POLE Mutation Characteristics in a Chinese Cohort with Endometrial Carcinoma.

OBJECTIVE: To study the characteristics of polymerase epsilon (POLE) exonuclease domain mutations in Chinese patients with endometrial carcinoma (EC).
METHODS: This study analyzed data from 529 patients with EC in The Cancer Genome Atlas (TCGA) and 467 EC patients evaluated at the Shanghai First Maternity and Infant Hospital (SFMIH). POLE mutation heterogeneity was analyzed in paired curettage and hysterectomy samples from 120 SFMIH patients. Sanger sequencing identified mutations in the POLE exonuclease domain, and correlations between POLE mutation status and various clinicopathological features were determined by chi-squared testing and Cohen's kappa analysis, with Kaplan-Meier survival curves generated to assess correlations between POLE mutation status and overall survival (OS).
RESULTS: Thirty-five mutations were identified in 467 samples (7.5%), and novel mutations were detected in the SFMIH cohort. Compared to the TCGA cohort, the SFMIH cohort had fewer POLE mutations when matched by age (<60) and histology (endometrioid) (p < 0.001 and p = 0.010, respectively). In our study cohort, POLE mutations were significantly associated with adjuvant treatment (p = 0.029), and patients with POLE mutations who underwent chemoradiotherapy had a poor OS (p < 0.0001). Notably, shorter OS was significantly associated with POLE mutations in hysterectomy samples from patients aged >60 years or with stage I disease in the paired curettage-hysterectomy group.
CONCLUSION: The significant difference in POLE mutation profiles between the TCGA and SFMIH cohorts, as well as the poor consistency between the curettage and hysterectomy samples, suggests that different parameters need to be applied to determine the prognosis of patients with EC in China.

Introduction

Endometrial carcinomas (ECs) are categorized into two subtypes according to their clinicopathologic characteristics. Approximately 70% to 80% of ECs are type I, with estrogen-dependent ECs often accompanied by a favorable prognosis. Type-II tumors account for ~10% to ~20% of ECs and exhibit more aggressive biological behaviors and poor outcomes.1,2 Treatments are selected according to the histologic subtype and other clinicopathologic features associated with prognosis;2–6 therefore, proper subtype classification is critical for selecting appropriate therapy. In 2013, The Cancer Genome Atlas (TCGA) project characterized ECs into four separate groups based on genomic features. A novel “ultramutated” subgroup harboring mutations in the exonuclease domain of the polymerase epsilon (POLE) gene was identified, often manifested as a microsatellite stable super-mutant phenotype with high somatic mutation rate. For example, Kandoth et al found a high mutant prevalence of PTEN (94%), PIK3R1 (65%), PIK3CA (71%), FBXW7 (82%) and KRAS (53%) in 17 ultramutated samples.7 Roberts et al detected 30 POLE mutations (5.6%) in 535 endometrioid carcinomas, which was lower than that in TCGA.8 In the Church study, POLE mutations were detected in 48 out of 788 (6.1%) endometrial cancers. In FIGO grade 3 endometrioid carcinoma, the POLE mutation rate is high, ranging from 15%-22%.7,9,10 Furthermore, Hussein et al observed that the endometrial carcinoma with POLE mutation was mainly inclined toward endometrioid differentiation, 60% of which were high-grade endometrioid carcinoma.11 However, POLE mutations, mainly referring to POLE exonuclease domain mutations in the present study, may be associated with outcomes in which studies involving Chinese patients are lacking.

Multiple studies identified POLE mutations in several types of malignant tumors, especially in ECs, colorectal carcinomas, and gliomas.9,12-14 POLE encodes DNA polymerase epsilon, which is responsible for chromosomal DNA replication during cell division.15,16 Polymerase proofreading represents one of the primary mechanisms ensuring DNA replication fidelity.17,18 Moreover, inactivation or suppression of the proofreading capacities of DNA polymerases can cause a dramatic increase in the number of spontaneous mutations.11,19 In ECs, mutations in the exonuclease domain of POLE were previously identified primarily in hotspot regions, including exons 9, 13, and 14.10,20 POLE mutations tend to occur in patients with FIGO grades 2–3, and POLE mutant FIGO grade 3 ECs are less prone to recurrence.10 A study of 534 endometrioid carcinomas showed that 30 (5.6%) were POLE mutations, and the recurrence rate was approximately 3.4%, while the POLE wild type had a recurrence rate of 17%.21 At the same time, Stenzinger et al found that the prognosis of colorectal cancer patients with POLE mutations was not significantly different from that of wild-type patients, but the mortality rate of patients with adjuvant or palliative chemotherapy was significantly higher than that of POLE wild type for terminal stage.22

