Next generation sequencing of carcinoma of unknown primary reveals novel combinatorial strategies in a heterogeneous mutational landscape
1Department of Palliative, Rehabilitation, and Integrative Medicine, University of Texas MD Anderson Cancer Center, Houston, TX, USA
2 Department of Investigational Cancer Therapeutics, Division of Cancer Medicine, University of Texas MD Anderson Cancer Center, Houston, TX, USA
3Department of Pathology, all at the University of Texas MD Anderson Cancer Center, Houston, TX, USA
Correspondence to: David S. Hong, email: firstname.lastname@example.org
Keywords: carcinoma of unknown primary, next generation sequencing
Received: March 24, 2017
Accepted: May 02, 2017
Published: June 23, 2017
Copyright: Subbiah et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License 3.0 (CC BY 3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Background: Advanced carcinoma of unknown primary (CUP) has limited effec-tive therapeutic options given the phenotypic and genotypic diversity. To identify future novel therapeutic strategies we conducted an exploratory analysis of next-generation sequencing (NGS) of relapsed, refractory CUP.
Methods: We identified patients in our phase I clinic where archival tissue was available for a targeted NGS CLIA-certified assay.
Results: Of 17 patients tested, 15 (88%) demonstrated genomic alterations (median 2 aberrations; range 0–8, total 59 alterations). Nine (53%) patients had altered cell signaling including the PI3K/AKT/MTOR (n=5, 29%) and MAPK pathways (n=3,18%); 7 (41%) patients demonstrated ≥1 alterations in tumor suppressor genes (TP53 in 5 patients), 8 (47%) had impaired epigenetic regulation and DNA methyla-tion, 8 (47%) had aberrant cell cycle regulation, commonly in the cyclin dependent kinases. Ten (59%) patients had alterations in transcriptional regulators. Concurrent mutations affecting cell cycle regulation were noted to occur with aberrant epigen-etic regulation (n=6, 35%) and MAPK/PI3K pathway (n=5, 29%).
Conclusion: Every patient had a unique molecular profile with no two patients demonstrating an identical panel of mutations. We identify two emerging novel com-binatorial strategies targeting impaired cell cycle arrest, first with epigenetic modi-fiers and, second, with MAPK/PI3K pathway inhibition.
Carcinoma of unknown primary (CUP) remains a unique challenge to the clinician in a landscape where al-gorithms for the diagnosis, management and outcomes of cancers are often histology-dependent. Indeed, CUP rep-resents a heterogeneous group of malignancies with a dis-tinct disease course and biology, often displaying aggres-sive behavior with a short period of clinical symptoms preceding diagnosis and early dissemination to multiple metastatic sites leading to advanced staging at presentation and [1–3]. CUP is defined as a metastatic cancer without a clearly identified primary site despite an adequate standard workup including an in-depth pathologic analysis with concurrent detailed history, physical examination, and laboratory and radiologic assessments .The preliminary categorization on microscopic evalu-ation classifies an overarching histology as adenocarcinomas (ranging from well, moderate, poor, or undifferentiated), which represent up to 90% of CUP cases; lesser prevalent characterizations include squamous cell carcinomas, undif-ferentiated neoplasm, melanoma, sarcoma, or lymphoma seen in the remaining 10% of cases [5, 6]. Next, after identi-fication of preliminary histology, the pathologic evaluation continues with a systematic immunohistochemistry algo-rithm to identify the general cellular subset, including adeno-carcinoma, neuroendocrine, germ-cell, etc. while a further testing based on cytokeratins (CKs) may reveal the detailed phenotypic expression of specific organ .
Progress in gene expression and molecular profiling further aids the sub-classification of the tissue of origin into predominant subtypes of cancer, such as gastrointes-tinal, gynecologic, etc [7–10]. However, the final defini-tive histologic identity of the cancer remains elusive. Re-cent rapid advances in genomic profiling present new opportunities to not only characterize CUP but also offer insight into pathways and cellular systems with aberra-tions that may be of value as a therapeutic target. To that end, we performed next generation sequencing on patients with advanced, relapsed CUP referred to our phase I clini-cal trials program to identify potentially future targets for drug development and clinical trial design.
