Clinicopathologic Features, Molecular Landscape, and Prognostic Implications of Ovarian Low-Grade Serous Tumors with Histologic Transformation

Abstract

Purpose:

To characterize the clinicopathologic features, molecular genetic landscape, and clinical behavior of ovarian low-grade serous tumors with histologic transformation (LGS-HT) to indeterminate/high-grade carcinoma.

Experimental Design:

LGS-HTs were retrospectively identified from an institutional cohort of ovarian cancer patients and underwent central pathology re-review. Data on clinicopathologic characteristics, including age, stage, surgical outcomes, systemic treatments, and overall survival (OS) were collected. Immunohistochemical profiling and next-generation sequencing were performed. OS comparisons were performed with our institutional cohorts of ovarian low-grade serous carcinoma (LGSC, n=109) and high-grade serous carcinoma (HGSC, n=1672).

Results:

From 4371 ovarian serous cancers, 40 (0.9%) LGS-HTs were identified: 30 with synchronous low-grade and higher-grade tumor components at initial diagnosis (LGS-HT-Sync) and 10 with an ovarian low-grade serous neoplasm that recurred as a higher-grade carcinoma (LGS-HT-Metac). The most common somatic driver mutations included TP53 (38.5%), KRAS (21.8%), NF1 (15.6%), BRAF (15.6%) and NRAS (12.5%), with co-existing TP53 and RAS/RAF mutations in 18.8% of cases. Alterations of DNA damage response genes (BRCA2, PALB2, CHEK2, ATM, NBN, RECQL4) were identified in LGS-HTs lacking TP53 genetic alterations. LGS-HT-Sync was associated with poorer OS (median 59.7 months) compared to LGSC (median 105.4 months; p=0.026), and was similar to HGSC (median 48.8 months, p=0.61). Severe nuclear atypia and absence of RAS/RAF-driver mutations were significant adverse prognostic factors.

Conclusions:

LGS-HT exhibit both low-grade and high-grade morphologic and molecular features, representing an exception to the dualistic classification of ovarian serous neoplasms. The presence of a definitive high-grade carcinoma component in a low-grade serous tumor portends aggressive clinical behavior.

INTRODUCTION

Ovarian serous carcinomas are subclassified into low-grade and high-grade subtypes, associated with distinct morphologic and genetic features, pathogenesis, and clinical behavior(1–3). High-grade serous carcinoma (HGSC) is the most common subtype of ovarian cancer, almost all of which carry a TP53 genetic alteration and is associated with chromosomal instability(4). The majority of HGSCs are thought to be derived from a fallopian tube precursor lesion, termed serous tubal intraepithelial carcinoma(5). While clinically aggressive and usually fatal, HGSC is initially sensitive to platinum-based chemotherapy(3). Furthermore, PARP inhibitor therapy is an effective maintenance option for tumors with homologous recombination deficiency (HRD), especially those with BRCA1/2 mutations. In contrast, low-grade serous neoplasms, encompassing low-grade serous carcinoma (LGSC) and its direct precursor, serous borderline tumor (SBT), are TP53-wildtype and have few copy number alterations and mutations(6–8). LGSCs are, in general, poorly responsive to platinum-based chemotherapy and other cytotoxic regimens(9). Approximately 60% of LGSCs harbor genetic alterations causing constitutive MAPK pathway activation, most commonly hotspot mutations affecting KRAS (G12), NRAS (Q61), and BRAF (V600E). MAPK-altered LGSCs are associated with better prognosis as well as response to emerging MEK inhibitor targeted therapies(6,10).

Ovarian serous tumors are diagnosed based on their morphologic features, using a two-tiered system, based on the degree of nuclear atypia and mitotic activity (11,12). While most serous carcinomas can be classified as either LGSC or HGSC, there are rare tumors, described in case reports and small case series, comprised of distinct low-grade and high-grade components which either co-exist synchronously, or as SBT/LGSC recurring in the form of HGSC or poorly-differentiated carcinoma(13–20). In addition, there is a separate subgroup of morphologically ambiguous serous carcinomas with moderate nuclear atypia, which could exhibit focal nuclear pleomorphism, and high mitotic activity(20,21). While those harboring TP53 mutations are likely misclassified HGSCs, the pathogenesis and clinical behavior of those without TP53 alterations have not been well characterized. Often, these “indeterminate”-grade serous carcinomas contain areas of low-grade morphology, suggesting that they may represent a transition state between LGSC and HGSC.

In the present study, “low-grade serous tumor with histologic transformation” (LGS-HT) refers to unconventional serous tumors with distinct low-grade and higher-grade components, co-existing at the time of initial diagnosis (i.e. synchronous, LGS-HT-S) or primary low-grade serous tumors associated with higher-grade malignant recurrence (i.e. metachronous, LGS-HT-M). Herein, we describe the histologic and immunophenotypic features, molecular genetic landscape, and clinical behavior of LGS-HT, providing support for a prognostically relevant and distinct pathologic entity exhibiting features of both low-grade and high-grade serous neoplasia.

MATERIALS AND METHODS

Case Selection and Central Pathology Review

This retrospective study was conducted in accordance with guidelines set forth by the Declaration of Helsinki and was approved by the Institutional Review Board at MSK; written informed consent for molecular profiling was obtained from all patients (IRB#12–245). LGS-HTs were identified by employing two separate approaches. First, a database search of pathology reports issued at Memorial Sloan Kettering Cancer Center (MSKCC) between Jan 2001 to July 2021 was conducted using the terms “ovary”, “low-grade”, “high-grade” and “serous” in the diagnosis and comment sections (Figure 1A). Second, within our institutional cohort of HGSCs subjected to clinical targeted next-generation sequencing (NGS), those lacking TP53 mutations and/or harboring characteristic genetic alterations of LGSC, namely KRAS, NRAS, or BRAF^V600E hotspot mutations, were identified. All available histologic slides or digitally scanned images were retrieved for central pathology review; cases lacking available diagnostic material were excluded from further analysis.

