Background

Endometrial carcinomas (ECs) arise from the inner epithelial lining of the uterus and can be classified into a number of discrete histological subtypes (reviewed in [1]). Endometrioid endometrial carcinomas (EECs) represent the vast majority of diagnosed cases [1]. They are generally estrogen-dependent tumors that are associated with a number of well-established epidemiological risk factors that lead to unopposed estrogen exposure including obesity, nulliparity, early age at menarche, and late age at menopause [2]. Most EECs are detected at an early clinical stage when surgery or surgery with adjuvant radiotherapy can often be curative [3, 4].

Serous and clear cell ECs are high-grade tumors that are rare at diagnosis but are clinically aggressive and contribute substantially to mortality from endometrial cancer (reviewed in [5]). For example, in a large retrospective study of 5,694 cases of endometrial cancer in the US, serous and clear cell tumors together represented 13% of diagnoses but accounted for 47% of deaths [6]. Historically, serous and clear cell ECs are considered to be estrogen-independent tumors with no well-established epidemiological risk factors other than increasing age [7, 8]. However, a recent large epidemiological study has suggested that increased body mass index may be a risk factor for serous endometrial carcinomas [9]. Current therapeutic approaches to treat patients with serous or clear cell ECs are variable but generally include surgery and adjuvant chemotherapy and/or radiotherapy [10, 11].

Alternative therapeutic options are being sought for patients with serous or clear cell EC and for patients with advanced-stage or recurrent endometrioid EC. Rationally-designed therapeutics targeting tyrosine kinases can be clinically efficacious against tumors that have somatically mutated, amplified, or rearranged the target kinase, and which are dependent on the aberrant kinase-mediated signaling for their survival [12–17]. Recently, the tyrosine kinase gene family has been sequenced in 133 serous ECs, 329 endometrioid ECs, 53 ECs of unspecified histology, and 13 mixed histology ECs either by targeted sequencing of the tyrosine kinome [18], or by comprehensive sequencing of all protein-encoding genes including the tyrosine kinome [19–24]. However, it has been estimated that at least 500 tumors of a given histology need to be sequenced to provide adequate statistical power to reliably detect mutations occurring at a frequency of at least 3% in a particular histotype [25]. Therefore, sequencing tyrosine kinase genes in additional serous ECs may shed further insights into the frequency and spectrum of mutations in potentially druggable targets in this clinically aggressive subtype. Moreover, the lack of a systematic search for mutations in the tyrosine kinome of clear cell ECs merits such an analysis for this histological subtype.

Here, we performed a mutation discovery screen to determine the incidence of somatic mutations in the catalytic domains of 86 tyrosine kinases in a series of 24 primary serous, 11 clear cell, and 5 mixed (serous-endometrioid) histology ECs. Somatically mutated genes were then resequenced from another 72 ECs, and evaluated for copy number alterations in all 112 tumors. We report low-frequency somatic mutations and copy number gains of the TNK2 (tyrosine kinase non-receptor, 2) and DDR1 (discoidin domain receptor tyrosine kinase 1) kinases among the three major histological subtypes of EC.

Methods

Ethics statement

The NIH Office of Human Subjects Research determined that this research activity was exempt from Institutional Review Board review.

Clinical specimens

Anonymized, fresh-frozen, primary tumor tissues and matched histologically normal tissues were obtained from the Cooperative Human Tissue Network (100 cases), which is funded by the National Cancer Institute, or from the Biosample Repository at Fox Chase Cancer Center, Philadelphia PA (1 case). DNAs from another 11 cases of fresh-frozen tissue, including all five mixed histology (endometrioid-serous) cases (Additional file 1), were purchased from Oncomatrix. To the best of our knowledge, the mixed-histology tumor tissues were not macrodissected to separate individual histological components prior to DNA extraction by Oncomatrix. The entire cohort of 112 cases consisted of 45 serous, 21 clear cell, 41 endometrioid, and 5 mixed histology ECs. The endometrioid cases consisted of grade 1 (n = 26), grade 2 (n = 12), grade 2/3 (n = 1), and grade 3 (n = 2) tumors (Additional file 1). All primary tumor tissues were collected prior to treatment. For tumor tissues (n = 100) procured from CHTN, a hematoxylin and eosin (H&E) stained section was cut from each tumor specimen and reviewed by a pathologist to verify histology and to delineate regions of tissue with a tumor cell content of ≥70%.

Nucleic acid isolation

Genomic DNA was isolated from macrodissected tissue with greater than 70% tumor cellularity using the Puregene kit (Qiagen).

