In the present retrospective analysis, 75 consecutive patients with uterine cervical squamous cell carcinoma (median age, 62 years; range, 32–87 years) who underwent concurrent platinum-based CRT or RT alone between August 2009 and November 2013 at the Gunma University Hospital (Maebashi, Japan) were enrolled. Herein, the abbreviation ‘RT’ represents both CRT/RT unless otherwise stated. Paired tumor specimens obtained from all patients before RT (pre-RT) and after 10 Gy RT (post-10 Gy) were used for pathological analysis and immunohistochemistry (IHC). Most specimens in both groups (~89% in the RT alone group and ~96% in the CRT group) were collected on days 8–9 (range, 5–11 days; Table SI). The number of days between application of 10 Gy RT and biopsy of the specimens was in the range of 0–4 days. Of these, 60/75 (80%) were performed on the same day or one day later (Table SII). Specimens were used to examine PD-L1 expression and stromal CD8⁺ TILs. Patient characteristics were recorded for tumor stage [stages IB, IIA, IIB, IIIA, IIIB and IVA according to the International Federation of Gynecology and Obstetrics [FIGO classification 2008 (20)], age and lymph node metastasis (Table I). The Institutional Review Board for clinical trials of Gunma University approved the study protocol.

All patients underwent definitive RT, involving a combination of EBRT and intracavitary brachytherapy (ICBT). EBRT was typically administered using a four-field technique and 10 MV X-ray. The most common EBRT dose and fractionation regimen was 50 Gy in 25 fractions. EBRT was performed with a combination of whole-pelvic irradiation (20-40 Gy) followed by 3-cm-wide central shielding irradiation. In patients with para-aortic lymph node metastases, pelvic irradiation fields were extended to include the gross metastatic region (2 Gy/fraction, 1 fraction/day, 5 fractions/week). For patients with lymph node metastases, additional boost irradiation of 6–8 Gy in 3–4 fractions was administered. ICBT was performed once per week, concurrently with the central shielding EBRT. EBRT was skipped on the day of ICBT administration. Three-dimensional IGBT was performed with a high-dose rate source using an ¹⁹²Ir remote afterloading system (microSelectron; Elekta Instrument AB) for all patients. The prescribed dose of each ICBT was determined to cover 90% of high-risk clinical target volume with a 6 Gy total dose. Interstitial brachytherapy was added along with ICBT for bulky and/or asymmetric tumors. ICBT was most commonly performed four times.

Concurrent chemotherapy was administered to 63.0% of the patients (48/75). Cisplatin was typically administered weekly at a dose of 40 mg/m² for patients treated with CRT.

After the completion of CRT/RT, patients were followed up every 1–3 months for the first two years and every 3–6 months for three subsequent years by radiation oncologists. During each follow-up examination, disease status was assessed in terms of locoregional control (LC) and progression-free survival (PFS). OS was defined as the term from initial RT until death as a result of any cause or the date of last follow up. LC was defined as no evidence of tumor regrowth or recurrence in the pelvic region. PFS was defined as no evidence of tumor regrowth or recurrence in the pelvic region or distant metastasis.

PD-L1 expression on tumor cells and density of stromal CD8⁺ TILs were evaluated with IHC using biopsy samples excised from the cervical cancer samples pre-RT and post-10 Gy. Biopsied samples were fixed in 10% buffered formalin for 24 h at room temperature, then dehydrated, degreased and paraffin-embedded. Paraffin sections (4-µm-thick) were dewaxed in xylene at room temperature and rehydrated using a graded ethanol series. Endogenous peroxidase activity was blocked with a 10 min incubation at room temperature in 0.3% hydrogen peroxide. Subsequently, PD-L1 sections were heated in 1 mmol/l ethylenediaminetetraacetic acid (pH 8.0) and CD8 sections were heated in 0.01 mol/l citric acid (pH 6.0) at 121°C for 10 min for antigen retrieval. After blocking by 10% goat normal serum for PD-L1 sections and 10% rabbit normal serum for CD8 sections with a 20 min incubation at room temperature, sections were incubated overnight with primary antibodies at 4°C. Next day, the sections were incubated with biotin-labeled secondary antibodies followed by peroxidase-labeled streptavidin (Histofine; SAB-PO (rabbit) cat. no. 424032 and SAB-PO (mouse) cat. no. 424022; Nichirei Biosciences Inc.; Nichirei Corporation), both for 20 min each at room temperature. Then, sections were incubated for 5 min with diaminobenzidine at room temperature for detecting the molecules. The following primary antibodies were used: Monoclonal anti-PD-L1 antibody (1:100; clone E1L3N; rabbit IgG; cat. no. 13684; Cell Signaling Technology, Inc.) and monoclonal anti-CD8 antibody (1:800; clone C8/144B; mouse IgG; cat. no. M7103; Dako; Agilent Technologies, Inc.).

