Posted: December 11, 2014

TCGA’s Pan-Cancer Analysis Shows New Possibilities for Classifying Tumors

Teagan Keating Kuruna

In the August 14, 2014 volume of Cell, The Cancer Genome Atlas (TCGA) Research Network published a multiplatform analysis of 12 cancer types, revealing that some tumors were more likely to be genetically and molecularly similar based on the type of cell from which it arose rather than its tissue site of origin. This study is part of TCGA’s Pan-Cancer project, an initiative designed to analyze TCGA data across tumor types and platforms and to make the analysis and interpretation freely available.

Researchers found that approximately 10 percent of tumors could be reclassified based on their cell of origin rather than their tissue site of origin. This finding may inform research that could, down the road, help clinicians increase prognosis accuracy, as well as improve targeted and standard therapies. It may also help clinicians more accurately determine the cancer’s aggressiveness and patient prognosis.

New subtypes emerge

Currently, cancer treatment is usually based on its tissue site of origin. Previous TCGA studies have shown that single-tissue tumors can be split into several subtypes based on their molecular profile. TCGA researchers were interested in identifying patterns of molecular changes that may be common across tumor subtypes regardless of organ of origin. Comparing the 12 tumor types, they discovered that some tumor types were molecularly heterogeneous (the converged-diverged subtypes) and others were molecularly homogeneous (same-tissue origin subtypes).

The converged-diverged subtypes include tumors that can show different kinds of molecular profiles—for example, not all lung adenocarcinomas are exactly alike. Sometimes a lung adenocarcinoma is more molecularly similar to a bladder tumor or head and neck tumor. Cancers that fall into these subtypes diverge from other tumors from the same tissue of origin and converge into subtypes with other, molecularly similar tumors arising from different tissues. The table below summarizes the converged-diverged subtypes’ clustering patterns.

Subtype name Tumor types found within this subtype





Head and neck


BRCA/luminal Breast
BRCA/basal Breast
BLCA Bladder




Breast cancer diverged into two subtypes, BRCA/luminal and BRCA/basal, while bladder cancer diverged into three major subtypes, LUAD-enriched, squamous-like, and BLCA. Colon cancer and rectal cancer merged into a single subtype, COAD/READ. The breast cancer and colorectal cancer findings echo TCGA’s previous papers, both published in Nature in 2011.

However, some tumor types showed consistent molecular characteristics across patients. These molecular similarities meant they grouped together during analysis, creating the same-tissue origin subtypes. These tumor types were:

    • Ovary,
    • Endometrium,
    • Kidney,
    • Glioblastoma, and
    • Acute myelogenous leukemia.


The squamous-like subtype

Lung squamous cell carcinoma, head and neck squamous cell carcinoma, some bladder cancers, and a few lung adenocarcinoma biospecimens coalesced into the squamous-like subtype. This means that within those four tumor types, some samples exhibited more molecular similarities between each other than with other samples from the same tissue-of-origin. A hallmark of this subtype is loss of 3p, the short arm of chromosome 3.

Researchers identified four subnetworks of mutated genes characteristic of the squamous-like subtype. These subnetworks all include genes that, if altered, can cause tumor growth. The frequency of these alterations is another distinguishing feature of the subtype.

More than 90 percent of the biospecimens in this subtype showed a mutated subnetwork that included well-known tumor suppressors TP53, CDKN2A, and PTEN. Tumor suppressor genes encode proteins that protect against the various genomic alterations that can cause the cell to become cancerous. Changes in tumor suppressor genes can lead to unchecked cell growth, uncontrolled division, or unnatural cell immortality.

Nearly 60 percent of squamous-like samples showed a mutated subnetwork including genes that regulate oxidative stress, an abnormal level of antioxidants that results in tissue damage. When cells are unable to contend with oxidative stress, their DNA is damaged. The damage in turn may promote cancer proliferation, invasiveness, and metastasis.

Over one-third of squamous-like samples also exhibited a mutated subnetwork that includes ASCOM complex and the gene KDM6A. The ASCOM complex is a transcriptional regulator that is involved with histone modifications1. The KDM6A protein, encoded by the KDM6A gene, is a histone demethlyase, an enzyme that modifies histones and is a tumor suppressor. Some somatic mutations of KDM6A lead to nonfunctional proteins, which means that they cannot function correctly as tumor suppressors.

RAC and RHO signaling were also elevated. These signaling pathways regulate cell division, survival, polarity, adhesion, and motility2. In particular, proteins in these pathways regulate the bonds between cells and, when disturbed by mutations, can allow cancer cells to achieve amoeba-like motility, which may allow the cancer cells to detach from one another and move more easily around the body, the end consequence of which is metastasis3.

One cancer, three subtypes

Bladder cancer was found to be the most molecularly diverse tumor type. Overall, bladder tumors were classified into seven of the subtypes. However, the majority of tumors originating in the bladder were sorted, based on their genomic profiles, into three different subtypes: LUAD-enriched, squamous-like, and BLCA.

TCGA scientists examined the survival patterns of patients diagnosed with a bladder cancer that was classified into one of these three subtypes. Patients whose tumors coalesced into the LUAD-enriched and squamous-like subtypes showed far worse survival outcomes than those with tumors in the BLCA subtype. This finding is consistent with survival patterns of patients with different tumors in these subtypes such as lung squamous cell carcinoma, head and neck squamous cell carcinoma, and a minority of lung adenocarcinomas.

Bladder tumors that fall into the squamous-like subtype also show the immune features that are often found in other tumors in the subtype. This may help explain how chronic cystitis and recurrent bladder infections can cause squamous dysplasia. The presence of these immune features may also illuminate the mechanisms involved in early-stage bladder cancers’ frequent responsiveness to the Baccillus-Calmette-Guerin (BCG) tuberculosis vaccine. No predictive test exists to determine whether a bladder tumor would respond to BCG4. Understanding the diverse molecular composition of bladder cancer may assist in the development of tests that can determine the efficacy of BCG for an individual patient.

Accessing the data

The breadth and depth of the data produced by TCGA made this study possible. Future researchers will be able to validate these results and build upon them to learn more about the intricacies of cancer and advance the development of precision medicine.

In order to achieve the goal of making this data freely available, the data sets and results are available in Synapse. Results are also available through various portals, including UCSC Genome Browser, Gitools, and MD Anderson’s Next Generation Heatmaps. TCGA data can be downloaded from the TCGA Data Portal and CGhub.


Hoadley, K.A., Yau, C., Wolf, D.M., Cherniack, A.D., Tamborero, D., Ng, S., Leiserson, M.D.M., Niu, B., McLellan, M.D., Uzunangelov, V., et al. (2014) Multiplatform analysis of 12 cancer types reveals molecular classification within and across tissues of origin. Cell. doi: 10.1016/j.cell.2014.06.049. Read the full article

Selected References

1 National Center for Biotechnology Information (2014) Gene: KMT2D lysine (K)-specific methyltransferase 2D [Homo sapiens (human)]. Gene database. Available from:

2 Parri, M. and Chiarugi, P (2010) Rac and Rho GTPasses in cancer cell motility control. Cell Commun Signal. doi: 10.1186/1478-811X-8-23. Read the full article.

3 Ibid.

4 Stallard, J. (2013) Study clarifies how bladder cancer treatment works. On Cancer. Available from: