CXCR2 Levels Correlate with Immune Infiltration and a Better Prognosis of Triple-Negative Breast Cancers

Simple Summary Tumor microenvironment is critical for cancer progression. The role of the chemokine receptors in breast cancers is still under investigation. The aim of this study was to focus on a retrospective cohort of triple-negative breast cancers (TNBCs) and analyze the involvement of CXCR2 and its link with immune infiltration and immune checkpoint markers. High densities of CXCR2-positive cells were associated with high-grade tumors. Higher quantities of CXCR2-positive cells were correlated with elevated density of tumor-infiltrating lymphocytes (TILs), CD8+ cytotoxic lymphocytes, expression of PD-L1 by tumor and stromal cells and of PD-1 by stromal cells. In univariate analysis, low levels of CXCR2 were correlated with poor OS and RFS. In multivariate analysis, low levels of CXCR2 were associated with poor OS. Overall, our data highlight the potential beneficial association of high levels of CXCR2 with a subgroup of TNBC patients characterized by a better prognosis. Abstract Chemokines and their receptors are key players in breast cancer progression and outcome. Previous studies have shown that the chemokine receptor CXCR2 was expressed at higher levels by cells of the tumor microenvironment in triple-negative breast cancers (TNBCs). The aim of this study was to focus our attention on a retrospective cohort of 290 TNBC cases and analyze the involvement of CXCR2, CD11b (a marker of granulocytes) and CD66b (a marker of neutrophils) and their link with immune infiltration and immune checkpoint markers. We report that high densities of CXCR2-, CD11b- and CD66b-positive cells were associated with high-grade tumors. Moreover, molecular apocrine TNBCs, defined here as tumors that express both AR and FOXA1 biomarkers, exhibited low levels of CXCR2 and CD11b. High CXCR2 and CD11b levels were correlated with elevated density of tumor-infiltrating lymphocytes (TILs), CD8+ cytotoxic lymphocytes, expression of PD-L1 by tumor and stromal cells and of PD-1 by stromal cells. On the other hand, CD66b levels were associated only with CD8+, stromal PD-L1 and PD-1 expression. In univariate analysis, low levels of CXCR2 were correlated with poor OS and RFS. In multivariate analysis, low levels of CXCR2 were associated with poor OS. Finally, in TNBC treated with adjuvant chemotherapy, CXCR2 density was associated with longer RFS. Overall, our data highlight the potential beneficial association of high levels of CXCR2 with a subgroup of TNBC patients characterized by a better prognosis.


Introduction
Four main subtypes of breast cancer have been identified, corresponding to triplenegative breast cancers (TNBCs, negative for estrogen receptor (ER), progesterone receptor (PR) and Her2), Her2-positive breast cancers and two types of luminal breast cancers (luminal A and B, positive for ER) [1]. TNBCs, which represent about 15% of all breast cancers, are characterized by poor clinical outcomes, with shorter relapse-free survival (RFS), overall survival (OS) and higher metastasis rate [1][2][3]. TNBCs have been further subclassified into different groups based on definitions that might differ between studies [4][5][6][7]. Even if the classification of TNBCs is still subject of debate, a consensus would be a division into six subtypes displaying unique gene expression profiles, including 2 basal-like (BL1 and BL2), an immunomodulatory (IM), a mesenchymal (M), a mesenchymal stem-like (MSL), and a luminal androgen receptor (LAR) subtype [5,8].
The tumor microenvironment (TME) is now considered as a key component of the tumor and includes a variety of cells (in particular B and T lymphocytes, neutrophils, macrophages, natural killer (NK) cells, dendritic cells, endothelial cells, cancer associated fibroblasts (CAFs)) but also extracellular components (cytokines, chemokines, hormones, growth factors, extracellular matrix) that surround and interact with tumor cells [9]. Tumor microenvironment can not only modulate the growth of the primary tumor but also the metastatic process [10]. Chemokines, which are soluble factors secreted by many types of cells, belong to a large superfamily and act through G-protein coupled receptors [11]. Chemokines are involved both in homeostatic functions and in inflammatory response and serve as mediators between cells [12]. Thus, the TME is involved in the inflammatory state of the tumor, which involves in particular a "chemokine storm" favoring or inhibiting tumor progression and metastasis [13][14][15][16].
CXCR2 has been shown in mouse to control a variety of functions including wound healing [24], resistance to infections [25], reproduction under microbiota influence [26], senescence [27], myelin repair [28] and metabolism [29]. The potential role of CXCR2 has been described in murine preclinical models of CXCR2 gene deletion or targeting of CXCR2 protein with antibodies. While a majority of these studies suggests that CXCR2 favors or has no effect on primary tumor growth [30,31], one study showed the contrary [32]. In human cancers, the function of CXCR2 has been scarcely investigated so far. We have shown recently that CXCR2 was mainly expressed by neutrophils in breast tumors and was present at higher levels in TNBC compared to luminal or Her2-positive breast cancers [33]. Interestingly, despite its high expression in high-grade tumors, CXCR2 predicted a lower risk of relapse [33].
The aim of this study was to use an independent, larger, cohort of TNBC patients to confirm these results and investigate whether CXCR2 levels were associated with clinicopathological variables, immune infiltration, checkpoint proteins and clinical outcome.

