Low plasma PD-L1 levels, early tumor onset and absence of peritoneal carcinomatosis improve prognosis of women with advanced high-grade serous ovarian cancer

A prospective analysis was performed on a cohort of one hundred patients with advanced-stage HGSOC recruited at the two Sicilian hospital centers: “Sicilian Regional Center for the Prevention, Diagnosis and Treatment of Rare and Heredo-Familial Tumors” of the Section of Medical Oncology of University Hospital Policlinico “P. Giaccone” of Palermo (Italy), and Department of Gynecologic Oncology of the Hospital ARNAS Civico “Di Cristina Benfratelli” of Palermo (Italy). The information concerning the personal history of tumor, age of cancer diagnosis, International Federation of Gynecology and Obstetrics (FIGO) stages, tumor histological subtype and tumor grading were anonymously recorded and coded for all enrolled patients who had previously signed and provided a written informed consent (Table S1). The study (Protocol “TIC-OC v.1.1”) was approved by ethical committee (Comitato Etico Palermo 1) of the University-affiliated Hospital A.O.U.P. (Azienda Ospedaliera Universitaria Policlinico) “P. Giaccone” of Palermo (Italy) [18].

Between May 2018 and July 2021 we prospectively collected blood samples from one hundred women with a confirmed histological diagnosis of advanced HGSOC (stage IIIB-IV) at baseline, before surgery (surgical staging or cytoreductive surgery as clinically indicated) and starting first-line chemotherapy treatment with Carboplatin AUC (area under the curve) 5 and Paclitaxel (175 mg/m2) based on the current therapeutic strategies. In summary, patients were considered eligible for this study based on the previously reported inclusion criteria [18]: i) Histologically confirmed and well documented diagnosis of advanced HGSOC; ii) Women older than 18 years; iii) Availability of peripheral blood from affected patients for the plasma isolation; iv) Therapeutic treatment-naïve patients at the moment of blood sampling. Patients with Eastern Cooperative Oncology Group (ECOG) Performance Status (PS) ≥ 3 were excluded from the study.

In order to confirm the obtained results, another independent cohort of 24 advanced HGSOC patients (validation cohort) recruited at the Section of Medical Oncology of University Hospital Policlinico “P. Giaccone” of Palermo (Italy) was used [18].

The peripheral blood samples from untreated patients with advanced HGSOC were collected at baseline, processed for plasma isolation and subsequently stored as previously reported [6, 7, 19].

The plasma sPD-L1, sPD-1, sBTN3A1, pan-sBTN3As, sBTN2A1, and sBTLA concentrations were determined by means of specific enzyme-linked immunosorbent assays (ELISAs). Because other commercially available assays showed some discrepancies, we used specific ELISAs, produced by the company DYNABIO S.A. (Parc de Luminy, Marseille, France), according to the previously reported specifications [18, 20, 21]. Table S2 shows all information regarding the characteristics of six ELISA tests, whose protocol has been previously described [19, 20].

An analysis by ROC curves was performed to determine the optimal concentration cut-offs for each soluble form of immune checkpoints and other examined patients characteristics (age at diagnosis and Body Mass Index), in order to discriminate HGSOC patients based on long (≥ 30 months) versus short PFS (< 30 months). The Kaplan–Meier method and log-rank test were used to calculate patient PFS. Univariate and multivariate Cox proportional hazard regression models were built to identify significant prognostic factors for PFS [21].

MedCalc software v.18.2.1 for Windows (MedCalc Software, Ostend, Belgium) and GraphPad Prism software v. 9.0.0 (GraphPad Software, San Diego, CA) were used to generate and graphically represent data [21]. P values < 0.05 have been considered statistically significant.

The plasma concentrations of sPD-L1, sPD-1, sBTN3A1, pan-sBTN3As, sBTN2A1, and sBTLA were measured in peripheral blood from one hundred advanced HGSOC patients, prior to surgery and starting first-line chemotherapy treatment, using specific ELISA tests.

