Intratumoral heterogeneity in a minority of ovarian low-grade serous carcinomas

In comparison to the more commonly occurring high-grade serous carcinomas (HGSC), ovarian low-grade serous carcinomas (LGSC) are characterized by a younger age at onset, lower mitotic rate and longer median overall survival [1‐6]. Whereas the vast majority (80%) of patients with HGSC are responsive to platinum-based chemotherapy, patients with LGSC are highly resistant to treatment in the neoadjuvant, adjuvant and recurrent setting, with response rates of 4-5% [1, 7, 8]. Women diagnosed with LGSC typically experience multiple recurrences over a protracted clinical course before ultimately dying of their disease, with an associated 10-year survival rate of <50% [2]. This suggests that despite having a less aggressive clinical course, women with LGSC have a poor long-term prognosis similar to HGSC patients; this is highlighted by a recent study reporting a similar hazard ratio for death in LGSC and HGSC patients with measurable residual disease after adjusting for additional variables [9].

In an effort to identify potential molecular targets, limited mutational studies in primary or recurrent LGSC samples have revealed an overall low mutation frequency, with exome sequencing by Jones et al. showing an average of 10 validated somatic mutations (or 7.5 somatic non-synonymous or splice site mutations) per tumor [10]. The mitogen-activated kinase (MAPK) pathway is most frequently mutated [11], with 19-35% of cases containing a KRAS mutation and 2-33% containing a BRAF mutation [3, 10, 12‐14]. KRAS and BRAF mutations are also frequently detected in serous borderline tumors (SBT), the histologic precursor to invasive LGSC [5, 6, 11, 15‐17].

The prevalence of KRAS/BRAF mutations in LGSC has resulted in clinical trials of inhibitors of MAP kinase kinase (MEK1/2), which lies immediately downstream of BRAF and upstream of ERK1/2 in the MAPK pathway [18, 19]. Previous studies have reported profound growth inhibition and apoptosis in ovarian cancer cells with mutated but not wildtype KRAS or BRAF upon treatment with CI-1040 [20] in tissue culture and xenograft studies [19, 21], suggesting that mutation status predicts sensitivity to MEK inhibition. A recent phase II study of selumetinib, another small molecular inhibitor of MEK1/2, in women with recurrent ovarian/peritoneal LGSC has shown an objective 15% response rate despite heavy pre-treatment; however patient response does not appear to be correlated with KRAS/BRAF mutation status [18]. The mutation status of the patients in this trial was based solely on a single sample of LGSC; most were obtained from the primary tumor and a small percentage were obtained from the recurrent tumor. In this study we aimed to assess the stability of targetable mutations over space and/or time by targeted sequence analysis of one or more tumor samples from both the primary and recurrence, to inform future clinical trial design. Herein we report our findings of mutational stability in the majority of cases, as well as remarkable instability in one case of ovarian LGSC, in presumed drivers of disease KRAS and BRAF. If validated in more cases this could impact clinical trial design for this patient population in the future.

Among the 11 cases of LGSC sequenced in our study, only 7 confirmed somatic mutations were identified in 5 cases from a targeted hotspot panel of 46 cancer-associated genes. This low mutation rate is consistent with the detection of only 10 mutations per tumor by exome sequencing by Jones et al.[10], and further suggests that few mutational events are required to achieve malignancy. The frequency of mutations in LGSC is much lower than in other subtypes of ovarian carcinoma such as HGSC (n = 61 mutations/tumor by exome sequencing) [29] and clear cell carcinoma (n = 34 mutations/tumor by exome sequencing) [30]. This likely suggests that: [1] there is limited replication of precursor cells prior to initiation of tumorigenesis, [2] there are few bottlenecks once initiation occurs, and [3] the ratio of driver to passenger mutations should be higher than in other tumor types [10]. Consequently, targeted agents would likely be particularly effective in women with LGSC if key mutations are shown to be stable.

The most commonly reported drivers in LGSC are KRAS and BRAF. We detected a KRAS mutation in three patients (including two stage IIIC and one stage IV) and a BRAF mutation in two patients (including one stage IIB and one stage IV). Previous studies have reported conflicting findings with respect to mutation of KRAS/BRAF and disease stage, with the Jones study [10] detecting KRAS or BRAF mutations in 4/13 (31%) and 3/13 (23%) of stage III LGSC patients respectively. Additional studies report BRAF mutations in only 3% [12] and 5% [13] of advanced stage LGSC. Grisham and Wong both reported that women with mutations in KRAS and/or BRAF[12, 13] experience a more favorable outcome than women without these mutations. This positive prognostic effect appears to be dominated by BRAF V600E mutations, with a lower incidence of stage III-IV disease, enrichment for SBT rather than invasive LGSC and reduced requirement for systemic treatment among women with this mutation [12, 13]. Possible explanations include reports that SBTs from women with BRAF mutations over-express genes with cell growth inhibitory effects [12] or that activating BRAF mutations induce cellular senescence and prevent progression to LGSC [12, 31‐33]. In our study we observed a trend for increased mean overall survival in study patients with a MAPK pathway mutation (KRAS, BRAF, NRAS) compared to patients with wildtype status (92 months vs. 60 months respectively; p = 0.23); however, this difference in outcome was largely influenced by the two cases originally presenting as a SBT (143 and 183 months) and disappeared when these cases were excluded from the analysis.

