读文献07-黑色素瘤与T细胞亚型细胞互作

• Date : [[2022-07-26_Tue]]
• 北野茶缸子
• 参考：Frontiers | Single-Cell Transcriptomic Analysis Reveals the Crosstalk Propensity Between the Tumor Intermediate State and the CD8+ T Exhausted State to be Associated with Clinical Benefits in Melanoma[1]
• 二区细胞通讯：黑色素瘤细胞互作案例剖析

同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，为哈医大Li Xia 实验室出品。也是完全依靠public data。

1-数据

• The processed scRNA-seq dataset was downloaded from the GEO database under the accession code GSE115978, where tumor cells and CD8+ T cells were extracted according to the cell labels defined in the original studies. Raw read counts were counts per million (CPM)-normalized and genes that were expressed in less than 10% of the CD8+ T and tumor cells were filtered out using Seurat4.0 R package, respectively (28).
• The gene expression data, mutation data, clinical data, and immune feature profiles of the TCGA-SKCM cohort were available in the article (29).
• Bulk expression profiles, namely, GSE22153 and GSE91061, were also obtained for survival analysis and immunotherapy resistance analysis, respectively.
• The bulk datasets for primary and metastatic analyses were gathered from the GEO database under the accession codes GSE8401, GSE46517, and GSE59455.
• The mRNA profiles of 48 melanoma cell lines before and after 6 hours of treatment with interferon-gamma were downloaded under the accession code GSE154996.
• The RNA-seq data of CD8+ tumor-infiltrating lymphocytes from wild-type and Prdm1 conditional knockout (cKO) mice bearing B16F10 melanoma were obtained by GSE113221. Moreover, spatial transcriptome data and H&Estained annotation information of a melanoma sample were obtained from previous research (30).
• The RNA-seq data of 222 histologically distinct micro-regions (~5–20 cells per region) extracted from a melanoma patient were downloaded under the accession code GSE171888.

2-鉴别CD8与肿瘤亚群

2.1-转录因子对肿瘤细胞分群

The defined transcriptional factor (TF) motif-based regulons related to tumor cell states were collected from the published research.

Among those regulons, we applied the AUCell method to calculate activities of regulons with normalized count profile using the AUCell R package (31) and scaled activities by the maximum difference normalization method.

Consensus unsupervised clustering result was obtained based on 1,000 k-means clustering of scaled regulons activities using the ComplexHeatmap R package (32[2]). The clustering results with K set to 4 best matched the TF regulons’ pattern of tumor cell states reported in previous research (3[3]).

文献筛选和黑色素瘤相关模型转录因子分析 >> 根据发表文献，将获得转录因子结果聚类计算 >> 选定聚类4。

image-20220726091841174

区分四种亚群：

• melanocytic state、intermediate、neural-crest、mesenchymal-like

黑素细胞状态表现出谱系特异性转录因子的调控活性的升高(如SOX10和MITF)和显著升高的色素沉着和增殖潜能(图1A、C、D)。与黑素细胞状态相比，间充质样状态失去了黑素细胞转录因子的调控活性，但具有显著的更高的干性和侵袭潜能(图1A，E，F)。大部分肿瘤细胞处于中间状态，受EGR3、NFATC2和SOX6的调控(图1A)。

2.2-图聚类和拟时序鉴别T细胞亚群

seurat 聚类分群：

image-20220726095558643

• C1亚群标记为耗竭状态，高表达耗竭标记物（如HAVCR2、PDCD1和LAG3）和细胞毒性相关基因（如GZMB、PRF1、GZMA和NKG7）。
• C2亚群被定义为一种过渡状态。
• C3亚群富集初始/记忆T细胞相关基因（例如IL7R、CCR7、SELL和 TCF7）。
• C4 expressed a higher level of interferon induction gene (e.g., IFI44L, IFIT1, and IFIT3).
• C5、C6 亚群数目相对较少。

3-细胞通讯看肿瘤与CD8T互作

3.1-整体通讯情况

选取了五个细胞通讯数据集整合，至少两个符合的配受体对用来进一步分析：

image-20220726100620606

作者自己设计了筛选策略：至少在20% 细胞中表达 >> 互作得分是该配体受体在对应细胞亚群的表达均值的积 >> 通过置换检验进行显著性检验：

image-20220726100926982

并绘制互作网络：

image-20220726101618091

并筛选 chemokines and ligands associated with antigen presentation and TGF-beta signaling pathway, while the corresponding receptors were widely expressed in CD8+ T cells，说明它们对影响黑色素瘤免疫细胞浸润有很强作用：

image-20220726101831243

而在T 细胞当中，proinflammatory cytokine TNF, and ligands related to tumor necrosis factor family and cytotoxicity were expressed in CD8+ T cells, indicating the killing potential of CD8+ T cells against tumor cells，说明他们潜在的杀伤效果。