Objectives
The primary objective was to assess of the impact of CXCR2+ cell infiltration on clinical outcome of patients with TNBCs. Secondary objectives were the evaluation of the association between CXCR2+ cell infiltration and clinicopathological variables, including markers of the tumor microenvironment.

Patients and Tumor Samples
The study was reviewed and approved by the Montpellier Cancer Institute Institutional Review Board (ID number ICM-CORT-2019-27). Samples were selected from a prospective collection of breast cancers surgically removed at the Institut du Cancer de Montpellier (France) between 2002 and 2010 and annotated in a dedicated prospective database (tumor biobank number BB-0033-00059). Only tumors from patients who did not have a history of another invasive cancer in the previous 5 years, without known germline BRCA mutation and with unifocal, unilateral, untreated, non-metastatic breast cancers and with ER and PR negativity (defined as <10% of tumor cells stained by IHC) [34] and HER2 negativity (defined as IHC 0/1+ or 2+ and negative fluorescent/chromogenic in situ hybridization), were selected for this study. Each tumor sample was sampled as two cores of 1mm diameter and arrayed in six tissue microarray (TMA) blocks using a manual arraying instrument (Tissue Arrayer 1, Beecher Instruments, Sun Prairie, WI, USA).
All patients were treated according to our institutional guidelines [35] and gave their written informed consent for the use of their specimens for research purposes. As part of the study evaluating the prognostic impact of biological markers, this manuscript followed the REMARK guidelines. The clinicopathological characteristics of the study cohort are summarized in Table 1. The features of the patients enrolled in our study are consistent with the classical characteristics of TNBCs. Median age was 57.7 years (range: 28.5-89.1). Most of the tumors were T1/T2 (46.2 % and 47.6% respectively), node-negative (65.2%), ductal carcinomas (82.9%) with high histological grade (76.9%). According to our institutional guidelines [35], 75.4% of the patients received adjuvant chemotherapy. Basal-like phenotype, based on the IHC detection of CK 5/6 and/or EGFR, was observed in 64.9% of the tumors. Molecular apocrine phenotype, defined as TNBC tumors displaying both AR and FoxA1 expression, was observed in 41.8% of the studied population.
The CD11b-, CD66b-and CXCR2-immunostained TMA sections were digitalized using a NanoZoomer slide scanner system (Hamamatsu Photonics, Hamamatsu City, Shizuoka Pref. Japan) with a 20× objective. Quantification of IHC staining was performed by image analysis using HistoLab ® Image Analysis Software (Microvision, Evry, France) as previously described [18] by delineating the invasive component in each sample core as a region of interest (ROI). Density of immune-reactive cells in the ROI, recorded as the number of positive cells per cm 2 , was calculated for each tumor sample and finally matched to clinicopathological data. Missing TMA cores, those with folded tissue or demonstrating significant artifacts were not scored.

