ABSTRACT
Classification of endometrial carcinomas (EC) based solely on histological features is not sufficient for the prognostic and therapeutic guidance of patients. Furthermore, the existence of EC in which the histological type cannot be determined clearly and the poor reproducibility of histological typing have led to difficulties in clinical management. However, molecular classification of EC is very promising because of the high reproducibility and good correlation with clinical outcome. Within the scope of “the Cancer Genome Atlas Project”, EC were divided into four different genomic subtypes, and molecular classification models for EC were developed based on these molecular subcategories. The prognostic differences between these molecular subgroups and the benefit for guidance for adjuvant therapy have been clearly demonstrated in studies. In this article, the importance of molecular classification for EC is discussed and its use in clinical practice is reviewed.
Introduction
More than 50% of endometrial carcinomas (EC) present at an early stage, are in the low-risk group and can be treated with surgery alone, but there is a significant proportion of patients with an aggressive disease course. There is a need for an accurate and useable risk stratification model and for the identification of predictive biomarkers that will determine the extent of surgery, the need for post-operative adjuvant therapy and, if needed, the type of adjuvant therapy for optimal management of these agressive cases.
According to the traditional dualistic model, defined by Bokhman (1) in 1983, ECs have been categorized into two main groups, type 1 and type 2 carcinomas, in terms of clinical, endocrine and histopathological features. Based on this classification, endometrioid type ECs (EEC) constitute the majority of type 1, are associated with excess unopposed estrogenic stimulation of the endometrium and have a favourable prognosis. However, type 2 ECs, which include non-endometrioid histotypes, such as serous and clear cell carcinoma, have a worse prognosis and don’t respond well to hormonal therapy. While some ECs will be prototypic examples of type 1 and type 2 ECs, many of them, particularly high grade [International Federation of Gynecology and Obstetrics-(FIGO) grade 3] ECs often do not fit into either category. Clear cell endometrial carcinomas (CCEC) generally fall into the type 2 category, but not all CCECs show the expected aggressive course of type 2 ECs. Histological type of the EC is an important factor in the selection of appropriate adjuvant therapy. However, there is a morphological overlap, especially in high-grade ECs, which even complicates the distinction between the two main clinically significant histological categories; endometrioid or non-endometrioid. Furthermore, it has been clearly seen that patient management models based solely on histological subtypes are inadequate, both for the very rare and little-known EC subtypes (2,3,4), and for the more common and well-known EC subtypes, which may show prognostic inter-patient heterogeneity.
The original dualistic model does not fully reflect the clinical diversity of ECs, and recent molecular developments have highlighted the importance of incorporating molecular features into risk grouping algorithms, including for patients with EC, as has been done for some other cancer types. The Cancer Genome Atlas (TCGA) project categorised ECs into four distinct genomic subtypes through integration of mutational analysis, copy number variation, and mRNA expression results in 2013 (5).
In this review, the role of the TCGA molecular classification in the management of patients with EC are discussed, including how it is applied in clinical practice and the difficulties that may be encountered.
Conclusion
Molecular classification provided a major improvement in the management of patients with EC across diagnostic, prognostic and therapeutic aspects. Molecular classification has also been integrated into the patient risk stratification guidelines for ECs. MMR and p53 analysis by IHC should be routinely performed in all ECs. Presence of any pathogenic POLE mutation in an EC plays a driver role in the determination of the molecular subgroup and constitutes the first step in the algorithmic approach, in which the MMRd/MSI and p53 tests are subsequent steps to be performed. However, as POLE mutation analysis is expensive and not widely available, this test may be reserved for patients with EC who will be given adjuvant therapy, until a cheaper and easier method, such as an IHC assay, is developed. The significance of molecular classification of ECs should be validated prospectively and improved with further studies.
The Cancer Genome Atlas genomic classification of endometrial carcinoma
TCGA performed a genome-wide analysis of 373 ECs, including EEC (n=307), serous endometrial carcinomas (SEC) (n=13), and mixed endometrioid and serous (n=13) carcinomas. Based on the data obtained from the genomic, transcriptomic, and proteomic analysis of these cases, four distinct molecular subgroups of ECs with different clinical, pathological and molecular features, have been identified (5). Then, molecular classification models based on TCGA molecular subcategories of ECs were developed for further adaptation for clinical practice (6,7). The prognostic differences between these molecular subgroups, and the benefit for guidance for adjuvant therapy have been clearly demonstrated (8).