Identity testing

Paired tumor-normal DNA samples were genotyped using the Coriell Identity Mapping kit (Coriell). Genotyping fragments were size separated on an ABI-3730xl DNA analyzer (Applied Biosystems). Alleles were scored using GeneMapper software.

Primer design, PCR amplification, nucleotide sequencing and variant calling

M13-tailed primer pairs (Additional file 2) were designed to target 577 of 591 exons that encode the catalytic domains of the 86 protein tyrosine kinases (Additional file 3), using previously published methods [26]. Sequence constraints precluded the design of primers for 14 of 591 exons. Primers were also designed to target the exons that encode the exonuclease domain (exons 3 to 13) of POLE (polymerase (DNA directed), epsilon, catalytic subunit) and are available on request. PCR amplification conditions are available upon request. Bidirectional Sanger sequencing of PCR products and subsequent nucleotide variant calling were performed as previously described [27]. Variant positions were cross-referenced to the dbSNP (Build 129) database to annotate and exclude known germline variants. To determine whether novel variants were somatic mutations or germline variants, the appropriate tumor DNA and matched normal DNA were re-amplified in an independent PCR followed by sequence analysis of the variant position. Primers used in the secondary screen of TNK2 and DDR1 are provided in Additional file 4.

Quantitative real-time PCR

Predesigned primers targeting TNK2 (VPH103-1002824A), DDR1 (VPH106-0859748A) and B2M (beta-2-microglobulin) (VPH115-0515670A) were purchased from SABiosciences (Qiagen). Reactions were assembled to contain either Taqman control genomic DNA (Applied Biosystems) or 2 ng of tumor genomic DNA, 2 μl of primers (diluted 1:4), 3.5 μl SYBR Green Rox qPCR mastermix (Qiagen), to a final 10 μl reaction volume. qPCR was preformed on a ABI 7900 HT Fast Real-Time PCR System (Applied Biosystems) with the following cycle conditions: 50°C for 2 min, 95°C for 10 min, and 40 cycles of 95°C for 15 sec and 60°C for 1 min. A standard curve was generated with Taqman control genomic DNA, to permit a determination of the absolute quantitation using SDS 2.4 software (Applied Biosystems). For each experiment, tumor samples were assayed in triplicate for the target gene and control gene (B2M). For each sample, the mean quantity of each target gene was normalized to the mean quantity of B2M. For tumors displaying copy number gains (defined here as a ≥3-fold increase of the target gene compared to B2M), the matched normal DNAs were analyzed to confirm that the copy number gain was somatic. Three independent experiments were performed for each tumor and normal pair. The fold change in somatic copy number was determined by dividing the normalized mean quantity of the target gene in the tumor sample by the normalized mean quantity of the target gene in the matched normal sample. In addition, a 2-tailed Student t-test was used to calculate statistical significance.

Estimation of statistical power of study design

The estimated power to detect one gene mutation in a set of 40 tumors was calculated as 1 - (1-X)^40, where X is the actual fraction of tumors with a mutation in that gene (Additional file 5).

Cell lines and immunoblotting

Serous endometrial cancer cell lines (ARK1 and ARK2) were kindly provided by Dr. Alessandro Santin (Yale School of Medicine). RL-95-2, HEC1A, HEC1B, KLE were obtained from the American Type Culture Collection, or the National Cancer Institute’s Developmental Therapeutics Program. RL95-2 was established from a grade 2 moderately differentiated adenosquamous carcinoma of the endometrium [28], KLE was established from a poorly differentiated endometrial carcinoma [29], HEC1A was established from a human moderately differentiated endometrial adenocarcinoma [30, 31], and HEC1B is a sub-line of HEC1A [31, 32]. Cells were washed in phosphate-buffered saline then lysed with ice-cold RIPA buffer (Thermo Scientific) containing 1 mM Na-orthovanadate, 10 mM NaF, and 1X protease inhibitor cocktail (Roche). Lysates were centrifuged and proteins were quantitated with the Bio-Rad protein assay (Bio-Rad 500–0006). Equal amounts (μg) of the cleared lysate were denatured at 95°C in 2X SDS sample buffer (Sigma) prior to SDS-PAGE and transfer to PVDF membranes (Bio-Rad). Primary and HRP-conjugated secondary antibodies were: αDDR1 (Cell Signaling), αTNK2 (Upstate), αβ-Actin (Sigma), goat anti-mouse HRP (Cell Signaling), and goat anti-rabbit HRP (Cell Signaling). Immunoreactive proteins were visualized with enhanced chemiluminescence (Pierce).

