To identify potential biomarkers, microarray analyses of ESCC tissue samples were first performed using 3D-Gene miRNA oligo chips (Toray Industries, Inc.), each containing 2,565 genes. miRNA expression in tumor tissues was collected from patients with ESCC who underwent curative resection at our institution, comparing miRNA expression between non-por and por ESCC within the same specimens. RNAs were labeled using the 3D-Gene miRNA labeling kit (Toray Industries, Inc.). Fluorescent signals were scanned with a 3D-Gene scanner 3000 (Toray Industries, Inc.) and analyzed using the 3D-Gene Extraction software (Toray Industries, Inc.). Next, our analysis was supplemented with a GEO dataset (GSE43732) (8), a publicly accessible repository that includes miRNA expression profiles from primary lesions of 119 patients with ESCC. miRNA expression was compared between patients with por and non-por ESCC and additionally performed overall survival (OS) analysis according to tumor differentiation status.

The study population comprised a retrospective cohort of patients with consecutive ESCC who underwent esophagectomy at the Department of Digestive Surgery, University of Yamanashi (Chuo, Japan) between February 2008 and January 2023. To eliminate any confounding effects of chemotherapy or radiotherapy, 169 patients who received neoadjuvant chemotherapy (n=144) or neoadjuvant chemoradiotherapy (n=25) were excluded. An additional 11 cases were excluded due to non-curative resection. Furthermore, 30 cases unsuitable for RNA extraction were removed because the tissue samples were too shallow for adequate RNA retrieval. As a result, a final cohort of 61 patients with ESCC were eligible for the present study (Fig. S1). Clinicopathological characteristics of patients (including sex and age distribution) are listed in Table I.

Tumor specimens and resected lymph nodes obtained at the time of surgery were immediately fixed in 10% neutral-buffered formalin at room temperature for 24 h and embedded in paraffin after fixation (FFPE). The clinical and pathological tumor stages of esophageal cancer were classified according to the UICC TNM staging, 8th Edition (3).

Patients were followed up with physical examinations, blood tests, and computed tomography every 4 months for at least 1 year, and every 6 months thereafter. The present study was approved (approval no. 2723) by the Ethics Committee of Yamanashi University (Chuo, Japan) and conducted in accordance with the ethical standards of the Declaration of Helsinki and its amendments.

Serial sections were obtained from FFPE tissue blocks for IHC staining and RNA extraction. FFPE sections (5 µm) were used for IHC. After deparaffinization with xylene, dehydration using a graded ethanol series, and antigen retrieval at 120°C for 15 min endogenous peroxidase activity was inhibited (Peroxidase-Blocking Solution; cat. no. S2023; Dako; Agilent Technologies, Inc.), and non-specific binding was blocked (Protein Block Serum-Free; cat. no. X0909; Dako; Agilent Technologies, Inc). The sections were then incubated with a polyclonal FKBP5 antibody (1:50; cat. no. 14155-1; Proteintech Group, Inc.) at room temperature for 1.5 h. Subsequently, slides were treated with an anti-rabbit secondary antibody (EnVision⁺ HRP; cat. no. K4003; Dako; Agilent Technologies, Inc.) at room temperature for 60 min. DAB working solution was used to stain positive cells. The results were observed and recorded using optical microscopy (BZ-X810; Keyence Corporation).

Each slide was scored according to the H-score, which is calculated by multiplying two parameters: the proportion of positive cells (1–100%) and the intensity of staining (1=weak; 2=moderate; 3=strong; range, 0–300).