The receiver operating characteristic (ROC) analysis was carried out to determine for each soluble biomarker the optimal concentration threshold (Youden index associated criterion) able to discriminate advanced HGSOC patients based on long (≥ 30 months) versus short PFS (< 30 months). At the same time, a further ROC analysis was performed to establish the optimal cut-offs of two different parameters: age at diagnosis and BMI. The ROC curve analysis showed that the optimal concentration threshold was 0.42 ng/ml for sPD-L1 (AUC = 0.71, P = 0.01), 2.48 ng/ml for sPD-1 (AUC = 0.60, P = 0.04), 4.75 ng/ml for sBTN3A1 (AUC = 0.64, P = 0.01), 13.06 ng/ml for pan-sBTN3As (AUC = 0.65, P = 0.008), 5.59 ng/ml for sBTN2A1 (AUC = 0.64, P = 0.02), and 2.78 ng/ml for sBTLA (AUC = 0.62, P = 0.02). Furthermore, the best cut-offs for age at diagnosis and BMI were 60 years (AUC = 0.67, P = 0.002) and 25 (AUC = 0.62, P = 0.01), respectively (Figure S1). Interestingly, these thresholds calculated through ROC analysis were close to median concentration values determined for each examined circulating immune checkpoint. Indeed, median concentration values were 0.64 ng/ml for sPD-L1 (range 0.13 to 2.64 ng/ml), 1.80 ng/ml for sPD-1 (range 0 to 6.82 ng/ml), 6.22 ng/ml for sBTN3A1 (range 0 to 40.0 ng/ml), 14.60 ng/ml for pan-sBTN3As (range 0 to 40.0 ng/ml), 6.35 ng/ml for sBTN2A1 (range 3.53 to 10.0 ng/ml), and 3.04 ng/ml for sBTLA (range 0 to 28.61 ng/ml). Also cut-offs previously obtained by ROC analysis for age at diagnosis and BMI were close to their median values: 61 years for age at diagnosis (range 27 to 79 years) and 24 for BMI (range 16 to 56). Afterwards, the plasma levels of each molecule, ages at diagnosis and BMIs were graphically represented, in order to discriminate and divide the advanced HGSOC patients into 2 groups at long versus short PFS based on each investigated factor (Fig. 1). The red dotted lines were used to indicate the concentration threshold of each circulating biomarker as well as cut-offs for age at diagnosis and BMI, previously obtained by ROC analysis. All analyzed factors showed high predictive power. Interestingly, most of advanced HGSOC patients with PFS ≥ 30 months showed lower plasma levels (under specific threshold) for each soluble biomarker (sPD-1, sPD-L1, sBTN3A1, pan-sBTN3As, sBTN2A1, and sBTLA), whereas patients with PFS < 30 months predominantly exhibited higher circulating levels of these molecules. In addition, advanced HGSOC women with PFS less than 30 months were over 60 years of age and had a BMI > 25.

Because the clinical impact of circulating immune checkpoints as predictive biomarkers of clinical outcome has yet to be defined in advanced HGSOC patients, a Kaplan–Meier survival analysis was performed to understand the potential prognostic value of circulating sPD-L1, sPD-1, sBTN3A1, pan-sBTN3As, sBTN2A1 and sBTLA concentrations in advanced HGSOC patients, suggesting that their plasma expression levels could be useful, in the future, to predict patient survival. Therefore, we discriminated advanced HGSOC women based on low and high plasma levels for each tested biomarker, using the thresholds previously determined by ROC analysis. Then, the correlation between PFS and circulating levels of the soluble biomarkers was plotted using Kaplan–Meier curves (Fig. 2a-f). For each tested biomarker, patients with plasma levels above and below the specific threshold showed statistically significant differences in PFS. Plasma concentration cut-offs associated with poor prognosis and shorter PFS were defined for sPD-L1 (> 0.42 ng/mL), sPD-1 (> 2.48 ng/mL), sBTN3A1 (> 4.75 ng/mL), pan-sBTN3As (> 13.06 ng/mL), sBTN2A1 (> 5.59 ng/mL) and sBTLA (> 2.78 ng/mL) (Fig. 2a-f). Conversely, advanced HGSOC women with plasma concentrations below the indicated thresholds showed a median PFS which was from 6 to 16 months longer than that of subjects with levels above the concentration cut-offs. Specifically, patients with high baseline levels of the soluble proteins exhibited the following median PFS values compared to subjects with lower baseline levels: 24 versus 40 months for sPD-L1 (95% CI: 14 to 28 vs 30 to 55; log-rank p-value < 0.0001); 24 versus 30 months for sPD-1 (95% CI: 17 to 30 vs 24 to 36; log-rank p-value = 0.02); 21 versus 37 months for sBTN3A1 (95% CI: 15 to 26 vs 32 to 45; log-rank p-value < 0.0001); 21 versus 35 months for pan-sBTN3As (95% CI: 15 to 26 vs 30 to 45; log-rank p-value < 0.0001); 25 versus 32 months for sBTN2A1 (95% CI: 20 to 29 vs 24 to 41; log-rank p-value = 0.004); and 24 versus 32 months for sBTLA (95% CI: 17 to 28 vs 25 to 44; log-rank p-value = 0.0002). Interestingly, baseline plasma levels of sPD-L1, sBTN3A1 and pan-sBTN3As below their respective cut-offs showed a significant benefit in terms of median PFS (14–16 months), whereas a lower advantage in median PFS (6–8 months) was associated with baseline concentrations of sPD-1, sBTN2A1 and sBTLA below their specific thresholds. Therefore, the soluble forms of all proteins investigated in this study have been shown to be potential predictive biomarkers of survival in advanced HGSOC patients.

