Next‐Generation Sequencing for Patients with Sarcoma: A Single Center Experience

This article describes result of targeted next‐generation sequencing in sarcoma patient samples. More than 400 cancerrelated genes were analyzed on a commercially available platform, with the goals of identifying patients who could be early phase clinical trial candidates, determining KIT or PDGFR‐a mutational status in gastrointestinal tumors, and characterizing tumor subtypes in sarcomas or cancers of unknown primary.

Abstract

Background.

Sarcomas comprise over 50 subtypes of mesenchymal cancers. For the majority of sarcomas, the driver mutations remain unknown. In this article, we describe our experience with a targeted next‐generation sequencing (NGS) platform in clinic patients.

Materials and Methods.

We retrospectively analyzed results of NGS using 133 tumor samples from patients diagnosed with a variety of sarcomas that were analyzed with targeted NGS covering over 400 cancer‐related genes (405 DNA, 265 RNA) on a commercially available platform.

Results.

An average of two gene alterations were identified per tumor sample (range 0–14), and a total of 342 DNA mutations were detected. Eight‐eight percent of samples had at least one detected mutation. The most common mutations were in the cell cycle, including TP53 (n = 35), CDKN2A/B (n = 23), and RB1 (n = 19). Twenty‐seven PI3‐kinase pathway alterations were observed, including PTEN (n = 14), PIK3Ca (n = 4), TSC1 (n = 1), TSC2 (n = 3), STK11 (n = 1), mTOR (n = 3), and RICTOR (n = 2). There were 75 mutations in genes that are targetable with existing drugs (excluding KIT in gastrointestinal stromal tumor) that would allow enrollment onto clinical trials. In general, the estimated tumor mutation burden was low, in particular for those with disease‐defining gene fusions or genetic alterations. Microsatellite instability (MSI) data were available for 50 patients, and all were MSI stable.

Conclusion.

Our study describes a single‐center experience with targeted NGS for patients with sarcoma. Mutations were readily detected and 75 (representing 40% of patients) were testable for therapeutic effect using existing drugs within the confines of a clinical trial. These data indicate that targeted NGS is a useful tool in potentially routing patients to mutation‐specific clinical trials. Further study will be required to determine if these mutations are clinically meaningful drug targets in sarcoma.

Implications for Practice.

The sarcomas are a heterogenous family of over 50 different mesenchymal tumors. Current practice for metastatic disease involves systemic chemotherapy or nonspecific kinase inhibitors such as pazopanib. Sarcomas typically lack the classic kinase alterations seen in many carcinomas. The role of next‐generation sequencing in sarcoma clinical practice remains undefined.

Introduction

Sarcomas are a rare, heterogeneous group of malignant mesenchymal cancers. It is estimated that there are at least 50 distinct subtypes [1]. Advances in molecular profiling of sarcomas allow for subgrouping into those with aneuploidy and those with molecularly defining events such as oncogene point mutations or gene fusions (e.g., c‐Kit in gastrointestinal stromal tumor [GIST], EWSR1‐FLI1 in Ewing sarcoma, etc.) [2], [3].

For metastatic sarcomas, other than a few subtypes (e.g., imatinib/sunitinib/regorafenib‐GIST and imatinib‐dermatofibrosarcoma protuberans), there are no U.S. Food and Drug Administration (FDA)‐approved therapies that are based on treating specific molecular targets. For most patients, current therapies largely focus around cytotoxic chemotherapy agents and tyrosine kinase inhibitors (i.e., pazopanib and olaratumab for soft tissue sarcomas). However, in the era of personalized medicine, it is the hope that other sarcomas will be identified where targeted therapies may be relevant.

Since 2012, at our high volume tertiary referral center, we have been analyzing selected sarcoma patient samples with next‐generation sequencing (NGS) through a commercial vendor (Foundation Medicine, Inc.). The primary goals included (a) identifying patients who could be early phase clinical trial candidates, (b) determining KIT or PDGFR‐α mutational status in GIST, and (c) characterizing tumor subtypes in sarcomas or cancers of unknown primary. Our experience with this emerging technology is described as follows.

