读文献06-纯生信挖掘找到与黑色素瘤预后相关的T细胞亚型及肿瘤激活相关基因集

• Date : [[2022-07-15_Fri]]
• 北野茶缸子
• 参考：Frontiers | Single-Cell Transcriptomic Analysis Reveals a Tumor-Reactive T Cell Signature Associated With Clinical Outcome and Immunotherapy Response In Melanoma (frontiersin.org)[1]

In this study, we utilized scRNA-seq profiles of CD8+ T cells in melanoma to derive a cluster of tumor-reactive T cells, and further developed a tumor-reactive signature (TRS) to indicate the tumor reactivity of tumor samples. We validated the ability of distinguishing tumor-reactive cells or cell groups of the TRS in multiple cohorts. Furthermore, we demonstrated significant correlation of the TRS with clinical outcomes and response to immunotherapy of melanoma patients

1-数据

we downloaded three Smart-seq2 datasets [GSE120575 (16), GSE72056 (6)and GSE115978 (15)] of single-cell RNA sequencing (scRNA-seq) in melanoma from GEO database. A total of 8262 CD8+ T cells from 80 samples

CCA 整合数据。

从各个黑色素瘤数据集整合的CD8+ T细胞如下：

image-20220715111440066

大部分主要是C0 和C1 细胞：

image-20220715111612682

也一定程度反应细胞的异质性。

2-亚群注释

image-20220715112446632

• 其他表达都不高，但表达IL7R，memory 相关的基因集也相比C4 显著上升（GSEA）：C0_memory
• high levels of cytotoxic molecules and numerous inhibitory checkpoints, including PDCD1, CTLA4 and HAVCR2：C1_exhausted
• besides C1 marker genes, also had the highest levels of cell cycle markers and highest proportion of proliferative cells: C2_cellcycle

看G2M 期细胞的比例（分裂增殖期）

image-20220715134224791

• CX3CR1, PRF1, GZMH and GZMB: C3_effector
• SELL, TCF7, CCR7 and LEF1 ：C4_naive
• the interferon signal and the signature genes of C5_interferon were involved in defense response to virus, response to interferon-gamma：C5_interferon

image-20220715134622635

• 表达部分naive gene 同时，还有effect gene GZMK：C6_transition

image-20220715133003480

3-耗竭T细胞肿瘤反应

Recent studies have shown that tumor-reactive T cells exhibit exhausted phenotype:

image-20220715134813196

针对这三个亚群。

看与T细胞活化相关基因表达：CD38 and HLA-DRA

image-20220715135259211

三个亚群显著表达升高。

以及肿瘤反应T细胞marker：

image-20220715135356233

而C3_effector 却不怎么表达，说明这些效应细胞可能是bystander。

另外设计了三个gene sig，we curated two tissue resident memory signatures (29942092_rm, 28930685_rm) and a T cell activation signature (31359002_act)，三个cluster 依然上升：

image-20220715135834939

这些证据都表明这三个耗竭亚型都可能是肿瘤反应T细胞。

4-T细胞克隆构建

TCR 结果来自GSE120575。

Previous studies have proven that the majority of TCR clones with high clonal expansion have been shown to be associated with tumor reactivity in melanomas。高度克隆扩张的T 细胞与黑色素瘤肿瘤反应相关。

with 1381 cells harboring unique TCRs, 500 cells harboring repeated TCRs, and 1197 cells with clonally expand TCRs：

image-20220715142351252

C0_memory 和C1_exhausted 有较多的TCR 克隆：

image-20220715142421747

不同病人的克隆异质性也很强：

image-20220715142602988

克隆扩张情况较多的病人，C1 亚型也越多。

image-20220715142844497

而TCR 样本中T 细胞的数目、克隆的数目也和C1 比例存在一定的相关性：

image-20220715142926509

image-20220715143018167

而更多比例的C1 还表现出较低的克隆性。

C1 与克隆扩张事件相关。

综合与肿瘤反应相关基因、基因集，与细胞毒性相关的基因，考虑C1_exhuasted 亚型可以反应T细胞的肿瘤反应性。

5-构建肿瘤反应T细胞基因集（signature）

image-20220715144551060

采用两步的策略：

• 差异分析，找到C1_exhuasted 差异基因；
• 利用选择的差异基因，来预测是否能很好的区分C1_exhuasted ；

d state from the others, resulting in 20 genes (Figure 3A). These genes were defined as the tumor reactive signature (TRS), including co-inhibitory receptors (CTLA4, PDCD1, TIGIT and HAVCR2), reactive markers (CD38 and ENTPD1), effector molecules (NKG7 and PRF1), tumor necrosis factor TNFRSF9 and critical exhaustion-related regulator TOX

image-20220715144756603

GO 富集显示The genes in TRS are widely involved in T cell activation,cellkilling,responsetotumorcell,chemokine production, cytokine secretion, and chronic inflammatory response，也与tumor reactive：

image-20220715144816378

6-验证基因集

下载其他癌症的单细胞数据：

we downloaded 4 scRNA-seq datasets of different cancer types (including hepatocellular carcinoma, non-small cell lung cancer, colorectal cancer and melanoma) from the GEO database under accession numbers GSE98638, GSE99254, GSE108989 and GSE123139,

we obtained 4 bulk datasets containing tumor-reactive T cells or cell groups from the GEO database under accession numbers GSE114944, GSE132810, GSE141878 and GSE145596.