Total proteins were obtained from the harvested KYSE70 cells, and protein concentrations were determined using the bicinchoninic acid (BCA) protein assay kit (Thermo Fisher Scientific, Inc.). Cell lysate proteins (20 µg/lane) were separated by 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and then transferred onto polyvinylidene fluoride membranes. After blocking with StartingBlock Blocking Buffer (Thermo Fisher Scientific, Inc.) at room temperature for 1 h, the membranes were incubated overnight at 4°C with the following antibodies: E-cadherin (1:1,000; cat. no. 20874-1-AP; Proteintech Group, Inc.), Vimentin (1:2,000; cat. no. 10366-1-AP; Proteintech Group, Inc.), phosphorylated (p-)PI3K (1:1,000; cat. no. 17366; Cell Signaling Technology, Inc.), PI3K (1:1,000; cat. no. 11889; Cell Signaling Technology, Inc.), p-AKT (1:1,000; cat. no. 4060; Cell Signaling Technology, Inc.), AKT (1:1,000; cat. no. 4691; Cell Signaling Technology, Inc.), GAPDH (1:50,000; cat. no. 60004-1-lg; Proteintech Group, Inc.). Then, the membranes were incubated with goat anti-rabbit or anti-mouse IgG conjugated to horseradish peroxidase (1:10,000; cat. nos. ab6721 and ab6789; Abcam) buffer for 1 h at room temperature. The membranes then were probed with the indicated antibodies, and proteins were detected by a western blot HRP Substrate (Takara Bio, Inc.). Densitometric analysis of the protein bands was performed using FIJI (ImageJ; version 1.54q, National Institutes of Health).

FFPE tissue samples were cut into 10-µm-thick sections, and total RNA was extracted using the miRNeasy FFPE kit (Qiagen GmbH) and NucleoSpin miRNA (Takara Bio, Inc.), following the manufacturer's protocols. A NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Inc.) was used to measure the total RNA concentration.

Reverse transcription was performed using the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific, Inc.) and the TaqMan MicroRNA Reverse Transcription Kit (Thermo Fisher Scientific, Inc.). Total RNA and miRNA levels were quantified by RT-qPCR following standard procedures.

RT-qPCR conditions were as follows: Initial denaturation at 95°C for 10 min, followed by 40 cycles of 95°C for 15 sec and 60°C for 60 sec. Total RNA levels were normalized to the endogenous control gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH), while miRNA levels were normalized to RNU6B.

The following primers were used for the TaqMan assay (Thermo Fisher Scientific, Inc.): human hsa-miR-100-5p (cat. no. 78224_mir; 5′-AACCCGUAGAUCCGAACUUGUG-3′), hsa-miR-203a-3p (cat. no. 478316_mir; 5′-GUGAAAUGUUUAGGACCACUAG-3′), FKBP5 (cat. no. Hs01561006_m1), GAPDH (cat. no. Hs02786624_g1) and RNU6B (cat. no. 001093; 5′-CGCAAGGATGACACGCAAATTCGTGAAGCGTTCCATATTTTT-3′). ΔCq values for all RNAs were calculated relative to the control genes GAPDH and RNU6B. Relative mRNA and miRNA expression levels were evaluated using the 2^−ΔΔCq method (9).

The poorly differentiated ESCC cell lines [KYSE70 (JCRB0190), KYSE110 (JCRB1064), and TE5 (RCB1949)] and the well- to moderately differentiated ESCC cell lines [TE4 (RCB2097), TE8 (RCB2098), and TE11 (RCB2100)] were used in the present study. KYSE70 and KYSE110 were obtained from the Japanese Collection of Research Bioresources (JCRB) Cell Bank, while TE series cell lines were obtained from the RIKEN BRC Cell Bank in 2024. All cell lines were confirmed to be free of mycoplasma contamination, as previously described (10–12).

KYSE70, TE4, TE8 and TE11 were cultured in Roswell Park Memorial Institute (RPMI)-1640 medium (Nacalai Tesque, Inc.) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin (Sigma-Aldrich; Merck KGaA). The other ESCC cell lines were cultured in Ham's F-12 medium (FUJIFILM Wako Pure Chemical Corporation) supplemented with 2% FBS and penicillin-streptomycin (Sigma-Aldrich, Merck KGaA). All cells were maintained in a humidified incubator at 37°C with 5% carbon dioxide.

The miRNA mimics for miR-100-5p (mirVana miRNA mimic; cat. no. MC10188; 5′-AACCCGUAGAUCCGAACUUGUG-3′), miR-203a-3p (mirVana miRNA mimic; cat. no. MC10152; 5′-GUGAAAUGUUUAGGACCACUAG-3′, and the miRNA mimic negative control (mirVana miRNA Mimic Negative Control #1; cat. no. 4464058) were purchased from Thermo Fisher Scientific, Inc. Transfections were performed using Lipofectamine 2000 (Thermo Fisher Scientific, Inc.) according to the manufacturer's protocol.