Figure 1:

Flow diagrams depicting (A) selection of cases included in the study. LGSC – low-grade serous carcinoma; HGSC – high-grade serous carcinoma; LGS-HT – low-grade serous tumors with histologic transformation (Sync – synchronous, Metac – metachronous); NGS – next-generation sequencing; (B) diagnostic algorithm for assigning histologic grade in ovarian serous neoplasms, as applied to homogenous tumors and individual tumor components within LGS-HTs.

Central review was performed by 3 gynecologic pathologists (M.H.C., E-Y.K, S.C.). Tumor grade was determined primarily based on assessment of nuclear features, with mitotic activity as a secondary criterion, as recommended by the WHO Classification of Female Genital Tumors(12), and further aided by p53 immunohistochemistry (IHC) for morphologically equivocal cases (Figure 1B)(22). Nuclei with mild atypia (Grade 1) were uniform (<3:1 size variation), round or oval, and occasionally had a central nucleolus. Moderate nuclear atypia (Grade 2) was characterized by enlarged, crowded nuclei with irregular borders, and equivocal with respect to 3:1 size variation, while severe nuclear atypia (Grade 3) denoted marked nuclear pleomorphism (>3:1 size variation) and was often associated with hyperchromatic chromatin and prominent large nucleoli. Low-grade tumors had mild or mild-to-moderate nuclear atypia and low mitotic activity (<5 mitoses per 10 HPF), while high-grade tumors showed severe nuclear atypia with high mitotic activity (≥12 mitoses per 10 HPF). Tumors with moderate atypia and high mitotic activity were morphologically ambiguous and classified as high-grade, if they exhibited an aberrant p53 immunohistochemical staining pattern (diffuse overexpression, complete absence of expression or cytoplasmic expression), and indeterminate-grade if they showed a wildtype p53 immunohistochemical expression pattern(22).

Only cases that were confirmed on central pathology review to exhibit both low-grade serous tumor and a higher grade (indeterminate or high-grade) carcinoma component either at the time of initial diagnosis (LGS-HT-Sync) or alternatively, low-grade serous neoplasm with higher grade recurrence (LGS-HT-Metac) were included in the study. Since “high-grade transformation” was often mentioned in the original clinical pathology report for cases that on central pathology review showed at most, moderate nuclear atypia and best classified as “indeterminate grade,” the decision was made to include these cases in the cohort so that the prognostic relevance of the degree of the nuclear atypia can be determined. However, for all cases, distinct low-grade serous and higher-grade tumor populations had to be identified to qualify as LGS-HT. A minimum size of 5 mm in greatest dimension (in aggregate) was necessary for the higher-grade component.

Histologic subtype of the higher-grade component was determined by morphologic assessment with supporting PAX8 and WT1 IHC. Since PAX8 and WT1 expression have been reported in most, but not all HGSCs, classification of a tumor as HGSC did not require these markers to be positive, provided that the morphologic features were sufficient for this diagnosis(12).

From our clinically annotated institutional database of ovarian cancer patients who had their primary surgical resection and post-operative treatments at MSKCC, those with LGSC or HGSC, diagnosed during the same time period (between January 2001 to July 2021), were included for comparison (Figure 1A). Clinical information, including age, stage at diagnosis (FIGO 2014), neoadjuvant chemotherapy, residual disease after surgery (complete gross resection, optimal debulking or suboptimal debulking), and overall survival (OS) were collected from the electronic medical record.

Immunohistochemistry

IHC was performed on all cases with available tissue using the following primary antibodies: PAX8 (clone BC12; Cell Signaling, 1:50), WT1 (clone WT49; Leica, pre-diluted), ER (SP1; Ventana, pre-diluted), PR (1E2; Ventana, pre-diluted), and p53 (clone DO-7; Dako, pre-diluted). All immunohistochemical stains were performed on the BOND RX platform (Leica), using the BOND Epitope Retrieval Solution 2 (Leica) and BOND Polymer Detection DAB kit (Leica).

Next-generation sequencing

NGS data was available for 32 LGS-HTs, which included cases profiled by MSK-Integrated Mutation Profiling of Actionable Cancer Targets (MSK-IMPACT)(23) assay (n=29) or extracted from clinical NGS results provided by Foundation Medicine (n=3). MSK-IMPACT is our institutional clinical NGS assay which involves paired tumor-normal targeted massively parallel sequencing analysis of 341–505 cancer-related genes. Whole exome sequencing was performed on 6 cases, and analyzed as previously described(24), and includes 3 cases previously reported in detail in another study(25): Case IDs 5, 6, and 7 from Chui et al.(25) correspond to Cases 28, 5, and 9 in the present study. For 4 cases in our cohort, low-grade and higher-grade tumor components were isolated separately by microdissection and extracted DNA from the paired samples were profiled by MSK-IMPACT to compare their somatic mutational profiles.

