Tumor immune microenvironment and epithelial–mesenchymal transition in recurrence of low-grade stage IA endometrial endometrioid carcinoma

Abstract

Background

Low-grade, low-stage endometrial carcinomas are generally associated with an excellent prognosis. However, some of these tumors recur, although the prognostic factors are largely unknown. Recently, reports have described associations among tumor immune microenvironment (TIME), epithelial-mesenchymal transition (EMT), and circulating tumor cells (CTCs), factors that have also been reported to be associated with recurrence in endometrial carcinoma. Therefore, we conducted the present study to clarify the association between recurrence and TIME- and EMT-related factors in low-grade, low-stage endometrial carcinomas.

Methods

This study included 80 Japanese patients with low-grade, stage IA endometrial endometrioid carcinoma (20 in the recurrence group and 60 in the no-recurrence group). The institutional review board of Kyushu University approved this study (project number: 24101-00). The median follow-up was 94.5 months (range, 10–298). Immunohistochemistry was performed to evaluate regulatory T cells, marginal and intratumoral tumor-associated macrophages, and tumor cell expression of C-X-C motif chemokine ligand 12 (CXCL12) and Secreted Protein Acidic and Rich in Cysteine (SPARC).

Results

In 13 of the 20 recurrent cases, recurrence occurred distant from the primary site, at lung or lymph nodes. Immunohistochemistry revealed that the numbers of FoxP3-positive cells, and marginal and intratumoral CD68-positive cells in the recurrence group were significantly higher than in the no-recurrence group. In addition, tumor cell expression of CXCL12, a chemokine involved in macrophage recruitment, was significantly higher in the recurrence group. Similarly, tumor cell expression of SPARC, an EMT-associated protein, was also significantly higher in the recurrence group. These findings suggest that a tumor microenvironment permissive of tumor cell dissemination may be associated with recurrence.

Conclusions

Recurrence in low-grade, stage IA endometrial carcinoma was associated with increased FoxP3-positive cells, increased marginal and intratumoral CD68-positive cells, and higher tumor cell expression of CXCL12 and SPARC. However, these findings suggesting a tumor microenvironment permissive of tumor cell dissemination should be interpreted as exploratory and hypothesis-generating, given that the involvement of CTCs was not directly evaluated in this study.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12885-026-16277-8.

Background

Endometrial cancer has conventionally been classified into two distinct types based on the criteria established by Bokhman [1]. Type 1 tumors, which are typically characterized by endometrioid histology, are generally associated with a more favorable prognosis than type 2 ones. In 2013, The Cancer Genome Atlas (TCGA) Research Network reported a comprehensive genomic and transcriptomic analysis of endometrial cancers [2]. From this analysis, endometrial cancers were categorized into four genomic classes: (1) POLE ultramutated tumors, (2) microsatellite-unstable tumors, (3) copy-number-low tumors, and (4) copy-number-high tumors. This molecular classification has also been incorporated into the World Health Organization (WHO) classification and clinical risk stratification of endometrial cancer [3]. Notably, more than 85% of endometrial cancer cases exhibit low-grade histological features, predominantly endometrioid carcinoma of grade 1 or 2. These low-grade, low-stage endometrial carcinomas demonstrate 5-year survival rates of up to 95% [4]. However, some of these tumors recur, although the prognostic factors are largely unknown. Recently, several studies have reported the association between circulating tumor cells (CTCs) and tumor recurrence [5]. CTCs are cells shed from the primary tumor that circulate in the bloodstream while exhibiting metastatic potential. CTC presence was found to be closely associated with multiple clinicopathological factors that predicted high metastatic risk in different solid cancers [6]. Similarly, associations among CTCs, the tumor immune microenvironment (TIME), and epithelial-mesenchymal transition (EMT) were reported [6, 7]. The TIME, the main components of which are lymphocytes and tumor-associated macrophages (TAMs), is widely accepted to play a key role in cancer development and progression. For example, CD8-positive cytotoxic T lymphocytes in the TIME recognize tumor antigens and attack tumor cells. Meanwhile, EMT increases cell motility and invasiveness and has been proposed to contribute to CTC generation [8].