All immunostaining images were obtained using a KEYENCE light microscope (BZ-9000; Keyence Corporation; magnification, ×400). PD-L1⁺ tumor cells and CD8⁺ cells were automatically counted using the visual inspection application software BZ-X analyzer JP ver.1.4.1 (Keyence Corporation). Percentages of tumor cells with cell surface staining for PD-L1 were recorded and presented as tumor proportion score (TPS). When the TPS was >1%, the sample was classified as PD-L1⁺. A minimum of 100 tumor cells was evaluated to calculate the TPS. The percentage of CD8⁺ TILs among total nucleated cells in the stromal compartments was defined as stromal CD8⁺ TIL density (21). To calculate stromal CD8⁺ TIL density, the CD8⁺ cells in hotspot areas of the specimens were counted; hotspot areas were defined as those containing the highest density of nucleated cells (22). The tumor area was manually excluded. The quality of tumor samples was carefully evaluated and validated independently by 2 co-authors (Department of Human Pathology, Gunma University Graduate School of Medicine) of the present study who were pathologists.

The OS, LC and PFS rates were calculated using the Kaplan-Meier method, and the log-rank test was used to confirm significant differences. Continuous data were compared with a non-parametric test (Wilcoxon signed-rank test for paired data). Receiver operating characteristic (ROC) curve analyses were performed to determine the optimal cut-off value of CD8⁺ TIL density. Univariate analyses were performed using Cox proportional hazards model. P<0.05 was considered to indicate a statistically significant difference. Bonferroni correction was used to adjust the familywise error rate from multiple comparisons for P-values. All statistical analyses were performed using SPSS 26.0 for Mac (IBM Corp.).

Median follow-up duration was 63 months (range, 8–120 months). The 5-year OS, LC and PFS rates for all patients were 80.0% (95% CI, 70.9–89.0%), 89.0% (95% CI, 81.6–96.2%) and 70.5% (95% CI, 60.1–80.9%), respectively (Fig. 1). Patient clinical characteristics are shown in Table I.

To investigate the alteration of PD-L1 expression post-10 Gy, the rate of PD-L1⁺ cells in the IHC samples pre-RT and post-10 Gy was analyzed (Fig. 2A). In pre-RT samples, 4 patients (5%) were positive and 71 patients (95%) were negative; in post-10 Gy samples, 39 patients (52%) were positive and 36 patients (48%) were negative (Fig. 2B). In all the patient groups, the percentage of PD-L1⁺ patients significantly increased from 5 to 52% between pre-RT and post-10 Gy (P<0.01; Fig. 2B). In patients treated with CRT, the percentage of PD-L1⁺ patients significantly increased from 6 to 48% between pre-RT and post-10 Gy (P<0.01; Fig. 2B). In patients treated with RT alone, the percentage of PD-L1⁺ patients significantly increased from 4 to 59% between pre-RT and post-10 Gy (P<0.01; Fig. 2B). Therefore, RT significantly upregulated PD-L1 expression in patients with uterine cervical squamous cell carcinoma. When compared with PD-L1 expression before RT, 38 patients (51%) exhibited an increase, two patients (3%) exhibited a decrease and 35 patients (47%) exhibited no change post-10 Gy RT (Fig. 2C).

Subsequently, the association between pre-RT and post-10 Gy PD-L1 expression and clinical outcome was examined. OS, LC and PFS did not display significant differences with positive and negative PD-L1 expression for both pre-RT and post-10 Gy samples (Fig. S1A-F). Furthermore, patients were categorized into three groups according to changes in PD-L1 expression. It was revealed that PD-L1 alteration exhibited no significant association with clinical outcome (Fig. S1G-I).

