Molecular profiling of metastatic colorectal tumors using next-generation sequencing: a single-institution experience

Background Recent molecular characterization of colorectal tumors has identified several molecular alterations of interest that are considered targetable in metastatic colorectal cancer (mCRC). Methods We conducted a single-institution, retrospective study based on comprehensive genomic profiling of tumors from 138 patients with mCRC using next-generation sequencing (NGS) via FoundationOne. Results Overall, RAS mutations were present in 51.4% and RAF mutations were seen in 7.2% of mCRC patients. We found a novel KRASR68S1 mutation associated with an aggressive phenotype. RAS amplifications (1.4% KRAS and 0.7% NRAS), MET amplifications (2.2%), BRAFL597Ralterations (0.7%), ARAFS214F alterations (0.7%), and concurrent RAS+RAF (1.4%), BRAF+RAF1 (0.7%), and rare PTEN-PIK3CA-AKT pathway mutations were identified and predominantly associated with poor prognosis. ERBB2 (HER2) amplified tumors were identified in 5.1% and all arose from the rectosigmoid colon. Three cases (2.2%) were associated with a hypermutated profile that was corroborated with findings of high tumor mutational burden (TMB): 2 cases with MSI-H and 1 case with a POLE mutation. Conclusions Comprehensive genomic profiling can uncover alterations beyond the well-characterized RAS/RAF mutations associated with anti-EGFR resistance. ERBB2 amplified tumors commonly originate from the rectosigmoid colon, are predominantly RAS/BRAF wild-type, and may predict benefit to HER2-directed therapy. Hypermutant tumors or tumors with high TMB correlate with MSI-H status or POLE mutations and may predict a benefit from anti-PD-1 therapy.


INTRODUCTION
Colorectal cancer (CRC) remains the third leading cause of cancer death in both men and women in the United States with an estimated 134,490 new cases and 49,190 deaths in 2016 [1]. Recent advances in the treatment of metastatic CRC (mCRC) have identified improved outcomes with the addition of epidermal growth factor receptor (EGFR)-targeting agents to conventional combination cytotoxic therapy in patients with extended RAS wildtype tumors. In contrast, activating mutations in the RAS gene (KRAS or NRAS, present in approximately 50% of cases of mCRC) and BRAF gene (present in about 5% of mCRC patients) have been associated with lack of clinically meaningful benefit or harm when anti-EGFR therapy is employed [2]. The identification of candidates for anti-EGFR therapy through the exclusion of RAS and BRAF mutations in mCRC serves as a model of selecting optimal therapy based on patient genomic profiles and molecular phenotypes.
Several decades of genomic studies, including the use of more recent next-generation sequencing (NGS), have expedited the search of genetic alterations for potential therapeutic targeting in CRC [3,4]. Recently, comprehensive molecular characterization of 224 colorectal www.impactjournals.com/oncotarget/ Oncotarget, 2017, Vol. 8, (No. 26), pp: 42198-42213 Research Paper www.impactjournals.com/oncotarget tumors was performed by The Cancer Genome Atlas (TCGA) Network [5]. Sixteen percent of colorectal tumors were found to be hypermutated and more commonly found in the right colon with 75% of these cases demonstrating expectedly high microsatellite instability (MSI-H). Twentyfour genes were identified to have significant mutations of interest including APC, SMAD4, TP53, PIK3CA, and KRAS mutations, as expected. Interestingly, mutations, deletions, or amplifications of the ERRB gene family were found in 19% of tumors. In sum, this genomic analysis identified several molecular alterations that are considered targetable, including mediators of dysregulated WNT, RAS, and PI3K pathways such as ERRB2, ERRB3, MEK, AKT, MTOR, IGF2, and IGFR.
The recent identification of gene mutations and amplifications of potential significance for therapeutic purposes has led us to investigate the genomic profiles of mCRC patients using NGS (FoundationOne). Here, we describe a single-institution experience in reporting results from comprehensive genomic analysis of tumors from 138 mCRC patients. We aim to characterize genetic alterations present in our study population that have known correlates to prognosis, therapeutic resistance, and potential therapeutic targets in mCRC. In this study, we also report the existence of concurrent gene mutations rarely described in the literature and novel mutations and amplifications that can lead to targeting outside of National Comprehensive Cancer Network (NCCN) standard treatments.

