Detailed Description
The present invention is described in more detail below to facilitate an understanding of the present invention.
Patient recruitment and sample cohort
GC patients of the study of this invention (> T2N + M0, UICC-AJCC version 8) were recruited in the university of beijing tumor hospital (beijing, china) in 2015 to 2018. The study was performed according to the declaration of helsinki and was approved by the ethical committee of the cancer hospital, beijing university (IRB approval No., 2019KT 05). All patients provided written informed consent prior to treatment, sample collection and analysis. Tumors were collected by biopsy and matched to adjacent non-tumor tissue prior to neoadjuvant therapy. For patients with no adjacent non-tumor tissue available, blood samples were taken for replacement. All patients then received a fluorouracil-based capecitabine/S-1 + oxaliplatin treatment regimen (XELOX [ oxaliplatin, 130mg/m2, intravenous, day 1; capecitabine, 1000mg/m2, oral, day 1 to 14 ] or SOX [ oxaliplatin 130mg/m2, day 1 intravenous drip; S-1, 40-60mg, 2 times daily, oral, day 1 to 14 ] for 2-4 cycles and evaluated the response to treatment before surgery. Cancer 73,2680-2686(1994). To reduce classification noise, only three pathologists were focused on consensus assessment patients, and finally 35 patients (17 responses and 18 non-responses) were included as a study cohort. Tumor samples were further collected post-operatively. For patients who responded, no post-treatment tumor samples were collected.
Multi-set mathematical data generation
For 35 GC patients, genomic DNA was extracted from tissue (or blood) samples using the AllPrep DNA Mini Kit (QIAGEN) and the QIAamp DNAbood Mini Kit (QIAGEN), respectively. For whole exome sequencing, 1. mu.g of DNA was cut into short fragments (150-250bp) using Bioruptor (Diagenode). Repairing the resulting DNA fragment. Adaptor fragments are then ligated to both ends of the fragment. DNA fragments of the target size are selected. Thereafter, Polymerase Chain Response (PCR) was performed, and the resulting mixture was purified. Exome capture was performed using SureSelect Human All exon v6(Agilent) according to the manufacturer's protocol. The hybridized mixture was then amplified by PCR. The validated DNA library was then sequenced on illumina novaseq 6000. For whole genome sequencing, 1. mu.g of genomic DNA was sheared to 150-250bp using Bioruptor (Diagenode). The DNA library was then generated using the standard protocol of Truseq nano dnakit (illumina). Paired-end sequencing (paired-end runs) was performed using Illumina NovaSeq 6000, and the library was sequenced to a minimum depth of 6 × base coverage. For RNA sequencing, total RNA was extracted from fresh tissue using the AllPrep RNA mini kit (QIAGEN). For each sample, a library was generated by TruSeq RNAv2 kit (illumina) using 3 μ g total RNA. The library was sequenced with Illumina NovaSeq 6000 and an average of 3,300 million 2 x 150bp paired ends per sample.
Analysis of variant data
The whole exome sequencing read pairs were trimmed and the remaining read pairs had < 3% N bases and > 50% high quality bases. The resulting high quality reads were aligned to the human reference genome (Homo _ sapiens _ assembly19) using Burrows-WheelerAligner 0.7.17 (h.li, r.durbin, Fast and acid short read alignment with Burrows-Wheeler transform. bioinformatics 25, 1754-. To improve The alignment accuracy, The BAM file was processed using a Genome Analysis Toolkit (version 3.8.1) (A.McKenna et al, The Genome Analysis Toolkit: a MapReduce frame for analyzing next-generation DNA sequencing data. Genome Res. 20,1297-1303(2010)) by: repeat entries are labeled, local re-alignments are performed around high confidence insertions and deletions and base quality is re-aligned. Based on about 7,000 high frequency SNP sites, matched pre-treatment, post-treatment and normal samples from the same patient were confirmed. A variant harvesting pipeline developed in the cancer Genomic map MC3 project was used to identify high confidence somatic base substitutions and insertions/deletions (K.Ellrott et al, Scalable Open science application for Mutation Calling of Tumor organs Using Multiple Genomic pipelines. cell st 6,271-281e277 (2018)). In short, this pipeline uses six callers to invoke the substitution mutation and three callers to identify the insertion/deletion with detailed annotation information. Only at least two caller-supported substitution mutations and insertions/deletions were retained for further analysis. WES somatic base substitutions were further validated based on RNA-seq data from the same samples. An alignment was generated using TopHat2 (D.Kim et al, TopHat2: acquisition alignment of transformations in the presentation of insertions, deletions and gene fusions. genome Biol 14, R36 (2013)). For each substitution site, the coverage and number of mutation reads were calculated from RNA-seq BAM files from the same sample. For