Twenty patients with low-grade (endometrioid carcinoma of grade 1 or 2), stage IA endometrial endometrioid carcinoma (EEC) with recurrence were encountered at our institution from 1992 to 2022. Among them, 13 experienced recurrence in the lung or lymph nodes. The consistency of this recurrence pattern raised the hypothesis that tumor cell dissemination may have contributed to the recurrence. Given that tests for CTCs have been incorporated into clinical practice for various cancers, such testing could be valuable to explore this possibility [9–11]. Against the above background, we conducted the present study to clarify the association among recurrence of low-grade, stage IA EEC, TIME, and EMT.

Methods

Case selection and clinicopathological characteristics

All samples used in this study were collected at the Department of Gynecology and Obstetrics, Kyushu University, from 1992 to 2022. In this period, 541 patients were diagnosed with low-grade EEC that had been classified as International Federation of Gynecology and Obstetrics (FIGO) 2009 stage IA at diagnosis, of whom 80 were included in the present retrospective analysis (20 in the recurrence group and 60 in the no-recurrence group). All recurrent cases that met the eligibility criteria (sufficient pathological material for immunohistochemical evaluation and adequate clinical follow-up) were included. For comparison, we first identified no-recurrence cases that met the same clinicopathological and follow-up criteria as the recurrence group and additionally had sufficient archival tissue and evaluable histological areas for quantitative immunohistochemical analysis. From this technically eligible pool, 60 cases were selected by simple random sampling within the study period. A 1:3 ratio of recurrent to non-recurrent cases was chosen to enhance the precision of effect estimates while maintaining the feasibility of the extensive immunohistochemical and image analyses. Because recurrence in low-grade, stage IA EEC may frequently become clinically evident after a relatively long follow-up, only patients with a follow-up sufficiently long to reliably rule out late recurrence were considered eligible for inclusion in the no-recurrence group. Consequently, the no-recurrence group was predominantly drawn from patients treated in the earlier years of the study period, whereas laparoscopic procedures were introduced and became more common in the later years (approximately 2013–2022); laparoscopic cases with shorter follow-up were therefore not eligible for the no-recurrence group. Lymphadenectomy, when performed, was defined as pelvic lymphadenectomy with or without para-aortic lymphadenectomy. No patients received adjuvant therapy in the recurrence and no-recurrence groups.

The medical records and hematoxylin and eosin (HE)-stained slides were re-reviewed to obtain clinical and pathological data regarding age, FIGO 2009 stage, metastasis, lymphovascular space invasion (LVSI), uterine cervical invasion, peritoneal cytology, myometrial invasion depth, therapeutic methods, recurrence, progression-free survival (PFS), and overall survival (OS).

In this context, individuals aged 60 years or older are categorized as elderly to create a dichotomous variable for the multivariable Cox models. Lymph node metastasis is classified based on the extent of tumor involvement as follows: isolated tumor cells, defined as clusters measuring ≥ 0.2 mm and containing ≤ 200 cells; micrometastases, measuring 0.2–2 mm and/or comprising more than 200 cells; and macrometastases, defined as lesions larger than 2 mm [12]. LVSI is further categorized into three levels: none, focal, and substantial. Focal invasion refers to the involvement of fewer than five vessels, whereas substantial invasion involves five or more vessels [3].

All methods were carried out in accordance with relevant guidelines and regulations. This retrospective study using archived clinical data and tissue samples was approved by the institutional review board of Kyushu University (project number: 24101-00). The need for informed consent was waived, and an opt-out approach was used in accordance with institutional regulations.

Immunohistochemical staining

Antibodies

The following primary antibodies were used: CD8 (mouse monoclonal, clone C8/144B, prediluted; Nichirei Biosciences, Tokyo, Japan), FoxP3 (mouse monoclonal, clone 236 A/E7, 1:100; Abcam, Cambridge, MA, USA), CD68 (mouse monoclonal, clone KP-1, 1:100; Dako, Glostrup, Denmark), CD163 (mouse monoclonal, clone 10D6, 1:100; Novocastra, Newcastle, UK), C-X-C motif chemokine ligand 12 (CXCL12) (mouse polyclonal, 1:200; Bioss, Beijing, China), and Secreted Protein Acidic and Rich in Cysteine (SPARC) (mouse monoclonal, clone ON1-1, 1:500; TAKARA, Tokyo, Japan).

Procedure

Four-micrometer-thick sections on silane-coated slides were stained using the universal immunoperoxidase polymer method (Envision kit; Dako), in accordance with the manufacturer’s instructions. After deparaffinization, rehydration, and inhibition of endogenous peroxidase, the sections were exposed to the primary antibodies at room temperature for 90 min. After incubation of the secondary antibody at room temperature, the sections were then incubated in 3,3′-diaminobenzidine (DAB), counterstained with hematoxylin, and mounted. They were subsequently subjected to microwave heating and pressure cooking for antigen retrieval. Positive and negative controls were also established.