Materials and Methods

Patients

Under an institutional review board‐approved protocol (#2007P002464), we retrospectively collected data from patients who were seen at the sarcoma clinic at Massachusetts General Hospital from August 2012 to October 2016. These patients had a wide variety of soft tissue and bone sarcomas. Their tumors were analyzed using NGS technology made commercially available by Foundation Medicine, Inc. The patients included in this study had metastatic or locally advanced/unresectable tumors, sarcomas with unknown primary site, or GIST where mutation analysis would impact the choice or dose of tyrosine kinase inhibitor. We collated the data by sarcoma subtypes and genes that were identified to have mutations.

NGS‐Based Genomic Profiling

A series of 133 sarcoma samples were assayed prospectively with a validated Clinical Laboratory Improvement Amendments‐certified comprehensive genomic profiling (CGP) platform (August 2012 to October 2016). DNA was extracted from 40µm of formalin‐fixed paraffin‐embedded (FFPE) sections, and CGP was performed on hybridization‐captured, adaptor ligation‐based libraries to a mean coverage depth of greater than ×550 for 236 or 315 cancer‐related genes plus select introns from 19 or 28 genes frequently rearranged in cancer, as described previously [4]. All classes of genomic alterations were identified, including base substitutions, insertions/deletions, copy number alterations, and rearrangements.

In a subset of samples, DNA and RNA were extracted from 40µm of FFPE sections of tissue. Adaptor‐ligated DNA underwent hybrid capture for all coding exons of 405 cancer‐related genes and select introns for 31 genes frequently rearranged in cancer. cDNA libraries prepared from RNA underwent hybrid capture for 265 genes and sequencing was performed to a median exon coverage depth of ∼3 million unique reads using Illumina sequencing.

Tumor Mutational Burden and Microsatellite Status

Tumor mutational burden (TMB) was calculated by using a novel algorithm as the number of somatic base substitution or indel alterations per megabase (Mb) of the coding region target territory of the test (currently 1.11 Mb) after filtering to remove known or likely somatic driver mutations and germline mutations and extrapolating that value to the exome or genome as a whole [5]. Tumors were classified as microsatellite unstable‐high (MSI‐H) or microsatellite stable (MSS) as previously described [6], [7].

Results

We identified 133 patient tumor samples (Table 1) analyzed using NGS. The patient subtypes sequenced reflect the referral scope of our practice. These also included patients seeking clinical trials and those without standard chemotherapy options. Sarcoma subtypes ranged from relatively common histologies (e.g., leiomyosarcoma [LMS]) and rare sarcomas on which we have a clinical focus (e.g., chordoma) to uncommon and exceedingly rare sarcomas (e.g., epithelioid sarcomas, extraskeletal myxoid chondrosarcoma). In this cohort of 133 patients, we detected 342 gene alterations (mean n = 1, median n = 2, range n = 0–14; analysis excluded variances of unknown significance). In all, 88% had at least one detectable gene alteration.

Table 1. Sarcoma patients and number of alterations per sample.

One hundred thirty‐three patient tumor samples underwent NGS.

Abbreviations: LGFMS, low‐grade fibromyxoid sarcoma; NGS, next‐generation sequencing; NOS, not otherwise specified; UPS, undifferentiated pleomorphic sarcoma.

Gene alterations arranged by disease and gene are displayed in Figure 1 (chordoma), Figure 2 (LMS), Figure 3 (soft tissue sarcomas [STS], including undifferentiated pleomorphic sarcomas [UPS]), and Figure 4 (remaining pooled sarcomas). Across all sarcoma subtypes, the most common alterations were in the cell cycle (Fig. 5), including TP53 (n = 35), CDKN2A/B (n = 23), and RB1 (n = 19). Most of the cell cycle alterations were concentrated in LMS, STS, and chordoma. For example, TP53 was altered in 15/23 LMS and 8/14 STS. RB1 was altered in 12/23 LMS but only 3 STS not otherwise specified (NOS). CDKN2A/B or CDKN2A were altered in 8/24 chordoma and 5/14 STS NOS.

Figure 1.

Tile plot of chordoma patient samples. Patients were labeled chronologically 1–24 and ranked by number of detected genetic alterations. Patient Chordoma‐1 was of the dedifferentiated chordoma subtype. Chordoma‐6 and Chordoma‐20 were of the poorly differentiated subtype.

Figure 2.

Tile plot of leiomyosarcoma (LMS) patient samples. Patients were labeled chronologically 1–23 and ranked by number of detected genetic alterations.