In this study, we compared POLE mutation frequencies and clinicopathologic features between a TCGA database cohort, a cohort of predominantly non-Asian people, and a Chinese cohort of patients with EC and analyzed the effects of POLE mutations on overall survival (OS) in these two cohorts. Additionally, we determined the consistency of POLE mutation status between curettage specimens and hysterectomy specimens and examined the associations between mutation statuses and OS.

Materials and Methods

Patient Cohorts

In this study, we used the following two cohorts: our local cohort of patients with EC from Shanghai First Maternity and Infant Hospital (SFMIH) and the TCGA cohort. Specimens from the SFMIH cohort, including 467 primary tumor samples and 120 paired curettage specimens, were derived from EC patients treated at the department of gynecology (SFMIH) between 2007 and 2018. Clinical data were obtained by retrospective chart review, disease-specific death was defined as death due to EC and excluded death from other causes, and OS was defined as the time from diagnosis to death. Ethical approval was received from the institutional research ethics board. Current guidelines for adjuvant chemotherapy and radiotherapy (NCCN clinical practice guidelines (2018)23) rely on the International Federation of Gynecology and Obstetrics (FIGO) grade, FIGO stage, histotype, and lymphovascular-invasion (LVI) status. For the TCGA cohort, we downloaded the clinical and POLE mutation data from the EC interface of the TCGA portal (), resulting in data from 529 patients with ECs, 51 of whom harbored POLE mutations.

DNA Isolation

Before DNA extraction, samples are enriched by macrodissection to increase the content of tumor cells. DNA was isolated using a GeneRead DNA kit (Qiagen, Hilden, Germany) from tumor-rich regions (at least 50% tumor cells) of formalin-fixed, paraffin-embedded samples, as assessed by two independent pathologists. In samples with mutations, nontumor endometrium was also extracted from the same slide and underwent the same determination of somatic mutations.

Targeted Sequencing and Analysis

Mutations in the exonuclease domain of POLE (amino acids 268–471) were previously identified mainly in exons 9, 13, and 14.10,20 The primer sets used covered these regions and are described later in this subsection. Polymerase chain reaction (PCR) amplifications were performed using 2×Taq master mix (Vazyme Biotech, Nanjing, China), and PCR products were purified using a QIAquick gel extraction kit (Qiagen) according to the manufacturer’s instructions, followed by Sanger sequencing. The primers used were as follows: POLE-Exon 9 forward, 5′-GTGTTCAGGGAGGCCTAATG-3′ and reverse, 5′-CCATCCCAGGAGCTTACTTC-3′; POLE-Exon 13 forward, 5′-CCTGGCTTCTGTTCTCATTCT-3′ and reverse 5′-GATGTGGCTCACATGCCT-3′; and POLE-Exon 14 forward 5′-GACCCTGGGCTCTTGATTT-3′ and reverse 5′-GGACATCCACCTCCATTCAG-3′. Sequencing reactions were performed as previously described,24 and all mutation-positive samples were resequenced to confirm the mutational status.

Statistical Analysis

Among the two study cohorts and the different tumor types, the chi-squared test or Fisher’s exact test was used to assess significant differences in POLE exonuclease domain-mutation status. Kaplan–Meier curves and log-rank statistics were applied to determine differences in the univariate OS analyses. Cohen’s kappa analysis was performed to determine the consistency of POLE mutation status between hysterectomy and curettage specimens. For all analyses, p < 0.05 was considered statistically significant, and all statistical tests were two sided. Statistical analyses were performed using SPSS (v.20.0; IBM Corp., Armonk, NY, USA).

