracted with TRIZOL reagent (Thermo Fisher, USA). The cDNA was prepared using HifairTM II 1st Strand cDNA Synthesis SuperMix Kit (YEASEN, Shanghai, China) according to the manufacturer’s protocol. Real time q-PCR was performed using ChamQ SYBR qPCR Master Mix (Vazyme, Nanjing, China) following instructions. 18S(Human) was used as an internal reference to normalize the expression of other mRNAs. Sequences used in this study were listed as follows:

Nine matched GC tissues and the adjacent normal tissues were selected from the 13 GC patients with LSM5, LSM8 high or low expression. After formalin fixed and paraffin embedded, the tissues were cut into 5 µm slides and then received IHC staining. Briefly, the slides were dewaxed with xylene and ethanol, and antigen repaired with EDTA buffer (pH 8.0) in a microwave oven. After being incubated with 5% goat serum diluted with TBS buffer containing 1% Tween-20, the slides were respectively probed with the primary antibody against CD3 (1:400), CD8 (1:1000), FoxP3 (1:400), CD68 (1:500), CD20 (1:500) and CD66b (1:500) overnight at 4°C and the corresponding second antibody for 1 hour at room temperature, and the DAB color development time of these immune cell markers was 3 minutes. To analyze the expression level of T cells, CD8+ T cells, regulatory T cells (Tregs), macrophage, B cells and neutrophilia in different gastric tissue samples, we counted the cells to compare the differences between chemotherapy sensitive and non-sensitive.

Statistical analyses were performed using GraphPad Prism (GraphPad, version 7). The statistical data of TCGA datasets were processed by R 3.6.3. Wilcoxon rank-sum test and Wilcoxon signed-rank test were applied for comparing the expression of LSMs between GC and the normal group. Chi-square test and t-test were applied for variance analysis, and Spearman rank correlation method was for correlation analysis. Each of the statistical tests was two- tailed, and a P-value <0.05 was identified statistically significant.

The transcriptional level of LSM family members was analyzed in 33 cancer datasets from the TCGA database. LSMs was up-regulated in most cancer tissues out of the 33 and all LSM members were significantly higher in GC tissues compared to the normal tissues ( Figure S1 ). Using the GEPIA dataset, we compared the mRNA expression of LSMs between GC and normal tissues. The results indicated that the expression levels of all LSM members except LSM11 elevated in GC tissues than their matched normal tissues ( Figure 1A ). Also, the comparison in unpaired GC tissues showed that expression of LSMs was greatly higher in GC tissues ( Figure 1B ), which was consistent with those from TCGA.

We then assessed the correlation between the expression of LSMs and the pathological stage of GC patients. All LSM members except LSM10 in GC patients with any pathological stage were markedly different from normal tissues. These data also showed that the expression of LSM1 was different between stage I and stage II, while that of LSM8 differed from stage II and stage III ( Figure S2 ).

We examined the correlation among the LSM members using the Pearson correlation analysis. Significantly positive correlations were observed between LSM1 and 2, 3, 10, 12; LSM2 and 3, 4, 6, 10; LSM3 and 5, 6, while negative correlation was observed between LSM8 and 10, 14B. Overall, the topmost Pearson coefficient (0.56) was observed between LSM2 and 3, LSM5 and 8 ( Figure 2A ). Moreover, we conducted a protein-protein interaction (PPI) network analysis of the differentially expressed LSMs with Cytoscape to explore the potential interactions among them ( Figure 2B ). Also, we analyzed the genetic alterations of LSMs in GC patients by using the cBioPortal online tool. Overall, two or more alterations were detected in different subtypes of GC, and amplification alterations were more common in GC ( Figures 2C, D ).

DNA methylation is frequent epigenetic event that relates to the viability of cancer cells, which gains significant resistance to anticancer drugs and escapes programmed cell death. Here, we investigated that LSM1, 2, 3, 5, 6, 7, 8, 10, 14A, 14B were negatively correlated with DNA methylation, while LSM4, 11, 12 showed positive correlations with it ( Figure 3 ).

We assessed the associations between the mRNA expression of LSM family members and OS using KM survival analysis in GC patients with 5-FU chemotherapy. As shown in Figure 4 , the results indicated that 9 of 13 members of the LSM family were significantly associated with poor OS of GC patients undergoing 5-FU based chemotherapy, such as LSM2 (HR = 1.52, p for trend = 0.019), LSM3 (HR = 1.57, p for trend = 0.01), LSM4 (HR = 1.75, p for trend = 0.0024), LSM5 (HR = 1.56, p for trend = 0.012), LSM6 (HR = 1.48, p for trend = 0.043), LSM7 (HR = 1.59, p for trend = 0.017), LSM8 (HR = 1.47, p for trend = 0.032), LSM11(HR = 3.23, p for trend = 0.019), and LSM14A (HR= 1.9, p for trend = 0.00026). In contrast, high expression levels of LSM12 indicated longer OS (HR = 0.54, p for trend = 0.0011). In addition, LSM1, LSM10 and LSM14B showed non-significant prognostic values.