Figure 3.

Tile plot of the soft tissue sarcoma (STS) patient samples. Patients were labeled chronologically 1–13 and ranked by number of detected genetic alterations. STS‐1, −3, −5, −8, and −11 are undifferentiated pleomorphic sarcomas (UPS); STS‐2 is a UPS with leiomyosarcoma features; STS‐4 is a low‐grade fibroblastic sarcoma; STS‐6 is a low‐grade sarcoma not otherwise specified (NOS); STS‐7, −12, and −13 were STS NOS per pathology. STS‐9 and STS‐10 are myofibroblastic sarcomas.

Figure 4.

Tile plot of the remaining miscellaneous sarcoma patient samples.

Abbreviations: ASPS, alveolar soft parts sarcoma; DDLPS, dedifferentiated liposarcoma; DSRCT, desmoplastic small round cell tumor; EHE, epithelioid hemangioendothelioma; GIST, gastrointestinal stromal tumor; LGFMS, low‐grade fibromyxoid sarcoma; M/RCLPS, myxoid liposarcoma; MPNST, malignant peripheral nerve sheath tumor (of note: MPNST‐3 may be an MPNST or UPS where there was differential classification between pathologists); RMS, rhabdomyosarcoma; Sarcoma‐1, a low‐grade bone neoplasm of the jaw; SFT, solitary fibrous tumor; WDLPS, well‐differentiated liposarcoma.

Figure 5.

Prevalence by percent age of patients of the top gene alterations detected.

There were 27 PI3Kinase pathway alterations, including PTEN (n = 14), PIK3Ca (n = 4), TSC1 (n = 1), TSC2 (n = 3), STK11 (n = 1), mTOR (n = 3), and RICTOR (n = 2). Also observed were 14 cases of ATRX mutations, which primarily focused in LMS (n = 9/23 cases), dedifferentiated liposarcoma (n = 2/3 cases), STS NOS (n = 2/13 cases; 1 of which was likely an LMS), and 1 of 2 epithelioid sarcomas.

We detected known and/or expected disease‐defining gene alterations that were included in the assay, including KIT (GIST), CDK4/MDM2 amplification (liposarcoma), IDH (chondrosarcoma), and various gene fusions such as EWSR1‐FLI1, TWSR1‐NR4A3, EWSR1‐ATF1, FUS‐DDIT3, etc. Of note, 29 patient samples were from sarcomas associated with gene fusions. In 18 of these, Fluorescent In Situ Hybridization (FISH) was not sent on the pathology specimens because it was not required for the clinical and pathological diagnosis. Of the remaining 11, the NGS was concordant with FISH in clear cell sarcoma (n = 1 of 1, EWSR1‐ATF1), desmoplastic small round cell tumor (DSRCT, n = 1 of 1, EWSR1‐WT1), Ewing sarcoma (n = 1 of 3, EWSR‐FLI1), synovial sarcoma (n = 3 of 3, SS18‐SSX1 and SS18‐SSX2), and extraskeletal myxoid chorndrosarcoma (n = 1 of 1, EWSR1‐NR4A3). For two patients, the NGS changed the clinical diagnosis. In one Ewing sarcoma patient, the tumor was originally classified as a salivary gland tumor. With NGS performed at our institution and Foundation Medicine, the EWSR1‐FLI1 transcript was detected in both assays. In one of the epithelioid hemangioendothelioma (EHE) patients the tumor was originally classified as a high‐grade epithelioid angiosarcoma. When the NGS returned with the WWTR1‐CAMTA fusion, the diagnosis was updated. Of the patients without FISH diagnostics on their clinical pathology samples, there were two Ewing sarcoma patients with EWSR1‐FLI1, one synovial sarcoma with SS18‐SSX1, one low grade fibromyxosarcoma with FUS‐CREB3L2, one myxoid liposarcoma with FUS‐DDIT3, one additional EHE patient with WWTR1‐CAMTA, and one endometrial stromal tumor with a JAZF1‐SUZ12 fusion detected.

