CXCR4-CXCL12-CXCR7 and PD-1/PD-L1 in Pancreatic Cancer: CXCL12 Predicts Survival of Radically Resected Patients

Pancreatic ductal adenocarcinoma (PDAC) is currently the most deadly cancer. Although characterized by 5–20% of neoplastic cells in the highly fibrotic stroma, immunotherapy is not a valid option in PDAC treatment. As CXCR4-CXCL12 regulates tumor invasion and T-cell access and PD-1/PD-L1 controls immune tolerance, 76 PDACs were evaluated for CXCR4-CXCL12-CXCR7 and PD-1/PD-L1 in the epithelial and stromal component. Neoplastic CXCR4 and CXCL12 discriminated PDACs for recurrence-free survival (RFS), while CXCL12 and CXCR7 discriminated patients for cancer-specific survival (CSS). Interestingly, among patients with radical resection (R0), high tumor CXCR4 clustered patients with worse RFS, high CXCL12 identified poor prognostic patients for both RFS and CSS, while stromal lymphocytic-monocytic PD-L1 associated with improved RFS and CSS. PD-1 was only sporadically expressed (<1%) in focal lymphocyte infiltrate and does not impact prognosis. In multivariate analysis, tumoral CXCL12, perineural invasion, and AJCC lymph node status were independent prognostic factors for RFS; tumoral CXCL12, AJCC Stage, and vascular invasion were independent prognostic factors for CSS. CXCL12’s poor prognostic meaning was confirmed in an additional perspective-independent 13 fine-needle aspiration cytology advanced stage-PDACs. Thus, CXCR4-CXCL12 evaluation in PDAC identifies prognostic categories and could orient therapeutic approaches.

Three-micrometer sections were cut from formalin-fixed paraffin-embedded (FFPE) tissue blocks. Sections were stained with hematoxylin and eosin for adequate tumor representation.

Statistical Analysis
Association between CXCR4, CXCL12, CXCR7, and PD-1/PD-L1 expression cancer cells and patients' clinic-pathological features were analyzed applying chi-square and Mann-Whitney U tests for categorical and continuous variables, respectively. Recurrencefree survival (RFS) was set as the time from diagnosis to the recurrence or last follow-up. Cancer-specific survival (CSS) was set as the time from diagnosis to death for cancer. Kaplan-Meier method was performed to estimate the survival curve and logrank test for statistical comparison. Cox proportional hazards regression was utilized to test the effect of multiple dichotomous covariates (risk factors) on RFS and CSS; The backward method for variable selection was applied in the final model with a conventional p-value threshold of p < 0.05 (enter variable), and p > 0.1 (remove variable). p-values less than 0.05 were considered significant. All statistical tests and graphs were conducted using SPSS version 20 (IBM Corp., Armonk, NY, USA) and MedCalc 12.3.0 (MedCalc. Software; Mariakerke, Belgium).

Validation Cohort and Tissues
A population of twenty patients diagnosed with advanced PDAC, evaluated through endoscopic-US-guided pancreatic fine-needle aspiration cytology (FNAC), was considered a validation cohort. The samples were collected and analyzed at Pathology Unit, Department of Mental and Physical Health and Preventive Medicine, University of Campania "Luigi Vanvitelli", Naples, Italy, from January 2020 to October 2021. Three-micrometer sections were stained for CXCL12 immunocytochemical evaluation following hematoxylin/eosin reviewing and histological confirmation. All cases were retrospectively reviewed, and the cytology on direct smears showed moderately to richly cellular samples composed of three-dimensional aggregates of epithelial cells with severe cytological atypia. The cyto-block (CB) sections were also re-examined, with a cytological morphology substantially similar to that observed on direct smears; when necessary, immunocytochemical (ICC) analysis was performed, which showed positivity in the neoplastic cells for CK19 and CA . In all 20 cases selected, the overall morphological analysis was consistent with the diagnosis of pancreatic adenocarcinoma (Category VI sec. Papanicolaou System). ImmunoCytoChemistry (ICC) was carried out on CB sections and stained for CXCL12 as previously described.