In addition, we also assessed the impact on PFS of age at diagnosis, baseline BMI and peritoneal carcinomatosis at onset by Kaplan–Meier analysis (Fig. 2g-i). Also for each of these factors, advanced HGSOC women with values above and below the specific thresholds showed statistically significant differences in PFS, suggesting their involvement in predicting prognosis. Poor clinical outcome and shorter PFS were associated with age at diagnosis over 60 years, BMI > 25, or presence of peritoneal carcinomatosis (Fig. 2g-i). In particular, median PFS values for patients with age at diagnosis > 60 years, BMI > 25, or peritoneal carcinomatosis compared to those with features below the respective thresholds were the following: 19 versus 32 months for age at diagnosis (95% CI: 13 to 25 vs 28 to 44; log-rank p-value < 0.0001); 22 versus 32 months for BMI (95% CI: 15 to 26 vs 25 to 37; log-rank p-value = 0.007); and 17 versus 38 months for peritoneal carcinomatosis (95% CI: 12 to 24 vs 31 to 45; log-rank p-value < 0.0001). Interestingly, the better survival advantage was observed in the absence of peritoneal carcinomatosis at the diagnosis.

Since previous analyses showed that each circulating immune checkpoint (sPD-L1, sPD-1, sBTN3A1, pan-sBTN3As, sBTN2A1, and sBTLA) and other clinic-pathological factors, such as age at diagnosis, baseline BMI and peritoneal carcinomatosis, have a high predictive power of clinical outcome in patients with advanced HGSOC, a multivariate analysis for PFS was performed in order to correlate these variables between them. The prognostic relevance of each biomarker/factor was assessed by an univariate survival analysis using the Cox regression model. The results of the multivariate analysis are reported in Table 1. Age at diagnosis, baseline BMI, peritoneal carcinomatosis at onset, and plasma levels of sPD-1, sPD-L1, sBTN3A1, pan-sBTN3As, sBTN2A1, and sBTLA were observed to be significantly associated with PFS in univariable analyses, whereas in the final multivariable Cox regression model, only the age at diagnosis > 60 years (HR: 1.70; 95% CI: 1.07 to 2.70; p = 0.024), presence of peritoneal carcinomatosis (HR: 1.87; 95% CI: 1.23 to 2.85; p = 0.003) and plasma levels of sPD-L1 > 0.42 ng/mL (HR: 2.23; 95% CI: 1.34 to 3.73; p = 0.002) were statistically significant. No statistically significant association was observed for other considered biomarkers/factors. Therefore, our analysis highlighted that age at diagnosis ≤ 60 years, absence of peritoneal carcinomatosis and expression levels of sPD-L1 ≤ 0.42 ng/mL were independent prognostic factors associated with a longer PFS in patients with advanced HGSOC.

Lastly, we used a further independent cohort of 24 peripheral blood samples from advanced HGSOC patients in order to confirm the previously observed correlations for each tested parameter. The previously determined concentration thresholds in leading cohort by the ROC curve approach were used to perform a Kaplan–Meier survival analysis. As expected, we observed a significant inverse correlation between PFS and high expression levels in plasma for each biomarker/factor (Fig. 3). This confirms and emphasizes our previous results obtained in the study leading cohort (Fig. 2). Therefore, advanced HGSOC women with plasma levels below the specified thresholds showed a gain in median PFS which was from 17 to 26 months longer than patients with levels above the concentration cut-offs.

Additionally, also for women from the validation cohort the impact on PFS of age at diagnosis, baseline BMI and peritoneal carcinomatosis at onset was evaluated by Kaplan–Meier curves. This further analysis showed similar results to those obtained in the leading cohort, suggesting that age at diagnosis > 60 years, BMI > 25, or presence of peritoneal carcinomatosis were associated with a shorter PFS. Finally, multivariate analysis performed on validation cohort confirmed that low plasma PD-L1 levels (≤ 0.42 ng/mL), age at diagnosis ≤ 60 years and absence of peritoneal carcinomatosis are favorable independent prognostic factors for PFS in women with advanced HGSOC (Table S3).