Investigation of Uterine Carcinosarcoma: Case Report and Cohort Analysis

Timothy N. Hembree, DO, PhD, Jamie K. Teer, PhD, Ardeshir Hakam, MD, and Alberto A. Chiappori, MD


Cancer Control. 2016;23(1):61-66. 

In This Article

Cohort Analysis


We accessed clinical and tissue data from the Total Cancer Care protocol, which was approved by the Institutional Review Board of the H. Lee Moffitt Cancer Center & Research Institute (Tampa, Florida); patients also provided prospectively written informed consent.[11] Analysis of our study patient cohort included staging by either clinical or pathological staging, with the composite stage being the higher of the 2 scores. We used the composite stage to describe our cohort.

Deidentified clinical data, including age, race, sex, smoking status, stage of disease, histology, and type of treatment, were obtained through an "honest broker" web-based system designed by Moffitt Cancer Center to protect confidential patient information following the regulations set forth by the Health Insurance Portability and Accountability Act. Tissue samples had been collected using common protocols for tissue preservation, processing, and data annotation.


Tumor samples from Total Cancer Care were subjected to genomic capture and massively parallel sequencing. Sequences were aligned to the hs37d5 human reference with the Burrows-Wheeler Alignment tool.[12] Insertion/deletion realignment, quality score recalibration, and variant identification were performed with the Genome Analysis ToolKit (Broad Institute, Cambridge, Massachusetts).[13] Sequence variants were annotated with ANNOVAR ([14] Additional contextual information was incorporated, including allele frequency in other studies (eg, 1000 Genomes, National Heart, Lung, and Blood Institute's "Grand Opportunity" Exome Sequence Project), in silico function impact predictions, and observed impacts from select databases such as ClinVar ( and the Collection of Somatic Mutations in Cancer (COSMIC; version 61; Wellcome Trust Sanger Institute, Cambridge, UK).[15] Mutation frequencies were compared with data from The Cancer Genome Atlas (TCGA) using the cBioPortal for Cancer Genomics tool (Memorial Sloan Kettering Cancer Center, New York, New York).[16,17]

To enrich for somatic mutations, we excluded any positions observed in 1,000 genomes or at a frequency larger than 5% in a local normal population. We eliminated any variants with a variant quality score recalibration tranche level equal to 100 and required a genotype quality of at least 10 to include the variant of a particular sample.


Clinical Analysis. The Total Cancer Care cohort included 100 study patients diagnosed with uterine carcinosarcoma seen at Moffitt Cancer Center or an affiliated hospital between 1996 and 2012. Most study patients at presentation were between the ages of 60 and 79 years, female, white, and diagnosed in an early composite stage (1/2). Tumor grade was most often poorly differentiated. The first course of treatment was distributed between surgery alone, surgery and chemotherapy, and combination surgery, chemotherapy, and radiotherapy. The strongest predictor of survival was stage at presentation, with stage 1 having a median survival rate of 33 months (95% confidence interval, 19–109) and stage 4 having a median survival rate of 6 months (95% confidence interval, 3–12; Fig 4).

Figure 4.

Characteristics of the 100-person cohort. TNM = tumor, node, metastasis.

Genetic Mutation Analysis Nine of the 100 study patients were part of a targeted sequencing study covering 1,321 genes.[10] We examined their mutation profiles, comparing observed frequencies to 31 patients with carcinosarcoma in a previous study of 15 cancer genes.[10] We focused on those positions seen more than 5 times in COSMIC.[15] Mutations in JAK2, KRAS, PIK3CA, CTNNB1, PTEN, FBXW7, TP53, ARID1A, and ERBB2 were identified (Fig 5; Supplemental Table). All 9 samples had at least 1 mutation in these genes, with TP53 being the most commonly mutated (6 of 9). If TP53 is removed, then 8 of the 9 samples had at least 1 mutation in the remaining genes, suggesting that these pathways play an important role in carcinosarcoma. Many of these genes are mutated at recurrent positions, including PIK3CA (R88Q, H1047R/Y), FBXW7 (R465H, R658X), ERBB2 (V842I), JAK2 (V617F), and CTNNB1 (S37F, T41A).[16,17] Truncating mutations in PTEN (R233X, Q97X), TP53 (2 different nonsense mutations, 1 splice-altering mutation), KMT2C (1 nonsense mutation), and ARID1A (2 nonsense mutation) were also observed, consistent with previous observations of truncating mutations in these genes in different tumor types (TCGA via cBio Portal).[16,17] A single KRAS G12D mutation was also identified.

Figure 5.

Mutations in commonly mutated cancer-related genes. Mutation matrix shows missense (blue) and truncating (black) mutations in each study patient (n = 9) for the listed genes.

Many of these genes (PIK3CA, KRAS, TP53, PTEN, and CTNNB1) were observed in an earlier study and were part of the 15-gene panel used by Penson et al.[10]ARID1A, KMT2C, and FBXW7 were identified in a recent study by Jones et al.[18] Two of these genes were observed to be commonly mutated in uterine carcinosarcoma (TCGA via cBio Portal): ARID1A at 18% and FBXW7 at 39%.[16,17] However, although we observed 1 mutation in PPP2R1A (the most significantly mutated gene in uterine carcinosarcoma via the cBio Portal, as of publication), this exact mutation was not observed in TCGA data.[16,17] Furthermore, CTNNB1 mutations were observed in our study as well as in Penson et al[10] but not in TCGA data.[16,17]

A single sample, DS–90250, had many mutations across these commonly mutated driver genes (Fig 5; see Supplemental Table). Upon further examination, we identified a POLE proofreading-domain mutation, V411L, in this sample. After reviewing uterine carcinosarcoma data from TCGA, we identified a single sample with a POLE mutation in the same region, P286R; the sample also had a high number of mutations.[16] Thus, although these mutations are uncommon (ie, less common than those reported in uterine corpus endometrial carcinoma[19]), our findings suggest that POLE mutations can be observed in uterine carcinosarcoma.