Results

The TNK2 and DDR1tyrosine kinases are somatically mutated in endometrial carcinomas

In a mutation discovery screen, we sequenced 577 exons that encode the catalytic domains of 86 tyrosine kinases (Additional file 3), from 24 serous, 11 clear cell, and 5 mixed (serous/endometrioid) histology endometrial carcinomas. We selectively sequenced the catalytic domain of each kinase because this domain can be preferentially mutated in other cancers [12, 15, 33]. For a gene that has kinase domain mutations at an actual frequency of 10%, we estimate that a discovery screen of 24 serous tumors has 92.0% statistical power to observe at least one mutation (Additional file 5). For a discovery screen of 11 clear cell tumors and 5 mixed histology tumors the corresponding statistical power is estimated to be 68.6% and 40.9% respectively (Additional file 5). Six serous tumors (T27, T33, T45, T56, T65, T75) in our discovery screen were previously subjected to whole exome sequencing [19].

We obtained high quality sequence data for 84% (11.8 Mb) of targeted bases (14.1 Mb). After excluding known germline variants, there were 24 nucleotide variants that represented potential somatic mutations. Sequencing of the matched normal DNA revealed that two of the 24 variants were bona fide nonsynonymous somatic mutations. The somatic mutations occurred in TNK2 (Tyrosine kinase non-receptor protein 2) and DDR1 (Discoidin domain receptor tyrosine kinase 1). We therefore extended our analysis of TNK2 and DDR1 to sequence the remaining coding exons from the 40 tumors in the discovery screen and to sequence all coding exons of TNK2 and DDR1 from another 72 primary endometrial tumors consisting of 10 clear cell, 21 serous, and 41 endometrioid tumors. The secondary screen revealed nine additional nonsynonymous somatic mutations localizing to the catalytic and non-catalytic domains of the encoded proteins (Figure 1, Additional file 6, Additional file 7).

Figure 1
figure 1

Localization of nonsynonymous, somatic mutations in TNK2 and DDR1 relative to important functional domains of the proteins. All the somatic mutations were uncovered in primary endometrial tumors. Individual missense mutations (black boxes) are distinguished from frameshift mutants (black diamonds). Abbreviations: CB, clathrin binding site; CRIB, Cdc42/Rac interactive binding; DS, discoidin; SAM, sterile alpha motif; SH3, Src Homology 3; TM, transmembrane; UBA, ubiquitin associated.

Full size image

Overall, among the 112 tumors in this study, TNK2 was somatically mutated in 2.2% (1 of 45) of serous, 4.8% (1 of 21) of clear cell, 7.3% (3 of 41) of endometrioid, and 20% (1 of 5) of mixed histology endometrial tumors. DDR1 was somatically mutated in 4.4% (2 of 45) of serous tumors and in 2.4% (1 of 41) of endometrioid tumors (Table 1). Of the three endometrioid tumors that harbored somatic DDR1 or TNK2 mutations, two cases (T88 and T117) were grade 1 and one case (T131) was grade 3. Overall, there was no significant difference in the frequency of TNK2/DDR1 mutations between low/intermediate-grade and high-grade endometrioid ECs; 2 of 38 (5.3%) low/intermediate-grade (grade 1 or grade 2) endometrioid ECs had a TNK2 or DDR1 mutation compared with 1 of 3 (33.3%) high-grade (grade 2/3 or 3) endometrioid ECs (P = 0.2086). The TNK2D572N, TNK2R849W, TNK2R256H, and DDR1R570Q missense mutants are predicted, by at least two in silico algorithms, to impact the function of the encoded proteins (Table 1). Immunoblotting confirmed that TNK2 and DDR1 are endogenously expressed in endometrial cancer cells (Figure 2).

Table 1 Somatic mutations of TNK2 , and DDR1 identified among 112 primary ECs
Full size table
Figure 2
figure 2

TNK2 and DDR1 are expressed in endometrial cancer cell lines. Immunoblots showing expression of the TNK2 and DDR1 proteins in a panel of endometrial cancer cell lines. Actin served as a loading control.

Full size image

Increased copy number of TNK2 and DDR1in endometrial carcinoma

We next used quantitative real-time PCR to determine whether TNK2 or DDR1 were affected by copy number alterations among the 112 endometrial tumors in this study. Somatic copy number increases of TNK2 were detected in 8.9% (4 of 45) of serous tumors and in 2.4% (1 of 41) of endometrioid tumors, but in none of the clear cell tumors (Table 2). The single endometrioid tumor displaying a copy number gain of TNK2 was a grade 2 tumor. Somatic copy number increases involving DDR1 were detected in 2.2% (1 of 45) of serous tumors (Table 2). For each gene, tumors that displayed copy number alterations were distinct from tumors that had somatic mutations (Additional file 8). Considering mutations and copy number alterations together, TNK2 was somatically altered in 11.1% (5 of 45) of serous, 4.8% (1 of 21) of clear cell, 9.8% (4 of 41) of endometrioid, and 20% (1 of 5) of mixed histology tumors and DDR1 was somatically altered in 6.7% (3 of 45) of serous and 2.4% (1 of 41) of endometrioid tumors but not in clear cell or mixed histology tumors (Additional file 8).