值得注意的是，在所有细胞状态中，无论肿瘤细胞作为发送者还是受体，中间状态肿瘤细胞具有最多的配体-受体互作（图3B，C，S1D)。这些结果表明，中间状态肿瘤细胞和神经嵴样肿瘤细胞可能更频繁地与CD8+ T细胞发生串扰。

image-20220726102252608

3.2-解读肿瘤与T细胞通讯的特性

主要研究：某些细胞通讯是否和某些细胞类型有显著特性。

ps：即是否某些细胞类型间的某种细胞通讯，相比其他细胞类型更加显著。

用NicheNet 预测配体活性。

肿瘤与CD8+T 细胞分别作为受体相关基因的活性：

image-20220726104404399

当肿瘤细胞作为受体时，不同肿瘤细胞状态下某些配体的活性存在明显差异，如中间肿瘤细胞状态接受最强的配体IFNG信号（图4A，C）。但CD8+ T细胞中这种现象不太明显。

这一步主要鉴别出不同cell state 其显著表达的细胞受体。

进一步研究细胞通讯与细胞状态的关系。

通过上一步构建的不同cell state 对应受体的signatures，看sender 细胞中哪些legend 表达情况跟这些signatures 显著关联；看receiver 细胞中哪些predicted activity 跟这些signatures 显著关联。

During signal transduction, the ligands both significantly related to a tumor state and a CD8+ T state were considered as the shared ligands between them. The hypergeometric test analysis was performed to explore whether the number of shared ligands has a significant over-occurrence, which could indicate a significant association between CCC and cell states.

比如研究肿瘤对T细胞作用，就看肿瘤中的legend 表达和T 细胞中的predicted activity 和这些cell state signatures(legend) 的关系。

image-20220726110517705

ps：这个相关性是如何计算的呢？

这个each single cell predicted activity 主要参考教程：

nichenetr/ligand_activity_single_cell.md at master · saeyslab/nichenetr (github.com)[4]

运行速度有点不敢恭维。

image-20220726112327838

image-20220726112320356

发现当CD8+T 作为受体时，intermediate tumor cells may affect CD8+ T exhausted state cells by those shared ligands。However, this phenomenon was not detected in the other case when CD8+ T cells act as sender cells

进一步验证，T 细胞作为sender 作用的legend 是否不具有特定细胞类型的选择特征。挑选了其他的一些legend 看（we repeated this process with functionally dependent ligands, such as cytotoxic, exhausted, and naive ones）：

image-20220726113213000

the cytotoxicity-related ligands in almost all CD8+ T cells significantly cooccurred with immune-related ligands in intermediate tumor cells. In addition, exhaustion-related ligands in CD8+ T exhausted cells significantly cooccurred with mesenchymal-related ligands in intermediate tumor cells

总结一下这里：

1）区分肿瘤 >> T细胞，T细胞 >> 肿瘤，看整个对应下的legend 和其对应的区分细胞类型的signature 之间是否显著相关。

2）针对T细胞作为sender 的情况，看每个细胞和选定的functionally dependent ligands 如cytotoxic, exhausted, and naive ones

说明肿瘤与CD8+T 细胞的通讯，更多的是发生在CD8+ T exhausted state and the tumor intermediate state。

ps：重点是从大的整体的各种肿瘤与T细胞的通讯，定位到肿瘤亚型和T细胞亚型的通讯。

4-验证CD8+ T exhausted 和tumor intermediate关系

4.1-构建伪空间关系

Considering that cell function is often influenced by neighboring cells（但是细胞的通讯不仅仅是位置上的）。

ps：这里的假设是，距离越近的细胞，相互影响越强吗？这个假设是成立的吗？

通过CSOmap 算法根据单细胞表达结果构建three-dimensional pseudo space：

image-20220726135609144

image-20220726141228658

显示CD8+ T exhausted 和tumor intermediate 有最紧密的空间关系。

image-20220726141435752

. (C) The difference in cell density between 10 cell clusters of CD8+ T cells and tumor cells

ps：展示了不同位置下的细胞的密度吗？证明CD8+ T exhausted 和tumor intermediate 在类似位置的密度关系。

image-20220726141806510

ps：这个connection 又是如何定义呢？

ps：细胞通讯的结果，竟然还能和CSOmap 关联，看对建立的伪空间的影响？

image-20220726142052958

4.2-空间数据

image-20220726142516497

基于state-specific signatures 构建gene set，对microregion sequencing data数据做ssGSEA，发现e invasive melanoma boundary (IB) region 的CD8+ T exhausted state as well as the tumor intermediate state 分数都显著的高。

ps：mrSEQ 这个技术和空转的差别？感觉很少报道。

The Spatial Landscape of Progression and Immunoediting in Primary Melanoma at Single-Cell Resolution | Cancer Discovery | American Association for Cancer Research (aacrjournals.org)[5]