sites with sufficient RNA-seq coverage (. gtoreq.10X), sites with at least two reads can be considered to have validated the base substitution of interest. The characteristics of mutations in the pre-treatment samples were determined using R package deconstructed Sigs v1.8.0 as a matrix of mutation characteristics based on COSMIC characteristics (R.Rosenthal, N.McGranahan, J.Herreo, B.S.Taylor, C.Swanton, deconstructed Sigs: delinating biological processes in single structures differentiation DNA repair details and patterns of cancer evolution.genome Biol 17,31 (2016)). The Microsatellite instability status of each tumor was assessed using MANTIS (E.A. Kautto et al, Performance evaluation for Rapid detection of pan-cancer specificity with MANTIS. on target 8,7452-7463 (2017)), and 2539 loci (S.J. Salipate, S.M. probes, H.L. Hampel, E.H. Turner, C.C. Pratcard, Microcuterite specificity detection sequencing. clin.M 60,1192-1199 (2014)) obtained from the mSINGS package were used for this analysis. A clearly mutated gene (FDR <0.05) was identified in all pre-treatment samples using MuSiC2(18.N.D. Dees et al, MuSiC: identifying mutation in cancer genes. genome Res 22,1589-1598 (2012)). According to the Fisher's exact test, Maftools (A. Mayakonda, D.C.Lin, Y.Assenov, C.Plass, H.P.Koeffler, Maftools: effective and comprehensive analysis of physiological variations in cancer. genome Res.28, 1747-. To detect the gene group rich in the base substitution change after the treatment, HotNet2 (29) was used.
M.D. Leiserson et al, Pan-cancer networks analysis identities combinations of random acidic events and protein complexes Nat Genet47,106-114 (2015.) identification of apparently mutated sub-networks based on the caloric value of each protein. Only genes with altered events were retained. An altered event is defined as a different non-silent variation of the same patient before and after treatment. The heat score is limited to major drivers or genes with at least two events of change. The consensus sub-network is visualized using the HINT + HI2012, MultiNet and iRefIndex interactive networks. The Fisher exact test was used to determine statistical significance.
Drug response determination and analysis
Cisplatin response data (AUC scores) were obtained for 21 gastric Cancer cell lines with CCLE mutation profile data from GDSC (W. Yang et al, Genomics of Drug Sensitivity in Cancer (GDSC): a resource for therapeutic biological marker discovery in Cancer cells. nucleic acids Res 41, D955-961 (2013)). In cell lines, 3 contained non-silent single nucleotide variations in C10orf71, while 18 were wild-type. The Student's t test was used to assess the difference between cell lines with the C10orf71 mutation and cell lines with the wild type.
Reverse phase protein array analysis
For the RPPA experiments, 18 gastric lines were first confirmed by short tandem repeats and then prepared. The antibody was verified as previously described (M.Ghandi et al, Next-generation Characterization of the Cancer Cell Line encyclopedia. Nature 569, 503-. As previously described, RPPA data was generated by the RPPA core agency of the MD anderson cancer center. RPPA sections were first quantified using arraypro (meda cybernetics) to generate signal intensity, then processed by SuperCurve to estimate relative protein expression levels, and then normalized by median modification. The RPPA slide quality was evaluated by a quality control classifier, and only those slides above 0.8 (range: 0-1) were retained for further analysis. Protein markers were divided into 11 signal pathways and pathway scores were calculated based on the orientation of the protein members. The Student's t test was used to assess the difference between the cell line with the C10orf71 mutation and the wild type cell line, and FDR was used for multiple test corrections.
Somatic copy number Change analysis
Whole genome sequencing data was used to infer copy number values in samples before and after treatment. The whole exome sequencing read pairs were trimmed according to the criteria used for WES data. Reads were aligned to the human reference genome (Homo _ sapiens _ assombly 19) using BWA 0.7.17. SCNAs were then estimated by Control-FREEC (version 11.5) based on matching normal tumor pairs with a window size of 5kb (v. boeva et al, Control-FREEC: a tool for assessing copy number and adaptive content using next-generation sequencing data. bioinformatics 28,423-425 (2012)). Regions with statistically significant copy number variation frequencies were identified using gist 2.0(c.h. memel et al, gist 2.0 defects sensitive and reliable localization of the targets of local acidic copy-number alteration in human cameras Biol 12, R41(2011)) based on the corresponding segmentation values. The amplification and deletion thresholds were set at 0.1. Using the same tools, whole exome sequencing data were analyzed in parallel as one independent dataset to validate the SCNA peaks identified by whole genome sequencing.