Results

Characteristics of the EC-Patient Cohorts

We retrospectively analyzed 467 EC patients at SFMIH, the majority of whom (69.6%) were <60 years of age (32.9% of the TCGA cohort were <60 years of age). In total, 91% of the patients in the SFMIH cohort had early-stage disease (defined as patients with FIGO stage I or II disease), which was greater than this proportion in the TCGA cohort (72%). The SFMIH cohort included a lower proportion of patients with high-grade tumors than the TCGA cohort (55.8% TCGA vs 13.3% SFMIH), including grade 3 endometrioid, serous, mixed, and clear-cell carcinomas. Other notable differences between the SFMIH cohort and the TCGA patient population were found in the tumor histotypes, with endometrioid and serous carcinomas accounting for 85.2% and 7.3% of the patients in the SFMIH cohort compared with 74.9% and 21.0% in the TCGA cohort, respectively (Table 1). These findings suggested that the clinical characteristics of EC patients in the TCGA cohort differed from those in the SFMIH cohort.

Table 1

Clinical Features of Patients in the TCGA and SFMIH Cohorts

Characteristic	Subcategory	TCGA (n = 529)	SFMIH (n = 467)	Chi-Squared Test p
Age (y)	≥60	352 (66.5%)	142 (30.4%)	<0.001
	<60	174 (32.9%)	325 (69.6%)
Clinical stage	I	330 (62.4%)	388 (83.1%)	<0.001a
	II	51 (9.6%)	37 (7.9%)
	III	121 (22.9%)	38 (8.1%)
	IV	27 (5.1%)	4 (0.9%)
Grade	1	96 (18.1%)	320 (68.5%)	<0.001a
	2	116 (21.9%)	58 (12.4%)
	3	295 (55.8%)	62 (13.3%)
	Mixed	22 (4.2%)	13 (2.8%)
Histotype	Endometrioid	396 (74.9%)	398 (85.2%)	<0.001a
	Serous	111 (21.0%)	34 (7.3%)
	Mixed	22 (4.2%)	13 (2.8%)
	Clear	0 (0.0%)	7 (1.5%)
	Carcinosarcoma	0 (0.0%)	14 (3.0%)
	Undifferentiated	0 (0.0%)	1 (0.2%)

Note: aBonferroni-corrected p-value.

Comparison of POLE Mutations Between the TCGA and SFMIH Cohorts

Despite commonalities in the frequent mutations found in the POLE exonuclease domain between both cohorts, we found notable differences in clinical features and outcomes. Among the 467 patients with EC in the SFMIH cohort and the 529 patients in the TCGA cohort, we identified 35 and 55 mutations, respectively, in POLE exons 9, 13, and 14 (Figure 1A). Twenty-four novel mutations were identified in the POLE exonuclease domain (amino acids 268–471) in the SFMIH cohort, including missense mutations and a truncating mutation. Only three hotspot mutations within the POLE exonuclease domain were found in both cohorts, suggesting that the POLE mutations in the SFMIH cohort differed from those in the TCGA cohort (Figure 1B and Table 2).

Figure 1

Mutations in the POLE exonuclease domain in ECs. (A) Mutations in the POLE exonuclease domain in the TCGA and SFMIH cohorts. (B) Hotspot POLE mutations in the SFMIH cohort.

Table 2

Sites of Mutations in the POLE Exonuclease Domain in ECs

	TCGA		SFMIH
Histology	POLE-mutated	POLE mutation (n)	POLE-mutated	POLE mutation (n)	Common mutation
Endometrioid	51/396	A456P, L424V, M444K, P286R,V411L (3) A465V, P286R (7), Q453R, S459F A426V, A456P, K429N, L424I (3), M295R, P286R (11), P436R, S297F (3), S461L, T278M, V411L (9)	28/398	V411L (1), T278S (1), T457M (2), V411M (1), W410* (1), T278P (1), S459Y (1), P441L (1), P286R (3), L455Q (1), K429R (1), G433D (1), F285S (1), D301N (1), A456T (2), E277V (1), F285L (1), P286S (1), R446W (1), P476L (1), Y434N/V411L (1)P286R (1), Q303R (1)	P286RV411LA426V
Serous	3/111	A480D, V411L, A428T	4/34	A426V (1), M487V (1), T279A (1), T457L (1)
Clear cell	0		1/7	P286R (1)
Undifferentiated	0		0/1
Mixed Carcinomas	1/22	P286R	1/13	Q303R (1)
Carcinosarcoma	0		1/14	H475R (1)
Total	55		35

Notes: Text in bold represents POLE mutations specific to the TCGA cohort, text with underline represents POLE mutations specific to the SFMIH cohort, and text in italics represents POLE mutations common to both the TCGA and SFMIH cohorts.