Driver genetic alterations were identified by considering only somatic mutations annotated as “oncogenic” or “likely oncogenic” by OncoKB(26). Cancer cell fractions of somatic mutations identified were determined using ABSOLUTE (v1.0.6), with mutations being classified as clonal if the probability of being clonal was >50% or if the lower bound of the 95% CI of the cancer cell fraction was >90%. Fraction of genome altered (FGA), and tumor mutation burden were determined from MSK-IMPACT sequencing data from cases with adequate tumor content (>25% as estimated by histologic review) for LGS-HT, as well as for comparator LGSC and HGSC cohorts, described in the Case Selection section, with available MSK-IMPACT data.

Statistical analyses

Data distributions between tumor subtypes were analyzed using Wilcoxon rank sum test for continuous variables and Fisher’s exact test for categorical variables, respectively. OS for LGS-HT-S was compared with LGSC and HGSC. OS was defined as the time from initial pathologic diagnosis to death or last follow-up. Survivors were censored at the last follow-up date. Left truncation methodology(27) as applied to address selection bias, using date of the patient’s first appointment at MSK. Survival curves were generated using the Kaplan-Meier survival method, and hazard ratios with 95% confidence intervals (CI) and p-values were obtained by the Cox proportional hazard model, accounting for left truncation. Variables significant on univariate analysis were entered into multivariate modeling. All tests were two-sided and a p-value <0.05 was considered statistically significant. Due to small sample size and the exploratory nature of the analyses, we did not adjust for multiple comparisons. All analyses were performed using R version 4.1.2 (https://www.R-project.org/).

Data Availability

Targeted sequencing data supporting the findings of this study can accessed at the cBioPortal for Cancer Genomics website, using the following link: https://www.cbioportal.org/study/summary?id=ovarian_msk_2025.

Additional data are available upon reasonable request to the corresponding author

RESULTS

Demographic and clinical characteristics

Following central pathology review, a total of 40 cases, representing 0.9% of the 4371 ovarian serous carcinomas included in the search, met criteria for qualification as LGS-HT: 30 with ‘synchronous’ low-grade and higher-grade components at primary diagnosis (LGS-HT-Sync) and 10 cases of SBT/LGSC with a subsequent ‘metachronous’ indeterminate/high-grade carcinoma (LGS-HT-Metac). The median age at the time of initial diagnosis was 56 (range, 26–78 years). Stage (FIGO 2014) distribution at presentation, was as follows: I (n=4, 10.0%), II (n=3, 7.5%), III (n=21, 52.5%), IV (n=12, 30.0%; Table 1). Initial surgical cytoreduction was performed in 30 (75.0%) patients; 10 received neoadjuvant chemotherapy, of which 9 subsequently underwent interval debulking surgery and 1 was unresectable after chemotherapy. Information on residual disease following surgery was available for 30 cases, as follows: complete gross resection (n=14, 46.7%), optimal debulking (n=14, 46.7%) and suboptimal debulking (n=2, 6.7%). Most patients (n=38, 95.0%) received post-operative chemotherapy, and a subset also received maintenance therapy, in the form of bevacizumab (n=7), endocrine therapy (n=6), or PARP inhibitor (niraparib, n=2). For LGS-HT-Metac cases, the median time from the primary low-grade serous neoplasm to higher-grade recurrence was 63 months (range 10–135 months).

Table 1:

Clinical characteristics of patients with ovarian low-grade serous tumors with histologic transformation

Characteristic	Entire cohort (n=40) No. of patients (%)	LGS-HT-Sync (n=30) No. of patients (%)	LGS-HT-Metac (n=10) No. of patients (%)
Age	56 (26–78)	56 (29–78)	58 (26–72)
Stage
I/II	7 (17.5%)	4 (13.3%)	3 (30.0%)
III/IV	33 (82.5%)	26 (86.7%)	7 (70.0%)
Primary treatment
Primary debulking surgery	30 (75.0%)	22 (73.3%)	8 (80.0%)
NACT	10 (25.0%)	8 (26.7%)	2 (20.0%)
Residual disease after surgery†
Complete gross resection	14 (46.7%)	7 (31.8%)	7 (87.5%)
Optimal debulking	14 (46.7%)	13 (59.9%)	1 (12.5%)
Suboptimal debulking	2 (6.7%)	2 (9.1%)	0 (0%)
Post-operative Treatment
Adjuvant chemotherapy	38 (95%)	30 (100%)	8 (80.0%)
Maintenance therapy
Endocrine therapy	6 (15.0%)	4 (13.3%)	2 (20.0%)
Bevacizumab	7 (17.5%)	7 (23.3%)	0 (0%)
PARP inhibitor	2 (5.0%)	2 (6.7%)	0 (0%)

LGS-HT – low-grade serous tumors with histologic transformation (S - synchronous low-grade and higher-grade components; M – metachronous low-grade and higher-grade components); NACT – neoadjuvant chemotherapy.

^†Data for residual disease after surgery were missing for 10 patients.

Six patients showed persistent/progressive disease despite chemotherapy/surgical resection. Another four were disease-free after median follow-up of 34.5 months (range 23–160 months) following primary treatment. For the remaining 30 patients who progressed, median time to first recurrence was 22.5 months (range 8–187 months).