Clonal structure analysis of sample pre-and post-treatment
To explore the new adjuvant chemotherapy-driven clonal evolution, dynamic changes in clonal structure were inferred using sciClone (c.a. miller et al, sciClone: inducing cyclic architecture and tracking the activity and temporal patterns of tumor evolution. plos computer Biol 10, e1003665 (2014)). Copy number values based on whole exome sequencing data inferred by Control-FREEC were used to exclude SNVs from the copy number change region for clonal analysis. The inferred subclone evolution was further validated using another tool, PyClone (A.Roth et al, static information of cyclic position structure in cancer. Natmethods 11,396-398 (2014)).
Analysis of Gene expression
Gene expression was quantified based on RNA sequencing data using the alignment-free tool Kallisto (N.L.Bray, H.Pimentel, P.Melsted, L.Pachter, Near-optimal genomic RNA-seq quantification. NatBiotechnol 34,525-527 (2016)). The output of Kallisto was converted to gene-level counts and TPM (C.Soneson, M.I.love, M.D.Robinson, Differential analytes for RNA-seq: transcript-level assays. F1000Res 4,1521(2015)) using a txoport. To identify genes that are differentially expressed between the responsive and non-responsive groups, a t-test was performed on each gene and they were ranked according to t-stability. The hallmark pathway of enrichment was determined using a pre-rank based Gene set enrichment assay (A. Subramanian et al, Gene set expression analysis: a knowledge-based approach for expressing genome-wide expression profiles. Proc Natl Acad Sci U S102, 15545 15550 (2005)). To identify genes differentially expressed between tumor samples at different treatment stages in non-responders, a multifactorial model was fitted with patient ID as blocking factor, and then a Wald test was performed for treatment stage effects using DESeq2 (m.i. love, w.huber, s.anders, modeled evaluation of fold change and dispersion for RNA-seq data with DESeq2.genome Biol 15,550 (2014)). Genes with FDR <0.05 and fold change > 2.0 or < 0.5 were selected as differentially expressed genes. Overexpression analysis of the MSigDB marker Gene set achieved by cluster Profile (G.Yu, L.G.Wang, Y.Han, Q.Y.He, cluster Profile: an R package for formulating biological enzymes systems clusters OMICS 16,284-287(2012)) was performed separately for genes up-regulated and down-regulated in the post-treated samples (E.I.Boyle et al, GO: (TermFinder- -open source for accessing Gene encoding and formulating design encoded genes with a list of genes. Bioinformatics 20, 3710-372004). The differential gene expression of MYC target genes before and after MYC-amplified non-responder treatment was further compared using Student's t-test. Tumor infiltrating immune cell abundance was inferred from mRNA expression data using TIMERs (b.li et al, Comprehensive analyses of tumor immunity: immunology for cancer immunity. genome biology 17,174 (2016)). Paired Student's t-test was performed to identify differences in cellular composition between matching pre-and post-treatment samples.
Biopsy tumor samples from 35 pre-neoadjuvant-chemotherapy GC patients were subjected to multicohort sequencing (detailed patient and sample information is shown in fig. 1A). For these cases, whole exome sequencing was performed on tumor samples as well as matched germline DNA samples, which were used primarily to identify somatic base substitutions and small insertions/deletions. In 32 of 35 cases (3 excluded due to insufficient DNA quantity), whole genome sequencing was also performed, as well as matched germline samples for identification of somatic copy number changes (SCNA). In addition, RNA sequencing was performed on all tumor samples to characterize the mRNA expression profile of the protein-encoding genes. Following biopsy, patients received neoadjuvant chemotherapy based on 5-fluorouracil + oxaliplatin for 2-4 cycles. Cases were then divided into two categories based on a strict assessment of radiology and pathology evidence by three independent pathologists: response (n-17) and non-response (n-18). Representative radiological and pathological images are shown in fig. 1B. In the response group, the necrosis rate was higher than 65% in all cases, while in the non-response group, the necrosis rate was lower than 15% (fig. 1C). The Mandard tumor regression rating for the response group cases was <2, whereas the non-response group cases were 4 or 5 (FIG. 1D). Typically, patients in the response group are younger than those in the non-response group (Wilcoxon rank-sum test, 17:18 for non-response and non-response, 0.047 for p); patients in the non-responsive group tend to be advanced (Wilcoxon rank-sum test, non-responsive and non-responsive 17:18, p 0.023).