PD-L1 Is Predominantly Expressed by TME in PDAC
PD-1 and its natural ligand PD-L1 were evaluated in 76 PDAC. While PD-1 was not detectable in cancer cells, it was only sporadically expressed (<1%) in focal lymphocyte infiltrate (data not shown). PD-L1 was predominantly identified at the membrane of cancer cells in 29/76 (38.2%) PDAC. PD-L1 mean expression was 3.4 ± 7.1 ( Table 2)

PDAC Cancer Cells Express CXCL12
CXCL12 was predominantly identified in the membrane in 20 out of 76 (26.3%) tumors. The CXCL12 mean expression was 3.7 ± 7.20% (Table 2). Representative CXCL12 expression is reported in Figure 3A
Median follow-up was 25

Prognostic Significance of Epithelial and Stroma CXCL12, CXCR4, CXCR7 and PD-L1 in PDAC
The CXCL12 and CXCR4 tumor expression positively correlated (Spearman's r = 0.42, p = 0.004). The relation among the evaluated markers and clinic pathologic characteristics revealed that CXCR4-positive tumor and CXCR4-positive tumor-infiltrating inflammatory cells were associated with vascular invasion (p = 0.0421 and 0.045) as expected [32] ( Table 3). Tumor-infiltrating inflammatory cells CXCR7 positive were mainly detected in smokers (p = 0.002) ( Table 3), as previously reported [33].       Of note, high CXCR4 predicted short RFS in R0 subgroup (radical resected, of neoplastic cells within 1 mm from the lesion) (n = 48) (RFS: 11.05 months v months p = 0.0071) ( Figure 6A). CXCL12 expression predicted RFS (positive CXCL While the expression of the CXCR4-CXCL12-CXCR7 axis or PD-L1 in stromal cells was not prognostic in the whole cohort (Supplementary Figure S1), we found a significant association of high stromal lymphocytic-monocytic PD-L1 expression with improved RFS (p = 0.016) and CSS (0.047) in R0 patients, (Figure 7). While the expression of the CXCR4-CXCL12-CXCR7 axis or PD-L1 in stromal cells was not prognostic in the whole cohort (Supplementary Figure S1), we found a significant association of high stromal lymphocytic-monocytic PD-L1 expression with improved RFS (p = 0.016) and CSS (0.047) in R0 patients, (Figure 7).

Tumoral CXCL12 Is an Independent Poor Prognostic Factor at the Multivariate Analysis
In multivariate analysis (Table 5), the pathological AJCC Stage 8th ed. (p = 0.049), perineural invasion (p = 0.0107), and CXCL12 (p = 0.00002) but not CXCR4, predicted shorter poor RFS (Table 5 left). The pathological AJCC Stage 8th ed. (p = 0.00007), vascular invasion (p = 0.0067), and CXCL12 (p= 0.0062) but not CXCR7 nor CXCR4 expression predicted shorter poor CSS (Table 5 right); thus, CXCR7 or CXCR4 were not retained in the final multivariate models of RFS and CSS.    To confirm the CXCL12 prognostic role, CXCL12 expression was evaluated in an independent cohort of FNAC (n = 20). The median age was 72 ± 10 years (range 50-85), with 8 (40%) male and 12 (60%) female, and 17 (85%) of patients ≥60 years old. At diagnosis, the majority of patients (85%) had an advanced stage. CXCL12 was detected in 70% (14/20) PDACs predominantly at the membrane and cytoplasm of cancer cells ( Figure 8) and in some surrounding monocytes/macrophages stromal cells. A subgroup of 8 out of 20 (40.0%) showed a consistent membranous cytoplasmic distribution of CXCL12 (mean expression 86.9 ± 11.9%) ( Figure 8A,B). CXCL12 was negative/low in 12 out of 20 (60%); absent in 6 out of 20 FNAC PDAC (30.0%) and low 6 out of 20 showed low broad distribution (36.66 ± 12.11%) ( Figure 8C,D). High CXCL12 was also detected in tumor emboli ( Figure 8E) and perineural infiltration ( Figure 8G). Follow-up was available for 13 patients. A total of 7 patients died (53.8%), showing a median overall survival of 6 months. The patients with a high sharp, consistent distribution of CXCL12 expression showed short CSS with a median survival of 3 months, while low/negative CXCL12 expressing displayed short CSS with a median survival of 12 Follow-up was available for 13 patients. A total of 7 patients died (53.8%), showing a median overall survival of 6 months. The patients with a high sharp, consistent distribution of CXCL12 expression showed short CSS with a median survival of 3 months, while low/negative CXCL12 expressing displayed short CSS with a median survival of 12 months (p = 0.029) (Figure 8).