Immunohistochemical scoring

Intraepithelial and stromal CD8-positive cells were counted separately, and stromal FoxP3-positive cells were counted in five high-power fields (HPFs, 40⋅) [13, 14]. The mean number of counted cells was obtained for further statistical analyses. To evaluate CD8 and FoxP3 status, representative fields were selected from areas with adequate tumor volume and an assessable tumor epithelium–stroma interface. Cases or fields lacking sufficient stromal tissue, lacking myometrial invasion, or showing limited tumor volume that precluded reliable compartment-based counting were not used for this part of the analysis. The FoxP3/CD8 ratio was calculated as the mean number of FoxP3-positive cells divided by the mean number of stromal CD8-positive cells in the corresponding case. The CD68- or CD163-positive cells that had infiltrated into cancer nests or stroma were defined as intratumoral TAMs, while those that had become distributed along the tumor–myometrial interface were defined as marginal TAMs. The number of CD68- or CD163-positive cells was counted at 200⋅ magnification, and the mean of the values of three representative fields was used for statistical analyses [15]. For evaluation of marginal and intratumoral TAMs, representative fields were selected from the tumor–myometrial interface (marginal area) and intratumoral regions with adequate tissue preservation. Areas with marked artifacts, necrosis, or insufficient evaluable tissue were excluded from field selection. Field selection was performed using the same predefined criteria in all cases.

A computer-assisted cytometric analysis of the immunoreactivity was performed using the PatholoCount image processing software program (Mitani Corp., Tokyo, Japan). Using this program, we were able to objectively assess the results of immunohistochemical analyses in our study. All image analyses were performed using predefined parameters and identical settings across all cases to ensure consistency of quantification.

Regarding CXCL12 expression, representative immunohistochemical images were captured from formalin-fixed paraffin-embedded sections under identical microscope and camera settings. For each case, 10 representative tumor glands distributed across the tumorous area were selected, while avoiding necrotic, crushed, or poorly preserved areas. Luminal spaces were excluded from the measurement area [16]. Quantitative image analysis was performed using ImageJ [17] with the H-DAB color deconvolution workflow related to the IHC profiler plug-in [18]. After color deconvolution, the DAB-enhanced image was used for intensity measurement, and the mean grayscale value of the tumor gland area was recorded (pixel intensity range, 0–255). Because lower grayscale values indicate stronger DAB staining, the values were inverted so that higher values corresponded to stronger CXCL12 immunoreactivity. The mean value per case was used for statistical analysis. The same image-processing and measurement settings were applied to all cases.

In addition, SPARC immunoreactivity was evaluated in tumor cells, defined as cytoplasmic staining with or without concurrent nuclear positivity [19]. The proportion of positive cells was assessed among grades 1–4 (< 5%, 5%–25%, 26%–75%, > 75%) and staining intensity among grades 0–3 (none, weak, moderate, strong). These two values were summed, and the total was categorized as follows: absent (< 5% regardless of intensity), weak (= 3), moderate (= 4–5), and strong (= 6–7). For statistical purposes, tumors were dichotomized into low reactivity (absent/weak) versus high reactivity (moderate/strong) [19]. The assessment of immunohistochemical expression was carried out by a pathologist and a physician specializing in gynecological oncology, both of whom were blinded to the clinicopathological details of the patients.

Statistical analysis

We used the statistical software program JMP^® 18 (SAS Institute, Cary, NC, USA) for all statistical analyses. Pearson’s χ² test, Fisher’s exact test, and Mann–Whitney U test were performed to analyze the data. Kaplan–Meier analysis and the log-rank test were performed for the survival analysis. Cox proportional hazards regression analyses were performed for PFS and OS. In univariable analyses, each clinicopathological and immunohistochemical variable was evaluated individually. Given the limited number of recurrence events, multivariable Cox models were prespecified to include age as a dichotomous variable (< or ≥ 60 years), LVSI, and lymphadenectomy in a base clinical model, and each immunohistochemical marker was added individually to this base model (one marker per model) to avoid overfitting. For PFS, we additionally applied the Benjamini–Hochberg procedure to control the false discovery rate (FDR) across nine immunohistochemical variables in the univariable Cox models. CD8, FoxP3, CD68, and CD163 were prespecified as primary immune markers of interest based on prior evidence of their roles in the TIME, whereas CXCL12 and SPARC were evaluated in an exploratory manner as EMT- and tumor dissemination-related factors.