Several studies, including TCGA, had shown that POLE mutations were associated with outcomes. Before assessing the effect of POLE mutations on overall survival, baseline variables were examined using multivariate analysis, the results of which showed no significant differences in the clinical characteristics within each group of the SFMIH cohort (Table 3). We performed a detailed stratified analysis to further explore the differences in POLE mutations between the two study cohorts. In the SFMIH cohort, younger patients (age <60) and those with the endometroid subtype of EC had significantly fewer POLE mutations than their counterparts in the TCGA cohort (p < 0.001 and p = 0.01, respectively) (Table 4). Because previous findings showed that POLE mutations were associated with a favorable prognosis, we compared the effects of POLE mutations on OS in both cohorts. The presence of POLE mutations in the TCGA cohort remained a significant favorable prognostic factor for OS (p = 0.010) in both the age <60 years subtype (p < 0.0001) and the endometrioid tumor subtype (p < 0.0001). In the SFMIH cohort, the OS analysis grouped according to age and histotype did not show significant differences (Figure 2A and B).

Table 3

Patient Characteristics at Baseline in the SFMIH Cohort

	SFMIH (n = 467)	Mutated (n = 34)	Not Mutated (n = 433)	Chi-Squared Test
Age
	median	54.5	56	ns
	IQR	51–59.5	49–61
Grade				ns
	1	19 (55.88)	302 (69.75)
	2	6 (17.65)	52 (12.01)
	3	7 (20.59)	56 (12.93)
	Mixed	2 (5.88)	11 (2.54)
Clinical stage				ns
	I	25 (73.53)	363 (83.83)
	II	5 (14.71)	32 (7.39)
	III	4 (11.76)	34 (7.85)
	IV	0 (0.00)	4 (0.92)
Histology				ns
	Endometrioid	27 (79.41)	371 (85.68)
	Serous	4 (11.76)	30 (6.93)
	Mixed	1 (2.94)	6 (1.39)
	Clear	0 (0.00)	1 (0.23)
	Carcinosarcoma	1 (2.94)	12 (2.77)
	Undifferentiated	1 (2.94)	13 (3.00)

Table 4

Demographic and Clinical Characteristics of Mutations in the POLE Exonuclease Domain in the TCGA and SFMIH Cohorts

Indicators	Group	TCGA			SFMIH			Chi-Squared Test
		Wild-Type	POLE Mutations	POLE Mutations (%)	Wild-Type	POLE Mutations	POLE Mutations (%)	p
Age (y)	≥60	336	16	4.55	133	9	6.34	0.411
	<60	140	34	19.54	300	25	7.69	<0.001
Clinical stage	I	299	31	9.39	363	25	6.44	0.142
	II	46	5	9.80	32	5	13.51	0.841a
	III	107	14	11.57	34	4	10.53	1.000a
	IV	25	2	7.41	4	0	0.00	1.000b
Grade	1	89	7	7.29	302	19	5.94	0.631
	2	106	10	8.62	52	6	10.34	0.711
	3	261	34	11.53	56	7	11.29	0.958
	Mixed	21	1	4.55	11	2	15.38	0.268
Histology	Endometrioid	348	48	12.12	371	27	6.78	0.010
	Serous	108	3	2.7	30	4	11.76	0.089a
	Clear cell	0	0	0	6	1	14.29	——
	Undifferentiated	0	0	0	1	0	0	——
	Mixed carcinomas	21	1	4.55	12	1	7.69	0.698
	Carcinosarcoma	0	0	0	13	1	7.14	——
Total cases		477	52	9.83	433	34	7.28	0.153

Notes: aCorrected p-value. bFisher’s exact test.

Figure 2

Kaplan–Meier estimates according to POLE mutation status in the TCGA and SFMIH cohorts. (A) OS for the younger cohort (<60 years) and (B) the endometrioid cohort. P-values were calculated using the Log rank test (two-sided).

These results suggested that POLE mutations of EC patients from China are unique and might not be a favorable prognostic factor for OS. The low incidence might weaken the positive effect of POLE mutation on the prognosis of EC.

