Next generation sequencing of carcinoma of unknown primary reveals novel combinatorial strategies in a heterogeneous mutational landscape

Background Advanced carcinoma of unknown primary (CUP) has limited effective 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 methylation, 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 epigenetic 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 combinatorial strategies targeting impaired cell cycle arrest, first with epigenetic modifiers and, second, with MAPK/PI3K pathway inhibition.


INTRODUCTION
Carcinoma of unknown primary (CUP) remains a unique challenge to the clinician in a landscape where algorithms for the diagnosis, management and outcomes of cancers are often histology-dependent. Indeed, CUP represents a heterogeneous group of malignancies with a distinct 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][2][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 [4].
The preliminary categorization on microscopic evaluation 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, undifferentiated neoplasm, melanoma, sarcoma, or lymphoma seen in the remaining 10% of cases [5,6]. Next, after identification of preliminary histology, the pathologic evaluation continues with a systematic immunohistochemistry algorithm to identify the general cellular subset, including adenocarcinoma, neuroendocrine, germ-cell, etc. while a further testing based on cytokeratins (CKs) may reveal the detailed phenotypic expression of specific organ [1].
Progress in gene expression and molecular profiling further aids the sub-classification of the tissue of origin into predominant subtypes of cancer, such as gastrointestinal, gynecologic, etc [7][8][9][10]. However, the final definitive histologic identity of the cancer remains elusive. Recent rapid advances in genomic profiling present new opportunities to not only characterize CUP but also offer insight into pathways and cellular systems with aberrations 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 clinical trials program to identify potentially future targets for drug development and clinical trial design.

Patients characteristics
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 diagnosis 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 moderately differentiated and one poorly differentiated). Remaining histologies described included moderate to well-differentiated epithelioid neoplasm (n=2, 12%), poorly differentiated neuroendocrine (n=2, 12%), and poorly differentiated 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 median of 2 mutations and a range of 0 to 8. Table 2 details the pathways and cellular process implicated by the identified molecular aberrations. Overall, the most common aberrations 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 pathways affected within each histologic subtype (detailed in Table 3) with an emphasis of the two most predominant histology, carcinoma and adenocarcinoma. Of the five patients 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 amplification. 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 cycle regulation including CDKN2A/B loss, and CCND1 and CCNE1 amplifications.
Next, 4 patients had the second predominant histologic 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, highlighting 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 implication in a signal transduction cascade. Specifically 5 patients (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 kinases. 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 [11]. Similarly two patients had a CCND1 amplification, which encodes Cyclin D1, which in turn interacts with the cyclin-dependent kinases Cdk4 and Cdk6, resulting 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 therefore lead to excessive proliferation [12,13]. Additional aberrations included three (18%) patients with SOX2 amplification, three (18%) each with amplification in FGF3, FGF4, and FGF19, two (12%) with CCND1 amplification, 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 encodes key regulators of cell-cycle progression [12].

Impaired epigenetic regulation and DNA methylation
Similarly, 8 (47%) had impaired epigenetic regulation and DNA methylation with mutations in most commonly 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 transcription), SETD2 which encodes a histone lysine-36 methyltransferase (both G1644* and N2071fs*17 in the same patient), ATRX R840fs*29 (which encodes a helicase protein and binds tightly to chromatin during chromosomal segregation at mitosis), and CREBBP S893L (which encodes proteins acting as intrinsic histone acetyltransferases and as stabilizers within the transcription complex) [16].

Other aberrations: Tumor suppressors and transcriptional regulators
Seven (41%) patients demonstrated one or more alterations in tumor suppressor genes while 10 (41%) patients had unique alterations in transcriptional regulators. Most commonly, five patients had 6 unique aberrations within TP53, specifically R273C, R248Q, R 196*, and www.impactjournals.com/oncoscience 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 concurrent 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 demonstrated 8 mutations overall including alterations in the PI3K, MAPK pathways (mutations in FBXW7, FGFR1) with concurrent epigenetic and transcriptional deregulation, 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 amplified, 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 mutations 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. Similarly, 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 amplification 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.

Clinical outcomes
Of the 17 patients, 11 (65%) elected to participate in a phase I clinical trial after having met all eligibility criteria. 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 patient 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-
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 concurrently in colorectal cancer [19].
Our early analysis suggests emerging patterns of aberrations where impaired cell cycle arrest has been observed concurrently with either epigenetic deregulation or activation of the cell signaling. Aberrations in signal transduction pathways including the PI3K and MAPK pathways were consistently observed. Specific mutations including 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%) [20]. These FBXW7 mutations have been showed to occur prior to the F-box domain and the highly conserved WD40 repeat region, an area targeted by proteasome-mediated degradation processes. Therefore, mutations in this region have implicated in the stabilization of oncogenic intermediates including the mTOR signaling protein. Early preclinical 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, suggesting 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 [22]. 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 pathway inhibitor (including mTOR inhibitors) [19].
Our analysis also highlights the emerging description of epigenetic deregulation in advanced CUP patients. In a subset of patients, the identification of somatic mutations in epigenetic genes encoding proteins which lead to alterations in specific DNA methylation and chromatin modification patterns, most commonly through inappropriate gene silencing [23,24]. The increased prevalence of somatic mutations noted in CUP patients that lead to epigenetic 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 mutations further hopes to identify combinatorial strategies for this malignancy. In our set of patients, 4 of 5 (80%) patients 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 associated 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 specifically deregulated in our larger subset of patients, specifically the patient with poorly differentiated carcinoma and adenocarcinoma. In the carcinoma subset, 4 of five patients demonstrated an aberration leading to activation of the PI3K/AKT/mTOR pathway, suggesting that activation of this cascade was more closely associated with this histologic subset when compared to the other 4 subsets. In-  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 clinical trial, leading to a small sample size with tumor histologic 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.