SMAD4 mutational status in case LGSC-12 was also consistent among 6 tumor samples from 4 different sites in the primary and recurrence, and despite multiple treatment cycles. Although found to be unstable in another case, all samples from LGSC-12 also contained a BRAF mutation at a similar allelic fraction. The observed SMAD4 mutation (chr18:48,591,918C > G, R361G) is at a highly conserved genomic position among placental mammals, and is situated within the C-terminus MH2 domain of the SMAD4 protein. This domain mediates protein-protein interactions and provides functional specificity and selectivity. It was previously reported as the most frequent target of SMAD4 missense mutations in human tumors, with a mutational hotspot corresponding to codons 330–370 [34]. Lassus et al. reported allelic loss at one or more loci at 18q12.3-q23 in 59% of ovarian serous carcinomas (or 7.1% of grade 1 tumors), with lost or weak expression of SMAD4 protein in a subset of these tumors [35]. Mutations in SMAD4 have been reported to frequently co-exist with KRAS mutations in colorectal cancer [36], and studies in pancreatic cancer suggest that wildtype SMAD4 blocks progression of KRAS G12D-initiated tumors [37]. In addition, mutation of KRAS, NRAS and BRAF[38‐46], and loss of functional SMAD4 [47], have all been reported to predict resistance to anti-EGFR therapy. Unfortunately we were unable to assess the impact of the SMAD4 R361G mutation on protein expression by IHC in our samples, therefore we cannot comment on the utility of SMAD4 mutation status as a predictive marker in women with LGSC without further study.

In contrast to NRAS and SMAD4, mutations in KRAS and BRAF were not stable in one patient (LGSC-10) in our study, despite traditionally being thought of as 'drivers’ of tumorigenesis. This is akin to our recent observation that mutations in other key 'drivers’ PIK3CA and CTNNB1 are only present in a subset of ovarian HGSC samples from the same patient [48]. These examples clearly defy the concept of oncogene addiction, which posits that the growth and survival of a tumor is dependent on a single dominant oncogene [49, 50]. Our findings in LGSC-10 suggest that even at the time of primary diagnosis three distinct tumors/clones were present (i.e. KRAS-mutation positive, BRAF-mutation positive and KRAS/BRAF-mutation negative). As neither KRAS nor BRAF were mutated in any of the recurrent samples, a different, as yet unidentified, dominant gene or pathway in the KRAS/BRAF-negative population was likely driving disease recurrence. One possibility is that we have missed a mutation in gene/s either directly or indirectly involved in the MAPK pathway that is not included on the targeted panel used to screen our samples. The KRAS and BRAF mutations were detected at an allelic fraction of 22-31% in the right ovary and 3-7% in the left ovary respectively, hence the clonal population containing an undetected driver mutation could have already been present in some or all of the tumor samples at primary debulking; expansion/recurrence of this population could then explain the absence of mutant KRAS/BRAF in the recurrent setting. In addition, mutations such as those in KRAS and BRAF that occur early in the development of SBT/LGSC [17] may not be required and/or advantageous for tumor maintenance once additional alterations are acquired. This phenomenon has previously been described in HGSC, in which secondary mutations in BRCA1/2 restore protein function and result in acquired resistance to treatment [51]; however, reversion of both a KRAS and BRAF mutation in the current scenario seems highly unlikely.

Of potential interest, LGSC-10 was the only study case diagnosed with stage IV disease and the only patient treated with radiation after primary diagnosis. While the presence of mutational instability in the primary setting (prior to treatment) argues against a direct impact of radiation, the possibility of instability exclusively in stage IV LGSC is an intriguing one that requires more study. To date, limited studies have reported on either temporal or spatial instability of BRAF/KRAS mutations in SBT and LGSC. Instability in KRAS mutation status was recently described in a subset of matched SBT-LGSC pairs (2/5 cases discordant) [52] and matched SBT-peritoneal implant pairs (3/37 discordant for KRAS, while 14/14 concordant for BRAF) [53]. A recent study by Heublein et al.[54] also noted instability in KRAS and BRAF in 2/5 cases of bilateral SBT. In one case, a KRAS G12V mutation was detected in one ovary and a BRAF V600E mutation was detected in the contralateral ovary, while the other case contained a KRAS G12V and BRAF V600E mutation in one ovary and only a KRAS G12V mutation in the other ovary. This is consistent with our finding of spatial heterogeneity in the primary setting in LGSC-10. Unfortunately, a detailed breakdown of disease stage in cases with discordant vs. concordant sample pairs was not provided in any of these studies. Instability in KRAS has also been described for metastatic colorectal cancer [55, 56]. Bossard et al.[55] reported several patterns of heterogeneity in KRAS mutation status in 22% of 18 colorectal carcinomas studied. This included exclusive presence in the primary tumor or metastatic site, presence in some metastases but not others, varied status among different samplings from the same metastatic site, and presence in the recurrent but not primary setting. Similarly, Otsuka et al.[56] reported the presence of a KRAS mutation in metastatic sites but not the primary colorectal tumor in 1 of 9 patients studied; BRAF mutation status was concordant in all cases, in contrast to what we observed.