The Clinical Significance of POLE Mutations in EC in the SFMIH Cohort

We then determined associations between mutations in the POLE exonuclease domain and patient clinicopathologic characteristics in the SFMIH cohort (Table 3). We followed the NCCN clinical practice guidelines (2018) to determine the treatment regimen for patients.23 After stratifying according to age group, we found no significant difference in POLE mutations between younger (<60 years) and older (≥60 years) subgroups. All POLE-mutated tumors were stage I, stage II, or stage III, with no POLE mutations identified in stage IV tumors. The POLE mutation rate according to the stage of EC was as follows: 6.44% in stage I, 13.51% in stage II, and 10.53% in stage III, with no significant difference among the groups (p = 0.275) (). Surprisingly, the frequency of POLE mutations was significantly higher in patients who received adjuvant chemoradiotherapy than in patients who did not, which was a novel characteristic of the SFMIH cohort (). We then investigated whether mutations in the POLE exonuclease domain correlated with differences in the OS of EC patients receiving adjuvant therapy. Among patients who underwent chemoradiotherapy, those harboring POLE mutations had a significantly worse prognosis than those without POLE mutations (p < 0.05; Figure 3).

Figure 3

Among patients who underwent chemoradiotherapy, those harboring POLE mutations had a significantly worse prognosis than those without POLE mutations.

Mutations in the POLE Exonuclease Domain in Hysterectomy and Curettage Specimens

We randomly selected 120 curettage specimens paired with hysterectomy specimens from EC patients from the SFMIH cohort for POLE mutation detection. Sixteen tumors (13.3%) were found to harbor mutations in the POLE exonuclease domain in the definitive hysterectomy specimens, whereas 23 curettage specimens (19.2%) harbored POLE mutations (Table 5). Comparison of the curettage samples with the hysterectomy samples revealed similarities with respect to histological subtype and grade, although identification of POLE mutations was inconsistent between the two groups (Table 5). We performed survival analysis using a univariate model according to POLE mutations in the subgroups of older patients, patients with stage I disease, patients with superficial myometrial invasion, and patients who underwent chemotherapy. Interestingly, patients in the older subgroup, as well as the subgroup with stage I disease and POLE mutations in the hysterectomy specimens but not in the curettage specimens, had significantly worse OS than patients in the same subgroup without POLE mutations (Figure 4A and B). Furthermore, OS was not associated with POLE mutations in either the curettage or hysterectomy specimens of patients with myometrial invasion and a history of chemotherapy (Figure 4C and D).

Table 5

Concordance of Tumor Features in Hysterectomy and Preoperative Curettage Specimens from the SFMIH Cohort

Characteristic	Subcategory	Curettage	Hysterectomy	Concordant	Disconcordant	Cohen’s Kappa
						κ-value	p
POLE	Mutated	23	16	5	11	0.118
	Not mutated	97	104	86	18
Grade	1	78	77	68	9	0.641	<0.001
	2	16	21	12	8
	3	17	21	16	6
Histology	Endometrioid	98	104	95	9	0.647	<0.001
	Serous	10	12	10	2
	Clear	3	3	2	1
	Sarcoma	1	1	1	0
	Atypical hyperplasia	8	0	0	0

Figure 4

Kaplan–Meier estimates according to POLE mutation status in curettage and hysterectomy specimens. (A) OS for the older cohort (≥60 years), (B) the stage I cohort, (C) the myometrial-invasion cohort, and (D) the cohort having previously received chemotherapy. P-values were calculated using the Log rank test (two-sided).

Discussion

In this study, we identified somatic mutations in the POLE exonuclease domain in 7.28% of a large independent cohort of 467 patients with EC from SFMIH, China. We identified characteristics distinguishing the SFMIH cohort from the TCGA cohort, including age, stage, grade, and histotype, that might be related to differences in race between the two cohorts.25 In the TCGA cohort, 55 POLE mutations were located in exons 9, 13, and 14, whereas 35 were detected in the SFMIH cohort. Furthermore, most hotspot mutations found in the TCGA cohort were not detected in our cohort, but we detected 24 novel somatic POLE mutations in the SFMIH cohort. Additionally, we explored differences in POLE mutations between subgroups of the TCGA and SFMIH cohorts, with the results confirming that in younger (<60 years) and endometrioid-histotype subgroups, the SFMIH cohort had fewer POLE mutations than the TCGA cohort. Moreover, previous studies investigating a TCGA cohort of patients with ECs showed that POLE mutations occurred predominantly in endometrioid tumors. In the SFMIH cohort, POLE mutations were found at higher frequencies in serous tumors than in endometrioid tumors, which was unexpected.

