文献抄读-可变剪切在头颈癌

我觉得每天看一篇文献的问题应该是不大的，三个小时应该足够了，如果不扣细节的话，当下理解不了的可以先放下，后面再碰到可能就会增加新的理解，不论学到了什么，都记录下来，以后某天一定能用上。昨天学习了可变剪切，今天就做一篇文献抄读吧。文章是今年10月份发表的，八分。原文：Comprehensive characterization of the alternative splicing landscape in head and neck squamous cell carcinoma reveals novel events associated with tumorigenesis and theimmune microenvironment.

image-20191201222547383

• 下面的解释中我觉得如果用英文更容易理解的话就用文章中表述

摘要

Alternative splicing (AS) has emerged as a key event in tumor development and microenvironment formation.However, comprehensive analysis of AS and its clinical significance in head and neck squamous cell carcinoma(HNSC) is urgently required.

paired HNSC

adjacent normal tissues

Cancer-associated AS events (CASEs)

Unsupervised clustering analysis

方法

• Generation of AS events profiling
  • RNA-Seq data were analyzed with SpliceSeq software to generate the AS profiles for eachpatient as previously described
    • SpliceSeq用法https://www.ncbi.nlm.nih.gov/pubmed/22820202
  • To generate a more reliable set of AS events, we implemented a series of stringent filters (percentage of samples withPSI values≥75, average PSIvalue≥0.05)
  • We alsoremoved the context-dependent AS events bymeasuring the significant association of the detected AS events with ESTIMATE scores. Spearman’s rankcorrelation analysis was performed (|R|≥ 0.4 and adjusted P < 0.05).
  • Interactive sets among the seventypes of AS were illustrated by UpSet plot created by UpSetR(version 1.3.3)
  • Circos plots weregenerated to visualize the detail of AS events and genes in chromosomes byCirclize (version 0.4.5)

结果

1.纵观TCGA数据集的AS数据集

1A-初步检测到来自13862个基因的事件95,700个，约占人类蛋白编码基因的67%

These AS eventswere classified into seven splicing modes: alternateacceptor site (AA), alternate donor site (AD), alternatepromoter (AP), alternate terminator (AT), exonskipping (ES), mutually exclusive exons (ME) andretained intron (RI).

这些事件被分为7种剪接模式:交替受体位点(AA)、交替供体位点(AD)、交替启动子(AP)、交替终止子(AT)、外显子跳越(ES)、互斥外显子(ME)和内含子(RI)。在这些拼接方式中，ES的发生频率最高。

image-20191201194949392

To ensure high stringency, we screened the AS events with a series of filters (percentage of samples with PSI values 75,average PSI value 0.05).

We also eliminated AS events with a significant association with stromal or immune cell content to specifically reflect cancer-associated AS alterations.

1B-最终共获得来自9844个基因的32,309个AS事件

image-20191201195637374

1C-Upset plots图展示七类AS事件的相交的基因

image-20191201195645805

1D- Circos plots图更直观展示 HNSC中的AS事件

image-20191201195654095

2.HNSC的癌症相关AS事件的确定

To identify the HNSC-specific AS events, wecompared the PSI values between 40 paired tumorand adjacent normal tissues. After screening, a total of 473 cancer-associated alternative splicing events(CASEs)from 420 genes were identified

为了鉴别hnsc特异性事件，我们将40对肿瘤和邻近正常组织的PSI值进行比较。经过筛选，420个基因共鉴定出473个癌症相关的可变剪接事件(病例)(补充表S3)。在80例标本中，仅有1例正常标本被误诊为肿瘤标本，准确率为98.6%，提示病例提供了准确区分肿瘤与正常组织的能力.

2A-B

image-20191201200944824

2C-Upset plots展示一部分AP事件也应该被认为是CAEs

image-20191201201456668

2D-There was an unevendistribution in the splicing patterns that suggested differed roles in cancer development. 此外，一些基因(如BCAR3、GNAS、ISLR和rasgrp3)在肿瘤组织和正常组织中在不同剪切变体上表现出相反的模式

image-20191201201508880

2D的补充如下图，此外，肿瘤TNC中的ESevents通常在肿瘤中更活跃

image-20191201202730629

下面的补充图片是说AT/AP事件中的基因和差异表达的基因是有重复的

image-20191201203624421

3.CASEs的潜在功能

Previousstudysuggestedthat alternativesplicing (AS) events may affect similar domainfamilies in cancer drivers and serve as potentialdrivers of cancer。

上面这句参考：Tumor-Suppressor Inactivation of GDF11 Occurs by Precursor Sequestration in Triple-Negative Breast Cancer.https://www.ncbi.nlm.nih.gov/pubmed/29161592