The final concentration for miR-100-5p and miR-203a-3p mimics, as well as the negative control, was 10 nM. For double transfection, both mimics were used at 10 nM each. Total RNA and protein were collected 48 h post-transfection. Transfection efficiency was assessed by quantifying miRNAs expression levels using RT-qPCR.

The siRNA for FKBP5 (Silencer Select siRNA; cat. no. s5215) (sense sequence: GAGAAAGGCUUGUAUAGGAtt) and the siRNA negative control (Silencer Select Negative Control #1; cat. no. 4390843) were purchased from Thermo Fisher Scientific, Inc. The transfection protocol and concentration were the same as those used for the mimics. Knockdown efficiency of FKBP5 was confirmed by western blotting.

The cell migration and invasion assays for KYSE70 cells were conducted using Transwell chambers. For invasion assays, the chambers were pre-coated with Matrigel (Corning, Inc.), while migration assays were performed without it. Cells (1×10⁵) suspended in serum-free medium were placed in the upper chamber, while 10% FBS was added to the lower chamber. After a 24-h incubation, cells were stained using the Differential Quik III Stain Kit (Polysciences, Inc.). Following washing and drying, images were captured and examined under the microscope.

To determine whether the effects of the double-transfection of miR-100-5p and miR-203a-3p mimics on cell migration and invasion were mediated through FKBP5 suppression, functional rescue experiments were performed by co-transfecting the double-transfection condition with FKBP5 siRNA. KYSE70 cells were transfected with the double-transfection of miR-100-5p and miR-203a-3p mimics (10 nM each), with or without FKBP5 siRNA (10 nM). After 48 h, migration and invasion assays were conducted as aforementioned. The number of migrated and invaded cells in the double-transfection + FKBP5 siRNA group was compared with that in the double-transfection group to assess whether additional FKBP5 knockdown further modified the miRNA-induced phenotype.

To assess cell proliferation, KYSE70 cells (1×10⁴/well) were passaged in 96-well plates in RPMI-1640 culture medium containing penicillin-streptomycin, supplemented with 10% FBS. Cell proliferation was measured by the OD values using Cell Counting Kit-8 (CCK-8; (Dojindo Laboratories, Inc.). A total of 100 µl of CCK-8 solution was added to each well, and the plates were incubated for 2 h at 37°C in a CO₂ incubator. The absorbance was measured at 450 nm on a microplate reader (Spectra Max ABS Plus, Molecular Devices, LLC).

Apoptosis was assessed in KYSE70 cells following FKBP5 knockdown. The siRNA transfection procedure and concentrations were identical to those aformentioned. A total of 48 h after transfection, apoptotic activity was evaluated by both flow cytometry and western blotting.

For flow cytometric analysis, cells were collected, washed twice with cold PBS, and resuspended in binding buffer. Annexin V-FITC and propidium iodide (PI) (Nacalai Tesque, Inc.) were added according to the manufacturer's instructions, and samples were analyzed using a flow cytometer (BD FACSCelesta; BD Biosciences) with BD FACSDiva software (v8.0.1.1; BD Biosciences). The proportions of viable, early apoptotic, and late apoptotic cells were quantified.

For western blotting, protein expression of apoptosis-related markers was examined using antibodies against Caspase-3 (1:1,000; cat. no. 14220; Cell Signalling Technology, Inc.) and Bcl-2 (1:1,000; cat. no. ab692; Abcam).

RNA-seq transcriptome data of ESCC was obtained from The Cancer Genome Atlas database (https://portal.gdc.cancer.gov/). Potential miRNA target genes were predicted using TargetScan Human 8.0 (https://www.targetscan.org/vert_80/). To investigate the biological functions of FKBP5, Pearson's correlation coefficients were first calculated between FKBP5 expression and all other genes across ESCC samples. Genes with |Pearson r|>0.3 and P<0.001 were considered significantly correlated with FKBP5 and subjected to enrichment analysis. Gene Ontology (GO) Biological Process and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment were performed using the R package clusterProfiler (version 4.8; http://bioconductor.org/packages/clusterProfiler), with multiple-testing correction by Benjamini-Hochberg and significance thresholds of adjusted P<0.05 and q-value <0.05.