Study population
The molecular results from FoundationOne testing of tumors from 138 mCRC patients are summarized in Table 1
In our patient population, 2 KRAS amplifications (2.8%) and 1 NRAS amplification (1.4%) were identified. One patient was a 51-year-old female with KRAS amplified rectal cancer with synchronous diffuse metastases (lung and liver). Her best overall response to standard firstline combination chemotherapy (5-fluorouracil (5-FU) and irinotecan or FOLFIRI) plus anti-EGFR therapy (panitumumab) was stable disease (SD) for 6 months. The other patient with KRAS amplification was a 51-yearold male diagnosed with right-sided colon cancer and synchronous metastases to the liver and peritoneum who had rapid progression on first-and second-line non-anti-EGFR based therapies. Our 74-year-old male patient with NRAS amplification presented with poorly differentiated rectosigmoid adenocarcinoma and synchronous diffuse metastases (liver, mesentery, and bones) and experienced progressive disease (PD) at 2 months on second-line FOLFIRI + cetuximab. Notably, a novel KRAS R68S1 alteration ( Figure 2) was identified in a 41-year-old female (1.4%) with rectal cancer and synchronous metastases to the liver and retroperitoneal and supraclavicular lymph nodes who experienced PD at 2 months on anti-EGFR therapy with second-line irinotecan + cetuximab.

RAF mutations
A total of 11 RAF mutations (1 concurrent BRAF+RAF1 mutation) were found in 7.2% of our patients ( Figure 3). Of these, BRAF V600E activating mutations (exon 15) were the most common single mutations present (40.0%). One activating BRAF L597R alteration (exon 15) was identified (10.0%) in a 56-year-old male with bulky rectal adenocarcinoma with synchronous metastases that progressed through 9 months of first-line anti-EGFR therapy. One activating ARAF S214F alteration was also identified (10.0%) in our series of RAF mutations. This 60-year-old male patient developed multiple recurrences of rectal adenocarcinoma including, most recently, metastatic disease to the lung treated with neoadjuvant 5-FU, oxaliplatin, and irinotecan (FOLFOXIRI) followed by metastatectomy; he remains in clinical remission. A dual BRAF V600E +KRAS A164V^sub alterationwas present (10.0%) in an elderly male (age 72) with poorly differentiated rightsided colon cancer with synchronous metastases on firstline systemic combination therapy without anti-EGFR agents. Here an oncogenic RAS alteration was paired with a known activating BRAF mutation.

ERBB2 amplifications
Seven patients (5.1%) were found to have ERRB2 amplified tumors with one having a concurrent KRAS G12D mutation ( Figure 4). The majority of these tumors were MSS (87.5%) with HER2 copy numbers that ranged from 9-190 (Table 2). Notably, all ERRB2 amplified tumors were located in the rectosigmoid colon as its primary disease site. Four patients with RAS wild-type ERBB2 amplification received anti-EGFR therapy, 3 experienced SD ≥ 4 months (2 first-line and 1 secondline) and 1 (second-line) experienced a PR lasting for 5 months as their best overall response to anti-EGFR therapy. The concurrent ERRB2 amplified and KRAS G12D mutated tumor was found in a 58-year-old male with moderately differentiated rectal adenocarcinoma with synchronous solitary liver metastasis treated with neoadjuvant 5-FU, oxaliplatin (FOLFOX) followed by hepatic resection and resection of primary -he is currently under surveillance and without evidence of disease.