为了更好地了解癌症是如何驱动癌症发展的，我们首先调查了影响癌症驱动基因的案例。在这些病例中，有26个事件被确定在22个癌症驱动因子中产生新的亚型，如KRAS、SMARCA4和fbxw7。

3A-曾经有个文章表示， including ES events in KRAS and RHOT1,indicating protein feature losses or changes that was frequently observed in cancer progression. 参考：Hallmarks of alternative splicing in cancer.https://www.ncbi.nlm.nih.gov/pubmed/24336324作者受这个启发，we studied the proteins encoded by the transcripts involved in AS events. Out of the 26 events, 22 caused altered protein sequence. As expected, annotated proteins encoded by CASEs tended to be shorter than proteins in normal isoforms. To be note,domain families mediating interactions or involved in regulation of protein activity were involved.

image-20191201204503212

作者进行一些关于PCSK5和HDAC9在其他癌症中的举例，然后说：这些发现提示，病例可能通过蛋白功能失调发挥重要作用，为探索肿瘤发生机制提供线索

3B-做富集分析，说明这些CASEs富集的GO通路和HNSC的发展有关，如regulation of apoptotic signaling pathway、epithelial cell migration

KEGG通路,如viral carcinogenesis 、apoptosis

image-20191201204949147

接下来作者说了和HPV和为环境相关，表述为： Intriguingly,immune-related pathways were also enriched, such as Human Papillomavirus infection (FDR < 0.001),chemokine signaling pathway (FDR < 0.001), T cellreceptor signaling pathway (FDR< 0.001), and primary immunodeficiency(FDR=0.010),which indicated that CASEs are also involved in HPVinfection and immune microenvironment formationin HNSC patients.

3C-GSEA分析表明HNSC中差异表达的事件在细胞周期、病毒致癌和免疫相关通路中显著丰富

image-20191201205425512

接下来作者对这些CASEs对肿瘤微环境的影响做进一步研究

Several CASEs were closely related to HPV infectionandviral carcinogenesis,including PRKACB,CREB3L4, PXN, RELA andIKBKG. Correlation analysis revealed that higher expression of AP in IKBKG was associated with depletion of CD8 T cells and impaired cytolytic activity, which may be explained by NF-κB signaling pathways dysregulation。

• 这里面有个疑问，怎么把这些基因和CD8 Tcells联系起来的呢？

image-20191201210431081

4A-蛋白质网络，分成shape、color、size三个部分，感觉有点小特别

image-20191201210840067

PPI network of CASEs generated by Cytoscape. Nodes represent genes with CASEs. The shape, color and size of nodes represent splicing modes, change patterns and |log2FC|, respectively. Edges represent the potential interactions between the corresponding protein. Alternate acceptor site (AA), alternate donor site (AD), alternate promoter (AP), alternate terminator (AT), exon skipping (ES), mutually exclusive exons (ME), and retained intron (RI).

4-B/C/D

B:10个hub基因，8个基因的功能域家族受到了这些CASEs的影响

C:泛素化

D:细胞黏附和迁移

image-20191201212103579

4.CASEs网络和剪切因子

剪接因子(SFs)是通过选择性包含外显子或去除内含子来调控AS events的关键因子。SF改变促进肿瘤中与正常组织相比的差异剪接模式，从而产生前致瘤亚型。因此接下来研究在HNSC中，SFs是如何调控CASE的

为了深入了解这一问题，分析了71例经实验验证的SFs的表达与病例的psi值之间的相关性，并建立了一个基于显著相关性的剪接调节器网络。

接下来作者这样描述：To gain insightsinto this issue, we analyzed the correlation betweenexpression of 71 experimentally validated SFs and thePSI values of CASEs, and built a splicing regulatorynetworkbasedon thesignificantcorrelations(Supplementary Figure S3). A total of 91 CASEs wereassociated with 40 SFs (Supplementary Table S7)。但是事实上所有的Supplementary Table都是没有的，而这个experimentally validated SFs是如何确定的呢？

SFs和CASEs的网络图展示如下，在补充图片中

image-20191201213720323

5A-展示SFs和CASEs高度相关的几组

image-20191201214155620

5B-SFs promoter methylation and SFs expressio的关系图

下图展示了TIAL1、HNRNPA3、TRA2B、hnrnpd启动子的甲基化水平与其表达水平的关系。

接下来作者就研究CNA。这里可以说明CNA和甲基化的关系

image-20191201214149176

5C-在hnsc队列中，85.6%的病例中存在拷贝数(CN)丢失，94.0%的病例中至少有1个SF存在就会有拷贝数增加事件。29个SFs的CN损失与较低的SF expression相关，而30个SFs使CN显著增加。

image-20191201215700350

上面的例子说明，基因和表观遗传调控可能导致SF表达的改变，并进一步调控异常事件。

5.HNSC中CASEs的预后相关价值

Previous studies haveindicated that aberrant AS events occur in the earlystages of cancer and are implicated as prognosticmarkers in many malignancies。上面这句参考https://www.ncbi.nlm.nih.gov/pubmed/24189147和https://www.ncbi.nlm.nih.gov/pubmed/29397868