We identified 17 consecutive patients with advanced, relapsed CUP on whom adequate tissue was available for molecular profiling. Patient characteristics are described in Table 1. The median age of this group at the time of diagno-sis was 49 years (range, 18-72), while the age at time of initial phase I referral was 54 years (range, 21-75). Patients had a preserved performance status with all having an ECOG PS of 0 (n=2, 12%) or 1 (n=15, 88%). Patients were heavily pretreated with the median number of prior systemic therapies prior to phase I referral being 3 (range, 0-8). On pathologic evaluation, five of 17 patients (29%) had tumor with a microscopic description of carcinoma, 4 of which was poorly differentiated and one undifferentiated. Four patients’ tumors (24%) were characterized as adenocarcinoma, 3 of which were poorly- and one moderately differentiated. Three patients had a squamous cell of unknown primary (two mod-erately differentiated and one poorly differentiated). Remain-ing histologies described included moderate to well-differ-entiated epithelioid neoplasm (n=2, 12%), poorly differentiated neuroendocrine (n=2, 12%), and poorly dif-ferentiated sarcomatoid neoplasm (n=1, 6%).
Molecular analysis by histologic subtype
Of the 17 patients who had adequate tissue for molecular profiling, 15 patients (88%) had one or more molecular aberrations identified on the NGS assay; two patients (one squamous cell and one neuroendocrine by histology) did not reveal any identifiable aberrations. The 17 specimen harbored a total of 59 alterations with a me-dian of 2 mutations and a range of 0 to 8. Table 2 details the pathways and cellular process implicated by the identi-fied molecular aberrations. Overall, the most common ab-errations led to impaired cell cycle arrest with 8 patients demonstrating 21 aberrations, including 3 patients with CDKN2A/B loss and 2 with CCND1 amplification.
We then examined the cellular processes and path-ways affected within each histologic subtype (detailed in Table 3) with an emphasis of the two most predominant histology, carcinoma and adenocarcinoma. Of the five pa-tients with a carcinoma, the median number of aberrations was 2 (range 1 -8), and the median age at time of phase I evaluation was 65.2 years (range 48.2 – 74.6). Four of the 5 patients demonstrated an aberration leading to activation of the PI3K/AKT/mTOR pathway, including one patient with a PIK3CA Q75E mutation and a second patient with both a PIK3CA E545K mutation as well as an amplifica-tion. Three of these five carcinoma patients demonstrated a mutation in the tumor suppressor TP53 (R273C, R248Q, R196*) while two patients had five aberrations in cell cy-cle regulation including CDKN2A/B loss, and CCND1 and CCNE1 amplifications.
Next, 4 patients had the second predominant histo-logic subtype of adenocarcinoma. Notably, all 4 patients had an aberration implicated in epigenetic deregulation, specifically in ARID1A (Y1211fs*5 and E2250fs*28) in two samples, a dual SETD2 mutation (G1644*, N2071fs*17) and a CREBBP S893L mutation, highlight-ing a role in targets cellular epigenetics in therapy. Two mutations including CDKN2A/B loss leading to impaired cell cycle regulation were also noted.
Signal transduction mechanisms: Activation of the PI3K and MAPK pathways
Overall 8 (47%) patients had an aberration implica-tion in a signal transduction cascade. Specifically 5 pa-tients (29%) had 6 mutations leading to aberrant activation of the PI3k/Akt/mTOR signaling pathways, including three patients with PIK3CA mutations, one in the more common site H1047R, and two in less prevalent sites, Q75E and E545K with a concurrent amplification. Three patients had a mutation with FBXW7 mutation (one splice 726+1 G>A, a second W244*, and a third R465H). Three patients (n=3, 18%) had aberrant activation of the MAPK pathway with one patient harboring a KRAS amplification and two with FGFR1 amplification.