After receiving neoadjuvant chemotherapy in 18 patients in the non-responsive group, fresh tumor samples were obtained surgically from 14 of these and these post-treatment samples were subjected to whole exome sequencing (fig. 1A). For these cases, whole genome sequencing (for 13 cases) and RNA sequencing were also performed, respectively. In contrast, for the patients in the response group, tumor tissue was not obtained due to lesions with low tumor cell content resulting from good response to neoadjuvant chemotherapy. In summary, this experimental design enables the identification of a variety of molecular aberrations (bulk base substitutions SCNA and gene expression) involving tumor response to neoadjuvant chemotherapy from different perspectives (response versus non-response, pre-and post-treatment).
The composition of six possible base pair substitutions was examined and T was found>The G substitution showed the most pronounced pattern between the responsive and non-responsive groups, especially when the substitution sites flank C and T (fig. 2A). Consistently, the COSMIC signature 17 (T) when breaking down the mutation profile into different mutation signatures>G) (11.L.B.Alexandrov et al, Clock-like biological processes in human gastric cells. Nat Genet47, 1402-1407 (2015)), is a feature previously observed in esophagus and stomach cancer and associated with byproducts of oxidative damage (12.M.Tomkova, J.Tomek, S.Kriaucinations, B.Schuster-Bockler, Mutamatic signal distribution with DNA replication timing and strand methodology. genome Biol 19,129 (2018)). The response group contributed much higher than the non-response group (Wilcoxon rank-sum test, p 0.049, response vs non-response 17:18, fig. 2B). Next, the distribution of microsatellite instability (MSI) scores in these tumors as calculated by MANTIS was examined (13.E.A. Kautto et al, Performance evaluation for rapid detection of pan-candidate microsaltite with MANTIS.
Oncostatt 8,7452-7463 (2017)). Surprisingly, MSI scores were found to be significantly higher in the non-responsive group than in the responsive group (Wilcoxon rank-sum test, p 0.022, non-responsive vs. 17:18, fig. 2C). Thus, the non-responsive group was tested for the presence of higher mutation burden and confirmed this pattern (Student's t test, non-response vs non-response 17:18, p 0.04, fig. 2D and fig. 3A). The pre-clinical data show that it is,MSI-H status of large intestine tumors resistant to 5-fluorouracil-based chemotherapy (M.Hewish, C.J.Lord, S.A.Martin, D.cunningham, A.Ashworth, Mismatch repair specific tumor in the same of patients treated with viral tumor treatment. Nat Rev Clin Oncol 7,197
208 (2010); M.Koopman et al, Predictive and qualitative markers for the outome of therapy in advanced clinical Cancer, a retroactive analysis of the phase III random CAIRO studyr, Eur J Cancer 45,1999-2006(2009). Clinically, only patients with MSI-negative colorectal Cancer benefit, while MSI-H Does not benefit (16.C.M. Ribic et al, Tumor microsatellite-instability status as a predictor of bed from fluorogenic-based adjuvant chemotherapy for colon Cancer. N Engl J Med, 247 @ (2003)), which makes MSI-H status a strong predictor of non-response to 5-fluorouracil-based chemotherapy (G.Des Guetz et al, Does microsatellite inactivation prediction of the efficacy of the treatment in clinical fashion review with a parameter of evaluation. Eur. J.45, 11896 (2009)). This result provides the first evidence that MSI-H status can also be used as a predictor of non-response to neoadjuvant chemotherapy in GC patients.