AKT1/2 mutations
Three patients (2.2%) had AKT1 E17K mutations while 1 patient (0.7%) had an AKT2 E17K mutation (Table 3). Of these, a majority had concurrent mutations (75%) and tumors located in the right colon (75%). One AKT1 E17K mutated tumor was found to have concurrent BRAF V600E +KRAS A164V^sub alterations with phenotype described above. One AKT1 E17K mutated tumor had concurrent alterations in KRAS A146T +PIK3CA G106V and was found in a 61-year-old male with initial right-sided colon cancer that recurred with metastases to the liver showing moderately differentiated colon adenocarcinoma. His tumor was characterized by aggressive features, including metastatic disease recurrence following a diagnosis of stage I disease, and development of bony metastases within the first year of recurrence. A concurrent AKT2 E17K +KRAS G12C altered tumor was found in a 57-year-old female with originally moderately differentiated sigmoid adenocarcinoma that was resected but recurred with metastases to the retroperitoneal lymph nodes currently on first-line FOLFIRI + bevacizumab.

MET amplifications
Three patients (2.2%) in our series had MET amplifications (Table 5). Two-thirds of these tumors were MSS, located in the right colon, and associated with concurrent mutations in RAS or RAF genes. One 67-year-old male was initially diagnosed with rightsided colon cancer (KRAS G13D +MET alterations present) and synchronous liver metastases. His course has been punctuated by recurrent metastases to the liver and lungs despite several systemic and regional therapies. Another right-sided colon cancer was identified with both a deactivating BRAF D594G mutation and MET amplification with aggressive phenotype described above. A third patient was a 27-year-old male with primary rectal adenocarcinoma that recurred with metastases to the liver and retroperitoneal lymph nodes and refractory to capecitabine + irinotecan + cetuximab. In particular, 2 of 3 patients with MET amplications and RAS/BRAF V600E wildtype tumors were refractory to anti-EGFR-based therapies.

Hypermutant status
The majority of our 138 patients with mCRC had tumors with <9 clinically significant alterations (121 or 87.7%) as described by FoundationOne reports ( Table 6). The majority of these tumors were located in the left colon and all were MSS. Fourteen patients (10.1%) had 9-16 total alterations while only 3 patients (2.2%) were allocated to the highest number of clinically significant alterations category (17)(18)(19)(20)(21)(22)(23)(24)(25). Notably, 2 patients with MSI-H tumors were identified in the highest number of alterations group. One 55-year-old female patient was found to have a dual PIK3CA and PTEN mutated rectal tumor with phenotype associated POLE V411L mutation. Interestingly, this patient was elderly (age 80), had a right colon tumor, and had a recurrence pattern consistent with locoregional recurrence. This patient demonstrated a TMB of 122 mutations/Mb, which was the highest among the cohort.