Hazard ratios (HRs) and 95% confidence intervals (CIs) were reported. For continuous variables, HRs are expressed per 1-unit increase in the original scale of each variable. In all analyses, a p value of < 0.05 was considered statistically significant.

Results

Clinicopathological characteristics

The median follow-up for the entire cohort was 94.5 months (range, 10–298). Elderly patients (≥ 60 years) accounted for 45.0% of cases in the recurrence group and 28.3% in the no-recurrence group. LVSI was identified in 2 cases (10.0%) in the recurrence group (focal LVSI, n = 1; substantial LVSI, n = 1), but in 6 cases (10.0%) in the no-recurrence group (all focal LVSI). Lymphadenectomy was performed in 19 cases in the recurrence group and in 54 cases in the no-recurrence group. Peritoneal cytology data were available for 16 of 20 cases in the recurrence group and all 60 cases in the no-recurrence group; missing values are due to unavailable historical records. Among evaluable cases, peritoneal cytology was positive in 1 recurrence case and 3 no-recurrence cases. Regarding surgical approach, laparoscopic surgery was performed in 6 of 20 cases in the recurrence group, but not in any cases in the no-recurrence group.

After carefully reviewing the clinical data of the 80 cases, we reclassified all cases based on the FIGO 2023 staging. Upon this reclassification, only one case in the recurrence group was classified as stage IIB, owing to substantial LVSI.

Immunohistochemical analysis

Representative immunohistochemical images of CD8, FoxP3, CD68, and CD163 are shown in Figs. 1 and 2. The results indicate associations among the number of immunohistochemically positive cells, the FoxP3/CD8 ratio, and recurrence, as shown in Tables 1 and 2.

Fig. 1.

Representative immunohistochemical images of low-grade, stage IA endometrial endometrioid carcinoma (A and B: marginal CD68, C and D: intratumoral CD68, E and F: marginal CD163, G and H: intratumoral CD163). Marginal CD68-positive (A, high; B, low) or CD163-positive macrophages (E, high; F, low) are observed at the invasive border. Intratumoral CD68-positive (C, high; D, low) or CD163-positive macrophages (G, high; H, low) are observed within the tumor

Fig. 2.

Representative immunohistochemical images of low-grade, stage IA endometrial endometrioid carcinoma. (A and B) CD8-positive cells are observed at the invasive border (stromal CD8 expression: A, high; B, low). (C and D) CD8-positive cells are observed within the tumor (intraepithelial CD8 expression: C, high; D, low). (E and F) Tregs are highlighted by FoxP3 expression (E, high; F, low)

Table 1.

The association between recurrence and the number of cells immunohistochemically positive for CD68 or CD163

	Marginal CD68-positive cell count	p value	Intratumoral CD68-positive cell count	p value	Marginal CD163-positive cell count	p value	Intratumoral CD163-positive cell count	p value
	Mean (SD)		Mean (SD)		Mean (SD)		Mean (SD)
recurrence
present	261.43 (40.27)		247.94 (42.73)		75.17 (32.47)		37.67 (8.57)
absent	40.66 (7.30)	< 0.0001	51.00 (6.56)	< 0.0001	52.26 (7.80)	0.7138	50.11 (6.67)	0.4334

Table 2.

The association between recurrence and the number of cells immunohistochemically positive for CD8 and FoxP3 or their ratio

	Stromal CD8-positive cell count	p value	Intraepithelial CD8-positive cell count	p value	FoxP3-positive cell count	p value	FoxP3/CD8 + ratio	p value
	Mean (SD)		Mean (SD)		Mean (SD)		Mean (SD)
recurrence
present	40.81 (24.16)		19.44 (14.82)		17.90 (14.29)		0.53 (0.52)
absent	27.07 (10.37)	0.0320	19.96 (11.75)	0.5484	8.18 (6.06)	0.0068	0.34 (0.32)	0.1035

The number of FoxP3-positive cells in the recurrence group was significantly higher than in the no-recurrence group (p = 0.0068). Meanwhile, there was no significant association between FoxP3/CD8 ratio and recurrence (p = 0.1035). In addition, the numbers of marginal and intratumoral CD68-positive cells in the recurrence group were significantly higher than in the no-recurrence group (both p < 0.0001). However, there were no significant associations between marginal or intratumoral CD163-positive cells and recurrence (p = 0.7138 and 0.4334, respectively). The number of stromal CD8-positive cells in the recurrence group was significantly higher than in the no-recurrence group (p = 0.0320). However, there was no significant association between the number of intraepithelial CD8-positive cells and recurrence (p = 0.5484).