These molecular subgroups are the ultramutated subtype with mutation in the exonuclease domain of DNA polymerase epsilon (POLE) (7% of cases); the hypermutated-microsatellite instable (MSI) subtype characterized by deficiency of ≥1 mismatch repair proteins (MMRd) (28%); the copy number-high subtype/serous-like characterized by TP53 mutations (26%) and the copy number-low subtype [no specific molecular profile (NSMP)] (39%), which doesn’t fit any of these molecular subclasses mentioned above. The molecular classification of ECs has provided a better clinicopathological approach to patients with EC in many regards, including tumor histological typing, prognostic and therapeutic guidance, and elucidation of hereditary carcinomas.
1. POLE-ultramutated endometrial carcinomas: During DNA replication, the synthesis of the DNA strand and the rereading of the synthesized DNA strand plays a crucial role in correcting errors. DNA polymerase is one of the key molecules in DNA replication. The POLEmut subgroup of ECs is characterized by pathogenic mutation in the exonuclease domain of the POLE gene, which is a DNA polymerase, leading to high tumour mutation load (exceeding 100 mutations per megabase) (9,10).
This category includes 7-12% of ECs. Patients tend to be younger and have a normal body mass index. Although POLEmut ECs often have high-risk pathological features, such as extensive lymphovascular invasion (LVI) and high tumor grade, they have a very favourable clinical course, regardless of the tumor histotype and histological grade. Most, but not all, tumors in this group are of the endometrioid histotype. In the TCGA tumor cohort, 6.4% of low-grade EEC and 17.4% of high-grade EEC, but none of the mixed histology or serous carcinomas were POLEmut EC (5). High tumour grade, scattered tumour giant cells and prominent lymphocytic infiltrate are characteristic histological features. They may show intratumoral morphological heterogeneity and ambigious morphology with coexistence of both endometrioid and serous-like histological features (11).
Detection of POLE mutation in EC appears to lead reduction in therapy, as these carcinomas have a very favourable prognosis. These patients may also be candidates for anti-PD1/PDL1 immune checkpoint inhibitory (ICI) therapies, when they are at advanced stage or have recurrent disease.
2. MSI/MMRd-hypermutated endometrial carcinomas: DNA mismatch repair (MMR) system repairs errors that occur during DNA replication, and has two heterodimers consisting of MutS protein homologue 2 (MSH2)/MutS protein homologue 6 (MSH6) and MutL homologue 1 (MLH1)/PMS1 homologue 2 (PMS2) (12). The MMR is responsible for the maintenance of genomic stability by correcting DNA replication errors (base mismatches or insertion-deletion errors) (13). MMR deficiency is characterized by deficiency of ≥1 MMRd. Microsatellites are short repeated DNA sequences found throughout the genome and DNA polymerases are more prone to make mistakes in these regions. Dysfunction of any MMR gene is first manifested by variations in the length of these microsatellite repeats, termed MSI. MSI is an indirect indicator of MMR dysfunction.
Initially, MSI/MMRd ECs were defined as hereditary cases, associated with Lynch syndrome (LS), a hereditary cancer syndrome characterized by autosomal dominant heterozygous germline mutations in one of the four major MMR genes and which has a 60% lifetime risk of developing EC (14). LS-related cancers develop following somatic loss of function of the other intact allele of the affected MMR gene. Only 10% of MMR deficiencies in ECs are inherited, and associated with LS. In the remaining sporadic cases, epigenetic deletion of the MLH1 promoter region by hypermethylation is the main mechanism preceding the majority of sporadic MMR deficiencies, and some are associated with somatic mutations in the MMR genes. MMR deficiency contributes to high tumor mutation load (>10 mutations/megabase), therefore ECs in this molecular subgroup are highly immunogenic tumors, though not as immunogenic as POLEmut ECs.