Our finding that mutations in genes such as KRAS or BRAF are not necessarily stable features could provide an alternative explanation, in some patients, for the lack of correlation between response to selumetinib and KRAS/BRAF mutation status observed by Farley et al.[18]. Targeted sequencing (i.e. codon 599 of BRAF and codons 12 and 13 of KRAS) using a single representative tumor sample from 34/52 (65%) patients revealed a BRAF and KRAS mutation in 2 (6%) and 14 (41%) cases respectively. A similar proportion of mutation positive vs. negative cases responded to selumetinib treatment, leading the authors to postulate that its activity may not depend on BRAF/KRAS mutational activation. Tissue used for mutational analysis was obtained from the primary tumor in 82% of sequenced cases, metastatic tumor in 6% and recurrent or persistent tumor in 12% of cases. It is therefore possible that targetable mutations detected in the primary tumor were not present in the metastatic or recurrent tumor, or vice versa, leading to altered treatment response. It is also possible that some of these patients had undetected mutations in NRAS, a stable feature in our study, which also predicts response to MEK inhibitors.

It is important to recognize the limitations of our study, most notably small sample size and use of a hotspot targeted gene panel. Firstly, the small number of cases used in this study (despite being a collaboration between three institutions) is illustrative of the challenge in identifying primary-recurrent pairs for a rare tumor type such as LGSC. Confirmation of our findings in a larger cohort of LGSC will therefore require participation by multiple institutions or establishment of a worldwide registry. Secondly, by limiting the sequencing discovery phase to a panel of hotspot mutations in 46 genes, it is highly likely that we have missed additional case-specific mutations in our study population. However, a closer look at the mutations discovered by Jones et al. through exome sequencing [10] revealed that only KRAS and BRAF were recurrently mutated in LGSC. This suggests that it is also unlikely that we have missed additional recurrent drivers of disease, although patient-specific drivers outside the normal patterns of LGSC may exist. Thirdly, we have not investigated potential alternative drivers of disease that may be important in cases without identified somatic mutations, such as copy number alterations, epigenetic changes or microRNAs. Singer et al.[57] previously reported a progressive increase in copy number alterations from SBT through to LGSC, most notably allelic imbalance of chromosomes 1p, 5q, 8p, 18q and 22q. This was confirmed by Kuo et al.[58] who reported an increased chromosomal instability index in LGSC relative to SBT, suggesting that amplifications, deletions and aneuploidy play a role in the malignant transformation of SBT. Hemizygous deletion of chromosome 1p36 was especially enriched in LGSC samples; this region contains the microRNA miR-34a, which was found to have an anti-proliferative and pro-apoptotic effect in an LGSC cell line [58]. Finally, several groups have reported on differential methylation patterns in SBT and LGSC [59‐61], suggesting that methylation-induced transcriptional silencing of tumor suppressor genes may play an undefined role in malignant transformation and progression and response to systemic or targeted therapy.

The extent of intratumoral heterogeneity in kidney, breast, leukemia and ovarian cancers has recently been described [48, 62‐64]. Most papers have focused on high-grade cancers with many somatic mutations, and most of the mutations described have no immediate clinical relevance. Herein we show that, in a cancer type known to have a sparse mutational landscape [10], heterogeneity in targetable mutations can be observed. While the vast majority of evaluable cases contained mutations that were detected in all samples, one case showed remarkable instability in hotspot mutations of presumed drivers of disease, despite not receiving treatment that could have driven the specific evolution of (KRAS/BRAF) mutant clones. In addition, as we looked within a limited mutational space, the possibility remains that more underlying heterogeneity may be revealed in more cases with further study. Investigation of additional cases is required to confirm whether a consistent minority of LGSC cases show clinically relevant mutational heterogeneity; this would necessitate a change in clinical trial design with contemporary samplings of a cancer required to guide treatment decisions. Alternatively, if not found to be a general phenomenon upon further study, confirmation of mutational status in a single sample would be sufficient.

Alicia A. Tone, PhD

Scientific Associate II

Melissa K. McConechy, BSc

Doctoral Candidate

David Huntsman, MD, FRCPC, FCCMG

Dr. Chew Wei Memorial Professor of Gynaecologic Oncology

UBC Professor, Departments of Pathology & Lab Medicine and Obstetrics & Gynaecology UBC Director of OvCaRe, Vancouver General Hospital, BC Cancer Agency

UBC Medical Director, Centre for Translational and Applied Genomics, PHSA Laboratories