Discussion
The role of the CXCR4-CXCL12-CXCR7 axis and PD-1/PDL-1 was addressed in tumor/stromal cells in 76 consecutive single-center patients undergone surgery between January 2014/March 2015 and followed for 5 years until January 2021. We demonstrated that the entire axis CXCR4-CXCL12-CXCR7 was overexpressed in PDAC neoplastic cells as compared to TME cells. Moreover, CXCL12 significantly correlated with poor prognosis in an unrelated cohort of 20 FNAC from PDAC patients. CXCR4 was expressed by acinar cells and islets of Langerhans, surrounded by extensive dense fibrotic stroma or desmoplasia [34]. Trefoil Factor 2, an exocrine gene early expressed during mouse embryonic development, induced CXCR4 activation in acinar cells promoting cell proliferation and preventing apoptosis [34]. CXCR4 was also detected in lymphocytic-monocytic infiltration and endothelial cells [35] associated with vascular invasion [36]. Thus, it is possible to hypothesize that CXCR4 imaging might detect inflammatory and vascular involvement in PDAC and surgical/clinical settings [37,38]. CXCR7 is upregulated in tumor and stromal PDAC. Schwann cells, CXCR7 positive, were reported in pancreatic nerves and CXCR7-neutralizing antibody (9C4), but not CXCR4 inhibitor AMD3100, induced significant dose-dependent attenuation of transmigration toward PDAC cell lines [39]. Herein, CXCR7 expression was reported in nerve bundle, and CXCR7 expression of nerve bundle, in inflamed endothelial cells [40]. Of note, CXCR7 was significantly more expressed in the PDAC stroma of patients with a smoking habit as nicotine induces ERK-dependent IL-8-activating stromal nicotinic receptors [33,41] and IL-8 induces CXCR7 expression [42,43]. CXCL12 expression was observed mainly in tumor cells and, in some cases, in cancer-associated fibroblasts. Hypoxia induces CXCL12-CAF release and recruitment of CXCR4-positive immune-suppressive/supportive stromal cells [39,44]. Chronic pancreatitis was previously reported to be a risk factor for PDAC [45,46]. Long-standing inflammation increases cell turnover and stellate cell proliferation, promoting carcinogenesis [47,48]. Nevertheless, it was not prognostic in our cohort (p ≥ 0.05) as for NCCN 2022 guidelines. In our analysis, chronic pancreatitis was a "potential risk factor" (univariate analyses p-value 0.0365) not retained, for irrelevance or redundancy, in the multivariate analysis. As for PD-L1, early studies showed prognostic significance for overall survival with >5 or 10% cut-off [49,50]. Although evidence suggests that PD-L1 is an indicator of poor prognosis, different patient selections and variable thresholds (1-10%) do not allow comparison among the reports. Basile Tessier-Cloutier reported three values for PD-L1 tumor cell membrane staining (≥1%, >5% and >10%) on tissue microarray evaluating cancer-specific survival on 252 PDACs with 12 PDACs-PD-L1 overexpressing (>10%) displaying poor survival [51].
Karamitopoulou reported PD-L1 expression in about one-third of 349 samples. The authors concomitantly analyzed PD-L1, PD-1, CD3, CD4, CD8, FoxP3, and CD68, reporting that PD-L1, present with T-cell infiltration, improved overall survival [52]. Herein, a similar pattern, although not significant, was reported for RFS and OS. The possible discrepancy may be ascribed to patients' subgroup analysis (mismatch repair (MMR) proteins) and sample size. In agreement with the herein reported data, Diana et al. examined the prognostic value of PD-1 and PD-L1, together with CD8+ tumor-infiltrating lymphocytes (TILs) and FOXP3+ Tregs in 145 PDAC samples, describing PD-L1 expression not prognostic [53]. Wang et al. reported that PD-L1 expression failed to predict prognosis in 77 unselected PDACs, although a further subgroup analysis (high/low PD-L1 and tumor density) identified patients with worse overall survival [54]. As related to PD-1, rare PD-1-positive isolated T cells were revealed as expected by the low mutation burden in PDAC. Thus, we did not pursue more combined staining.