To assess pathway-level associations, single-sample Gene Set Enrichment Analysis (GSEA) was applied using the GSVA package (version 1.42; http://bioconductor.org/packages/GSVA). Each tumor received a score for the following two gene sets from MSigDB: HALLMARK_EPITHELIAL_MESENCHYMAL_TRANSITION, KEGG_PI3K_AKT_SIGNALING_PATHWAY (hsa04151).

Finally, pre-ranked GSEA was conducted with cluster Profiler: GSEA by ranking all genes in descending order of Pearson's correlation with FKBP5. Only the EMT (Hallmark) and PI3K-Akt (KEGG) gene sets were evaluated for enrichment, using an FDR cutoff of 0.25. All analyses were performed in R (version 4.4.0; R Foundation for Statistical Computing).

Statistical analysis was performed using GraphPad Prism version 10 (GraphPad Software, Inc.; Dotmatics). Unless otherwise stated, quantitative values were expressed as the mean ± standard deviation (SD). Statistical significance was assessed using the Mann-Whitney U test, Fisher's exact test, and one-way analysis of variance (ANOVA) for each time point, followed by Tukey's post hoc test. OS was calculated from the date of surgery to either death or the last follow-up. Relapse-free survival (RFS) was measured from the date of surgery to recurrence, death, or the last follow-up. Survival curves were generated using the Kaplan-Meier method and compared using the log-rank test. P<0.05 was considered to indicate a statistically significant difference.

First, miRNA candidates were identified using a miRNA array-based approach in combination with the GEO datasets. The expression levels of each miRNA were compared between por and non-por ESCC. miR-203a-3p was selected, which exhibited a four-fold difference in expression during microarray analysis, ensuring that the signal intensity of each miRNA was adequate (Fig. 1A). Additionally, miR-100-5p was selected, which exhibited a two-fold difference in expression during microarray analysis, supported by the GEO datasets (Fig. 1A and B).

Furthermore, analysis of a publicly available GEO dataset revealed a consistent trend toward poorer prognosis in patients with por ESCC (Fig. S2). Although this difference was not statistically significant, por ESCC is generally considered more aggressive, and this trend may have biological relevance. Consequently, miR-100-5p and miR-203a-3p were chosen for further analyses in the present study.

Next, the clinical significance of miR-100-5p and miR-203a-3p was examined in patients with ESCC. The expression levels of miR-100-5p and miR-203a-3p were significantly lower in patients with por ESCC than in patients with non-por ESCC (P=0.02 and P=0.04, respectively) (Fig. 2A). Based on these results, a receiver operating characteristic curve was constructed according to histological differentiation and divided the expression levels of miR-100-5p and miR-203a-3p into high-expression and low-expression groups for further analysis.

In the low miR-100-5p group, tumor differentiation was poorer, and the depth of invasion was greater compared with the high miR-100-5p group (P<0.001 and P=0.012, respectively). Similarly, in the low miR-203a-3p group, tumor differentiation was poorer, the depth of invasion was greater, and lymph node metastasis occurred more frequently compared with the high miR-203a-3p group (P=0.01, P=0.008, and P=0.03, respectively) (Table I).

As demonstrated in Fig. 2B, OS and RFS rates were significantly worse in the low miR-100-5p group compared with the high miR-100-5p group (P=0.02 and P=0.04, respectively). Additionally, the OS rate was significantly worse in the low miR-203a-3p group than in the high miR-203a-3p group, while the RFS rate was not statistically significant (P=0.05 and P=0.12, respectively; Fig. 2C).

Furthermore, the function of miR-100-5p and miR-203a-3p was examined in ESCC cell lines. First, their expression levels were measured in ESCC cell lines and it was found that miR-100-5p and miR-203a-3p were significantly lower in poorly differentiated ESCC cell lines compared with well- and moderately differentiated cell lines (Fig. 3A). This expression pattern was consistent with our clinical findings that patients with high expression of these miRNAs had higher degree of differentiation and more favorable prognosis.