Representative immunohistochemical images of CXCL12 and SPARC are shown in Fig. 3.

Fig. 3.

Representative immunohistochemical images of low-grade, stage IA endometrial endometrioid carcinoma (A and C: CXCL12, B and D: SPARC). A Diffuse and strong cytoplasmic CXCL12 expression is observed in the tumor cells. B High SPARC reactivity is observed in the tumor cells. C Weak cytoplasmic CXCL12 expression is observed in the tumor cells. D Low SPARC reactivity is observed in the tumor cells

Statistical analysis revealed that CXCL12 expression was significantly higher in the recurrence group than in the no-recurrence group (p = 0.0003) (Fig. 4).

Fig. 4.

Quantitative analysis of immunohistochemical images using ImageJ with the IHC profiler plug-in. Tumor cell expression of CXCL12 is significantly higher in the recurrence group than in the no-recurrence group

In addition, the recurrence group included 9 cases (45.0%) with high SPARC reactivity, while the no-recurrence group had 12 (20.0%), which were significantly different (p = 0.0278) (Table 3).

Table 3.

The association between the recurrence and SPARC immunohistochemical expression

	High SPARC reactivity (%)	p value
recurrence
present	9/20 (45.0%)
absent	12/60 (20.0%)	0.0278

As a sensitivity analysis, we repeated the main immunohistochemical comparisons after excluding all laparoscopic cases from the recurrence group, restricting the analysis to patients who underwent open surgery (14 recurrent and 60 no-recurrent cases). In this restricted cohort, the recurrence group continued to show significantly higher marginal and intratumoral CD68-positive cell counts (both p < 0.0001) and higher CXCL12 expression (p = 0.0125) than the no-recurrence group, whereas the differences in stromal CD8, FoxP3, and SPARC were directionally similar to those in the overall cohort but were no longer statistically significant, although the difference for SPARC showed a tendency for significance (p = 0.3515, 0.1632, and 0.0727, respectively) (Supplementary Table S1).

Survival analysis

Figure 5 shows the results of Kaplan–Meier analysis for OS. In the recurrence group, six cases resulted in tumor-related mortality. The overall survival periods were significantly longer in the no-recurrence group than in the recurrence group (p < 0.0001). Multivariable Cox regression analyses for PFS and OS were performed using the prespecified base clinical model including age (≥ 60 years), LVSI, and lymphadenectomy, with each immunohistochemical marker added individually to the base model (one marker per model) to avoid overfitting. For PFS, stromal CD8-positive cell count, FoxP3-positive cell count, marginal CD68-positive cell count, intratumoral CD68-positive cell count, CXCL12 expression, and SPARC expression remained significantly associated with outcome when added to the base clinical model (Table 4). For OS, stromal CD8-positive cell count, marginal CD68-positive cell count, and intratumoral CD68-positive cell count remained significantly associated with outcome (Table 5). Univariable Cox regression analyses for PFS and OS are shown in Supplementary Tables S2 and S3, respectively. In PFS analysis, FDR-adjusted p-values obtained by the Benjamini–Hochberg procedure confirmed significant associations for stromal CD8, FoxP3, marginal and intratumoral CD68, and CXCL12, whereas the association with SPARC showed only a tendency for significance (FDR-adjusted p = 0.0614) (Supplementary Table S4).

Fig. 5.

O verall survival analysis between the recurrence group (red line) and the no-recurrence group (blue line). The overall survival periods are significantly longer in the latter (p < 0.0001)

Table 4.