PD-L1 was recently reported in stromal cells in PDAC [55], and also, here, PD-L1 was reported on nerve bundle of desmoplastic stroma and affected the prognosis in a subgroup of R0 patients. Although previous studies reported a prognostic role of tumor PD-L1 expression, we could not observe it [55]. PDAC tumor microenvironment (TME) plays a central role in impairing the efficacy of immune checkpoint inhibition (ICI) [56]. Adding CXCR4 blockade to PD-1 targeted therapy increased tumor cell death concomitantly with lymphocyte expansion [56], thus ongoing trials are evaluating the efficacy of targeting CXCR4-CXCL12 axis in potentiating the immune checkpoint inhibitors-based therapy in pancreatic cancer (NCT03277209, NCT02907099 NCT04177810). This manuscript has several limitations. The retrospective nature and the patients number do not allow definitive conclusions. Although the patients were consecutively enrolled, they represent a selected group of patients for staging and fitness for surgery. A comparison with unresectable patients could represent the natural prosecution of the study. Nevertheless, this is the first time that the CXCR4-CXCL12-CXCR7 axis discriminates patient outcomes while PD-L1 expression both in cancer or stromal cells did not have significant predictive ability for survival; further, PDACs did not express PD-1. CXCL12 expression was an independent prognostic factor for RFS and CSS. Moreover, CXCL12 in an unrelated cohort of 20 FNAC from PDAC patients confirmed a significant correlation with poor prognosis. CXCL12 expression was an independent prognostic factor for RFS and CSS. Moreover, CXCL12 in an unrelated cohort of 20 FNAC from PDAC patients confirmed a significant correlation with poor prognosis. Notably, high CXCR4 and high CXCL12 identified high-risk patients among the radically resected patients, regarded as having good prognosis; thus, CXCR4 CXCL12 evaluation could be useful for R0 patient risk assessment and potentially allow to discriminate after a prospective validation of our results, those who could potentially better benefit from neoadjuvant chemotherapy among resectable stages and, conversely, identify those who can better benefit from upfront surgery. Herein high stromal lymphocytic-monocytic PD-L1 expression is associated with improved RFS in the R0 subgroup. It is hypothesized that the exhausted immune population is still exerting control over tumor growth, as recently reported [52,57], suggesting a potential benefit for PD-1/PD-L1 antagonists in this subgroup. To the best of our knowledge, this is the first study that systematically investigates on tumor vs. stromal prognostic role of CXCR4-CXCL12-CXCR7 and PD-L1/PD-1 in PDAC. Targeting the axis CXCR4-CXCL12-CXCR7 thus represents a suitable tool to improve diagnosis and obtain prognostic and/or predictive data for optimizing combined therapy in PDAC patients.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/cells11213340/s1, Figure S1: Stromal CXCR4, CXCL12, CXCR7 and PD L1 expression not predicts poor survival in unselected PDAC patient.  Institutional Review Board Statement: The study was conducted in accordance with the Institutional guideline (or Ethics Committee) of Pederzoli Hospital, Peschiera del Garda, 37019 (Verona), Italy.
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study. We used descriptive statistics to summarize the characteristics of the entire data set. We did not include individual patients' identifying information.