Evaluation of TILs
For each sample, TIL density was assessed on corresponding HES-stained digitalized TMA sections, following the guidelines issued by the International TIL Working Group [39]. Stromal TILs were reported as the percentage of area occupied by TILs relative to the whole stroma area.

Statistical Analysis
The categorical variables were described by the number of observations and the frequency of each modality. They were compared with Pearson's chi-square test or Fisher's exact test when the theoretical numbers were less than 5. Continuous variables were described by the median, the minimum and the maximum. OS represents the time between the date of surgery and the date of death, regardless of the cause. Details of patients alive or lost to follow-up were recorded at the last documented visit. The RFS represents the time between the date of surgery and the date of recurrence. Deaths prior to recurrence were recorded on the date of death. The most recent details for patients alive without recurrence and those lost to follow-up were also recorded. The Kaplan-Meier method was used to estimate median survival rates and times. Survival distributions were compared with the log-rank test. The hazard ratios (HRs) and their 95% confidence intervals (95% CIs) were estimated using a Cox proportional hazard model. Statistical analyses were performed with STATA 16.0 (StatCorp, College Station, TX, USA).

Correlations of CD11b, CD66b and CXCR2 Expression with Clinicopathological Features
In order to study the role of CXCR2 and its possible link with immune infiltration in TNBC patients, we analyzed the densities of CD11b (a marker of granulocytes), CD66b (a marker of neutrophils) and CXCR2-positive cells on a cohort of 290 samples from patients with TNBCs that we had previously characterized for both their phenotypic characteristics and their immune microenvironment [37,38]. The main characteristics of the studied population are described in Table 1 and are consistent with the classical characteristics of TNBCs (for further details, refer to Materials and Methods). This section may be divided by subheadings. It should provide a concise and precise description of the experimental results, their interpretation, as well as the experimental conclusions that can be drawn.
Densities of CD11b-, CD66b-and CXCR2-positive cells ( Table 2) were not correlated to the size of the tumors, their lymph node status or basal status, but a high density of these cells was significantly associated with high-grade tumors (p < 0.001 for the three populations). Conversely to CD11b, which was more frequently observed in young patients (p = 0.018), high density of CD66b was more often observed in older patients (p = 0.040); no significant difference of CXCR2-positive cell density was observed according to age. When looking at the types of TNBC, high CXCR2 levels were correlated with ductal tumors (p = 0.041). Moreover, tumors with molecular apocrine phenotype, i.e., expressing both AR and FOXA1 biomarkers, exhibited significantly lower density of intra-tumor CD11b-or CXCR2-positive cells (p = 0.001 and p = 0.006, respectively). There was no correlation of any of the three markers with the basal-like TNBC phenotype. Finally, high density of CD11b and low density of CD66b were more frequently associated with prescription of adjuvant chemotherapy (p = 0.041 and p= 0.019, respectively). Footnote: Basal-like phenotype was considered in the case of positive staining for cytokeratin 5/6 and/or EGFR (>10% of tumor cells stained in IHC); Molecular apocrine phenotype was defined in TNBC tumors that express both AR (androgen receptor) and FOXA1 (Forkhead box protein A1) biomarkers.