To investigate the alteration of CD8⁺ TILs in tumor tissues post-10 Gy, the density of stromal CD8⁺ TILs in the IHC samples both pre-RT and post-10 Gy was analyzed (Fig. 3A). IHC staining revealed that 31 patients (41.3%) exhibited an increase, while 44 patients (58.7%) exhibited a decrease in stromal CD8⁺ TIL density after 10 Gy (Fig. 3B). Stromal CD8⁺ TIL density was significantly decreased in post-10 Gy samples (median density, 23.1% pre-RT vs. 16.9% post-10 Gy; P=0.038; Fig. 3B and E). According to the treatment type, stromal CD8⁺ TIL density was significantly decreased after CRT (median density, 29.6% pre-RT vs. 13.4% post-10 Gy; P=0.001; Fig. 3C and E). By contrast, stromal CD8⁺ TIL density tended to increase after RT alone (median, 17.7% pre-RT vs. 24.0% post-10 Gy; P=0.400; Fig. 3D and E). No significant difference was observed in the density of pre-RT CD8⁺ TILs between the CRT and RT alone groups (data not shown). Furthermore, to evaluate the optimal cut-off value for CD8⁺ TIL density in the present study, a ROC curve analysis was performed. The association between clinical outcome and stromal CD8⁺ TIL density was assessed in all 75 patients. ROC curve analysis revealed that both pre-RT and post-10 Gy CD8⁺ TIL density exhibited significant prognostic value for predicting death and tumor recurrence, with optimal cut-off values of 32.2 and 16.9%, respectively (Fig. S2). Notably, patients with high stromal CD8⁺ TIL density (based on the ROC optimal cut-off values) both pre-RT and post-10 Gy exhibited significantly improved OS, LC and PFS compared with patients with low stromal CD8⁺ TIL density (Fig. 4A-F).

Furthermore, to analyze the effects of each factor on prognosis, univariate analyses using Cox proportional hazards model were performed (Tables II–IV). Univariate analyses revealed that FIGO stages III–IVA and lymph node metastasis in the pelvis were significantly associated with unfavorable OS and PFS (Tables II and IV). Univariate analyses revealed that para-aortic lymph node metastasis in the pelvis were significantly associated with unfavorable LC (Table III). Notably, high stromal CD8⁺ TILs in both pre-RT and post-10 Gy groups exhibited a significant association with improved OS and PFS (Tables II and IV).

Finally, patients were classified into four groups according to stromal CD8⁺ TIL density pre-RT and post-10 Gy [i) High (pre-RT CD8 density)-high (post-10 Gy CD8 density) defined as CD8 High-High; ii) high (pre-RT CD8 density)-low (post-10 Gy CD8 density) defined as CD8 high-low; ii) low (pre-RT CD8 density)-high (post-10 Gy CD8 density) defined as CD8 low-high and iv) low (pre-RT CD8 density)-low (post-10 Gy CD8 density) defined as CD8 low-low) (Fig. 5A). Notably, the group displaying low stromal CD8⁺ TIL density both pre-RT and post-10 Gy (CD8 Low-Low group) exhibited a significantly lower OS, LC and PFS compared with the other groups (Fig. 5B-D). Overall, the current data suggested that low stromal CD8⁺ TIL density either pre-RT or post-10 Gy may be a critical predictive biomarker for patients with cervical cancer treated with definitive RT.

To clarify the association between PD-L1 expression and CD8⁺ TIL density, the changes in PD-L1 expression and CD8⁺ TIL density in the 10-Gy irradiated samples were analyzed. It was revealed that 68% (21/31) of the increased CD8⁺ TIL density group also exhibited an increase in PD-L1 expression, whereas 61% (27/44) of the decreased CD8⁺ TIL density group exhibited no increase in PD-L1 expression (Table V). Thus, there was a significant association between an elevation in CD8⁺ TIL density and the induction of PD-L1 expression (P=0.013; Table V). Furthermore, the present study investigated whether changes in CD8⁺ TILs and PD-L1 expression in response to 10 Gy affected clinical outcomes in patients. However, a log-rank test did not indicate a significant association between changes in CD8⁺ TIL density and PD-L1 expression with prognosis (Fig. S3).