Table 2 Copy number gains of TNK2 and DDR1 among 112 primary ECs
Full size table

Copy number gains of TNK2 and DDR1 could reflect either targeted gene amplification of these kinases or gain of a multigenic genomic region encompassing these genes. To discriminate between these two possibilities, we interrogated the copy number status of TNK2, DDR1, and their flanking genes within The Cancer Genome Atlas (TCGA) catalogue of somatic alterations in serous and endometrioid ECs [21], via the cBIO Cancer Genomics Portal [34]. In the serous and endometrioid ECs within the TCGA cohort, copy number gains involving TNK2 and DDR1 were not focal but extended to flanking genes.

A subset of TNK2 and DDR1 mutated tumors are POLE-mutant or microsatellite unstable

Somatic mutations in the exonuclease domain of POLE and/or microsatellite instability (MSI) occur in a subset of ECs and are associated with elevated mutation rates [21]. We therefore sought to determine whether any of the TNK2- or DDR1-mutated cases were coincident with POLE mutations or MSI-positivity. We sequenced exons 3–13 of POLE, which encode the exonuclease domain, from all 112 tumors in our study; the MSI status of tumors in this study has previously been reported [19]. Three tumors had somatically mutated POLE (T3 (c.C890T;p.S297F), T24 (c.1096delT; p.F367Lfs*15), and T97 (c.C857G; p.P286R), Additional file 9). Overall, somatic mutations within the exonuclease domain of POLE were detected in 2.2% (1 of 45) of serous, 4.8% (1 of 21) of clear cell, and 2.4% (1 of 41) of endometrioid tumors in our cohort. The frequency of POLE mutations in TNK2-DDR1 mutated cases (1 of 7; 14%) compared with TNK2-DDR1 non-mutated cases (2 of 105; 2%) was not statistically significantly different (P = 0.1775). Of the seven tumors with TNK2 or DDR1 mutations, one case (T3) had a somatic mutation within POLE (POLES297F) and another three cases (T77, T88, and T117) were MSI-positive (Table 1). One of the two frameshift mutations in T117, an MSI-positive tumor, occurred at a polynucleotide (Cn) tract (Additional file 6), suggesting that this mutation (TNK2P761Rfs*72) may have arisen as a consequence of defective mismatch repair.

Discussion

Herein we report the occurrence of low-frequency somatic mutations in the TNK2 and DDR1 kinases among serous, clear cell, and endometrioid ECs. The TNK2 non-receptor tyrosine kinase is activated in response to a variety of stimuli including ligand-dependent stimulation of receptor tyrosine kinases [35], Cdc42 [36], and integrin-mediated cell adhesion [37]. TNK2 activation has been implicated in the regulation of cell growth, survival, and integrin-mediated cell adhesion and migration [37–41], and overexpression of TNK2 in cultured cells promotes a metastatic phenotype [37]. The DDR1 receptor tyrosine kinase is activated by triple-helical collagens [42] and has been implicated in the regulation of cell adhesion, survival, proliferation, differentiation, migration, invasion, morphogenesis and development [43–53].