Survival Analyses
We next evaluated the prognostic value of all parameters previously quantified on the survival of the patients ( Table 4). The median follow-up of our cohort was 10 (Figure 2).   OS was significantly associated with younger age (p = 0.001). High expression of PD-L1 by tumor cells or stromal cells was associated with longer RFS (p = 0.034 and p = 0.028, respectively). Neither CD3, CD8, CD11b nor CD66b was associated with clinical outcome.
In multivariate analysis (Table 5), tumor size, nodal involvement, ductal histology, lack of adjuvant chemotherapy and low densities of TILs or of CXCR2+ cells were independent poor prognostic factors of OS. All of these factors except histology and CXCR2 were also associated with a shorter RFS.  OS was significantly associated with younger age (p = 0.001). High expression of PD-L1 by tumor cells or stromal cells was associated with longer RFS (p = 0.034 and p = 0.028, respectively). Neither CD3, CD8, CD11b nor CD66b was associated with clinical outcome.
In multivariate analysis (Table 5), tumor size, nodal involvement, ductal histology, lack of adjuvant chemotherapy and low densities of TILs or of CXCR2+ cells were independent poor prognostic factors of OS. All of these factors except histology and CXCR2 were also associated with a shorter RFS.  In addition, 14 cases in our population displayed a weak HR expression, in the 1% to 9% positivity range. We performed a separate multivariate analysis excluding these 14 cases (Table S1). The results show that high levels of CXCR2 continue to be significantly associated with a better OS (HR 0.6; 95% CI 0.38-0.97, p = 0.033). In addition, in this second analysis, high levels of CXCR2 were significantly associated with a better RFS (HR 0.58; 95% CI 0.34-0.97, p = 0.036), suggesting a greater impact of CXCR2 on clinical parameters in the subpopulation displaying no HR expression (Table S1).
To evaluate the specific predictive or prognostic values of CXCR2+ cell infiltration, we analyzed OS and RFS in patients with TNBCs treated or not with adjuvant chemotherapy, respectively. In patients who had received chemotherapy, there was no effect of high density of CXCR2+ cells on OS (p = 0.194, Figure 3A), but CXCR2 density was significantly associated with longer RFS (p = 0.034, Figure 3B). In untreated patients (n = 70), we observed a trend between low CXCR2+ cell density and shorter OS (p = 0.070, Figure 3C) and RFS (p = 0.088, Figure 3D).  (Table S1).
To evaluate the specific predictive or prognostic values of CXCR2+ cell infiltratio we analyzed OS and RFS in patients with TNBCs treated or not with adjuvant chemothe apy, respectively. In patients who had received chemotherapy, there was no effect of hig density of CXCR2+ cells on OS (p = 0.194, Figure 3A), but CXCR2 density was significant associated with longer RFS (p = 0.034, Figure 3B). In untreated patients (n = 70), we o served a trend between low CXCR2+ cell density and shorter OS (p = 0.070, Figure 3C) an RFS (p = 0.088, Figure 3D).

Discussion
The aim of this study was to focus on the potential roles of CXCR2, CD11b and CD66 in TNBCs, with a particular attention to immune cell infiltration and immune checkpoi markers. We first observed that high levels of the three markers were correlated to hig grade TNBCs, confirming our previous study on a smaller cohort of breast cancer patien [33]. CXCR2 stained immune granulocytic cells and not cancer cells. This includes in pa