GSE14210 dataset has gene expression data of pretreatment and posttreatment endoscopic biopsy samples collected from cisplatin and fluorouracil combination chemotherapy in GC patients. Through GEO2R, we found 147 different expression genes as shown by volcano plot ( Figure 5A ), and analysis of 10 hub genes revealed the existence of LSM5, 7 and 8 ( Figure 5B ). Moreover, the AUC values of LSM5, 7 and 8 in the TCGA GC cohort were 0.937, 0.903, and 0.946 respectively, which were interestingly higher than the AUC values of the rest LSM members ( Figures 5C ; S3A ). And a time-ROC curve of these three members showed relatively high prognostic ability ( Figure S3B ). In summary, LSM5, 7 and 8 were considered primarily correlated with 5-FU chemoresistance in GC.

We then identified differentially expressed genes (DEGs) in the high LSM5, 7 and 8 expression group in GC compared to the low expression group respectively ( Figures 5D–F ). The thirty most significant DEGs in GC were shown in the single gene co-expression heat map ( Figures 6A–C ). Most importantly, to elucidate the certain role of LSM5, 7 and 8, we conducted qPCR in 13 GC patients that had paired tumor and normal tissues. These patients were divided into two groups according to the curative efficiency of neoadjuvant chemotherapy, including sensitive and nonsensitive. The results showed that the mRNA expression level of LSM5 and LSM8 was lower in the sensitive group compared to non-sensitive group while LSM7 showed no significance ( Figures 6D–F ), which suggested that LSM5 and LSM8 may promote 5-FU chemoresistance in GC.

Tumor microenvironment (TME) is associated with relapse and chemoresistance of cancer cell. In this study, the correlation between immune cell infiltration, microsatellite instability (MSI) statues, tumor mutational burden (TMB) with LSM members were explored, and due to the above validations, we mainly focused on LSM5 and LSM8. The expression of LSM5 was in negative correlation with the infiltration of B cells, CD8+ T cells, Tregs, neutrophils and macrophages in GC patients ( Figure 7A ). However, LSM8 was positively associated with the infiltration of CD8+ T cells and negatively associated with the infiltration of B cells, Tregs, neutrophils and macrophages in GC patients ( Figure 7B ). With regard to the rest LSM members, multiple immune cells have correlation in GC patients, including CD4+ T cells, neutrophils, B cells, CD8+ T cells, macrophages, and dendritic cells ( Figure S4 ). Then we found that all LSM family members were positively associated with TMB in GC patients ( Figures 7C ; S5 ). Also, most members of LSM family were positively related to the statues of MSI in GC patients, excepting LSM2 and LSM10 ( Figures 7D ; S6 ).

In this study, we have firstly identified LSM5 and LSM8 as favourable biomarkers related to 5-FU chemotherapy resistance and investigated their relationship with tumor microenvironment and immune therapy in GC patients. Furthermore, to determine the relationship between immune cell infiltration and chemotherapy effect in higher level of LSM5 and LSM8 patients, we verified the immune infiltration level of B cells, CD8+ T cells, Tregs, neutrophils and macrophages in GC patients selected for qPCR experiment. Finally, 4 GC patients in the non-sensitive group and 5 cases in the sensitive group were included for IHC test for expression of LSM5 and LSM8. As shown in Figure 8 , the infiltration of CD8+T cells showed no differences between the two groups. However, higher expression levels of B cells, T cells, Tregs, neutrophils and macrophages were found in the sensitive group compared to non-sensitive group, which was highly consistent with the results of TIMER database.

The functions of LSMs and the genes significantly associated with LSM alterations were predicted by analyzing gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG). GO enrichment analysis predicted the functional roles of target host genes based on three aspects. In GC, we found that RNA splicing, DNA biosynthetic process, positive regulation of canonical Wnt signalling pathway, regulation of stem cell differentiation and innate immune response activating cell surface receptor signalling pathway were significantly regulated by LSM alterations on biological processes ( Figure 9A ). In addition, condensed chromosome, centromeric region, Sm-like protein family complex and U2-type spliceosomal complex on cellular components ( Figure 9B ), singlestranded DNA binding, unfolded protein binding and threonine-type peptidase activity on molecular functions ( Figure 9C ) were also significantly controlled by these LSM alterations. KEGG analysis defined the pathways related to the functions of LSM alterations and the frequently altered neighbour genes. Totally 9 pathways related to the functions of LSM alterations in GC were found through KEGG analysis ( Figure 9D ), containing spliceosome, proteasome, DNA replication, cell cycle, RNA degradation, nucleotide excision repair, homologous recombination, mismatch repair and base excision repair.