Survival analysis showed that no significant prognostic impact of POLE mutations on OS was identified within the young and endometrioid subgroups of the SFMIH cohort, which differed from previous reports, which presented better outcomes for patients with POLE mutations.26–28 The higher frequency of low-risk (Grades 1 and 2) patients in the Chinese cohort (typically with good clinical outcomes), together with the pathogenicity of POLE EDM, could account for the discrepancies with previous literature regarding survival of patients with POLE mutated EC.

Interestingly, we found that among patients who received adjuvant chemoradiotherapy according to the treatment guidelines, those with POLE mutations had a significantly higher mortality rate than those without POLE mutations. This result is consistent with an in vitro study that showed that POLE mutant tumor cells have a significantly higher resistance rate to platinum drugs than the POLE wild-type cell line (p < 0.004),29 indicating that POLE mutations contribute to the insensitivity to traditional chemotherapy.30 These data support the possibility that POLE-mutated tumors, due to their defective polymerase proofreading capability, continue to acquire mutations during cell division; with the high tumor mutation burden, large amounts of tumor neoantigens are produced and may lead to insensitivity to adjuvant therapy. Studies have shown that POLE mutations do not play a significantly positive role in adjuvant treatment, and our results revealed a worse OS with adjuvant therapy.31 In summary, further studies should investigate whether POLE mutations have prognostic or predictive implications in these patients and, if so, the nature of the underlying biological mechanism. In the present study, we did not find an association between POLE mutation status and progression-free survival (PFS) due to the relatively small number of patients who experienced recurrence; therefore, a longer follow-up time will be needed to address the significance of the prognostic impact of POLE mutations on PFS.

One novel objective of this study was to determine whether curettage and hysterectomy specimens were consistent in terms of the identified somatic mutations in the POLE exonuclease domain and whether such mutations in the curettage specimens could be used for predictions of prognosis. Various studies showed inconsistencies in tumor typing and histological grading between preoperative and hysterectomy specimens.32,33 The present study showed good consistency for both pathological grade and histological type in paired samples from 120 patients with ECs. Unexpectedly, we identified more POLE mutations in curettage specimens than in hysterectomy specimens by Sanger sequencing, although a recent report showed that using DNA-based techniques to evaluate molecular alterations had high concordance rates between preoperative specimens and definitive hysterectomy specimens.34,35 To determine whether the POLE mutations in curettage specimens differed from those in hysterectomy specimens, we grouped patients according to clinical characteristics for detailed stratified analyses. Our data indicated that POLE mutations result in preoperative tumors not accurately predicting the final POLE mutation results, especially in patients aged >60 or in those with stage I tumor, exhibiting superficial myometrial invasion, or having received adjuvant chemotherapy. Although this result is somewhat puzzling and counterintuitive, expression inconsistencies between curettage and hysterectomy specimens were not uncommon. For example, β-catenin staining was observed in the curettage specimens but not in the hysterectomy specimens36 from patients with sporadic ECs. The reason behind this discordance needs to be further clarified. Recently, many researchers have tried to utilize molecular biomarkers in curettage specimens to predict the behavior of EC cells. However, the high inconsistency rate shown here indicates that this approach should be used very cautiously.

Furthermore, we also observed that identification of POLE mutation status could facilitate accurate patient counseling and help determine further treatment, thereby enabling tailoring of the extent of surgery and/or adjuvant therapies to the individual patient profile. Based on these findings, we explored whether POLE mutations in curettage or hysterectomy specimens correlated with OS. Importantly, we found that POLE mutations in hysterectomy specimens but not in curettage specimens were a significant predictor of poor OS among older patients and those with stage I disease. These findings suggested that identifying POLE mutations in curettage specimens was not helpful for determining the extent of treatment. However, this is the first study investigating the correlation of POLE mutations with prognosis in curettage specimens from subgroups of patients with different clinical features, and more studies are needed to determine their clinical significance.