DISCUSSION
Comprehensive molecular characterization of 138 tumors from patients with mCRC was performed via NGS  (FoundationOne) in this single-institution retrospective study. Overall, 51.4% and 7.2% of our patients with mCRC were shown to carry RAS and RAF mutations, respectively, which is concordant with frequencies historically reported in mCRC [2]. The majority of our RAS mutations were KRAS mutations of exon 2 (codons 12 and 13), which represent those identified in initial phase III trials that predicted lack of benefit from anti-EGFR therapy in mCRC [6,7]. We also found that 15.5% of all RAS mutations in our population comprised a panel of extended RAS mutations. This is also consistent with recent data from the PRIME and CRYSTAL clinical trials, where exon 3 and 4 KRAS and exons 2, 3, and 4 NRAS mutations reflected 14-17% of RAS mutations [8,9]. Identifying these rare RAS mutations has major clinical significance, given their association with anti-EGFR resistance [10].
Notably, we identified 2 KRAS amplifications and 1 NRAS amplification that are extremely rare and poorly characterized. These were found in 3 patients with diffusely metastatic CRC progressive through several lines of systemic therapy including anti-EGFR therapy.
Putative high-level amplifications of NRAS were observed in <1% of cases in TCGA dataset though its significance in CRC remains poorly described [5]. KRAS amplifications have been associated with acquired resistance to EGFR inhibitors cetuximab or panitumumab in CRC preclinical models [11]. To our knowledge, we are the first to report a novel KRAS R68S1 alteration that was associated with a particularly aggressive phenotype and PD at 2 months on anti-EGFR therapy with cetuximab.
The majority of RAF mutations found in our population were BRAF V600E activating mutations (exon 15), which have been historically associated with poorer survival, resistance to chemotherapy, and lack of clinical benefit with anti-EGFR therapy in mCRC [12][13][14][15]. We also identified a lone BRAF L597R alteration (exon 15), which is poorly described in CRC but has been shown to similarly activate RAF-MEK-ERK signaling in melanoma in vitro [16]. Of note, this patient received 9 months of first-line anti-EGFR therapy though our sample size of 1 precludes any meaningful generalizations. One ARAF S214F alteration was also identified in a patient whose course has been characterized by multiple recurrences of rectal    cancer. Mutations in ARAF have been linked as oncogenic drivers in lung adenocarcinoma, and are exceedingly rare in CRC and comprise approximately 2% of cases in the CRC dataset from TCGA [5,17]. Treatment with the oral RAF inhibitor, sorafenib, has demonstrated prolonged response in a case of refractory non-small-cell lung cancer and rapid responses in patients with refractory histiocytic neoplasms bearing somatic ARAF mutations [17,18]. Despite a previous conception that KRAS and BRAF mutations are mutually exclusive, we found 1 dual BRAF V600E +KRAS A164V^sub mutated tumor that, in our case, was associated with poor prognostic features [19]. One case of concurrent BRAF G466V +KRAS G12S mutation and one patient with a concurrent BRAF G469E +RAF1 S257L mutation were present in our cohort. BRAF mutants G466V and G469E have been shown to represent variants with impaired or complete loss of kinase activity in vitro [20,21]. Nevertheless, it has been shown that tumorigenesis is promoted in the presence of deactivating BRAF mutations through oncogenic RAS mutation and/or CRAF (or RAF-1) signaling [21,22]. In our study, one deactivating BRAF G466V mutation was paired with an oncogenic KRAS G12S mutation, and one deactivating BRAF G469E mutation was paired with an oncogenic RAF1 S257L alteration, supporting the notion of an evolutionary adaptation in the cancer genome to overcome BRAF mutations with impaired function. In both cases, there were associated features of poor prognosis though the dual BRAF G466V +KRAS G12S mutated tumor has seen disease control recently on 41 cycles of FOLFIRI and bevacizumab, which may argue for varying degrees of relative contribution from each mutation on tumor phenotype. Interestingly, one patient with deactivating BRAF D594G mutation was refractory to all lines of treatment, including anti-EGFR therapy, and ultimately died of aggressive disease. This is at odds with recent reports suggesting that BRAF D594G mutation may be an indicator of good prognosis [23]. It is unclear whether this patient's concurrent MET amplification may have contributed to his overall poor prognosis and therapeutic resistance.
ERBB2 (HER2/neu) amplifications were found in 5.1% of our mCRC patients with the majority in KRAS wild-type tumors (except for 1 with a concurrent ERRB2+KRAS G12D alteration). Another FoundationOne analysis of >10,000 cases of gastrointestinal malignancies identified HER2 amplifications and mutations in 3.0% and 4.8%, respectively, of cases from the CRC cohort [24]. Our patients with HER2 amplified tumors appeared to have shortened clinical benefit with anti-EGFR therapy, which is consistent with the recent phase II HERACLES trial where none of the patients with HER2 amplified, RAS/BRAF wild-type metastatic colorectal tumors had a response to anti-EGFR therapy [25]. Similar to the preponderance of left colon primary tumors in the HERACLES trial, all of our HER2 amplified tumors were located in the rectosigmoid colon. In short, the identification of HER2 amplifications in patients with RAS/BRAF wild-type metastatic colorectal tumors is of major significance given the clinical benefit derived from dual HER2-directed therapy including trastuzumab + lapatinib (HERACLES) or trastuzumab + pertuzumab (MyPathway) [25,26].