To identify the single mutant genes that may play a role in influencing the therapeutic response, the apparently mutated genes (SMGs) (p 2.9 10) were subsequently identified using MuSiC2(18.N.D. Dess et al, MuSiC: identifying mutation signaling in cancer genes. genome Res 22,1589--4FDR ═ 0.05). The most mutated genes included TP53, PI3KCA, RNF43, ARIDA1 and KRAS as previously reported in other gastric cancer cohorts (19.N. cancer Genome Atlas Research, Comprehensive molecular characterization of organic Adenocercinoma. Nature 513,202-209(2014)) (FIG. 3A). Among the SMGs identified in this and previous studies (19.N. cancer Genome Atlas Research, Comprehensive molecular characterization of structural adonnancinoma. Nature 513,202-209(2014)), C10orf71 was the only indication between the responding and non-responding groupsGenes for dominant patterns: mutations occurred in all 5 non-responsive samples, but not in the responsive samples (Fisher exact test, p 0.04, FDR)<0.25), and no recurrent mutations were detected (fig. 3B). To validate this observation, based on drug response data using a panel of 21 gastric cancer cells (20), it was indeed observed that cell lines with the C10orf71 mutation were more resistant to cisplatin (the equivalent platinum drug included in the neoadjuvant chemotherapy regimen) than these wild-type cell lines (Student's t test, nmut vs.nWT=3:18,p=1.1*10-4Fig. 3C). To gain more mechanistic insight, functional proteomic data of gastric cell lines were analyzed using reverse phase protein arrays. Cell lines with the C10orf71 mutation were found to show significantly lower cell cycle scores (based on 8 protein markers) than those of the wild type (Student's t test, n)mut vs.nWT3: 18, p ═ 0.015, fig. 3D, E). As cell cycle arrest is the major mechanism of action of platinum-based drugs, without wishing to be bound by theory, the inventors speculate that the C10orf71 mutation therein confers resistance to neoadjuvant chemotherapy by causing a less active cell cycle state (fig. 3F). These results indicate that mutations in this gene are potential biomarkers of resistance to neoadjuvant chemotherapy.
Key SCNA associated with neoadjuvant chemotherapy response
Using low-pass whole genome sequencing data (tumors, range: 5-9, median, 7; normal, range: 5-9, median 7), SCNA that could distinguish between the responsive and non-responsive groups was next studied. Distinct amplified or missing peaks were determined for these two groups using gist 2.0(c.h. memel et al, gist 2.0 defects sensitive and reliable localization of the targets of biological activity in human tumors Biol 12, R41(2011)) (FDR ═ 0.1). Although the missing peak profiles of responsive and non-responsive tumors were similar, the response group contained two distinct amplification peaks: one at 8q24.21, comprising a major driver gene MYC; the other is 19q12 with cancer gene CCNE 1. Meanwhile, the non-responsive group contained a unique amplification peak at 12q15, which contained the negative regulator MDM2 of TP53 (P. Chene, inhibition of the p53-MDM2 interaction: an antigen target for Cancer therapy Nat Rev Cancer 3,102-109 (2003)) (FIG. 4A). Similar peaks were observed using parallel whole exome sequencing data. This analysis implies further analysis of the original SCNA candidates.
If these amplification peaks play a role in influencing the response to neoadjuvant chemotherapy, corresponding signals will be observed in their associated downstream pathways. Thus, using RNA sequencing data, mRNA expression profiles of these pre-treatment samples were next examined. Although no significant differences in MYC expression were observed between the two groups (fig. 4B, t-test, p ═ 0.6), MYC target genes showed a significant abundance of up-regulated genes in the responsive group relative to the non-responsive group (fig. 4B, t-test, p ═ 0.6). In FIG. 4C, gene set enrichment analysis, nominal p 0.0, FDR < 10-3) Indicating that MYC signaling was indeed activated in the response group. A relationship between MYC expansion and better response to similar chemotherapy has been reported in a variety of diseases including breast cancer, small cell lung cancer and colorectal cancer. Consistent with the amplification peak of the non-responsive group, MDM2 showed significantly higher mRNA expression levels in this group (fig. 4D, Student's t test, response vs non-responsive 17:18, p 0.033). Furthermore, the DNA repair pathway showed a corresponding up-regulation in the response group relative to the non-response, supporting a negative regulatory effect of MDM2 on DNA repair (fig. 4E, gene set enrichment analysis, nominal p 0.0, FDR<10-3). In breast and pancreatic cancer, MDM2 amplification has been reported to correlate with poor clinical outcome and chemotherapy. No significant changes in the CCNE1 gene itself or its associated pathway were observed. Taken together, these results suggest that some key SCNAs may contribute to sensitivity (e.g., MYC amplification) or resistance (e.g., MDM2 amplification) to GC neoadjuvant chemotherapy.