Morphologic and immunohistochemical features

Representative photomicrographs of LGS-HT are shown in Figures 2A-I, 3A-F, and Supplementary Figures S1, S2). Low-grade serous tumor components consisted of SBT only (n=11), or LGSC, with or without a residual SBT component (n=29). The higher-grade tumor components were classified as being of indeterminate grade in 14 (35.0%) cases and high grade in 26 (65.0%) cases. Comparison of the pathologic features between indeterminate-grade versus high-grade tumors within the LGS-HT cohort revealed significant differences only with respect to nuclear grade and p53 IHC results, which were expected given that these were the definitional features that distinguished between these categories (Supplementary Table S1; see Case Selection and Central Pathology Review in the Materials and Methods section for details). Of note, indeterminate-grade serous carcinoma components often showed an intimate admixture of tumor cells with small-to-medium-sized nuclei, along with occasional large, bizarre nuclei, making assessment of nuclear pleomorphism challenging (Figure 3F, Supplementary Figure S1A-D). It is relevant to re-iterate that these tumors were diagnosed as or considered suspicious for “high-grade” in their original pathology reports, likely due to the equivocal nuclear atypia and high mitotic rates (≥12 mitoses per 10 HPF), which met traditional grading criteria for the diagnosis of HGSC(11).

Figure 2:

Low-grade serous tumors with histologic transformation to TP53-mutated high-grade carcinoma components. (A,B) Case 26, composed of low-grade serous carcinoma (left) and high-grade serous carcinoma (right), separated by dashed line (A, 200X magnification). LG - low-grade; HG - high-grade. Immunohistochemical analysis of p53 shows a wildtype (wt) expression pattern in the low-grade area and aberrant (abn) cytoplasmic expression in the high-grade area (B, 200X magnification). (C) Case 14, with the higher-grade component showing moderate (Gr. 2) nuclear atypia, but aberrant diffuse p53 expression (inset), consistent with a high-grade carcinoma (400X magnification). (D-F) Case 18, with high-grade carcinoma (delineated by dashed lines) emerging from a background of low-grade serous carcinoma (D, 200X magnification). Immunohistochemical stain for p53 shows a subclonal aberrant pattern, with foci of overexpression, corresponding with the morphologically more atypical areas (E, 100X magnification). Higher magnification of high-grade focus, with marked nuclear atypia. Representative “large” (red arrowheads) and “small” nuclei (blue arrowheads) demonstrate >3:1 nuclear size variation (F, 400X magnification). (G-I) Case 24. Immunohistochemical staining for p53 reveals a heterogeneous subclonal aberrant expression pattern. (G, 40X magnification). Morphologically, there are areas of low-grade and high-grade tumor, separated by dashed line (H, 200X magnification). These areas correspond to wildtype and aberrant diffuse p53 expression patterns, respectively (I, 200X magnification).

Figure 3:

Low-grade serous tumors with histologic transformation to TP53-wildtype higher-grade carcinoma components. (A-C) Case 13, showing admixture of low-grade (LG) and high-grade (HG) tumor components (A, 200X magnification). Higher magnification of low-grade serous carcinoma component with round, uniform nuclei showing mild atypia (B, 400X magnification) and high-grade serous carcinoma component, with severe nuclear pleomorphism and atypia, and abundant mitotic figures (black arrowheads). Representative “large” (red arrowheads) and “small” nuclei (blue arrowheads) demonstrate >3:1 nuclear size variation (C, 400X magnification). (D-E) Case 16, Low-grade serous carcinoma component (D, 400X magnification), and high-grade serous carcinoma component (E, 400X magnification), with immunohistochemistry showing p53 wildtype expression pattern (inset). (F) Case 1, with an indeterminate-grade serous carcinoma component, showing moderate nuclear atypia, with occasional large nuclei and abundant mitotic figures (400X magnification).

Immunohistochemical expression of markers of serous differentiation, PAX8 and WT1, was observed in the higher-grade tumor component in 27/29 (93.1%) and 27/31 (87.1%) cases, respectively; in all these cases, the low-grade serous component was positive for both markers. Based on morphologic features and PAX8/WT1 expression profiles, tumors were considered to retain serous differentiation post-histologic transformation in most cases (36/40, 90.0%), though 4 cases were classified as high-grade carcinoma, not otherwise specified (n=4, including 1 case with sarcomatoid differentiation). ER and PR positivity (at least focal, weak expression) was observed in 19/27 (70.4%) and 8/27 (29.6%) cases, respectively. IHC for p53 in the higher-grade tumor components for 31 evaluable cases revealed an aberrant pattern of expression (including subclonal aberrant) in 12 (38.7%), wildtype pattern in 14 (45.2%), and an equivocal result in 5 (16.1%; Figures 2A-I, 3E, Supplementary Figure S1D; see Molecular genetic landscape below for more details).

In LGS-HT-Sync cases, the low-grade and higher-grade components did not necessarily co-exist within the primary tumor: in 4 cases, the ovarian tumor was low-grade, while high-grade carcinoma was only present in metastases. In LGS-HT-Metac cases, comparative review of primary and recurrent tumors revealed striking morphologic differences, yet in most cases, retained immunohistochemical expression of Mullerian markers (PAX8, WT1, ER/PR) and/or “serous-like” histologic features (e.g. papillary/micropapillary growth, tufting), were consistent with tumor evolution and metastatic recurrence (Supplementary Figure S2A,B).