Of the detected alterations, the assessment that these findings can lead to clinical trials or commercially available therapies remains challenging and controversial. Excluding the oncogene KIT in GIST, a target with FDA‐approved drug options (KIT mutations were observed in eight of nine patients with GIST) we defined clinically testable or actionable alterations as those that would be acceptable for eligibility for recent or ongoing clinical trials. Examples include the ongoing NCI‐MATCH (NCT02465060) basket study, ASCO‐TAPUR (NCT02693535), or studies enrolling specific alterations for patients with sarcomas or solid tumors (alterations and trials listed in Table 2). We were able to identify 75 actionable alterations in 133 patients (some patients had multiple actionable alterations). In the PI3Kinase pathway, we identified 27 mutations/deletions (PTEN n = 14, PIK3CA n = 4, mTOR n = 2, TSC1/2 n = 4, STK11 n = 1, RICTOR n = 2) across a range of sarcomas including angiosarcoma (mTOR n = 1), chondrosarcoma (PTEN and TSC1 n = 1), chordoma (mTOR n = 1; TSC2 n = 1; PI3KCa n = 1; PTEN n = 1), GIST (PTEN n = 1), LMS (PTEN n = 5; PTEN and RICTOR n = 1; PI3KCa n = 1), myxoid liposarcoma (PTEN n = 1, PTEN and PI3KCa n = 1, PI3KCa n = 1), osteosarcoma (TSC2 n = 1; RICTOR n = 1), STS (PTEN n = 1, PTEN and TSC2 n = 1), MPNST (PTEN n = 1) and endometrial stromal tumor (STK11 n = 1).

Table 2. Detected gene targets and example clinical trials.

Seventy‐five potentially actionable alterations were detected across 342 genetic alterations in 133 patient samples.

Clinically testable cell cycle alterations were relatively common. These included 26 of 133 patients with loss of CDKN2A or CDKN2A/B including angiosarcoma (n = 1), chondrosarcoma (n = 3), chordoma (n = 8), clear cell sarcoma (n = 1), liposarcoma (n = 1), Ewing sarcoma (n = 2), GIST (n = 2), MPNST (n = 2), osteosarcoma (n = 1), and STS (n = 5). Additional patients had CCND1, CCNE1, CCND2, CCND3 amplifications including: chordoma, DDLPS, LMS (n = 2), RMS, and STS NOS.

In addition, we observed MET amplification (chordoma n = 1, DDLPS n = 1, WDLPS n = 1), and mutations in NF1 (MPNST n = 1, pleomorphic RMS n = 1, UPS n = 1), NF2 (myxofibrosarcoma n = 1), BRCA2 (chordoma n = 1, osteosarcoma n = 1), IDH1/2 (chondrosarcoma n = 2), NRAS (UPS n = 1), and SMARCB1 (poorly differentiated chordoma n = 1).

Tumor mutational burden was calculated for each sample in our dataset and reported in the number of mutations per Mb (Fig. 6). In nearly all samples, the TMB was in the low or intermediate range. Two soft tissue sarcomas were identified with levels in the high range (one undifferentiated pleomorphic sarcoma, one high‐grade soft tissue sarcoma with leiomyosarcoma features).

Figure 6.

Estimated tumor mutation burden per megabase by median and quartile.

Abbreviations: GIST, gastrointestinal stromal tumor; LMS, leiomyosarcoma; MPNST, malignant peripheral nerve sheath tumor; STS, soft tissue sarcoma.

Microsatellite instability data were available on 50 of the patient samples. Microsatellite instability was stable for all tested samples (data not shown).

Discussion

The sarcomas remain a challenging field for drug development. This is due to multiple factors, including disease heterogeneity, rarity, absence of classically targetable kinase drivers, and poorly understood pathophysiology, among others. In an effort to link patients to clinical trials, determine KIT status for GIST patients, and aid pathology diagnosis, we have been offering selected patients targeted NGS of archived pathology specimens. These patients typically had metastatic disease, received one or more prior chemotherapy regimens for systemic therapy, and had an expected lifespan of greater than 6 months.

In the 133 patients sequenced, we found primarily alterations in the cell cycle and known driver mutations/gene fusions. Oncogene amplifications were readily detected but were of unclear significance. The findings in general were consistent with what has been reported in sarcoma literature and sequencing databases (for examples, see references [8], [9], [10], [11], [12], [13], [14], [15], [16]). Whether these detected gene alterations identify patients who may benefit from molecularly targeted therapy remains controversial. Excluding KIT in GIST, in our actionable gene list that was derived from ongoing or reported clinical trials, we found 75 alterations with potential clinical trial options. Few of these patients were realistically able to enroll in trials due to various clinical and logistic scenarios including trial slot availability, disease status, and competing nontargeted therapy trials.