Impaired cell cycle regulation
Overall, 8 (47%) patients had mutations in cell cycle regulators, most commonly in the cyclin dependent ki-nases. Three patients (18%) harbored a CDKN2A/B loss, both genes encode the tumor suppressors p15 and p16, therefore loss of these genes leads to the deregulation of the p16-CDK4/Cyclin/Rb pathway and loss of cell cycle control . Similarly two patients had a CCND1 ampli-fication, which encodes Cyclin D1, which in turn interacts with the cyclin-dependent kinases Cdk4 and Cdk6, result-ing in the phosphorylation and inactivation of Rb and the progression of the cell cycle. Studies have shown that overexpression or amplification of Cyclin D1 may there-fore lead to excessive proliferation [12, 13]. Additional aberrations included three (18%) patients with SOX2 am-plification, three (18%) each with amplification in FGF3, FGF4, and FGF19, two (12%) with CCND1 amplifica-tion, and one with CCNE1 amplification. FGF 3, 4 and 19 encode various fibroblast growth factors and are known to be located a region of chromosome 11q13 that also en-codes key regulators of cell-cycle progression .
Impaired epigenetic regulation and DNA methylation
Similarly, 8 (47%) had impaired epigenetic regula-tion and DNA methylation with mutations in most com-monly in ARID1A which encodes the AT-rich interactive domain-containing protein 1A, a member of the SWI/SNF chromatin remodeling complex. Three different mutations in ARID1A were reported (the frameshift mutation Y1211fs*5, S1929fs*25, and E2250fs*28) [14, 15]. Other genes with reported aberrations included MLL2 R4904* (which is histone methyltransferase that is involved in the transcriptional response to progesterone signaling), KD-M6A S466 (implicated in the epigenetic regulation of tran-scription), SETD2 which encodes a histone lysine-36 meth-yltransferase (both G1644* and N2071fs*17 in the same patient), ATRX R840fs*29 (which encodes a helicase pro-tein and binds tightly to chromatin during chromosomal segregation at mitosis), and CREBBP S893L (which en-codes proteins acting as intrinsic histone acetyltransferases and as stabilizers within the transcription complex) .
Other aberrations: Tumor suppressors and transcriptional regulators
Seven (41%) patients demonstrated one or more al-terations in tumor suppressor genes while 10 (41%) pa-tients had unique alterations in transcriptional regulators. Most commonly, five patients had 6 unique aberrations within TP53, specifically R273C, R248Q, R 196*, and R248W including one patient with an adenocarcinoma who had a dual mutation within TP53 with a relatively uncommon N-terminal missense mutation L45P and a con-current Q38fs*79, which is frameshift mutation leading to the truncation of the p53 protein prior to the conserved DNA-binding domain region. This patient’s tumor demon-strated 8 mutations overall including alterations in the PI3K, MAPK pathways (mutations in FBXW7, FGFR1) with concurrent epigenetic and transcriptional deregula-tion, specifically ARID1A, ETV1 rearrangement, NOTCH1 APIP-NOTCH1 fusion, and MYST3 amplification. One patient had an MDM2 amplification while a second showed an MDM4 amplification, both of which, when am-plified, act as tumor suppressors acts to inhibit the activity of p53 [17, 18].
Concurrent mutations: Deregulation of cell cycle with PI3K pathway activation
Of the 17 patients, we further analyzed the incidence of concurrent mutations affecting two or more pathways and cell processes (Figure 1). Of the 8 patients with muta-tions leading to deregulation of the cell cycle, we identi-fied 4 (50%) patients with a concurrent mutation in the PI3K cascade and one with ERBB2 amplification, part of the epidermal growth factor receptor (EGFR) family. Sim-ilarly, of the 5 patients with mutations linked to the PI3K pathway, 4 (80%) occurred concurrently with an aberrance in cell cycle regulation (including CDK12, CCNE1 ampli-fication in 2 patients, CDKN2A/B loss). Furthermore, of the 8 patients with cell cycle aberrations, six (75%) had a concurrent mutations associated with epigenetic regulation and DNA methylation (including mutations in KDM6A, MLL2, ARID1A in 2 patients, and SETD2). Outcomes of molecular profiling for each patient are detailed in Table 4.