PIK3CA, PTEN, and AKT mutations were identified in 18.1% (25), 10.9% (15), and 2.9% (4) of our mCRC patients, respectively. Many of these patients had metastatic tumors associated with aggressive features. In addition, 75% of AKT mutated tumors were located in the right colon, almost half (48.0%) of PIK3CA mutated tumors were right-sided colon cancers, and concurrent mutations in RAS-RAF-MAPK or PTEN-PIK3CA-AKT signaling were common. For example, 19 patients (76.0%) with PIK3CA mutations also had concurrent KRAS mutations while 5 (20.0%) and 1 (4.0%) with PIK3CA altered tumors also had concurrent PTEN and AKT1 mutations, respectively. Mutations in mediators of the PTEN-PIK3CA-AKT signaling pathway in CRC have been associated with poorer prognosis and lack of clinical response to anti-EGFR therapy [27,28]. For PIK3CA mutations, in particular, prior studies have demonstrated that exon 9 mutations had no effect while exon 20 mutations were associated with resistance to anti-EGFR therapy [29]. However, this differential effect by exon has not been supported by recent meta-analysis [30]. Given the high rate of concurrent RAS mutations seen with PIK3CA and related pathway mutations, a definitive association between resistance to EGFR inhibition and MET amplification and increased c-MET expression have also been associated with an aggressive phenotype and therapeutic resistance, particularly to MEK inhibition, in mCRC [31,32]. Interestingly, we have observed anti-EGFR refractoriness in 2 of our patients with MET amplifications despite the presence of a RAS-wild-type phenotype and lack of activating BRAF mutations. This is consistent with preclinical data suggesting MET activation as a mechanism of resistance to anti-EGFR therapy [33].
We lastly identified 3 patients (2.2%) with tumors categorized in the highest number of clinically significant alterations group (17)(18)(19)(20)(21)(22)(23)(24)(25) that also demonstrated high TMB as per FoundationOne. TMB categories per FoundationOne testing have been validated in melanoma patients treated with PD-1 blockade [34]. Response to PD-1 inhibitors was significantly superior in patients with high TMB (>23.1 mutations/MB) compared to intermediate or low TMB (3.2-23.1 mutations/MB and <3.2 mutations/ MB, respectively). Furthermore, a recent phase II study showed that patients with advanced urothelial cancer who responded to the programmed death ligand 1 (PD-L1) inhibitor atezolizumab had a significantly higher TMB (median 12.4 mutations/Mb) than non-responders (median 6.4 mutations/Mb, p < 0.0001) [35]. Two patients had MSI-H tumors while 1 hypermutant tumor was MSS and harbored a POLE mutation. Interestingly, 42.9% of tumors with 9-16 clinically significant alterations were located in the right colon while one-third of tumors with 17-25 total alterations were located in the right colon; tumors with <9 number of alterations were predominantly located in the left colon. In the CRC dataset from TCGA, 75% of hypermutated tumors arose from the right colon yet not all of them were MSI-H [5]. Mutations in polymerase ε or POLE were found among 25% of hypermutated tumors in this cohort. Mutations in POLE have been shown to contribute to an ultramutated yet MSS phenotype in colorectal tumors [36]. A recent NGS study confirmed that increasing mutational load correlated with MSI yet colorectal tumors with the highest mutational burden that were distinct from MSI tumors all harbored POLE mutations [37]. Furthermore, mismatch repair-deficiency or MSI has recently been shown to predict clinical benefit to immune checkpoint blockade with anti-PD-1 therapy in mCRC [38]. The characterization of mutational load in CRC may serve as a better indicator than MSI status in determining a hypermutant profile that could predict benefit from immunotherapy. Our findings are hypothesis generating and offer support to consider molecular analysis of tumors to determine the total number of alterations as a potential correlate to MSI and candidacy for anti-PD-1 therapy in mCRC.
Future studies of larger size and, ideally, prospective design will be helpful in corroborating associations between molecular alterations of interest described in our study and prognosis, resistance to EGFR inhibition, and/or ability to be targeted for therapy in mCRC. Comparative genomic analyses have identified a high level of concordance particularly for RAS, BRAF, and PIK3CA mutations between colorectal primary and metastatic tumors [39,40]. However, other molecular alterations may differ based on the site of tumor and/or exposure to chemotherapy [41][42][43][44]. Although such mixed results are likely dependent on the specific mutation that is profiled, other factors including specimen integrity and sampling method may also contribute to heterogeneity. Indeed, further analyses are needed to describe the concordance or discordance of other mutations across tumor sites and treatment effects in mCRC, and careful consideration in design will be needed in order to account for confounding factors as described above.
In conclusion, comprehensive genomic profiling can uncover gene alterations beyond conventional RAS or RAF mutant subtypes that predict resistance to anti-EGFR therapy and in identifying potential therapeutic targets outside of NCCN standard treatments in mCRC. ERBB2 amplified tumors commonly originate from the rectosigmoid colon, are predominantly RAS/BRAF wildtype, and may predict benefit to HER2-directed therapy. Hypermutant tumors or tumors with POLE mutations may predict benefit to anti-PD-1 therapy. Our findings are hypothesis generating and warrant further investigation in larger datasets and in prospective settings.