Of the seven endometrial tumors that had somatic mutations in TNK2 and/or DDR1 in our study, one tumor (T3) was POLE-mutant and three tumors (T77, T88, and T117) were microsatellite-unstable, raising the possibility that the TNK2 and DDR1 mutations in these cases may have arisen as a consequence of replicative and mismatch repair defects respectively. A determination of whether the TNK2 and DDR1 mutations uncovered in this study are pathogenic driver mutations or incidental passenger mutations will ultimately rely on functional studies. In the interim, the potential effects of the TNK2 and DDR1 mutations on protein function can be postulated based on their positions relative to known functional domains of the encoded proteins and on in silico predictions. In this regard, the TNK2R256H mutant occurs within the catalytic loop of TNK2, at a conserved residue that forms a hydrogen bond with an ATP analog [54], and is predicted, in silico, to impact protein function. The TNK2D572N mutant occurs within a motif (LIDF) that is essential for binding to the clathrin heavy chain [55], and is predicted to be deleterious. Because a synthetic mutation (TNK2D572A) at this precise residue results in loss of clathrin binding [55], we speculate that the somatic TNK2D572N mutant might likewise alter the TNK2-clathrin interaction. The TNK2P761Rfs*72 mutant was recurrent in our study occurring in two endometrioid ECs one of which was MSI-positive. The TNK2P633Afs*3 frameshift mutation may also be recurrent: we observed TNK2P633Afs*3 (chr3:195,595,228-195,595,229 insC; Hg19) in a POLE-mutant serous EC and this variant has been catalogued by others in cancer cell lines and tumors although in those instances it has not been subjected to technical validation (URL: http://www.cbioportal.org/public-portal/). Both TNK2P633Afs*3 and TNK2P761Rfs*72 are predicted to encode truncated forms of TNK2 that lack the UBA (ubiquitin associated) domain, which has been implicated in ligand-dependent proteasomal degradation of TNK2 [38, 56]. An earlier observation that deletion of the UBA domain of TNK2 results in elevated protein levels [56], together with a report that synthetic C-terminal deletion mutants of TNK2 retain catalytic activity [57, 58], raises the possibility that the naturally occurring TNK2P633Afs*3 and TNK2P761Rfs*72 mutants found in this study might encode elevated levels of truncated but catalytically active proteins.

The three DDR1 mutations we identified in EC consisted of two missense mutations (DDR1R570Q and DDR1R574S), and a frameshift mutation (DDR1N740Ifs*10) that occurred in an MSI-positive tumor. The DDR1R570Q missense mutation, which we identified in a case of serous EC that was microsatellite-stable and POLE-wildtype, has been identified by others in an endometrioid EC [21], and in a case of metastatic melanoma [59]. Thus, the recurrent nature of the DDR1R570Q mutation across studies suggests it may be a pathogenic event that provides a selective advantage in tumorigenesis, including endometrial tumorigenesis.

In the recent catalogue of genomic alterations reported by TCGA for endometrioid and serous ECs, somatic mutations of TNK2 were documented in 2% of serous ECs and in 1% of endometrioid ECs, and somatic mutations of DDR1 were noted in 4% of serous ECs and 2% of endometrioid ECs [21, 34, 60]. The eight mutations we uncovered in TNK2 are different to the three TNK2 mutations previously described in EC by TCGA. Similarly, two of the three mutations we describe in DDR1 are unique to this study whereas, as discussed earlier, the third mutation (DDR1R570Q, CCDS4690; alternatively annotated as DDR1R607Q, CCDS34385) was present in a case of serous EC in this study and in a case of endometrioid EC by TCGA. Therefore, our observations not only validate the recent findings of low frequency somatic mutations in TNK2 and DDR1 in serous and endometrioid ECs by TCGA [21], but extend upon those findings by refining knowledge of the mutation spectrum of TNK2 and DDR1 in EC. Moreover, to our knowledge this is the first systematic search for somatic mutations in the tyrosine kinome of clear cell ECs.

In addition to somatic mutations, we also uncovered copy number gains involving TNK2 at an appreciable frequency in serous and endometrioid ECs (8.9% and 2.4% respectively), and copy number gains involving DDR1 at low frequency (2.2%) in serous ECs in our study. However, from an analysis of the TCGA endometrial cancer data, increased TNK2 and DDR1 copy number appears to reflect regional gains rather than focal amplification, thus making their potential biological relevance in endometrial cancer difficult to predict.

It is worth noting that our study has several limitations. First, our mutation discovery screen was restricted to the exons encoding the catalytic domains of tyrosine kinases and would not have detected mutations present in other exons. Second, our discovery screen did not have high statistical power to detect moderately to infrequently mutated genes (Additional file 5). Third, the use of Sanger sequencing for mutational analysis, in both the discovery screen and subsequent secondary screens of TNK2 and DDR1, may have precluded the identification of sub-clonal variants that are below the sensitivity of detection by this methodology.

Conclusions

In conclusion, we have identified rare somatic mutations and copy number alterations involving the TNK2 and DDR1 kinases amongst serous, clear cell, and endometrioid ECs. Our findings validate and extend the observation of TNK2 and DDR1 mutations in serous and endometrioid ECs catalogued by TCGA. To our knowledge, this is the first systematic search for somatic mutations in the tyrosine kinome of clear cell ECs. The recurrent nature of the TNK2P761Rfs*72 and DDR1R570Q mutants raises the possibility that these may be pathogenic events that bestow a selective advantage in endometrial tumorigenesis. Future mechanistic studies of the somatic mutations reported herein are warranted.

Availability of supporting data

All data supporting the somatic mutations reported in the manuscript are provided in Additional files 6, 7 and 9. Sanger sequencing files for the entire study will be made available through dbGAP with controlled access.