作者将CASEs的 median PSI value将HNSC 的患者分为两组。单因素Cox回归分析显示，59和53的事件分别与OS和DFS显著相关。多因素Cox回归分析显示分别有27例和26例是OS和DFS的独立预后因素。

6A-AS与OS和DFS相关如下图

image-20191201220719641

6-B/C:以egfl7中的AP和IL1RL1中的AT为代表

image-20191201220905287 image-20191201220916173

作者说他们还研究了这些CASEs是否独立于TP53或EGFR突变状态和分子亚型，但是因为没有补充的表格下载，所以无法展示。

18个CASEs中，11个意味着好的预后，7个代表不良预后。以往的感觉应该是CASEs应该是大多数代表差预后，而这这里却出乎意料的是好的预后。作者在文中这样描述：They expressed higherwith an alternate promoter and lower with an exon skipping. Out of the eleven CASEs associated withbetter prognosis, seven events had a gain, loss or alterationin at leastoneprotein feature。

下面是作者举了两个例子：

Lower expressionof exon skipping in TOP1MT resulted in an earlytermination of translation and therefore loss of DNA topoisomeraseI feature(IPR001631)and topoisomeraseI-related domains(IPR013499,IPR013500, IPR025834). Consequently, TOP1MT maylose its oncogenic function.Besides, AT in IL1RL1may lead to gain of Toll/interleukin-1 receptor homology (TIR) domain and promote of T Cell Responses, which may explain the positive correlation between AT in IL1RL1 and cytolytic activity we observed above。

6-D/E接下来作者猜想能否搜集出可以代表好预后和差预后的两组CASEs，筛选出后又引用了其他文献，说明挑选出的一些CASEs是有预测功能的，就是和其他结果呼应一下。

image-20191201223139788

(D) The correlation network of survival-associated CASEs. The color of edge represents correlation coefficient. (E) The correlation plot of 18 survival-associated CASEs. The color of dot represents correlation coefficient

上面结果直接表明，病例不仅具有生物学意义，而且具有潜在的临床价值

6.AS-based clusters significantly associated withprognosis, molecular characteristics andimmune features

7A-作者进行了unsupervised consensusanalysis无监督共识分析，以探索是否呈现出可识别的模式。基于一致的矩阵热图，识别出以下不同的基于AS的分子簇.原文表述：the following distinct AS-based molecular clusters were identified.

• 疑问：这个可以把他们区分开的无监督共识分析，是什么呢？怎么真名就能分开呢？是不是就是我们常做的热图呢，只不过这么说而已，就是根据相关性COR函数来看聚类？下图是464个病人被分成了4组。

image-20191201223615572

7E-研究了这18个CASEs与临床病理分期的关系

image-20191201224147205

Heatmap of 473 CASEs ordered by clusters. The association with clinical, molecular and immune features was annotated.

7B-生存相关,可以看到C3组预后最好

image-20191201224730052

7D-Immune and stromal scores were calculated based onthe ESTIMATE algorithm to quantify the presence of stromal cells and the infiltration of immune cells in tumor samples.

C3组与其他不同：C3 showed enhanced cytolytic activity compared withotherclusters

但是这个cytolytic activity 应该是与癌症的免疫浸润有关吧。

image-20191201225612161

为了评估AS与其他分子特征之间的相关性，我们进一步比较了HNSC的四种分子亚型(经典型、基底型、间充质型和非典型型)、TP53突变和簇间EGFR改变。

7C-C3组免疫活跃，同时这个C3大部分是分子亚型中的经典型。

image-20191201225217932

这个文章看起来还是有点吃力的，感觉作者的思路大部分是根据比如前面有其他人有过什么研究，然后自己基于那个基础也做了这个研究，还有就是根据自己上一步得到的结果继续下一步分析。比如涉及到了甲基化，作者就分析下CNA。然后所有的生信分析的基本的都有了，但是感觉还是主要是思路，一层一层扒，直到最后得到和临床相关的信息。还是有很大收获的。里面涉及的方法用的一些包还是要学的。然后这篇文章是纯数据挖掘，就是TCGA的数据。还有作者并没有展示补充材料中的表格，仅仅是补充的图片。感觉有收获。

最后友情宣传生信技能树

• 生物信息学“义诊”

• 生物信息学”拍卖会”

• 全国巡讲:R基础，Linux基础和RNA-seq实战演练 : 预告：12月28-30长沙站

• 广州珠江新城GEO数据挖掘滚动开班

• DNA及RNA甲基化数据分析与课题设计