Multivariable Cox regression models for progression-free survival

Variable	Base clinical model	Base+stromal CD8	Base+FoxP3	Base+marginal CD68	Base+intratumoral CD68	Base+CXCL12	Base+SPARC
Age (≥ 60 years vs. <60 years)	1.99 (0.82–4.84), 0.1278	2.28 (0.90–5.73), 0.0811	1.61 (0.64–4.10), 0.3144	2.64 (0.97–7.19), 0.0568	1.25 (0.45–3.52), 0.6710	1.91 (0.78–4.75), 0.1587	1.99 (0.82–4.84), 0.1271
Lymphovascular space invasion (present vs. absent)	0.81 (0.18–3.53), 0.7766	0.48 (0.11–2.19), 0.3445	0.23 (0.04–1.32), 0.0998	0.85 (0.19–3.89), 0.8384	1.08 (0.23–5.10), 0.9181	1.06 (0.23–4.85), 0.9379	1.18 (0.25–5.54), 0.8363
Lymphadenectomy (present vs. absent)	1.97 (0.28–14.82), 0.5110	2.01 (0.26–15.64), 0.5062	1.67 (0.22–12.75), 0.6225	2.35 (0.30-18.47), 0.4167	1.07 (0.13–8.56), 0.9492	2.51 (0.31–20.57), 0.3915	1.70 (0.22–12.88), 0.6072
Stromal CD8-positive cell count	-	1.05 (1.02–1.07), < 0.0001*	-	-	-	-	-
FoxP3-positive cell count	-	-	1.09 (1.05–1.14), < 0.0001*	-	-	-	-
Marginal CD68-positive cell count	-	-	-	1.01 (1.00-1.01), < 0.0001*	-	-	-
Intratumoral CD68-positive cell count	-	-	-	-	1.01 (1.00-1.01), < 0.0001*	-	-
CXCL12 expression	-	-	-	-	-	1.03 (1.02–1.05), < 0.0001*	-
SPARC expression (high vs. low)	-	-	-	-	-	-	2.61 (1.03–6.59), 0.0430*

HR hazard ratio, CI confidence interval, * p < 0.05

Values are shown as HR (95% CI), p value

Table 5.

Multivariable Cox regression models for overall survival

Variable	Base clinical model	Base+stromal CD8	Base+marginal CD68	Base+intratumoral CD68
Age (≥ 60 years vs. <60 years)	2.21 (0.44–11.18), 0.3345	12.35 (0.42–13.07), 0.3274	2.89 (0.47–17.68), 0.2513	0.93 (0.12–7.03), 0.9419
Lymphovascular space invasion (present vs. absent)	1.40 (0.16–12.22), 0.7628	0.91 (0.10–8.38), 0.9321	1.51 (0.16–14.30), 0.7183	2.68 (0.12–7.03), 0.9419
Lymphadenectomy (present vs. absent)	NE (NE), 0.9992	NE (NE), 0.9998	NE (NE), 0.9996	NE (NE), 0.9999
Stromal CD8-positive cell count	-	1.05 (1.00-1.09), 0.0079*	-	-
Marginal CD68-positive cell count	-	-	1.00 (1.00-1.01), 0.0085*	-
Intratumoral CD68-positive cell count	-	-	-	1.00 (1.00-1.01), 0.0077*

HR hazard ratio, CI confidence interval, NE not estimable, * p < 0.05

Values are shown as HR (95% CI), p value

Discussion

Low-grade, low-stage endometrial carcinomas occasionally recur, but the factors associated with their prognosis remain unclear. CTCs have been reported as a prognostic factor for metastasis and recurrence in various cancers [20, 21] and the clinical application of testing to identify the presence of CTCs is being explored [9–11]. It was also reported that there are associations among CTCs, TIME, and EMT [6, 7].

In a previous study of stage I endometrial carcinoma that included cases with deep myometrial invasion (> 1/2) or high-grade histology, a substantial proportion of recurrences were reported to occur in the vagina [22]. In contrast, in our study, most recurrences with low-grade, stage IA EEC occurred at distant sites in the lung or lymph nodes, rather than locally. This pattern suggests that recurrence occurred as a result of conditions that promoted tumor cell dissemination. However, direct evidence of the presence of CTCs was not available in the present study. Therefore, we conducted the present study to clarify the associations of recurrence of low-grade, stage IA EEC with TIME and EMT.