Major somatic base substitution changes following neoadjuvant chemotherapy
Genomic evolution of GC under neoadjuvant chemotherapy was examined using parallel whole exome sequencing data of 14 matched pre-and post-treatment sample pairs (mean coverage > 200). Overall, no significant change in mutation burden or mutation characteristics was found. For the known cancer drivers, some tumors showed critical mutational changes (loss or gain of mutation) before and after treatment, while other tumors showed the same mutational signature (fig. 5A). To systematically identify genes or pathways that exhibit the most frequently occurring mutational changes, HotNet2 was used to search for protein interaction networks that are rich in such signals (29.M.D. Leiserson et al, Pan-cancer network analysis identities associations and protein complexes Nat Gene 47,106-114 (2015)). One top subnet identified consisted of IRS1, IRS2, PIK3CA, JAK1 and IL6ST (fig. 5B). In particular, IRS1 plays a key role in signaling insulin and insulin-like growth factor-1 (IGF-1) receptors to the PI3K/AKT pathway, which continually acquired new mutations in five post-treatment samples (Fisher's exact test, p 0.041), suggesting the existence of a recurrent resistance mechanism. Analysis of responsive and non-responsive pretreatment samples indicated that mutations in C10orf71 contribute to improved treatment resistance. If so, it would be desirable to increase the frequency of allelic mutations or obtain new mutations in the post-treatment sample, since tumor cells that are susceptible to chemotherapy (i.e., those tumor cells that do not have such mutations) would likely be removed by selection during the course of treatment. Of the 14 pairs of samples investigated, 4 showed a C10orf71 mutation in the pre-or post-treatment samples, two of which acquired a new mutation after neoadjuvant chemotherapy. A significant increase in mutant allele frequency was indeed observed at the eight mutation sites detected compared to the pre-treatment samples (figure 5C, Student's st test, n-8 pairs, p < 0.05; the same pattern was true even after tumor purity was adjusted). Fig. 5D and 5E show schematic diagrams of the putative evolution of the two cases, in which the obtained C10orf71 mutation gradually increased its allele frequency. Although the true evolutionary trajectory is difficult to validate, this result suggests that the C10orf71 mutation has a key role in influencing tumor response.
Key gene expression and cellular composition changes following neoadjuvant chemotherapy
Finally, 14 pre-and post-treatments were usedParallel RNA sequencing data (3,300 million reads per sample) for sample pairs, the gene expression profile of perturbation after neoadjuvant therapy was studied. 869 differentially expressed genes were identified (DESEQ2, n-14 pairs, fold change>2,p<5*10-3,FDR<0.05, fig. 6A). These genes were significantly enriched in several downregulated pathways, including inflammatory response, allograft rejection, KRAS signaling, and IL6/JAK/STAT3 signaling (fig. 6B). Analysis of responsive and non-responsive pre-treatment samples indicated that MYC expansion and subsequent MYC signaling activation sensitized tumor cells to treatment. If so, it is expected that in tumor samples with MYC amplification, MYC signals will be "inactivated" by selection during the purification evolution. Indeed, it was found that for 7 MYC-amplified tumors, nearly half of the MYC target genes (47.2%, 93/197) were significantly down-regulated in tumors after non-responsive treatment (Student's t test, n-7 pairs, p)<0.05) and this ratio is much higher compared to other genes (Fisher's exact test, p)<10-3). In contrast, only 3% of MYC target genes were significantly upregulated in these samples (fig. 6C). These results further support a link between MYC expansion and sensitivity to neoadjuvant chemotherapy. Expression of several therapeutic targets currently used for GC treatment was also examined and found to be significantly down-regulated for HER2, VEGFR1 and VEGFR2 (paired Wilcoxon rank-sum test, n ═ 14 pairs, p ═ 14 pairs<0.05,FDR<0.1; fig. 6D and 5A).
To investigate the effect of neoadjuvant chemotherapy on the tumor microenvironment, tumor infiltrating immune cell abundance (including B cells, CD 8T cells, CD 4T cells, neutrophils, macrophages, and dendritic cells) was inferred from mRNA expression data using TIMERs. Surprisingly, significant reductions in the cellular composition of neutrophils and dendritic cells were found (fig. 6E, Student's T test, n-14 pairs, p <0.05, FDR <0.12), with a slight reduction in CD 8T cells (n-14, p <0.1, FDR <0.2) post-treatment. In summary, RNA sequencing analysis indicates that neoadjuvant chemotherapy can not only remodel immune signals in GC cells, but also alter their immune microenvironment.
The foregoing describes preferred embodiments of the present invention, but is not intended to limit the invention thereto. Modifications and variations of the embodiments disclosed herein may be made by those skilled in the art without departing from the scope and spirit of the invention.