Study patients and tumor samples
Patients with advanced or metastatic (stage IV) colorectal cancer treated at the Gastrointestinal Medical Oncology Clinic at City of Hope National Medical Center (Duarte, CA) between April 2013 and February 2016 were screened for this study. Eligibility criteria was limited to those who underwent expanded genomic tumor analysis by FoundationOne. There were no exclusions to tumor histology, medical comorbidities, previous treatment or lines of prior therapy, or performance status. Comprehensive genomic profiling was conducted through NGS via FoundationOne (Foundation Medicine, Inc., Cambridge, MA) with reports generated from April 2013 to February 2016. The study was approved by the Institutional Review Board (IRB).

Next-generation sequencing
Comprehensive genomic analysis was conducted on tumor samples (formalin-fixed paraffin-embedded) retrieved from surgical resection, core needle biopsies, or excisional biopsies and delivered to Foundation Medicine, Inc. The NGS assay performed by FoundationOne has been previously described and validated [45]. The initial whole-genome shotgun library construction and hybridization-based capture of 4,557 exons from 287 cancer-related genes and 47 introns from 19 genes with frequent DNA rearrangements has since been expanded to identify genetic alterations across the coding regions of 315 cancer-related genes and introns from 28 genes commonly rearranged in solid cancers.

Study design
Retrospective analysis of genetic mutations, amplifications, or alterations present in our cohort of 138 patients with mCRC was performed through test results provided in an integrative report available via FoundationICE (Interactive Cancer Explorer). Patient demographics including age, sex, ethnicity, site of primary, stage at diagnosis, and number of previous treatments were obtained from chart abstraction of each patient's electronic medical record (EMR). Microsatellite instability classified as stable (MSS), low (MSI-L), or high (MSI-H) were abstracted from pathology reports and response to anti-EGFR therapy, when available, was described according to Response Evaluation Criteria in Solid Tumors (RECIST) criteria and obtained from medical records [46]. The total number of clinically significant alterations for each patient was determined by tallying the sum of alterations included in the panel of clinically significant variants provided by FoundationICE reports and arbitrarily categorized into 3 groups (<9, [9][10][11][12][13][14][15][16], and > 16 total number of alterations). We defined hypermutant tumors as those in the highest number of mutations group that were also found to have high TMB as validated by FoundationOne (high >23.1 mutations/MB, intermediate 3.2-23.1 mutations/MB, and low <3.2 mutations/MB) [34].

Statistical analyses
All statistical analyses performed were descriptive and no formal statistical hypotheses were assessed. The sample size was determined by the total number of mCRC patients with FoundationOne results available. All descriptive statistics were conducted in Excel with associated formulas and functions.