The MSI/MMRd EC subgroup accounts for 25-30% of all ECs, and consists predominantly of high grade EECs (15). In the TCGA tumor cohort, 28.6% of low-grade EEC and 54.3% of high-grade ECC were MSI/MMRd EC (5). Mucinous differentiation, “microcystic, elongated and fragmented” pattern of myometrial invasion and LVI are common histological features. Dense peritumoral and intratumoral lymphocytic infiltration usually accompany the tumor cells (16). This tumor type tends to involve the lower uterine segment and occur in a wide age range; those associated with LS occur at a younger age than sporadic cases.
MSI/MMRd ECs constitute the group with intermediate prognosis, among the four molecular subgroups, in which POLEmut ECs have the best prognosis and those with TP53 mutation have the worst. CCECs with MMR deficiency have been reported to have a more favourable prognosis than MMR proficient CCECs. Although they are in the non-endometrioid EC group which is expected to have a poor prognosis, they behave more like MMRd EECs and thus patients with MMRd CCEC are recommended to be managed in the same manner as MMRd EECs (17). Hormonal therapy, given in the context of a fertility-preserving approach, is not suitable for the MMRd group of ECs. As MMRd ECs have high propensity for LVI, sentinel or other lymph node procedures may be required (18,19). They respond well to radiotherapy (RT), but patients do not benefit from platinum-based chemotherapy (CT). They may be good candidates for ICI therapy. PD-1/PD-L1 inhibitor therapy has been approved by The Food and Drug Administration for patients with recurrent or advanced MMRd EC and who do not have any other treatment option (20).
3. P53 abnormal EC (P53abn EC)/copy number-high/serous-like: P53abn EC is characterized by the mutation of the tumor supressor gene-TP53 and constitutes the most aggressive group among the four molecular subgroups, with a high number of somatic copy-number alterations. Most of the tumors in this group are high-grade tumors and have serous morphology. In the TCGA study, 97.7% of SECs, 75% of ECs with mixed histology, 19.6% of high-grade EECs, and 5.0% of low-grade EECs were included in this group (5). A relationship between SEC and hereditary breast and ovarian cancer syndrome, which is associated with germline mutations in the BRCA1 or BRCA2 genes, has been reported (21). For women with SEC, who have a family history of hereditary breast and ovarian cancer syndrome-related malignancy, this relationship should be considered and patients should be referred for germline BRCA1/2 screening.
P53abn ECs tend to be seen at an older age in comparison to other molecular subgroups and are more likely to present at an advanced stage. They can spread to the adnexa and peritoneum without deep invasion of the uterine wall.
P53abn ECs have the worst prognosis within the four molecular subgroups. However, no significant prognostic difference was observed between p53 wild type and p53abn CCECs, and thus similar management of these patients is recommended (17).
P53 abn ECs respond well to the combination of platinum-based CT and RT. Targeted therapies based on Human epidermal growth factor receptor 2 (HER2) (trastuzumab) or homologous recombination defects (HRD) poly-ADP ribose polymerase [(PARP) inhibitors] may provide treatment options (11).
4. No specific molecular profile endometrial carcinomas/copy number-low: Tumors in this molecular group do not harbor specific molecular features of other EC molecular groups, any pathogenic POLE mutation, MMR defect, or p53 abnormality. Hence, after the exclusion of these molecular features, the tumor should be included in this molecular category, which is also the most common one, accounting for almost half of all ECs. Somatic copy number alterations and mutation load are low in this group (22).
The majority of NSMP ECs are typically low grade (grade 1 or 2), early stage EECs, that develop on the basis of endometrial atypical hyperplasia/endometrioid intraepithelial neoplasia and may respond to hormone therapy.
NSMP ECs are the largest and most heterogeneous molecular group. The lack of biomarkers to identify tumors with a high propensity for disease recurrence and thus requiring aggressive treatment, complicates the management of this patient group. A potential biomarker is β-catenin (CTNNB1) mutation status. Studies have shown that low-grade EECs harboring mutations in exon 3 of the CTNNB1 gene have more aggressive outcome, with higher recurrence rates (23). Further studies are needed to elucidate predictive biomarkers that can identify patients with NSMP EC who require adjuvant therapy. NSMP ECs may be suitable candidates for hormonal therapy and they respond well to adjuvant therapy, which is given when there are poor prognostic features.
