Single-cell analysis of gastric signet ring cell carcinoma reveals cytological and immune microenvironment features

Gastric signet ring cell carcinoma (GSRC) is a special subtype of gastric cancer (GC) associated with poor prognosis, but an in-depth and systematic study of GSRC is lacking. Here, we perform single-cell RNA sequencing to assess GC samples. We identify signet ring cell carcinoma (SRCC) cells. Microseminoprotein-beta (MSMB) can be used as a marker gene to guide the identification of moderately/poorly differentiated adenocarcinoma and signet ring cell carcinoma (SRCC). The upregulated differentially expressed genes in SRCC cells are mainly enriched in abnormally activated cancer-related signalling pathways and immune response signalling pathways. SRCC cells are also significantly enriched in mitogen-activated protein kinase and oestrogen signalling pathways, which can interact and promote each other in a positive feedback loop. SRCC cells are shown to have lower cell adhesion and higher immune evasion capabilities as well as an immunosuppressive microenvironment, which may be closely associated with the relatively poor prognosis of GSRC. In summary, GSRC exhibits unique cytological characteristics and a unique immune microenvironment, which may be advantageous for accurate diagnosis and treatment.


Responses to the comments
Reviewer #1, expertise in scRNAseq, TME, gastric cancer (Remarks to the Author): In this paper, authors performed single-cell transcriptome profiling of cancer tissues from patients diagnosed with gastric signet ring cell carcinoma (GSRC). They Cell Ranger software (10X Genomics) was used to analyze the sequencing data and gene expression information was obtained for each cell. The Cell Ranger output was loaded onto Seurat 3.1.1 software for dimensionality reduction, clustering, and analysis of scRNA-seq data. Overall, 149,782 cells passed the quality control threshold, in which all genes expressed in less than three cells were removed，the number of genes expressed per cell > 500 was considered as low cut-off and <5000 as high cut-off, with a unique molecular identifier (UMI) count less than 500, and the rate of mitochondrial-DNA derived gene-expression was <25%. DoubletFinder was used to remove multiple cells from sequencing data.
To visualize data, the LogNormalize method of the "Normalization" function of the Seurat software was utilized to calculate the expression levels of genes. PCA was performed using the normalized expression levels, and the top 10 PCs were used to carry out clustering and t-distributed stochastic neighbor embedding (t-SNE) analysis.
Due to the obvious batch effect among samples, Harmony was used to correct the batch effect.
We analyzed the impact of cell cycle and found that the overall sample was less affected by cell cycle. From the tSNE diagram, the sample clustering was not significantly affected by the cell cycle. Therefore, we did not remove the impact of cell cycle, so as not to lose the difference between different cell types.

2.
The authors should provide more information or references about these genes used to annotation since each subgroup, for example, why VWF, CDH5, and PECAM were used to annotate enterochromaffin cells not the endothelial cell?

Response:
We have provided references about these genes used to annotation in the revised manuscript. VWF, CDH5, and PECAM were used to annotate endothelial cells, but not enterochromaffin cells. Enterochromaffin cells were written wrongly because of a clerical error. figure 1 and figure 2 were overlapped and figure 2 were missing some parts, which made confusion to read this paper.

Response:
We apologize for this mistake. Due to our typographical error, images were overlapped, and this mistake was corrected in the revised manuscript.

