The early-onset OC (diagnosed at < 30 years), represents a distinct subgroup exhibiting striking differences from late-onset OC in many aspects, including germline cancer predisposition. We undertake the most extensive germline analysis of early-onset OC patients so far, employing comprehensive approaches encompassing DNA WES complemented by RNA WES and PRS analysis.
Mutation profiles in our early-onset OC patients were dissimilar to unselected predominantly HGSC patients that we analyzed previously [6], with BRCA1/BRCA2 GPV identified in only two (1.6%) early-onset OC patients. Our observation is in agreement with findings from the few studies investigating at least a few early-onset OC patients that consistently reported either the absence, or unusually low frequency of GPV in established OC predisposition genes, as reviewed in [11]. It is noteworthy that GPV in established high-penetrance OC predisposition genes in our study significantly associated with multiple primary malignancies in early-onset OC patients who tended to develop invasive OC, and were more likely diagnosed with non-epithelial OC. The increased risk of double primary tumors was observed in a large study by Casper et al. who described the inverse correlation between age at OC diagnosis and double primary cancer risk [44]. Moreover, the risk of double primary tumors was not increased in patients with BTO in a SEER population-based study [45]. Notably, early-onset OC patients with double primary tumors in our study had a significantly more likely hematological malignancies in their family cancer history which was in agreement with the initial and so far, the only early-OC patient study by Stratton et al., who observed increased risk of non-Hodgkin lymphoma and malignant myeloma in first degree relatives of early-onset OC patients with invasive disease [7].
Most GPV were identified uniquely in our early-onset OC cohort and were enriched in only five genes compared to unselected controls, with the CHEK2 gene coding for checkpoint kinase 2 ranking first. Interestingly, CHEK2 GPV identified in 6/123 (4.9%) early-onset OC patients were significantly associated with earlier age at diagnosis compared to previously analyzed OC patients negative for GPV in HBOP cancer predisposition genes [6]. Furthermore, GPV in CHEK2 have been identified in early-onset OC patients by other studies [6, 10, 46, 47], the most prevalently by Carter et al. [10] who identified CHEK2 GPV in 5/147 (3.4%) early-onset OC patients. Although CHEK2 have not been acknowledged as the OC predisposition gene, some studies pointed to an association of CHEK2 with the OC risk [6, 48, 49]. This hypothesis further supports an indirect evidence potentially influencing prognostic and therapeutic considerations. It has recently been shown that CHEK2 is a master regulator of oocyte survival and modifier of the ovarian cellular response to damage [50]. The mechanistic link to these functions involves phosphorylation of the transcription factor FoxM1 whose overexpression is associated with a poor OC prognosis [51, 52]. GPV in CHEK2 are associated with moderate BC risk [53]. Interestingly, our overrepresentation analysis, which has greater power when multiple genes of a predefined set are associated with a small effect, identified a BC gene set enrichment in carriers of GPV in our early-onset OC patients. In addition, PRS313 developed for BC risk stratification significantly differed in early-onset OC patients compared to PRS controls (see below). Moreover, 5/13 early-onset OC patients with a second primary tumor were diagnosed with BC. Consequently, it is plausible to speculate that there may be a shared germline predisposition factor(s) or functional underlying mechanisms common to both BC and early-onset OC, particularly non-HGSC OC.
Besides CHEK2, LY75-CD302 was the second most GPV-enriched gene revealed by the gene burden analysis. This gene consists of LY75 and CD302 that are alternatively transcribed in a readthrough way leading to translation of fusion proteins with high similarity, acting as receptors involved in endocytosis-mediated immune responses including HLA class I-mediated antigen presentation [41, 54]. In addition, LY75 was shown to modulate cellular phenotype of epithelial OC cells and their metastatic potential through mediation of mesenchymal-epithelial transition [55, 56]. Although the significance of GPV in LY75-CD302 (and other immunity-related genes with identified private GPV) for early-onset OC risk is unknown, it can indicate an attractive direction in early-OC development, especially in view of the results of HLA analysis.
HLA molecules are coded by multiple highly polymorphic loci and are crucial for immune reaction activation and progression including anti-tumor immunity [57]. Previously, specific HLA alleles have been described to predispose to certain cancer types, including OC, as observed by Kubler et al. who identified a significantly higher frequency of carriers of HLA class II haplotypes HLA-DQA1*05:01-DQB1*02:01-DRB1*03:01 and HLA-DQA1*01:01-DQB1*05:01-DRB1*10:01) in OC patients from Germany [22]. In addition, HLA-DRB1*03:01 (homozygous or heterozygous) and HLA-DQB1*02:01 (only homozygous) were individually enriched in their patients compared to controls. However, we did not observe significant enrichment of either these haplotypes or genotypes in our cohort; however, we identified another HLA-DRB1 (*11:01) allele enriched in our early-onset OC patients. This allele was associated with BC in Italian cohort of patients previously [58]. The high abundance of HLA-DRB1*11:01 carriers among our early-onset OC patients might indicate a genetic link between BC, immune system, and early-onset OC. In addition, we identified significantly associated HLA-DQA1*01:03 risk allele and HLA-DQA1*03:03 protective allele. However, the cancer risk associations of HLA-DQA1 are contradictory and not very well understood yet [22, 59, 60]
In addition to the specific disease-related risk HLA alleles, also heterogeneity of inherited HLA alleles seems to be important with regard to tumor-associated antigen presentation, cancer cell recognition and elimination, as well as to immunoediting in early cancer development through e.g. stronger selective pressure on driver mutations in tumors [57]. Having analyzed the HLA zygosity, the early-onset OC patients were significantly more frequently homozygotes compared to super-controls. Interestingly, homozygotes in HLA class II were more abundant than homozygotes in HLA class I loci when compared to super-controls. HLA class II homozygosity was previously associated with increased risk of lung and head and neck cancer, and non-Hodgkin lymphomas, thus, particularly with tumors with high mutational burden or infectious etiology, whereas the lower diversity at HLA class I locus was associated with increased risk of Hodgkin lymphoma [61]. In addition, increased HLA homozygosity rate was also described in colorectal patients by Tsai et al. who also noted that the patients with higher HLA heterozygosity were more likely to manifest higher tumor infiltrating lymphocytes in their tumors [62].