Distribution of EC histological types based on molecular subgroups is illustrated in Figure 1 (24). Most of the low-grade (FIGO grades 1 and 2) EECs correspond to the NSMP and MMRd molecular subgroups. In contrast, high-grade EECs show a heterogeneous molecular profile with a similar distribution rate across all genomic categories. All SECs are encompassed by the p53abn group.
Application of molecular classification in clinical practice
Molecular classification models based on TCGA molecular subcategories of ECs were developed to be adapted to clinical practice (6,7). Talhouk et al. (6) proposed a clinically applicable method for the molecular classification of ECs, the proactive molecular risk classifier for endometrial cancer (ProMISE). In the context of their study, consisting of 319 ECs, four distinct prognostic molecular subgroups with significantly different survival profiles were identified using methods of MMR immunohistochemistry (IHC) to identify the MMRd group, p53 IHC to identify p53abn group, and sequencing for POLE exonuclease domain mutations to identify POLEmut group (6). Thus, the molecular subtype of ECs can be successfully determined with high interobserver reproducibility (25), and a high level of compatibility is achieved between the molecular subclasses defined in endometrial biopsy and hysterectomy specimens (26). Studies of ECs with dual or multiple molecular classes (consistent with more than one molecular subgroup) have suggested a sequential algorithmic approach to identify the exact molecular subtype that will provide prognostic and therapeutic guidance in these tumors. This approach starts with POLE mutation analysis as the first step, MMRd/MSI analysis follows this and p53 mutation analysis is the last step (Figure 2) (11). Since POLE mutation analysis, the first step, is an expensive and not widely available test, it is usually not possible in routine practice to apply this algorithmic approach in this sequence. Hence, the POLE mutation test can be used in a limited fashion, only for patients in whom adjuvant therapy is planned. This may remain the situation until a cheaper and easier method, such as an IHC assay, is developed to investigate pathogenic POLE mutations.
While the majority of EC are compatible with one of these four molecular categories, 3-6% show features of more than one molecular group and are termed “multiple classifier ECs” (27). Multiple classifier ECs harbor molecular features of different combinations of two molecular groups (POLEmut + MMRd/MMRd + P53abn/POLEmut + P53abn) or a combination of three groups (POLEmut + MMRd + P53abn). Concomitant TP53 mutation in POLEmut or MMRd ECs has been shown to be a passenger mutation that does not affect biological behavior of the tumor. It has been elucidated that these tumors do not have such a poor prognosis as single-classifier p53abn ECs and thus intensive treatment is not required. Since MMRd ECs with pathogenic POLE mutations are also found to have a good prognosis, like single classifier-POLEmut ECs, it is recommended to classify these tumors as single classifier POLEmut ECs (27). Therefore, POLE mutation analysis is the first step in the recommended algorithm to define the molecular subgroup of ECs.
a. POLE mutation testing: POLEmut ECs are diagnosed by detection of one of the 11 different pathogenic somatic missense mutations in the exonuclease domain of the POLE gene (Table 1) (28), P286R and V411L being the most common hot spot mutations. The method currently in use is DNA extraction from the tumor and sequencing of exons 9, 13, and 14 (or exon 9 through 14) by next-generation sequencing (NGS) or Sanger sequencing. Currently, there is no immunohistochemical assay to detect pathogenic POLE mutations, for the diagnosis of POLEmut ECs.
b. MMR/MSI testing: It is recommended to peform MMR/MSI testing for all ECs, due to its diagnostic, prognostic and therapeutic contributions in the management of patients with EC (Figure 3). One of the purposes of MMR/MSI testing is the detection of the MMRd EC molecular subgroup, which has its own characteristics, in terms of treatment response and alternative treatment approaches. Since ICI inhibitor therapy has been approved for all advanced MMRd or MSI-high solid tumors, detection of MMR deficiency or high MSI in an EC raises the option of targeted therapy with ICI. Screening for LS is another indication. EC is often the first carcinoma type detected in patients with LS. Therefore, screening by MMR/MSI testing in ECs enables earlier detection of this syndrome, and provides follow-up of these patients, in terms of the development risk for more fatal carcinomas in the future (frequently colorectal carcinoma), and thus reduces the cancer-related mortality (29,30).