The authors used inferCNV analysis to identification of mucous cells and SRCC cells, could the authors explain why the cells having lower degree of variation and CNV score were defined as SRCC cells?
Response: We identified mucous and SRCC cells mainly from tissue origin ， sub-clusters of 1, 4, 8, and 15 were SRCC cells, and sub-clusters of 0, 2, 3, 5, 7, 10, and 17 were mucous cells (Fig. 1a). In order to verify the correctness of the identification, we conducted cancer-related score (Fig. 1b), and GSVA of inferCNV and CNV score (Fig. 1c).
According to inferCNV analysis, we could distinguish non-malignant epithelial cells from M/PDA cells (Fig. 1d). The CNV scores of sub-clusters of 6, 11, 12, and 13 were relatively high, which were identified as M/PDA cells and were consistent with the typical characteristics of malignant tumors. The CNV scores of each sub-cluster were calculated based on the inferCNV. In addition, CNV score gave corresponding confirmation. This was consistent with previous GA-related studies, and further confirmed the correctness of our verification method. Based on the above-mentioned analysis, we applied the same method to the identification of mucous and SRCC cells.
It was revealed that CNV score of SRCC cells was relatively lower than M/PDA and mucous cells, while higher than non-malignant epithelial cells (t-test, P < 0.001)( To sum up, inferCNV and CNV score provide a favorable support for the regrouping of epithelial cells. SRCC may originate from sub-cluster 7. ScRNA-seq results showed that MSMB expression level was relatively higher in SRCC cells than mucous and M/PDA cells (Fig. 3f). IHC results confirmed that MSMB expression level in the gastric foveal proliferation area was relatively high, not in all non-malignant epithelial cells.
Therefore, MSMB was found as a potential marker gene of GSRC, which could be   (Fig. 5).

Fig. 4 Differential plot of infiltration of follicular B and MALT-B cells/CD4-Treg and CD8-Teff cells in M/PDA and GSRC by immunohistochemistry score. a
Compared with M/PDA, the immunohistochemical score of follicular B cells (expressing CD74) in GSRC was higher, and the difference was statistically significant (Wilcoxon rank-sum test, P=0.012); The immunohistochemical score of MALT B cells (expressing JCHAIN) in GSRC was higher, and the difference was not statistically significant (Wilcoxon rank-sum test, P=0.154). b Compared with M/PDA, the immunohistochemical score of CD4-Treg cells (expressing FOXP3) in GSRC was higher, and the difference was statistically significant (Wilcoxon rank-sum test, P=0.032); The immunohistochemical score of CD8-Teff cells (expressing KLRD1) in GSRC was lower, and the difference was statistically significant (Wilcoxon rank-sum test, P=0.021).

Response:
We have re-written this part according to your comment as follows:

The text require revision for thorough correction of typos and grammar, there
was no space or more than one space before some words.

Response:
Thanks for your comment. The manuscript was fully edited.

Reviewer #2, clinical expertise in gastric cancer (Remarks to the Author):
This is a single-cell RNA-seq study of SRCC. The results are novel and this is a significant contribution to this understudied field.
Several points should be addressed: All specimens were collected from patients who gave written informed consent. This research was conducted according to the principles of the Declaration of Helsinki.

3.
Although there was brief mention of CDH1, can the authors confirm that the GSRC are sporadic, with no known germline mutations in CDH1,

alphaE-cadherin, etc?
Response: Studies have found that the gene change of CDH1 is closely related to familial gastric cancer, and is closely related to the occurrence and evolution of GSRC 7,8 . According clinicopathological characteristics of GSRC specimen, all cases in our study may occur sporadically with no family history of the condition. However, we found that CDH1 expression level was basically the same between SRCC cells and adenocarcinoma cells. One explanation could be the frequency of CDH1 mutations increased in GSRC 9 . Furthermore, our study found that the immunoglobulin superfamily was crucial in the cell adhesion of SRCC cells. Compared with adenocarcinoma cells, the intercellular adhesion function decreased in SRCC cells.
However, the intercellular adhesion function decreased or destroyed, which is the key in causing cancer cells to break away from the parent tumor to initiate metastasis and may also be associated with higher risks of invasive growth and metastasis, especially for implanted metastasis of GSRC.

4.
No statements are made with respect to funding, conflict of interest, or data availability. While this might be hidden as part of blinded review, this is critical information.

Response:
We had corresponding data in supplementary materials (This might be hidden as part of blinded review), and uploaded according to relevant regulations.