Regarding TAMs, CD68 can be used as a pan‑macrophage marker, whereas CD163 can be used as an M2 marker, as described previously [23]. Polarized M1 and M2 macrophages represent the extremes of a continuum of functional states for TAMs [24]. The classically activated M1 macrophages are potent effector cells that kill microorganisms and tumor cells and produce copious amounts of proinflammatory cytokines. Meanwhile, M2 macrophages tune inflammatory responses and adaptive Th1 immunity to promote angiogenesis as well as tissue remodeling and repair. In our study, recurrence was significantly associated with the numbers of marginal and intratumoral CD68-positive cells, whereas no significant associations were observed for marginal or intratumoral CD163-positive cells. Matsukawa et al. previously demonstrated that a high density of CD68-positive macrophages and a low density of CD163-positive macrophages at the tumor margin were associated with worse prognosis [25].

This finding is consistent with our results about the association between prognosis and CD68-positive cell count. However, it has also been reported that a high CD163-positive cell count was independently associated with an inferior outcome or aggressive behavior in hepatocellular carcinoma [26], HER2-positive breast cancer [27], and endometrial carcinoma [28]. Considering these conflicting findings, further investigation will be needed to determine the significance of M1/M2 macrophage clustering. In addition, Wei et al. demonstrated that TAMs were the primary source of IL6, which in turn significantly upregulated Vimentin expression and downregulated E-cadherin expression. They also demonstrated that TAMs induce EMT to enhance the migration of colorectal carcinoma cells [6]. Elsewhere, Pollard et al. demonstrated that TAMs are recruited to the tumor by CXCL12 [29], while it was also reported that SPARC expression in tumor cells enhanced their migration and invasion [30, 31]. In this study, tumor cell expression of CXCL12 and SPARC was significantly higher in the recurrence group. These findings support an association between recurrence and EMT-related factors in low-grade, stage IA EEC, but they do not establish the causal mechanism behind this relationship.

As for FoxP3 expression, Kolben et al. reported that the presence of regulatory T cells (Tregs) in the endometrial cancer TIME is associated with a poorer outcome [32], which is consistent with our results. Tumor-associated immunosuppression is primarily driven by elevated TGF-β secretion or the direct infiltration of Tregs [33]. In contrast, the FoxP3/CD8 ratio did not reach statistical significance in the present study, so caution should be taken when interpreting the clinical value of this ratio. One possible explanation for this lack of significance is that stromal FoxP3-positive cells and CD8-positive cells may show spatial and functional heterogeneity, and simple numerical integration of these variables may not adequately capture the biological balance between immunosuppression and anti-tumor immunity. In addition, the functional state of CD8-positive cells cannot be determined by cell counts alone.

Meanwhile, the stromal CD8-positive cell count was significantly higher in the recurrence group than in the no-recurrence group in this study, whereas the intraepithelial CD8-positive cell count was not significantly associated with recurrence. This counterintuitive finding should also be interpreted with caution because elevated stromal CD8-positive cell counts in the absence of data on potential functional exhaustion may represent unexplained heterogeneity rather than effective anti-tumor immunity. Indeed, in a previous report, the association between the prognosis of breast cancer patients and T-cell exhaustion was reported [34]. Further investigation of exhausted T cells in those with recurrence, compared with those without it, is thus needed to validate these associations.

Notably, in sensitivity analyses restricted to open surgery cases, the recurrence group still demonstrated substantially higher marginal and intratumoral CD68-positive cell counts and higher CXCL12 expression than the no-recurrence group, indicating that the strongest TIME- and EMT-related signals were preserved even after excluding laparoscopic cases. In contrast, the differences in stromal CD8, FoxP3, and SPARC were directionally consistent with those in the overall cohort but less statistically robust, which is likely attributable to the reduced number of recurrence events in the restricted analysis.

We encountered 541 patients with low-grade, stage IA EEC from 1992 to 2022, for whom the recurrence rate was 3.7%. It was previously reported that the recurrence rate was 7.65% in grade 1, stage IA EEC [35], which is reasonably consistent with our result. In addition, six cases (30%) in the recurrence group in our study resulted in tumor-related mortality. Recurrence of low-grade, stage IA EEC is rare, but when it occurs, tumor-related mortality is often the outcome. Of the 20 cases of recurrence in this study, all but one were classified into the low-risk group, with the remaining case classified into the high-intermediate-risk group in the current staging for endometrial cancer [36]; adjuvant therapy is not recommended in the low-risk group.

Regarding the association between CTCs and adjuvant chemotherapy, in colorectal cancers, Yu et al. reported that the average OS of the CTC-negative group without adjuvant chemotherapy was significantly longer than that of the CTC-positive group also without adjuvant chemotherapy [37].

