Clinical and molecular characterization of ovarian carcinoma displaying isolated lymph node relapse

BACKGROUND: Disease relapse is the primary cause of death from ovarian carcinoma. Isolated lymph node relapse is a rare pattern of ovarian carcinoma recurrence, with a reported median postrelapse survival of 2.5 to 4 years. To date, investigations have not compared isolated lymph node relapse ovarian carcinoma directly to a matched extranodal relapse cohort or performed molecular characterization of cases that subsequently experience isolated lymph node relapse. OBJECTIVE: Here we seek to compare the clinical outcome, tumor-inﬁltrating lymphocyte burden, and frequency of known prognostic genomic events in isolated lymph node relapse ovarian carcinoma vs extranodal relapse ovarian carcinoma. STUDY DESIGN: Forty-nine isolated lymph node relapse ovarian carcinoma patients were identiﬁed and matched to 49 extranodal relapse cases using the Edinburgh Ovarian Cancer Database, from which the clinical data for identiﬁed patients were retrieved. Matching criteria were disease stage, histologic subtype and grade, extent of residual disease following surgical debulking, and age at diagnosis. Clinicopathologic factors and survival data were compared between the isolated lymph node relapse and extranodal relapse cohorts. Genomic characterization of tumor material from diagnosis was performed using panel-based high-throughput sequencing and tumor-inﬁltrating T cell burden was assessed using immunohistochemistry for CD3 þ and CD8 þ cells. RESULTS: Isolated lymph node relapse cases demonstrated signiﬁcantly prolonged postrelapse survival and overall survival vs extranodal relapse upon multivariable analysis (HR multi ¼ 0.52 [0.33 e 0.84] and 0.51 [0.31 e 0.84]). Diagnostic specimens from high-grade serous ovarian carcinomas that subsequently displayed isolated lymph node relapse harbored signiﬁcantly greater CD3 þ and CD8 þ cell inﬁltration compared to extranodal relapse cases ( P ¼ .001 and P ¼ .009, Bonferroni-adjusted P ¼ .003 and P ¼ .019). Isolated lymph node relapse high-grade serous ovarian carcinoma cases did not show marked enrichment or depletion of cases with BRCA1/2 mutation or CCNE1 copy number gain when compared to their extranodal relapse counterparts (24.4% vs 19.4% and 18.2% vs 22.6%, P ¼ .865 and P ¼ .900). CONCLUSION: Isolated lymph node relapse ovarian carcinoma represents a distinct clinical entity with favorable outcome compared to extranodal relapse. There was no clear enrichment or depletion of BRCA1/ 2 mutation or CCNE1 gain in the isolated lymph node relapse ovarian carcinoma cohort compared with extranodal relapse cases, suggesting that these known prognostic genomically deﬁned subtypes of disease do not display markedly altered propensity for isolated lymph node relapse. Diagnostic tumor material from isolated lymph node relapse patients demonstrated greater CD3 þ and CD8 þ cell inﬁltration, indicating stronger tumor engagement by T cell populations, which may contribute to the more indolent disease course of isolated lymph node relapse.