Discussion
The aim of this study was to focus on the potential roles of CXCR2, CD11b and CD66b in TNBCs, with a particular attention to immune cell infiltration and immune checkpoint markers. We first observed that high levels of the three markers were correlated to high-grade TNBCs, confirming our previous study on a smaller cohort of breast cancer patients [33]. CXCR2 stained immune granulocytic cells and not cancer cells. This includes in particular neutrophils but possibly other types of cells as well, as there was a better correlation of CXCR2 with CD11b than with CD66b. Interestingly, low expression of CXCR2 and CD11b was also associated with molecular apocrine-like TNBCs, expressing both AR and FoxA1, which represent a subgroup of TNBCs with a worse prognosis characterized by late relapses [7].
TNBCs are considered as an immunogenic breast cancer subtype and show a higher degree of infiltration by lymphocytes compared to luminal breast tumors [41], even though this infiltration tends to be lower in more advanced stages of the disease compared to early stages [42]. This could be the result of their higher production of multiple cytokines [16,17,19] or their genomic instability leading to high rates of mutations and to the production of neoantigens, which in turn increase their immunogenicity [43]. We observed that high levels of CXCR2 and CD11b were correlated with higher densities of TILs, CD3+ and CD8+ cytotoxic lymphocytes. In TNBCs, high TIL levels are generally correlated with a better prognosis [44]. In early TNBC patients treated with anthracycline-based adjuvant chemotherapy, the levels of TILs are positively correlated with a better outcome [45]. Similar results were obtained for untreated early TNBCs, showing that higher TIL densities correlated with better survival [46,47]. Similarly, high quantities of cytotoxic CD8+ TILs are positively associated with a longer survival of early TNBC patients [48] and better response to neoadjuvant chemotherapy [49]. Moreover, high levels of total T cells (based on CD3 expression) correlate with a better response to adjuvant chemotherapy of early breast cancer patients [50]. It is important to point out that more advanced stages of breast cancers display reduced immune infiltration [42]. These TILs may lose some of their immunosuppressive properties [51].
We next analyzed the presence of PD-L1-and PD-1-positive cells in our cohort. Immune checkpoints are upregulated to inhibit the cytotoxic response and prevent an excessive immune reaction [52]. PD-1 is normally expressed at the surface of T cells, B cells, natural killer cells and myeloid cells [53] whereas PD-L1 can be expressed both by tumor cells and cells from the TME, including activated T-cells, macrophages [54] and CAFs [55]. We report that high levels of CXCR2 and CD11b were correlated to higher levels of PD-L1 expression on both tumor and stromal cells as well PD-1 by stromal cells. CD66b was only correlated to PD-L1 and PD-1 expression by stromal cells. The correlation of CXCR2 levels with immune checkpoint markers might represent an attempt of the body to counteract tumor progression.
We have also analyzed the prognostic value of CXCR2 in our cohort. In univariate analysis, there was a correlation between poor OS and RFS and low levels of CXCR2. Multivariate analysis confirmed that low levels of CXCR2 were associated with poor OS. Moreover, when focusing on TNBCs treated with adjuvant chemotherapy, we observed that CXCR2 density was associated with longer RFS. A potential pitfall of our study could be linked to the choice of a European definition of TNBCs, considering a <10% negativity threshold for the determination of the hormone receptor status. To circumvent this potential issue, we analyzed our cohort using the ASCO/CAP threshold of 0 to define HR negativity [34,56] and found a greater impact of high levels of CXCR2 expression on survival parameters. This is in accordance with published evidence showing that these two groups of tumors are closely related in term of clinical behavior and mutational spectrum [57]. All this suggests that high expression of CXCR2 could favor a better outcome of TNBCs, which is also in agreement with our previous results obtained for all types of breast cancers, not just TNBCs [33]. This is consistent with the fact that CXCR2 is expressed in fewer cells in molecular apocrine-like tumors, which have a worse long term prognosis than other subtypes of TNBCs, and that CXCR2 levels correlate with TILs, CD8+ and CD3+, which are associated with a better prognosis. On the other hand, the presence of a high number of CXCR2 cells in tumors characterized by high expression of PD-L1 on both tumor and stromal cells, but also in tumors exhibiting high level of PD-1, could represent an attempt of the immune cells to counteract the immune escape that is promoted by the PD1/PD-L1 pathway.

Conclusions
In TNBCs, the expression of the chemokine receptor CXCR2 is associated with higher immune infiltration and a more favorable outcome, despite its correlation with the presence of PD-L1TC, PD-L1sc or PD-1sc cells. Its expression varies between TNBC subtypes, with a lower expression rate in molecular apocrine-like tumors, a subtype of TNBC characterized by later recurrence risk and worse prognosis. Whether these results indicate a direct role of CXCR2 in the control of the effects of immune checkpoint proteins, such as PD-L1 and PD-1, will require further investigation.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/cancers13102328/s1, Table S1: Multivariate analysis of survivals with exclusion of the 14 patients with ER and/or PR expression between1 and 9%.  Informed Consent Statement: All patients were informed before surgery that their surgical specimens might be used for research purposes.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

Acknowledgments:
We acknowledge the Clinical Resources Center of the Montpellier Cancer Institute (CRB-ICM. no. BB-033-00059) and the imaging facility MRI, a member of the national infrastructure France-BioImaging infrastructure supported by the French National Research Agency (ANR-10-INBS-04. "Investments for the future").

Conflicts of Interest:
The authors declare that they have no conflict of interest to disclose.