In this study, the recurrence group demonstrated significantly higher numbers of FoxP3-positive and CD68-positive cells than the no-recurrence group. In addition, tumor cell expression of CXCL12 and SPARC was significantly higher in the recurrence group than in the no-recurrence group. Taken together with the fact that most recurrences occurred at lung or lymph nodes, not locally, these results may reflect a tumor microenvironment permissive of tumor cell dissemination. However, no direct evidence of the presence of CTCs is presented here.

Here, 18 cases in the recurrence group showed no LVSI, although Maltoni et al. also reported that, even among postoperative CTC-positive cases, no LVSI occurred [38]. Accordingly, given the present data, the value of assessing CTCs postoperatively should be validated in future work, rather than presented as a clinical recommendation.

This study has various important limitations. First, the number of recurrence events was small, which limits statistical power and the stability of association estimates. Second, this was a retrospective sampled comparative analysis from a larger institutional cohort, and the no-recurrence group was selected from eligible cases with sufficient tissue and evaluable histological areas for immunohistochemical assessment; therefore, the possibility of selection bias cannot be ruled out. Third, multiple immunohistochemical variables were examined in an exploratory manner, and despite applying FDR adjustment for PFS, the limited number of events means that the multivariable Cox models are still at risk of overfitting; accordingly, the hazard ratio estimates should be regarded as hypothesis-generating and require validation in independent cohorts. Moreover, the long study period may have introduced heterogeneity in surgical practice, pathological processing, and follow-up. Finally, the surgical approach also differed between groups (laparoscopic surgery in 6/20 recurrent cases vs. 0/60 non-recurrent cases), which may have contributed to heterogeneity. Although sensitivity analyses restricted to open surgery cases suggested that key associations involving CD68-positive TAMs and CXCL12 were preserved, residual confounding by surgical factors cannot be completely ruled out. A previous report discussed potential effects of surgical manipulation and CO₂ pneumoperitoneum in minimally invasive surgery on tumor cell dissemination [39], although direct evidence regarding CTC dynamics in endometrial cancer remains limited. Therefore, our findings should be interpreted as hypothesis-generating and require validation in independent cohorts.

Conclusions

In this study, recurrence in low-grade, stage IA EEC was associated with TIME- and EMT-related factors. These findings may reflect a tumor microenvironment permissive of tumor cell dissemination; however, they do not provide direct evidence of CTCs. Therefore, these findings should be interpreted as exploratory and hypothesis-generating, and prospective studies with direct CTC assessment and independent validation are warranted.

Abbreviations

TCGA

The Cancer Genome Atlas

WHO

World Health Organization

CTC

circulating tumor cell

TIME

tumor immune microenvironment

EMT

epithelial-mesenchymal transition

TAM

tumor-associated macrophage

EEC

endometrial endometrioid carcinoma

HE

hematoxylin and eosin

FIGO

International Federation of Gynecology and Obstetrics

LVSI

lymphovascular space invasion

PFS

progression-free survival

OS

overall survival

CXCL12

C-X-C motif chemokine ligand 12

SPARC

Secreted Protein Acidic and Rich in Cysteine

DAB

3,3′-diaminobenzidine

HPF

high-power field

FDR

false discovery rate

HR

hazard ratio

CI

confidence interval

Treg

regulatory T cell

Authors’ contributions

**Kazuhisa Hachisuga: ** Conceptualization; data curation; formal analysis; investigation; methodology; resources; validation; visualization; writing – original draft. **Yoshihiro Katayama: ** Resources. **Yusuke Inomata: ** Resources. Minoru Kawakami: Resources. **Shoji Maenohara: ** Resources. **Hiroshi Yagi: ** Formal analysis. **Masafumi Yasunaga: ** Formal analysis. **Ichiro Onoyama: ** Formal analysis. **Kazuo Asanoma: ** Formal analysis. **Hideaki Yahata: ** Formal analysis. **Yoshinao Oda** : Conceptualization; project administration; supervision; writing – review and editing. **Kiyoko Kato: ** Conceptualization; project administration; supervision; writing – review and editing.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sector.

Conflict of interest statement.

The authors have no competing interests to declare that are relevant to the content of this article.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

All methods were carried out in accordance with relevant guidelines and regulations. This retrospective study using archived clinical data and tissue samples was approved by the institutional review board of Kyushu University (project number: 24101-00). The need for informed consent was waived, and an opt-out approach was used in accordance with institutional regulations.

Competing interests

The authors declare no competing interests.