MMR defect is mainly detected using two methods. The first is IHC in which four major MMR protein expressions (MLH-1, MSH-2, MSH-6, PMS-2) are evaluated in tumor cells. The second method is a polymerase chain reaction (PCR) based technique, in which MSI analysis is performed. These two methods have approximately 95% compatibility, and IHC is the leading method with several advantages, such as being cheaper and more widely accessible, and with the capability to identify which MMR gene is likely defective.
MMR proteins exist as heterodimers in which MSH2 pairs with MSH6 and MLH1 pairs with PMS2. MLH1 and MSH2 can maintain their stability by forming heterodimers with other proteins in the cell, even in the absence of their own counterpart. However, PMS2 and MSH6 can maintain their stability only in the presence of MLH1 and MSH2 in the cell. Therefore, in order to reduce the cost of MMR IHC, it is recommended to perform a first-line panel consisting of only two antibodies first (PMS2 and MSH6), and, if a defect is detected in any of these, to perform the second step in which the other two antibodies (MLH1 and MSH2) are added. Two-step IHC has been reported to have similar accuracy to a single-step four-antibody test (31).
Accordingly, isolated loss of MSH6 or isolated loss of PMS2 indicates hereditary or acquired MSH6 or PMS2 defects, respectively. On the other hand, the co-deficiency of MLH1 and PMS2 indicates hereditary or acquired MLH1 defect; co-deficiency of MSH2 and MSH6 indicates hereditary or acquired MSH2 defect. When reporting the MMR IHC result in the pathology report, the terminologies “normal” or “abnormal/defective/deficient” should be used. The use of ambiguous terminologies such as “positive, negative, present, absent, preserved, lost” should be avoided.
MSI PCR testing is a highly accurate and sensitive test, but in terms of its higher cost and poorer availability, it cannot be the first choice in routine practice, as the MMR IHC test is a cheaper and easily accessible test, and has a similar accuracy to the MSI test. As there is a high concordance of these two tests, co-administration of these two tests in all cases is unnecessary. It is recommended that the MSI PCR test is performed when an unexpected or unclear result is obtained with MMR IHC.
c. P53 testing: Immunohistochemically different p53 results are related to different types of TP53 mutations (missense, frameshift, truncating mutations) (32). TP53 missense mutations result in degradation resistant mutant proteins which accumulate in the tumor cell nucleus and reveal p53 overexpression immunohistochemically (strong nuclear expression in more than 80% of tumor cell nuclei) (Figure 4). In contrast, non-sense or frameshift mutations result in premature termination codons that terminate translation and appear as complete loss of p53 expression (null pattern) in tumor cells. The much rarer cytoplasmic p53 expression pattern is usually caused by TP53 mutations that impair the nuclear localization of the protein. In the absence of TP53 mutation, a “normal, wild type” staining pattern is observed immunohistochemically. The “wild-type” staining pattern is characterized by varying rate (from a few positive tumor cells to the positivity of most tumor cell nuclei) and varying intensity of p53 staining in tumor cells (Figure 4). In the “wild-type” staining pattern, the extent of p53 staining varies from a few positive tumor cells to the positivity of most tumor cell nuclei and unlike the mutation immunophenotype of TP53 gene, the intensity of p53 staining differs intercellularly. The level of “wild type” expression depends on the differentiation status and proliferative activity of tumor cells. Highly proliferating tumors may show high levels of wild-type p53 expression, and this profile may be difficult to distinguish from IHC p53 overexpression seen in TP53 missense mutations. The sensitivity of p53 IHC in the detection of TP53 mutation is quite high. The concordance of NGS and IHC for the detection of TP53 mutation is 88% (33). A small percentage of ECs harbouring TP53 mutation (truncating mutation) show “wild type” p53 expression pattern immunohistochemically (32).