Molecular genetic landscape

Somatic molecular genetic profiling by targeted panel NGS was performed for 32 cases (MSK-IMPACT assay, n=29; Foundation Medicine, n=3; Figure 4A-C). For tumors that met sufficient tumor purity criteria for assessment of global copy number changes, the median FGA, a measure of chromosomal instability, was 0.30 (range 0.02–0.60, n=25; Figure 4B). Comparison with cases of LGSC and HGSC with evaluable MSK-IMPACT data from our reference cohorts revealed LGS-HTs to have significantly higher FGA than LGSC (n=100; median FGA 0.17, range 0.0–0.74, p<0.001), while lower than HGSC (n=809; median FGA 0.44, range 0–0.98, p=0.011; Figure 4B). Tumor mutational burden was higher than LGSC, but not significantly different from HGSC (median for LGS-HT: 2.6, LGSC: 1.8, HGSC: 3.5 mutations/Mb, LGS-HT vs LGSC, p<0.001, LGS-HT vs HGSC, p=0.061; Figure 4C). Frequently altered genes included known drivers of low-grade serous tumors, such as components of the MAPK signaling pathway, which included hotspot mutations in KRAS (n=7, 21.8%, including G12V, n=3, G12D, n=2, G12C, n=1, G12A, n=1), NRAS (Q61R, n=4, 12.5%), and BRAF (n=5, 15.6%, including V600E, n=3; D594G, n=1; G469V, n=1), as well as loss-of-function mutations/deletions of NF1 (n=5, 15.6%) and NF2 (n=3, 9.4%). CDKN2A deletion/truncating mutations (n=3, 9.4%) and EIF1AX mutation (n=1, 3.1%), which have also been implicated in low-grade serous carcinogenesis(6–8), were also observed (Figure 4A).

Figure 4:

Molecular landscape of low-grade serous tumors with histologic transformation. (A) Oncoprint showing summary of somatic pathogenic genetic alterations, annotated as specified in the legend. LGS-HT – low-grade serous tumors with histologic transformation. SNV – single nucleotide variant. (B) Fraction of genome altered (FGA) and tumor mutation burden (TMB) in LGS-HT, compared to reference cases of ovarian low-grade serous carcinoma (LGSC, n=100) and high-grade serous carcinoma (HGSC, n=809).

TP53 mutations were identified in 12/32 (38.5%) LGS-HTs. There was no significant difference in FGA between TP53-wildtype and TP53-mutated tumors (median 0.39 vs 0.28, p=0.084). Of 27 cases with both NGS and p53 IHC performed on the same specimen, p53 status by both methods were concordant in 20 (74.1%) and discordant in 7 (25.9%): TP53-wildtype by NGS with equivocal p53 IHC (n=5; Supplementary Figure S1D), TP53-wildtype with aberrant diffuse p53 overexpression (n=1), and TP53-mutant with wildtype p53 expression pattern (n=1). Within the concordant group, there were 4 tumors which exhibited foci of diffuse p53 overexpression in a wildtype heterogeneous background (i.e. “subclonal” mutation pattern(28), Figure 2E,G,I). Of 4 cases with p53 IHC but without NGS results, 2 showed a wildtype expression pattern and 2 showed subclonal aberrant p53 overexpression. Overall, 15/35 (42.9%) LGS-HTs had “p53-abnormal” status (TP53 mutation by NGS or aberrant p53 IHC pattern).

Interestingly, pathogenic alterations in genes involved in DNA damage response/homologous recombination (BRCA2, PALB2, ATM, NBN, CHEK2, RECQL4) were observed only in cases of LGS-HT that were TP53-wildtype (Cases 3, 6, 12, 16, 20, 23, 37). Interestingly, pathogenic germline variants of CHEK2 (n=1) and RECQL4 (n=1) were observed. One TP53-wildtype LGS-HT (Case 13) harbored a CCNE1 amplification, a known driver of HGSC(4).

We sought to identify novel drivers of histologic transformation by performing WES on 5 of the cases lacking TP53 genetic alterations. WES identified a median of 30 total mutations (range 19–289) per tumor; however, with exception of USP9X mutations, we did not observe any other recurrent genetic alterations or pathogenic alterations in cancer driver genes that were not already detected by targeted sequencing. For Case 16, mutational signature analysis of WES data revealed a dominant homologous recombination deficiency (HRD) Signature 3, consistent with the presence of a somatic BRCA2 mutation (Supplementary Figure S3A,B). This patient was treated with niraparib maintenance therapy, and despite its aggressive morphologic and molecular features (high-grade nuclear atypia and extensive chromosomal instability), the patient remained disease-free at last follow-up (52 weeks since initiation of treatment).

For 4 cases of LGS-HT, the low-grade and high-grade tumor regions were individually microdissected and subjected to targeted panel sequencing. In all cases, shared mutations between the morphologically distinct components, which included RAS/RAF driver mutations, were consistent with a clonal relationship (Supplementary Figure S4A-D).

Response to targeted therapies

Seven LGS-HT patients received MAPK pathway targeted therapy for recurrent disease (Supplementary Table S1), with median duration on therapy of 7 months (range 0.25 – 24 months). Of these, 4 had HGSC and 3 had indeterminate-grade serous carcinoma components present at treatment initiation. Therapy was discontinued due to adverse effects (n=3) or disease progression (n=4). For the latter, median duration on therapy was 14 months (range 7 – 24 months). The 2 patients that remained on therapy for the longest duration (15 months and 24 months) had HGSCs harboring KRAS-G12V and BRAF-G469V mutations, respectively, which arose from SBTs.

Six patients received PARP inhibitor therapy (Supplementary Table S2), as frontline maintenance (n=2) or in the recurrent setting (n=4, of which 3 were given as maintenance therapy). Two patients, both with BRCA2 somatic mutations (Case 12 and aforementioned Case 16), remain on therapy with no evidence of disease at last follow-up (48 months and 52 months). Three patients, without HRD-related genetic alterations, discontinued therapy due to progression of disease at 4, 6 and 18 months, and 1 patient discontinued due to adverse effects after 1 month.

