利用多種 "omics"分析方法建立的腫瘤的分子病理分類 - Cell 2014

Multiplatform Analysis of 12 Cancer Types Reveals Molecular Classification within and across Tissues of Origin

Cell. 2014 Aug 14;158(4):929-44.

以往腫瘤病理是依據病灶位置來做分類及診斷，例如乳癌可以分類為luminal或ductal、肺癌可分為SCLC或NSCLC。這些病理學上的診斷都是依據病灶的原發位置來去為腫瘤下最終的診斷。此篇研究的新穎之處在於破除原發病灶處的觀念(tissue-of-origin)，全部以腫瘤在分子層次上的特性去做分類。作者宣稱這種新的腫瘤分子病理分類將大幅提升對罹患癌症的臨床預後以及能更精確地提供適合的治療，而每10個人中就會有1個會因為用這新的分類方式而有不同診斷。

GBM       - glioblastoma multiforme

OV          - serous ovarian carcinoma

COAD     - colon adenocarcinoma

READ     - rectal adenocarcinoma

LUSC      - lung squamous cell carcinoma

BRCA     - breast cancer

AML       - acute myelogenous leukemia

UCEC     - endometrial cancer

KIRC       - renal cell carcinoma

BLCA      - bladder urothelial adenocarcinoma

Research Aim

- to investigate whether tissue-of-origin categorization splits into subtypes based upon multiplatform analysis

- extend to see whether there is convergence between tissue-of-origin tumors

推進分子病理分類學

利用大量tumor sample的data base，以cluster analysis來分析

歸納出11個依據 "omics" platform results 的tumor subtypes (independent of tumor origin)

也就是說，不再依據病灶位置，還是依據腫瘤的特性來做分類，甚至進一步決定治療

Tumor sample

- 10 TCGA tumor types

-  lung adenocarcinoma (LAUD) + head and neck squamous cell carcinoma (HNSC)

- total 12 tumor types =   "Pan-Cancer-12"
- total tumor samples = 3527

RESULTS

- identified 11 integrated tumor subtypes

- 10% of cases originally classified by tissue origin now reclassified by molecular taxonomy

- claimed new molecular taxonomy (11 integrated subtypes) to be accurate for clinical outcomes prediction

whole-exome DNA sequence
DNA copy-number variation
DNA methylation
genome-wide mRNA levels
microRNA
protein/phosphorylated level (Reverse Phase Protein Arrays)

sample-wise clustering to derive subtypes

1st level cluster analysis

input data into 2nd level cluster analysis, specifically called cluster-of-cluster assignments (COCA)

the COCA algorithm identified 13 clusters of samples, 11 of which contained more than 10 samples

5 showed simple, near one-to-one relationships with tissue site of origin

• C5- KIRC
• C6- UCEC
• C9- OV
• C10-GBM
• C13-LAML

C1-LUAD (lung adenocarcinoma)

• lung adenocarcinoma  (258/306)
• squamous cell              (28/306)
• 其中有趣的是，原本28個squamous sample因為被算出來在C1-LUAD而被重新去檢視frozen biopsy，發現有11個其實病理上是adenocarcinoma
• 更加顯示molecular diagnosis的優勢

Clinical Importance of the COCA subtypes  (COCA分類的臨床意義)

利用clinical features + tissue origin + COCA subtypes 可以提高statistical likelihood

genomic instability (CNV) - C9-OV, C4-BRCA/basal, C1-LUAD-enriched
AML, UCEC least

C1-LUAD - KEAP1 and STK11
C2-squamous - CDKN2A, NOTCH1, MLL2, NFE2L2

--> distinct somatic mutations exist within different COCA subtypes


Convergence of the Squamous-like Subtype 
Divergence of Bladder Cancer Subtypes
- diversed into 7 of 11 COCA subtypes
- n = 120 samples, 10 in C1-LUAD, 31 in C2-squamous-like, 74 in C8-BLCA- survival: C2-squamous & C1-LUAD < C8-BLCA 

-C8-BLCA EMT markers較低, C2-Squamous-like EMT較高

C2-Squamous-like
SOX2,Np63, TP53 mutation previously reported to be involved in normal squamous tissue development
similar to C4-BRCA/basal and C9-ovarian = high TP53 mutation
-higher TP63 and TP73 expression

Bladder Cancer (BLCA) most heterogenous
其中C2-squamous-like和immune features相關
1. chronic cystitis / recurrent bladder infections cause squamous metaplasia, predispose to bladder squamous cancer

2. BCG會induce T-cell --> 去攻擊early-stage bladder cancer

"Cell-of-Origin" rather than pathway-based features dominate molecular taxonomy
-相較於pathways (上百), 現在mRNA, proteins, miR(上千上萬)的profiling更能顯示cell-type specificity
- pathways常常已經和cell-type 掛上關聯，所以pathway應該是cell-of-origin內的一項


參考文獻

1. Hoadley KA et al. Cell. 2014 Aug 14;158(4):929-44.