BACKGROUND: Disease relapse is the primary cause of death from ovarian carcinoma. Isolated lymph node relapse is a rare pattern of ovarian carcinoma recurrence, with a reported median postrelapse survival of 2.5 to 4 years. To date, investigations have not compared isolated lymph node relapse ovarian carcinoma directly to a matched extranodal relapse cohort or performed molecular characterization of cases that subsequently experience isolated lymph node relapse. OBJECTIVE: Here we seek to compare the clinical outcome, tumorinfiltrating lymphocyte burden, and frequency of known prognostic genomic events in isolated lymph node relapse ovarian carcinoma vs extranodal relapse ovarian carcinoma. STUDY DESIGN: Forty-nine isolated lymph node relapse ovarian carcinoma patients were identified and matched to 49 extranodal relapse cases using the Edinburgh Ovarian Cancer Database, from which the clinical data for identified patients were retrieved. Matching criteria were disease stage, histologic subtype and grade, extent of residual disease following surgical debulking, and age at diagnosis. Clinicopathologic factors and survival data were compared between the isolated lymph node relapse and extranodal relapse cohorts. Genomic characterization of tumor material from diagnosis was performed using panel-based highthroughput sequencing and tumor-infiltrating T cell burden was assessed using immunohistochemistry for CD3þ and CD8þ cells. RESULTS: Isolated lymph node relapse cases demonstrated significantly prolonged postrelapse survival and overall survival vs extranodal relapse upon multivariable analysis (HR multi ¼ 0.52 [0.33e0.84] and 0.51 [0.31e0.84]). Diagnostic specimens from high-grade serous ovarian carcinomas that subsequently displayed isolated lymph node relapse harbored significantly greater CD3þ and CD8þ cell infiltration compared to extranodal relapse cases (P ¼ .001 and P ¼ .009, Bonferroni-adjusted P ¼ .003 and P ¼ .019). Isolated lymph node relapse high-grade serous ovarian carcinoma cases did not show marked enrichment or depletion of cases with BRCA1/2 mutation or CCNE1 copy number gain when compared to their extranodal relapse counterparts (24.4% vs 19.4% and 18.2% vs 22.6%, P ¼ .865 and P ¼ .900). CONCLUSION: Isolated lymph node relapse ovarian carcinoma represents a distinct clinical entity with favorable outcome compared to extranodal relapse. There was no clear enrichment or depletion of BRCA1/ 2 mutation or CCNE1 gain in the isolated lymph node relapse ovarian carcinoma cohort compared with extranodal relapse cases, suggesting that these known prognostic genomically defined subtypes of disease do not display markedly altered propensity for isolated lymph node relapse. Diagnostic tumor material from isolated lymph node relapse patients demonstrated greater CD3þ and CD8þ cell infiltration, indicating stronger tumor engagement by T cell populations, which may contribute to the more indolent disease course of isolated lymph node relapse.
O varian carcinoma (OC) is the most lethal gynecologic malignancy, accounting for over 180,000 deaths per year worldwide. 1 OC is now recognized to comprise 5 core histologic subtypes: high-grade serous (HGS), endometrioid, clear cell, low-grade serous, and mucinous OC-each displaying distinct molecular landscapes and clinical behavior. 2 Within HGS cases, homologous recombination deficiency by virtue of BRCA1 or BRCA2 mutation has been associated with favorable outcome, greater sensitivity to platinumbased chemotherapy, and marked benefit from poly (ADP-ribose) polymerase inhibitors. 3e6 Conversely, CCNE1 copy number gain has been associated with chemoresistance and poorer survival in this group. 3,7 Though patients in most OC casesparticularly HGS OCs-are typically sensitive to chemotherapy in the firstline setting, the majority of patients will experience disease relapse, which acquires resistance to chemotherapy. 8,9 The most common sites of recurrence are the pelvis and peritoneum. 10 Involvement of lymph nodes (LNs) at relapse is common; however, recurrence confined solely to LNs is a rare event, accounting for 5% of relapsed Original Research ajog.org so as to identify potential subgroups of disease with a propensity to experience this distinct pattern of disease relapse.
Here, we report clinical and molecular characterization of a matched ILNR and ENR cohort with contemporary pathology review to compare the clinical outcome and molecular landscape of ILNR and ENR OC.

Materials and Methods
Isolated lymph node relapse patient identification ILNR OC cases were identified from the Edinburgh Ovarian Cancer Database (Appendix: Supplementary Figure S1), wherein the clinical variables, treatment details, and follow-up data of OC patients treated within the Edinburgh Cancer Centre are collected prospectively as part of routine care. Potential ILNR cases were identified using the search terms "lymph node" or "groin node" as the dominant site of relapse, yielding 161 results. Nonepithelial tumors (n ¼ 1), tumors of borderline malignancy (n ¼ 1), and primary LN serous carcinomas (n ¼ 2) were excluded. Patients with concurrent extranodal disease (n ¼ 50), lack of cross-sectional imaging to confirm sole ILNR (n ¼ 13), or coexistence of other malignancies leading to uncertain origin of LN disease (n ¼ 2) were excluded. Patients with residual disease (RD) after completion of first-line treatment (n ¼ 19) or insufficient clinical data for eligibility assessment (n ¼ 24) were also excluded, leaving 49 ILNR cases.
Matching of isolated lymph node relapse to extranodal relapse ILNR cases were electronically matched to ENR cases with complete response to first-line therapy using the Edinburgh Ovarian Cancer Database. Matching criteria were as follows: (1) diagnostic histologic subtype and grade, (2) stage at diagnosis, (3) extent of RD following debulking surgery, and (4) closest age at diagnosis following matching of criteria 1e3. Criteria were relaxed to facilitate matching of all ILNR cases as detailed in Supplementary Table S1 (Appendix).

Nucleic acid isolation
Up to 10 10-mm formalin-fixed paraffinembedded sections, macrodissected using marked hematoxylineeosin-stained slides as a guide to enrich for tumor purity (Appendix :  Supplementary  Table S2), were used for DNA extraction. DNA was extracted using the QIAamp DNA FFPE Tissue Kit and Deparaffinization Solution (Qiagen, Venlo, the Netherlands).
Panel-based sequencing of BRCA and non-BRCA homologous recombination deficiency genes High-throughput sequencing was performed using an 83-gene custom Integrated DNA Technologies gene capture panel with unique molecular indices, as described in the Appendix. Gene targets, centered around the homologous recombination DNA repair pathway, are detailed in Supplementary Table S3 (Appendix). The median per-sample mean target coverage achieved was 386X.