Overall survival outcomes

To characterize the clinical behavior of LGS-HT, we performed comparisons of survival outcomes with our institutional cohort of conventional LGSCs (n=109) and HGSCs (n=1672) diagnosed within the same period. We compared conventional LGSCs and HGSCs to only LGS-HT-Sync (n=30) cases to ensure the histologic subtypes for all cases were known at initial diagnosis. There were significant differences in clinicopathologic features across subtypes (Supplementary Table S3). The median age at diagnosis for LGSC, LGS-HT-Sync, and HGSC was 52, 56, and 63, respectively (p<0.001). LGSC and HGSC were more frequently diagnosed at advanced stage (FIGO Stage III/IV: 95.4% and 98.8%, respectively) compared to LGS-HT-Sync (86.7%, p<0.001). Neoadjuvant chemotherapy was more frequently given in patients with LGS-HT-Sync (26.7%) and HGSC (19.2%), compared to LGSC (4.6%, p<0.001). Complete gross resection was achieved at higher rates in LGSC (60.6%) compared to HGSC (44.9%) and LGS-HT-Sync (31.8%, p=0.023).

Significant univariate associations were observed between OS with age (p<0.001), treatment group (neoadjuvant chemotherapy versus primary debulking, p<0.001), residual disease after surgery (p<0.001), and tumor subtype (p<0.001; Supplementary Table S4). The median OS of LGS-HT-Sync was 59.7 months (95% CI 28.4–83.5), which was significantly worse compared to LGSC (median OS: 105.4 months [95% CI 61.1-not estimable]; p=0.026) and not significantly different from HGSC (median OS: 48.8 months [95% CI 45.8–51.2]; p=0.61; Figure 5A). Significant associations between age, treatment group, residual disease following surgery, and tumor subtype were retained on multivariate analysis, however, the difference between LGS-HT-Sync with LGSC was no longer significant.

Figure 5:

Prognosis of low-grade serous tumors with histologic transformation. (A) Overall survival associated with ovarian low-grade serous tumors with histologic transformation, with synchronous low-grade and higher-grade components (LGS-HT-Sync) relative to reference cohorts of ovarian low-grade serous carcinoma (LGSC) and high-grade serous carcinoma (HGSC). (B, C) Overall survival of LGS-HT-Sync cases stratified by RAS/RAF mutation status (B) and degree of nuclear atypia of higher-grade component (C). Swimmer’s plot of low-grade serous neoplasms with higher-grade tumor recurrences (LGS-HT-Metac), annotated as specified in the legend.

Considering only LGS-HT-Sync tumors with transformation to high-grade carcinoma (defined by nuclear grade 3 and/or p53-abnormal status, n=19), the median OS was 49.3 months (95% CI 4.6–78.6), accentuating the significant difference compared to LGSC (p<0.001), and overlapping with HGSC (p=0.21; Supplementary Figure S5A). In contrast, restricting the analyses to LGS-HT-Sync tumors in which the higher-grade component was of indeterminate-grade with nuclear grade 2 and TP53/p53-wildtype (n=9) revealed OS outcomes comparable to LGSC (median OS: 131.0 months [95% CI 59.7-not estimable], p=0.52; Supplementary Figure S5B).

We next evaluated the prognostic impact of individual tumor morphologic and genetic features within LGS-HT-Sync. The presence of the characteristic RAS/RAF driver mutations of low-grade serous neoplasms (BRAF-V600E, KRAS/NRAS hotspot mutations) were associated with prolonged OS (131.0 months [95% CI 59.7- not estimable] versus 49.3 months [95% CI 4.6–78.6] for tumors lacking these driver alterations; p=0.011; Figure 5B), while those with severe nuclear atypia had worse OS (Grade 3: 49.7 months [95% CI 4.6–78.6], versus Grade 2: 85.2 months [95% CI 3.6-not estimable]; p=0.038; Figure 5C). Interestingly, p53-abnormal status was not associated with OS (p=0.10).

Amongst the patients with LGS-HT-Metac tumors (n=10), there were 4 deaths (high-grade carcinoma, n=3; indeterminate-grade, n=1; Figure 5D). Two of these had tumors that harbored a TP53 mutation and 2 were RAS/RAF-mutated. The median time from diagnosis of the higher-grade recurrence to death was 10.2 months (range 0.3– 39.0). For the remaining cases which remain alive at last follow-up, median follow-up from time of the higher-grade recurrence was 39.6 months (range 26.2–80.1).

DISCUSSION

While LGSC and HGSC generally have non-overlapping molecular genetic profiles and progress along independent pathogenic pathways, occasionally low-grade serous neoplasms (SBT or LGSC) undergo high-grade transformation. However, the diagnostic criteria, prognostic significance and treatment implications pertaining to this rare phenomenon are ill-defined.

The two-tiered grading system for ovarian serous tumors relies predominantly on the degree of nuclear pleomorphism, with mitotic activity as a secondary criterion(11,12). In most cases, the severe nuclear atypia of HGSC is easily distinguishable from the mild nuclear atypia of LGSC. However, some tumors have a moderate degree of nuclear atypia (Grade 2), and for these, the mitotic rate has been proposed as a “tie-breaker,” with those exceeding 12 per 10 HPF warranting the diagnosis of HGSC, while those below this threshold, are classified as LGSC(11). This proposal is controversial.