The significance of molecular classification in the clinical management of endometrial carcinomas and aspects that require improvement
1. It has been elucidated that accurate and reproducible histotyping, and even grading, of ECs, is not always possible by an approach based solely on histological features, and this issue is more problematic in high-grade ECs. The inclusion of molecular features in the risk stratification scheme appears to make a significant contribution to the clinical approach to patients with ECs, specifically the decision about whether any adjuvant treatment is needed or determination of the appropriate treatment approach (Figure 5). However, prospective validated clinical data are needed to provide therapeutic guidance from molecular classification in routine clinical practice.
2.Molecular subgroups also guide the therapy of patients with ECs. Since MMRd ECs tend to have high LVI, a conservative approach with hormonal therapy is not a good option. In addition, RT should be preferred to CT as a choice of adjuvant therapy in MMRd ECs, as these tumors do not respond well to CT. These patients may also be candidates for anti-PD-1/PD-L1 ICI therapies, when they are at advanced stage or have recurrent disease.
As POLEmut ECs often show aggressive histological features, such as higher histological grade, deeper invasion, or LVI, most of the studies regarding POLEmut ECs report results from patients who have already received adjuvant therapy. Despite the aggressive features of these tumours, patients with POLEmut EC show almost no recurrence or death. It is not clear whether this good clinical course is a result of a good response to treatment or regardless of the treatment (34). However, a recent meta-analysis revealed that most of the POLEmut ECs did not exhibit any recurrence or death, and neither type of adjuvant therapy (RT or CT) was associated with clinical outcome in these patients (35). For now, it has been suggested to reduce the treatment of POLEmut ECs and thus protect the patient from the toxicity of an unnecessary treatment. However, this approach again needs to be supported by evidence from prospective studies. Anti-PD-1/PD-L1 ICI therapy may be a good treatment option in recurrent or advanced POLEmut ECs, as in MMRd ECs.
The p53abn group benefits from platinum-based CT and RT. Among the four molecular subgroups, p53abn ECs get most benefit from the addition of CT to RT in adjuvant therapy, even at an early stage (36). HER2/neu amplification is closely related to the p53abn EC group, regardless of the histology. Therefore, tumors with TP53 abnormality and HER-2/neu amplification may benefit from the addition of trastuzumab to therapy, even in non-serous histology. Moreover, the success of PARP inhibitors in ovarian carcinomas has prompted the consideration of their use in the treatment of p53abn ECs. Studies are ongoing to identify the appropriate patient group with HRD for this treatment.
3. Current findings indicate that the presence of a pathogenic POLE mutation in an EC is the most important prognostic determinant among these molecular features, and in the presence of MMRd or POLEmut, p53 mutation is a passenger mutation that does not appear to affect prognosis. In addition, MMRd ECs also harboring pathogenic POLE mutations (multiple classifiers) were found to have a good prognosis, similar to single classifier POLEmut tumors. Therefore, the accepted best current knowledge is that the presence of any of the reported 11 POLE pathogenic variants in an EC can be considered the driver genomic feature for molecular classification and take precedent over other added molecular features as the prognostic determinant. Thus, there is a great need for more widely available and cheaper methods, such as IHC, to detect POLE mutations. Currently, the only method available for POLE mutation analysis is sequencing by NGS or Sanger, and this remains relatively expensive and difficult. Therefore, POLE sequencing is not suitable for all patients with ECs and its use may be limited to patients who are scheduled to receive adjuvant therapy. On the other hand, the presence of any POLE mutation other than the 11 reported pathogenic mutations has no prognostic effect and these tumors cannot be considered in the POLEmut EC category.
4. NSMP ECs are the largest and most heterogeneous molecular group. The lack of biomarkers to identify those with a high propensity for disease recurrence and thus requiring aggressive treatment, complicates the management of this patient group. it has been shown that NSMP ECs containing β-catenin (CTNNB1) mutations show a more aggressive course. Low-grade EECs harbouring mutations in exon 3 of CTNNB1 gene have more aggressive outcome, with higher recurrence rate (23). Therefore, further studies are needed to identify prognostic subcategories in the molecular group of NSMP ECs. The presence of the CTNNB1 mutation provides a therapeutic option as well as these patients have been shown to benefit most from Bevacizumab treatment (37).