而空转数据也证明这一点：

image-20220726143127032

4.3-构建TF调控网络

通过scMLnet 构建细胞互作-转录因子调控网络：

image-20220726143726263

ps：这里scMLnet 在细胞互作层面，和调控网络层面，不会和之前的工具分析的数据库存在重叠吗？（不过最开始基于TF 的肿瘤细胞分型也是基于前人文章中的结果）

PRDM1可以调节CD8+T 耗竭细胞分泌的IFNG配体的表达水平，并与中间肿瘤细胞上的IFNGR1/IFNGR2 受体结合，进而激活下游的间充质相关转录因子，如FOS和NR3C1。

4.4-TCGA与细胞系数据验证

TCGA 结果也可以验证，IFNG 基因表达高低组别，对应的CD8+T 和肿瘤中间相关基因得分都有显著差别：

image-20220726150146378

这条调节的通路基因，也都存在显著：

image-20220726150552734

而在敲除PRDM1 的老鼠中，CD8+ T exhausted specific signature 在CD8+ 细胞中的表达水平也明显降低。

48 melanoma cell lines 施加干扰素，发现这些细胞系的中间肿瘤细胞得分也都有所上升：

image-20220726151556694

5-肿瘤中间和CD8耗竭T细胞和临床获益相关

5.1-生存分析

原发、转移病人的肿瘤中间和CD8耗竭T细胞得分没有差别：

image-20220727101653099

而区分得分高低，生存结果有显著差异：

image-20220727101909971

image-20220727102150308

5.2-细胞免疫响应

对不同样本计算activity scores，并看结果和inferon gamma 响应的相关性。

image-20220727103054201

区分免疫治疗组别，Intermediate tumor cells and exhausted CD8 + T cells were both significantly enriched in immunotherapy untreated samples：

image-20220727103812520

且二者共同的legend 大部分表现出treat 与untreat 之间的差异：

image-20220727104257521

且两个state 在anti-CTLA4 组别具有相关性：

image-20220727104449172

疗效组别的cd8+T 丰度相对更高：

image-20220727104606242

6-风险群体的突变及免疫相关特征

基于the median values of ssGSEA scores of the tumor intermediate state and the CD8+ T exhausted state 综合得分，将TCGA 的数据区分为高低风险组别的样本：

image-20220727105303656

the C2-interferon-gamma dominant subtype was particularly dominant in the low-risk group, while the high-risk group was enriched in C1-wound healing and C4lymphocyte depleted subtypes

从突变结果来看，BRAF在样本中发生高频突变，且突变率在低风险组中显著高于高风险组（突变频率分别为59.4%和45%，p=0.006）。这与先前观察的BRAF突变亚型在高低风险组的分布一致，预示了更长的生存：

image-20220727105325493

在免疫浸润方面，低风险组的抗肿瘤免疫反应相关淋巴细胞浸润明显升高，如CD8+ T、CD4+ T和活化的NK细胞。低风险组具有较高的BRAF突变和较强的抗肿瘤免疫反应，提示风险分组能够反映肿瘤基因组信息和抗肿瘤免疫微环境特征。

参考资料

[1]

Frontiers | Single-Cell Transcriptomic Analysis Reveals the Crosstalk Propensity Between the Tumor Intermediate State and the CD8+ T Exhausted State to be Associated with Clinical Benefits in Melanoma: tr

[2]

32: https://www.frontiersin.org/articles/10.3389/fimmu.2022.766852/full#B32

[3]

3: https://www.frontiersin.org/articles/10.3389/fimmu.2022.766852/full#B3

[4]

nichenetr/ligand_activity_single_cell.md at master · saeyslab/nichenetr (github.com): https://github.com/saeyslab/nichenetr/blob/master/vignettes/ligand_activity_single_cell.md

[5]

The Spatial Landscape of Progression and Immunoediting in Primary Melanoma at Single-Cell Resolution | Cancer Discovery | American Association for Cancer Research (aacrjournals.org): https://aacrjournals.org/cancerdiscovery/article/12/6/1518/699151/The-Spatial-Landscape-of-Progression-and