Of the 17 patients, 11 (65%) elected to participate in a phase I clinical trial after having met all eligibility crite-ria. The remaining 6 proceeded to receive conventional chemotherapies or a clinical trial closer to home. Table 5 details the specific clinical trials and outcomes of each pa-tient in our study. One patient received monotherapy while the remaining 10 received a combination of 2 or more drugs. All 11 patients were treated with a regimen that in-cluded a targeted therapy agent. Of the 11 patients treated on phase I clinical trials, over half (n=7, 64%) received a therapy matched with their mutational profile. The best tumor responses noted were stable disease lasting 4 or more months seen in 4 patients, 3 of whom had therapy matched to their mutational profile. Two of the 11 treated patients currently remain on therapy for over 8 months, first on a combination of carboplatin, bevacizumab, and temsirolimus and the second on crizotinib and pazopanib.
Our exploratory analysis provides insight into the molecular fingerprint of advanced, relapsed carcinoma of unknown primary. Indeed, no two patients had an identical molecular profile, highlighting the need for truly personal-ized therapy for this patient population with few effective therapeutic options particularly in the relapsed setting.The role for molecular profiling to identify the tissue of origin has been explored in depth in CUP; however data into identification of somatic alterations as identifiers of tissue of origin continues to be slowly developed where coexistence patterns in concurrent mutations may suggest the organ system origin. Existing large scale molecular profiling analysis notes that PIK3CA H1047R and KRAS mutations appear to coexist in genitourinary cancers and breast cancers when noted, and are less prevalent concur-rently in colorectal cancer .
Our early analysis suggests emerging patterns of ab-errations where impaired cell cycle arrest has been ob-served concurrently with either epigenetic deregulation or activation of the cell signaling. Aberrations in signal trans-duction pathways including the PI3K and MAPK path-ways were consistently observed. Specific mutations in-cluding FBXW7 have been implicated in the activation of PI3K/Akt/mTOR signaling pathways. This mutation has been identified in a variety of solid tumors including colorectal cancer (14%), squamous cell cancer of head and neck (11%), liver (8%), and ovarian cancers (3%) . These FBXW7 mutations have been showed to occur prior to the F-box domain and the highly conserved WD40 re-peat region, an area targeted by proteasome-mediated deg-radation processes. Therefore, mutations in this region have implicated in the stabilization of oncogenic interme-diates including the mTOR signaling protein. Early pre-clinical data has revealed response to the mTOR inhibitor rapamycin in cell lines with the inactivation of Fbxw7 and confers sensitivity to rapamycin, an mTOR inhibitor, sug-gesting a potential role for the more widely used mTOR inhibitors temsirolimus and everolimus; however their sensitivities to rapalogs remain to be studied prospectively as retrospective reviews of patient outcomes have not shown a definitive signal of activity [20, 21]. Furthermore, early data also suggests that inactivation of Fbxw7 may confer a resistance to anti-tubulin chemotherapies . Similarly patients harboring deregulation of the PI3K pathway through PIK3CA mutations including H1047R have been shown to have a response rate (defined as stable disease ≥6 months or partial response) of 45% in patients with all advanced cancers when treated with a PI3K path-way inhibitor (including mTOR inhibitors) .
Our analysis also highlights the emerging descrip-tion of epigenetic deregulation in advanced CUP patients. In a subset of patients, the identification of somatic muta-tions in epigenetic genes encoding proteins which lead to alterations in specific DNA methylation and chromatin modification patterns, most commonly through inappropri-ate gene silencing [23, 24]. The increased prevalence of somatic mutations noted in CUP patients that lead to epi-genetic changes reinforces the role for studying classes of drugs, specifically the histone deacetylase inhibitors such as vorinostat and panobinostat in this patient population, particularly in combination.