Assessment of CCNE1 copy number
Copy number variants in CCNE1 were characterized by TaqMan Genotyping

AJOG at a Glance
Why was this study conducted? A number of investigators have reported a relatively indolent disease course in ovarian carcinoma patients experiencing isolated lymph node relapse. However, none have systematically compared these to extranodal relapse or performed molecular characterization of patients who go on to experience this distinct pattern of recurrence.

Key findings
Isolated lymph node relapse patients demonstrated significantly prolonged overall and postrelapse survival compared with extranodal relapse cases. Isolated lymph node relapse cases demonstrated greater tumor-infiltrating lymphocyte burden at diagnosis, but did not demonstrate significant enrichment or depletion of BRCA1/2 mutation or gain of CCNE1, both known to be prognostic in ovarian carcinoma.

What does this add to what is known?
This is the first report demonstrating significantly improved clinical outcome in isolated lymph node relapse ovarian carcinoma when compared directly with extranodal relapse, and represents the first study to perform molecular characterization of patients who go on to experience isolated lymph node relapse.

Statistical analyses
Statistical analyses were performed using R version 3.5.1 (R Foundation, Vienna, Austria). Disease-free interval (DFI) was calculated as time from end of first-line chemotherapy to disease recurrence. Comparisons of OS and PRS were conducted using Cox proportional hazards regression models within the Survival R package 21 and presented as hazard ratios (HRs) alongside their 95% confidence intervals (CIs). Frequency comparisons were made using the c 2 test and Fisher exact test as appropriate. Comparisons of TIL density were made using the ManneWhitney U test. Analyses were adjusted for multiplicity of testing using the Bonferroni correction, where specified.

Cohort characteristics
Demographics of the ILNR and ENR cohorts are summarized in Table 2. There was no significant difference in age at diagnosis, RD following primary surgical debulking, histology or grade of disease at diagnosis, or disease stage at diagnosis between the ILNR and ENR groups. These data indicate good fidelity of the ILNReENR matching process. Patterns of ILNR are described in Table 3.
Clinical outcome in isolated lymph node relapse vs extranodal relapse  Longer disease-free interval is associated with prolonged postrelapse survival in isolated lymph node relapse ovarian carcinoma The importance of DFI on clinical outcome in ILNR OC remains controversial, with some authors reporting no association between DFI length and PRS or OS in this setting 11,16,18 and others reporting significant associations. 12,14,15 Within the ILNR cohort, DFI !12 months was associated with markedly prolonged PRS when accounting for patient age (HR multi

Impact of isolated lymph node relapse pattern on outcome
There was no clear differential PRS between multiregion ILNR and singleregion ILNR (  Original Research GYNECOLOGY ajog.org between the ILNR and ENR cohorts (Figure 2, A). The CD3þ and CD8þ TIL burden was greater in diagnostic tumor specimens from HGS OC patients who went on to experience ILNR when compared with their ENR counterparts (median CD3þ cell density 1.94% vs 1.13%, P ¼ .001 and median CD8þ cell density 0.90% vs 0.45%, P ¼ .009; Bonferroniadjusted P ¼ .003 and P ¼ .019) (Figure 2, B).

Principal findings
The principal findings of this study are as follows: (1) ILNR represents a distinct pattern of OC relapse with prolonged survival vs ENR cases; (2) longer DFI prior to ILNR is associated with prolonged PRS in ILNR; (3) ILNR OC do not demonstrate significantly differential composition of known genomic subtypes associated with prognosis, namely BRCA1/2 mutation or gain of CCNE1; (4) cases that go on to experience ILNR demonstrate greater TIL burden at diagnosis compared with ENR cases.

Study strengths and limitations
A key strength of this study is the direct comparison of ILNR OC with matched ENR cases: a number of studies have reported ILNR as a distinct pattern of OC relapse with a relatively indolent disease course but have not systematically compared ILNR cases directly to a matched ENR cohort. 11e18 Moreover, these studies did not perform pathology review of identified cases, precluding the ability to characterize ILNR outcome in the context of contemporary OC histotypes, which are now known to display markedly differential clinical outcome. 22 Critically, we characterize ILNR OC following contemporary histologic subtyping to facilitate investigation of ILNR in a histotypespecific manner.
The majority of previous studies investigating ILNR have identified fewer than 20 OC cases of serous histology that go on to experience this rare relapse pattern; moreover, previous reports have not performed molecular characterization of OC cases that demonstrate ILNR. 11e18 We identified 49 ILNR OC patients treated within the Edinburgh Cancer Centre, including 34 cases reviewed as HGS OC. This study represents the largest ILNR OC series from a single center and the only report investigating the molecular landscape of ILNR OC to date.
Though this study does represent one of the largest reported ILNR OC cohorts, case numbers were still restricted owing to the rarity of ILNR OC. In particular, power to detect differential outcome between distinct patterns of ILNR was limited, and we could not perform meaningful analysis comparing rates of rare genomic events present in both ILNR and ENR cohorts, including mutational events in RB1, NF1, and PTEN, as well as gene-specific analysis of BRCA1 and BRCA2. Other limitations of this study include heterogeneous treatment of OC patients across the time period in which these cases were diagnosed, though diagnosis periods were comparable between the ILNR and ENR cohorts ( Table 2).