To investigate the functional roles of these miRNAs in ESCC progression, cell migration and invasion assays were conducted using miRNA mimics in KYSE70 cells. Transfection efficiency for each mimic is shown in Fig. S3A. Cell migration was significantly inhibited by the miR-203a-3p mimic alone and by the combination of miR-100-5p and miR-203a-3p mimics compared with the negative control mimic (Fig. 3B and C). This result indicates that miR-203a-3p plays a suppressive role in the migratory ability of ESCC cells. Similarly, cell invasion was significantly suppressed by the miR-100-5p mimic and by the double transfection, indicating that miR-100-5p mainly contributes to the inhibition of invasive potential (Fig. 3D and E). These results suggested that both miRNAs cooperatively suppress ESCC progression through inhibition of migration and invasion.

FKBP5 was identified as a putative downstream target of both miR-100-5p and miR-203a-3p using TargetScan Human 8.0 (Fig. S4). To clarify the functional relevance of FKBP5 in ESCC, comprehensive bioinformatics analyses were first performed. GO and KEGG enrichment of FKBP5-associated genes revealed significant overrepresentation of pathways involved in cell cycle control, nuclear division and extracellular matrix organization (Fig. 4A and B). Moreover, Pearson correlation (Fig. 5A-D) and pre-ranked GSEA (Fig. 5E and F) demonstrated robust positive associations between FKBP5 expression and both epithelial-mesenchymal transition (EMT) and PI3K/AKT signalling pathway in ESCC.

Experimentally, transfection of miR-100-5p and miR-203a-3p mimics significantly decreased both FKBP5 transcript and protein levels, with the combined transfection of both miRNAs producing the most pronounced reduction (Fig. 6A and B). These results suggested that FKBP5 is a downstream target of both miRNAs and that the two miRNAs may act cooperatively to suppress its expression. To further confirm the functional involvement of FKBP5 in the miRNA-mediated suppression of ESCC cell motility, rescue experiments were performed using the double-transfection of miR-100-5p and miR-203a-3p mimics with or without FKBP5 siRNA. Adding FKBP5 siRNA to the double-transfection did not produce additional reductions in migration or invasion compared with the double-transfection alone, indicating that the effects of the mimics are largely mediated by FKBP5 suppression (Fig. S5). In clinical ESCC samples, the expression levels of miR-100-5p/miR-203a-3p and FKBP5 protein exhibited a negative correlation trend; however, this association did not reach statistical significance (Fig. S6).

Functional assays further demonstrated that FKBP5 knockdown, whose efficiency was confirmed by western blotting (Fig. S3B), significantly impaired ESCC cell migration, invasion, and increased cytotoxicity compared with negative controls (Fig. 7A-E), suggesting that FKBP5 promotes malignant phenotypes in ESCC cells. Additionally, apoptosis assays were performed. Annexin V/PI flow cytometry showed no significant differences between FKBP5 siRNA and controls (Fig. S7A). Western blotting revealed slight changes in Caspase-3/Cleaved Caspase-3 and a marked reduction in Bcl-2 (Fig. S7B). These results suggested that FKBP5 knockdown alone did not provide definitive evidence for apoptosis involvement.

Furthermore, the effect of FKBP5 on EMT and PI3K/AKT pathway activity was investigated by performing western blotting for key EMT markers and phosphorylated signaling proteins. FKBP5 knockdown led to upregulation of the epithelial marker E-cadherin and downregulation of the mesenchymal marker vimentin (Fig. 8A). In addition, knockdown of FKBP5 reduced the levels of p-PI3K and p-AKT, indicating suppression of the PI3K/AKT signalling pathway (Fig. 8B-D). These findings indicated that FKBP5 facilitates EMT and activates the PI3K/AKT pathway in ESCC cells, and that its suppression may help reverse the mesenchymal phenotype and inhibit oncogenic signaling associated with invasion and metastasis.

Finally, the prognostic significance of FKBP5 expression was evaluated in patients with ESCC using IHC. Representative IHC images are shown in Fig. 9A. Based on the median H-score, patients were categorized into high- and low-FKBP5 expression groups. Survival analysis revealed that patients in the high FKBP5 expression group exhibited significantly poorer OS and RFS compared with those in the low expression group (Fig. 9B). These findings suggested that elevated FKBP5 expression is associated with adverse clinical outcomes in ESCC and may serve as a potential prognostic biomarker.