We detected a number of amplifications in known oncogenes. Of these, only MET is currently an eligibility criterion for open clinical trials. It remains unknown if KIT, KDR, EGFR, IGF1R, and PDGFR‐α amplification is meaningful, although it is certainly of interest given the number of drugs being tested in clinical trials.

Tumor mutational burden is a calculated estimate of the number of mutations per Mb. Estimates of mutation burden are of emerging interest because they are hypothesized to be predictive of genomic instability and perhaps even response to immune therapies. The numbers are still being defined, but, in general, 20 mutations per Mb and 30 mutations per Mb are considered intermediate and high levels of burden in this assay [5]. In our cohort of sarcomas, TMB scores were low. This is expected for sarcomas with genetically defined lesions such as translocations (e.g., Ewing sarcoma), oncogene disease‐defining deletions (epithelioid sarcoma), or low‐grade sarcomas. It is surprising, however, that these scores were also low in sarcomas that have been characterized as genomically unstable, such as osteosarcoma [17], [18]. Feasible explanations include bias from limited coverage of the genome with the assay and small sample size.

The immune checkpoint inhibitor pembrolizumab was recently approved by the FDA for tumors that are MSI‐H. We had data on 50 of our sample set and all were MSS. It is possible that there are subsets of sarcomas that were not highly represented in our data that are MSI‐H. Further research will be needed to determine if this is a feasible strategy for identifying sarcoma patients for immunotherapies.

There are limitations to our study. First, over the 5 years we collected data, the sequencing assay evolved. This can lead to a false impression of increasing mutation burden rather than just a feature of which genes were included. Secondly, with this heterogeneous population, we cover a breadth of disease but there is limited depth within certain sarcoma subtypes. Thus, one should not overinterpret the frequency of alterations in specific sarcomas. Finally, even though the gene set is extensive, we are missing known disease‐defining alterations that were not included in the assay design. For example, it has been established that solitary fibrous tumor is known to frequently have NAB2‐STAT6 fusions. Two patients in our cohort with this disease had no detected mutation, but this is clearly an artifact of assay undercoverage.

The high cost for these assays and as of yet unvalidated utility in many cancer types suggests we should limit use of these types of tests. In our patient population of orphan diseases with limited therapeutic options, we feel that targeted NGS still maintains a role. Our preference is to be conservative to avoid misinterpretation or over‐interpretation by exposing patients to toxic but ineffective therapies. We believe it is reasonable to perform NGS for c‐KIT/PDGFR‐α assessment in metastatic GIST, cancer or sarcoma of unknown primary to aid diagnostics, and for potential clinical trial candidates. We favor the model of the NCI‐MATCH and ASCO‐TAPUR (sequencing as a research tool) or as part of internal research studies when it is available. At our institution, we have moved to an internal assay (not included in this article), in which our policy allows for targeted NGS as an option for rare tumors, such as the sarcomas. The additional burden to the cost of health care will be limited as the sarcomas make up only 1% of malignancies.

Conclusion

We performed targeted NGS on 133 sarcoma patients in the context of routine clinical care. We were able to detect a number of mutations, mostly concentrated in the cell cycle and known oncogene drivers. A significant portion of molecular targets that were identified is covered by clinical trials.

Author Contributions

Conception/design: Gregory M. Cote, Edwin Choy

Provision of study material or patients: Gregory M. Cote, Edwin Choy

Collection and/or assembly of data: Gregory M. Cote, Jie He

Data analysis and interpretation: Gregory M. Cote, Jie He

Manuscript writing: Gregory M. Cote, Edwin Choy

Final approval of manuscript: Gregory M. Cote, Jie He, Edwin Choy

Disclosures

Jie He: Foundation Medicine, Inc (E, OI). The other authors indicated no financial relationships.

(C/A) Consulting/advisory relationship; (RF) Research funding; (E) Employment; (ET) Expert testimony; (H) Honoraria received; (OI) Ownership interests; (IP) Intellectual property rights/inventor/patent holder; (SAB) Scientific advisory board