Clinical outcome in isolated lymph node relapse ovarian carcinoma
The median PRS and OS of ILNR cases was approximately 2.7 and 6 years, consistent with previous reports of ILNR OC. 11e18 ILNR cases displayed significantly prolonged OS and PRS compared to their ENR counterparts upon multivariable analysis (HR multi ¼ 0.51 and 0.52 for OS and PRS). Critically, this difference was maintained in a histotype-specific analysis of HGS cases, which account for the majority of OCs. To our knowledge, this is the first report directly demonstrating a significant difference in outcome between ILNR and ENR OC.
Only half of the reports investigating the impact of DFI length on ILNR outcome to date have identified associations with OS or PRS. 12,14,15 Here, we demonstrate that DFI !12 months is associated with a substantial PRS benefit (median PRS approximately 3.9 vs 1.7 years), largely reflective of established associations in unselected OC cases. 23 Although this contradicts reports from some investigators, 11,16,18 2 of these studies reported specifically in the context of ILNR undergoing  16,18 and the other compared cases using a cut-off DFI of 24 months, rather than 12 months as described here, 11 potentially explaining this discrepancy. Notably, the intervals considered in our study are akin to those used clinically to define platinum sensitivity in unselected relapsed OC. 23 We show no significant difference in clinical outcome between patients with ILNR at multiple sites vs those with single-site ILNR, or between distinct patterns of ILNR. Although univariable analysis suggested that supraclavicular LN involvement may confer inferior PRS, this trend was not apparent when accounting for DFI and patient age, suggesting that this is not a genuine phenomenon of supraclavicular ILNR. Notably, the number of patients with supraclavicular LN involvement was low (n ¼ 6). Together, these data support the consideration of ILNR OC as a single disease entity, regardless of the number and location of involved sites.
The genomic landscape of isolated lymph node relapse ovarian carcinoma Until now, the molecular landscape of ILNR has been completely uncharacterized. It has therefore been unclear as to whether OC cases that go on to experience ILNR demonstrate enrichment of tumors belonging to known favorable genomic subgroups. Within unselected cohorts of HGS OC, inactivation of BRCA1 or BRCA2 has been associated with favorable outcome, 3,4 while copy number gain of CCNE1 has been associated with poor survival and chemoresistance. 3,7 Genomic characterization of this cohort did not identify significant depletion or enrichment of these molecular events in ILNR HGS OC cases versus their ENR counterparts. These data suggest that the survival benefit of ILNR OC is not underpinned by large-scale enrichment for BRCA1/2mutant cases with favorable prognosis or absence of CCNE1-gained cases that have poorer prognosis, and suggest that these genomic subgroups do not display markedly differential propensity for ILNR.

ajog.org
Greater tumor-infiltrating lymphocyte burden at diagnosis in patients who subsequently experience isolated lymph node relapse Intriguingly, assessment of the CD3þ and CD8þ cell burden in ILNR and ENR tumor material-reflective of whole T cell and cytotoxic T cell populationsuncovered significantly greater TIL burden in diagnostic tissue from patients who subsequently experienced ILNR (2fold enrichment for CD8þ cells, approximately 1.7-fold enrichment for CD3þ cells). These data suggest that active engagement of the immune system at diagnosis impacts upon the nature of disease at relapse, and that immune-mediated control of cancer cells may contribute to the indolent disease course of ILNR OC. Indeed, these data may well be of interest in relation to the use of immune-directed therapies in cancer treatment. 24,25 However, though many ILNR cases displayed high TIL burden, some cases demonstrated relatively low levels of TILs, alluding to mechanisms beyond effective T-cell engagement at diagnosis underpinning some ILNR cases.

Conclusion
Collectively, the data presented heresupported by previous descriptions of apparent ILNR in the literaturedemonstrate that ILNR represents a distinct pattern of OC with favorable clinical outcome when compared with ENR. Patients that go on to experience ILNR harbor greater TIL burden at diagnosis, but they do not show marked enrichment or depletion of known genomic subgroups associated with differential outcome. n

SUPPLEMENTARY TABLE S4
Multivariable analysis for overall survival in isolated lymph node relapse vs extranodal relapse in ovarian carcinoma