Polygenic inheritance, arising from the cumulative effect of numerous genetic low-risk variants, might elucidate a portion of the missing heritability in the predisposition to early-onset OC [63]. However, PRS analysis has currently its limitations, given the lack of consensus on a specific SNP set and uncertain clinical efficacy in OC risk stratification [13, 64]. We conducted PRS analysis using 10 different SNP sets [12–21], but none demonstrated the ability to distinguish early-onset from histology/stage-matched OC patients, or from PRS controls. Conversely, PRS based on four SNP sets [13, 15, 16, 21] were able to discriminate between HGSC patients and PRS controls. This implies that these four SNP sets are specifically associated with the risk of HGSC, the most prevalent OC type in GWAS focused on identifying OC risk loci. Interestingly, PRS313, designed specifically for BC [37], significantly differed in early-onset, predominantly non-HGSC OC patients but not in HGSC. This suggests a potential pleiotropic effect, indicating a common mechanism underlying the development of multiple phenotypes associated with some common variant susceptibility loci. However, evidence supporting a biological function has only been identified for certain loci, highlighting a significant portion of biology that remains unclear.
Considering polygenic inheritance from an alternative perspective, we observed significant variability in GPV burden among different OC types and super-controls. Higher GPV burden was identified in patients diagnosed with invasive epithelial OC and, remarkably, GPV burden tended to increase with increasing somatic genomic instability characteristic for each histological OC types [39, 65]. Interestingly, Qing et al. noticed strong negative correlation between GPV burden and age, suggesting a greater contribution of GPV to the transformation process in early-onset OC patients compared to their late-onset counterparts, where somatic mutations were hypothesized to play a more predominant role [66]. This observation supports our hypothesis that, the cumulative GPV burden may potentially elevate the cancer risk rather than GPV in certain genes, particularly when concomitant with reduced HLA diversity, influencing the efficiency of neoantigen recognition.
Regarding to the association of clinicopathological and genetic factors with the survival, we observed an improved survival of early-onset OC patients compared to previously analyzed late-onset OC patients [6]. In addition, our findings revealed survival advantage in early-onset OC patients compared to histology/stage-matched OC patients lacking gBRCA1/2 GPV. This suggests that age is an independent positive prognostic factor and that the survival advantage in early-onset OC primarily stems not solely from the distinct distribution of histological OC types compared to late-onset OC, particularly the lack of prognostically unfavorable HGSC histological subtype in early-onset OC patients. A positive correlation between survival and age as well as improved survival among non-HGSC epithelial OC patients has been described in prior research [67, 68]. Nevertheless, certain investigations have delineated a less favorable prognosis and lower 5-year survival in LGSC OC in early-onset OC [69]. However, we did not observe any difference in survival of LGSC early- and late-onset OC patients. Additionally, our research unveiled a significant initial survival advantage in gBRCA1/2 positive OC patients, consistent with earlier findings [70–72].
While our study stands out as the most intricate and the third-largest investigation focused on early-onset OC patients diagnosed before the age of 30, still a noteworthy limitation lies in the restricted number of patients. The low number of patients hinders our ability to pinpoint potential private causal alleles effectively. These limitations underscore the need for comprehensive data to better understand the complex landscape of early-onset OC and its associated risk factors.
In conclusion, our comprehensive germline analysis of early-onset OC patients revealed two divergent trajectories of potential germline susceptibility. Overrepresentation analysis highlighted an association to BC, supported by the enrichment of GPV in CHEK2 and the presumably BC-specific PRS313, which successfully stratified early-onset OC from PRS controls. The second avenue pointed towards the impaired immune response, indicated by GPV in the LY75-CD302 gene, coupled with diminished HLA diversity. Furthermore, we found a significantly higher GPV burden in early-onset OC patients compared to super-controls.
In summary, the genetic predisposition to early-onset of OC appears to be a very heterogeneous and complex process beyond the conventional Mendelian monogenic understanding of hereditary cancer predisposition with a modifying role of the immune system. Based on our results, we speculate that rather a cumulative GPV burden than GPV in specific genes may increase early-onset OC risk, especially when it is concomitant with reduced HLA diversity, which affects the efficiency of neoantigen recognition. However, it cannot be definitively excluded that the occurrence of early-onset OC is a random event influenced by the chance and varying values of random variables.