The diagnostic category of indeterminate grade serous carcinoma has previously been proposed by Zarei et al. to describe serous carcinomas with morphologic features of both LGSC and HGSC(20). In their study, indeterminate-grade serous carcinomas had enlarged nuclei with irregularities and overlapping, but lacked the overt, generalized pleomorphism of conventional HGSC. Most of their indeterminate-grade serous carcinomas had a discernable low-grade tumor component and lacked TP53 genetic alterations. While numbers were small, indeterminate-grade serous carcinomas were associated with worse overall survival compared to classic LGSC, and more similar to classic HGSC. Ayhan et al. reported TP53 mutations in 10/11 (91%) serous carcinomas with Grade 2 nuclear atypia(21). These studies highlight the intrinsic interobserver subjectivity in the interpretation of nuclear atypia.

Since TP53 mutation is widely accepted to be a distinguishing feature of HGSC, we relied on p53 status to adjudicate tumors with Grade 2 nuclear atypia, characterized by nuclear size variation from 2-to-3:1, with only rare “bizarre” nuclei permitted, as opposed to the overt pleomorphism typically seen in HGSC. According to our diagnostic algorithm, “high-grade” designates tumors with either morphologically obvious severe nuclear pleomorphism (Grade 3 nuclear atypia) or those with Grade 2 nuclear atypia together with aberrant p53 IHC or TP53 mutation confirmed molecularly. We reserve the term “indeterminate grade” to refer to tumors with Grade 2 nuclear atypia and high mitotic rate with “non-aberrant” p53 status. Notably, this approach departs from the traditional two-tiered grading system that relies solely on mitotic rate to classify tumors with Grade 2 nuclear atypia(11).

We acknowledge the inherent subjectivity in morphologic assessment. Consensus agreement by a review panel of 3 gynecologic pathologists in this study provided assurance of diagnostic reproducibility. Since morphologic heterogeneity was a defining feature of all LGS-HTs in our study, comparison with areas of tumor showing conventional low-grade morphology (within the same tumor for LGS-HT-Sync or a prior specimen for LGS-HT-Metac cases) facilitated identification of “higher-grade” tumor components. The presence of distinct low-grade and higher-grade components was essential for the diagnosis of LGS-HT, as opposed to a homogeneous difficult-to-grade tumor; though even in LGS-HTs, there were often areas that show an intimate admixture of low-grade and high-grade tumor cells. A subclonal aberrant p53 IHC staining pattern was often helpful to delineate distinct tumor subpopulations for those cases with high-grade transformation associated with TP53 mutation.

Molecular profiling of low-grade and higher-grade components isolated by microdissection for a subset of our cases showed shared genetic alterations, consistent with a clonal relationship, as previously described(17–19,25). Moreover, we observed that KRAS, BRAF and NRAS hotspot mutations, the primary genetic driver alterations of low-grade serous neoplasia, were clonal (i.e. present in all tumor cells) in the higher-grade components of almost all LGS-HTs harboring these genetic alterations, consistent with their origin from the corresponding low-grade serous tumors.

Recent work suggests that evolution of a low-grade serous neoplasm to high-grade carcinoma is a relatively early pathogenic event(25). Accordingly, in 11/40 (27.5%) cases within our cohort, the low-grade component was only SBT, the precursor of LGSC. Furthermore, most of our cases were LGS-HT-Sync (30/40, 75.0%), in which histologic transformation already occurred by the time of initial diagnosis. It is likely some LGS-HT-Metac cases may have been missed during case selection due to sampling bias, however. Since tissue material for recurrent lesions is usually less available than for primary tumor resections, the true incidence of histologic transformation occurring late in the natural history of low-grade serous neoplasia remains unknown.

Consistent with prior studies, TP53 genetic alteration was the most consistently observed driver of high-grade transformation in our LGS-HT cohort and correlated with aberrant p53 expression by IHC(17,18). High-grade components were often morphologically and immunophenotypically indistinguishable from conventional HGSC (PAX8-positive/WT1-positive/p53-aberrant), though notably, only 12/21 (57.1%) of sequenced high-grade carcinomas of low-grade serous origin harbored TP53 mutations, indicating that high-grade transformation can occur by TP53-independent mechanisms.

Interestingly, mutations in genes implicated in DNA damage repair/homologous recombination were observed exclusively in tumors with a TP53-wildtype genomic background, providing an explanation for the high chromosomal instability (reflected by FGA and marked nuclear atypia) for some of these cases. However, this is contrary to the common co-occurrence of HRD with TP53 dysfunction in most cancers, particularly HGSC(29,30). It is generally well accepted that genomic instability due to BRCA1/2 mutations, the archetypical drivers of HRD, results in p53-mediated cell cycle arrest and apoptosis(30,31). The underlying mechanism by which HRD TP53-wildtype LGS-HTs evade p53-mediated checkpoints remains unknown. With exception of Case 16, which harbored a CDKN2A homozygous deletion (encoding p16, involved in the G1/S checkpoint), none of the other cases had co-existing genetic alterations involving known cell cycle checkpoint regulators. WES performed on a subset of TP53-wildtype LGS-HT did not reveal additional insights on potential pathogenic drivers. Further work is needed to elucidate the mechanisms that substitute for TP53 mutation to result in LGS-HT.

Notably, loss of PR expression was observed in ~70% of LGS-HTs, regardless of p53 status. A recent study has shown that loss of PR is associated with increased FGA and poor prognosis in LGSC(32), and the role of hormone receptors on maintenance of genomic integrity, through a variety of mechanisms, including PARP1 activation has been demonstrated in cell models (33). It is likely that loss of PR is involved in driving aggressive clinical behavior in low-grade serous neoplasia, and future studies are warranted to investigate morphologic features of PR-negative LGSCs. Identifying the morphologic overlap and differences between PR-negative LGSC and LGS-HTs, and establishing the “upper-limit” of nuclear atypia or FGA acceptable for LGSC will be critical for improving diagnostic reproducibility.