Furthermore, the observation of concurrent muta-tions further hopes to identify combinatorial strategies for this malignancy. In our set of patients, 4 of 5 (80%) pa-tients with a PIK3CA mutation had a concurrent aberrance in cell cycle regulation. While 6 of 8 (75%) patients with a cell cycle aberration had a coexisting mutation associ-ated with epigenetic regulation and DNA methylation, highlighting two potential combinations to be explored with therapeutic intent. Additionally histology-specific subset analysis reveals early insight into pathways specifi-cally deregulated in our larger subset of patients, specifi-cally the patient with poorly differentiated carcinoma and adenocarcinoma. In the carcinoma subset, 4 of five pa-tients demonstrated an aberration leading to activation of the PI3K/AKT/mTOR pathway, suggesting that activation of this cascade was more closely associated with this his-tologic subset when compared to the other 4 subsets. In-deed all 4 samples of adenocarcinoma, which was the sec-ond most predominant subtype among our CUP patients, demonstrated an aberration implicated in epigenetic de-regulation.
Our preliminary analysis highlights areas of future study but also does raise limitations. Clinical outcomes have yet to be explored longitudinally, particularly given that efforts were made to match each patient to a clinical trial with evidence of target inhibition based on their mutation profile. However the primary limitations of this effort were availability of specific clinical trials at time of patient need as well as time and travel commitments to meet the clinical trial participation requirements. Moreover, there exists a selection bias and its associated implications on tumor biology given that the molecular profiling was completed only on patients who were well enough to be considered for a referral to a phase I clini-cal trial, leading to a small sample size with tumor his-tologic heterogeneity. Nonetheless, these early patterns reveal cellular processes that are deregulated concurrent with signaling mechanisms and epigenetic modulations and represent novel combination therapies to be pursued in preclinical models and ultimately early phase clinical trials.
MATERIALS AND METHODS
Patient selection and treatment
We identified consecutive patients with CUP referred to the Clinical Center for Targeted Therapy (Phase I Clinical Trials Program) at MD Anderson Cancer Center starting from January 1, 2011 to December 31, 2013. Eligibility criteria for participation in phase I clinical trials included age >18 years, presence of metastatic or unresectable disease, measurable disease per Response Evaluation Criteria in Solid Tumors (RECIST) 1.0, Eastern Cooperative Oncology Group (ECOG) performance status (PS) of 0-1, and a life expec-tancy >3 months. Premenopausal women were required to have a negative pregnancy test and patients of childbearing potential to use contraception. Further eligibility criteria var-ied according to the particular study and all patients gave in-formed consent. All clinical trials were approved by the MD Anderson Institutional Review Board. Descriptive statistics summarized the patients’ characteristics.
Pathologic evaluation and mutational analyses and detection
Original hematoxylin and eosin slides were reviewed by an institutional pathologist to confirm CUP. Additional immunohistochemical staining to assist in the identification of tissue of origin was conducted as per pathologist’s dis-cretion. Archival formalin-fixed paraffin-embedded (FFPE) slides were then obtained and cut into 10 separate 5-mm sections. Next-generation sequencing from FFPE sections was completed in a Clinical Laboratory Improvement Amendments (CLIA)-certified laboratory using the Illu-mina HiSeq2000 platform (Foundation Medicine, Cam-bridge, MA, USA). Over 3230 exons of 236 cancer-related genes, plus over 47 introns from 19 genes often rearranged in cancer were fully sequenced for point mutations, inser-tions/deletions, copy number alterations (CNAs), select gene fusions, and variants of unknown significance. Then, the aberrations were classified into 10 encompassing cate-gories based on the pathway or cellular processes where they have been implicated, i.e. aberrations that affect apop-tosis, cell cycle regulation, epigenetic modulation/DNA methylation, signal transduction mechanism (PI3K, MAPK, Wnt or a receptor tyrosine kinase[RTK]), tran-scription regulation, or tumor suppressor.
CONFLICTS OF INTEREST
The authors report no conflicts in relation to this manuscript.
Financial support provided by NIH/NCI under award number P30CA016672 Cancer Center Support Grant (CCSG).
Last Modified: 2017-07-10 03:36:48 EDT