6.1-富集方法

bulk 是如何区分tumor-reactive T cells group的？

image-20220715145527011

single 呢？

image-20220715145821121

使用了多种GSVA、ssgsea (22), zscore (23) and plage (24) 方法计算score。

6.2-CIBERSORT

接着还用了多个bulk 黑色素瘤数据，使用cibersoft 估计其中T 细胞比例，并同样使用gsva 计算基因集得分：

image-20220715152011932

表现出二者的相关性。

7-tcga临床数据测试基因集

将tcga 黑色素瘤病人按照bulk 结果计算基因集得分（gsva），区分病人：

image-20220715152431253

7.1-差异基因

differentially expressed genes were mostly upregulated in the TRS-high group compared to the TRS-low group, including chemokines and cytotoxic-related genes

富集结果也显示上调基因和免疫激活相关：immune activation, such as lymphocyte activation, cytokine signaling in immune system, and inflammatory response

image-20220715152750156

7.2-差异突变

分组查看对应外显子结果的突变高频基因：

image-20220715152916258

突变情况显著差异的基因包括：

image-20220715152950954

而且发现高TRS组burden 显著高：

ps：这么看其实power 也不是很足啊。仅仅是significant

image-20220715153035137

7.3-其他通路得分比较

In addition, we found that TRS_high group exhibited higher scores of intratumor heterogeneity and Th17 cell, and lower scores of wounding healing and homologous recombination defects：

image-20220715153319556

7.4-生存分析

区分TRS 高低得分病人：

image-20220715153556349

采用performed stepwise Akaike’s Information Criterion (AIC) 对基因集进行精简，得到CTLA4, CXCR6, LYST, CD38, GBP2 and HLA-DRB5，并利用这些基因计算TRS 区分病人生存分析：

image-20220715153751469

其他富集方法（zscore、plage等）也都得到了显著的结果：

image-20220715154823915

这个B 图，咋回事？

并且结合其他临床数据，和refined trs 做风险森林：

image-20220715154958046

同时也在先前验证使用的黑色素瘤bulk基因集做生存分析：

image-20220715155221388

8-与其他基因signature 比较

For the 8 published signatures, we calculated their risk scores as summation of the product of coefficient and expression level of each gene：

image-20220715155426384

相比其他的signature，trs 表现最好。

还有其他的评价标准：significance of patient stratification (E), time-dependent AUC (F), C-index (G) and restricted mean survival time (RMST) ratio between high-risk and low-risk groups (H).

image-20220715155556210

以及重新算其他几个出版的signature 的gsva 得分结果去和TRS 比较。依然打不过后者。

9-TRS较好预测免疫治疗响应

Three datasets were used for predicting the response to immunotherapy, including ERP105482 from ENA (https://www.ebi.ac.uk/ena/browser/home[2]), GSE35640 from GEO (https://www.ncbi.nlm.nih.gov/geo/[3]), and the dataset Allen2015 which was kindly provided by the corresponding author (PMID: 26359337).

ICB therapies were designed to reinvigorate efficacious antitumor immune responses by targeting inhibitory receptors on T cells.

Therefore, we next examined whether the refined TRS could predict ICB clinical response utilizing two cohorts

image-20220715160138482

TRS 得分和ICB therapy targets 表达情况。

而且按照 all patients were classified as responders or nonresponders according to the RECIST criteria 是否相应区分病人，TRS 得分也有显著差异：

image-20220715160247180

用TRS 来预测病人的免疫治疗相应结果也不错。

参考资料

[1]

Frontiers | Single-Cell Transcriptomic Analysis Reveals a Tumor-Reactive T Cell Signature Associated With Clinical Outcome and Immunotherapy Response In Melanoma (frontiersin.org): https://www.frontiersin.org/articles/10.3389/fimmu.2021.758288/full

[2]

https://www.ebi.ac.uk/ena/browser/home: https://www.ebi.ac.uk/ena/browser/home

[3]

https://www.ncbi.nlm.nih.gov/geo/: https://www.ncbi.nlm.nih.gov/geo/