While LGS-HT comprises a heterogeneous cohort, as a group, LGS-HT-Sync was associated with worse OS than LGSC, and similar to HGSC. However, several caveats should be considered in the interpretation of OS comparisons between the different histologic subtypes. Given the rarity of LGS-HT, to maximize sample size, we included cases that were referred following initial diagnosis and treatment at outside institutions but have applied left truncation methodology to account for selection bias. We cannot exclude the possibility that some of the patients in the LGSC reference group may in fact developed histologic transformation during disease progression that went undetected. Like HGSC, the higher rates of neoadjuvant chemotherapy and lower rates of complete gross resection in the LGS-HT-Sync group suggest a greater extent of disease at presentation, compared to LGSC.

Despite these limitations, the worse OS in LGS-HT-Sync compared to LGSC, at least on univariate analysis, is consistent with prior reports of individual cases describing poor outcomes in patients with low-grade serous neoplasms with high-grade transformation(13–16). LGS-HT-Metac cases were excluded from survival analyses, given their unique natural history which precluded meaningful comparison of OS with the other groups, their small numbers and limited clinical follow-up. Nevertheless, it is notable that for the 4 of 10 patients that died, this occurred shortly after histologic transformation. As a caveat, we cannot exclude the possibility that a higher-grade focus was present at initial presentation, but not represented on the histologic slides for some of these presumably metachronous cases.

Within the LGS-HT-Sync group, the only prognostically relevant histologic feature was nuclear grade. TP53 mutation status was not significantly associated with survival. While this may, in part, be due to insufficient statistical power from small sample size, there were TP53-wildtype tumors with high-grade nuclear features in our cohort, which is an unusual combination in ovarian serous neoplasia. Through the decoupling of high-grade histology from TP53 mutation status, we observed that the former, rather than the latter, is the stronger determinant of poor prognosis. Notably, mitotic rate, a feature assessed for grading ovarian serous carcinoma, was not associated with OS. These findings caution against the over-diagnosis of indeterminate grade serous carcinomas as HGSC or high-grade transformation of LGSC. Indeed, OS of the LGS-HT-Sync cases with only indeterminate grade as the higher-grade component was comparable to patients with conventional LGSC.

Consistent with prior work demonstrating MAPK pathway alterations are associated with improved survival outcomes in LGSC(6), we also observed that in LGS-HT, KRAS/NRAS/BRAF^V600E hotspot mutations were associated with longer OS. Since LGS-HT are derived from low-grade serous neoplasms, it is not surprising that this property is retained even after evolution to higher-grade carcinoma. Furthermore, it suggests retained functional dependence on MAPK signaling pathway.

Indeed, MAPK-directed targeted therapy appeared effective in the few patients in our cohort who continued treatment until disease progression, particularly for 2 patients who harbored somatic KRAS and BRAF mutations, respectively. Similarly, 2 patients with BRCA2 somatic mutations have not developed recurrences following initiation of PARP inhibitor maintenance therapy. While these numbers are admittedly small, these observations support consideration of personalized therapy based on the tumor’s genetic profile, rather than eliminating therapeutic options based solely on histologic subtype. For clinical management, our data support molecular profiling of all LGS-HTs, as these tumors may harbor therapeutically relevant genetic driver alterations of either low-grade or high-grade serous tumors. Moreover, underlying pathogenic germline variants in 2 of our patients (involving CHEK1 and RECQL4, respectively) underscore the importance of comprehensive germline testing in all ovarian cancer patients, regardless of histologic subtype.

Despite the limitations of small size and heterogeneity of the cohort, our work provides compelling evidence that histologic transformation to indeterminate/high-grade carcinoma occurs rarely in low-grade serous tumors through a variety of possible mechanisms, including, but not limited to acquisition of TP53 mutations. As a group, LGS-HTs are associated with overall survival rates worse than conventional LGSC, and more comparable to conventional HGSC. Thorough histologic assessment of low-grade serous tumors for morphologic heterogeneity is important, as identification of a definitive high-grade tumor component, defined by severe nuclear atypia and/or p53-abnormal status is particularly associated with poor prognosis. Further work is necessary to determine the clinical significance and refine the diagnostic reproducibility of “indeterminate-grade” serous carcinomas. In addition, promising observations of clinical benefit to MAPK pathway-directed and PARP inhibitor therapies in LGS-HTs that harbor targetable genetic alterations support consideration of these therapeutic options in this ultra-rare patient population.

STATEMENT OF TRANSLATIONAL RELEVANCE.

Low-grade serous and high-grade serous carcinomas are generally considered biologically distinct. Rarely, low-grade serous tumors can undergo histologic transformation into high-grade serous carcinoma or other poorly differentiated tumors. In a clinically annotated cohort of low-grade serous tumors with histologic transformation to indeterminate/high-grade carcinoma, next-generation sequencing revealed a heterogeneous molecular landscape comprised of genetic alterations involving known drivers of low-grade and high-grade serous carcinogenesis and suggest MAPK pathway-directed or PARP inhibitor therapies as potential histology-agnostic treatment options. Histologic transformation identified at initial diagnosis was associated with poor overall survival, with severe nuclear atypia and RAS/RAF-wildtype status being adverse prognostic factors. This study supports the origin of a subset of high-grade serous carcinomas from low-grade serous neoplasm precursors and provides insights to refine the diagnostic criteria, prognostication, and treatment of these rare tumors.