1. Introduction
The implication of the tumor microenvironment has gained growing interest from the past years and it is now well admitted that targeting tumor cells only could be insufficient to achieve optimal therapeutic responses. Tumor cells are surrounded by a stromal compartment including different types of extracellular matrix, signaling molecules and stromal cells, among which immune cells, fibroblasts, mesenchymal stem cells and endothelial cells will be found [
1,
2]. A variety of immune cells can be involved, including in particular, T lymphocytes, B lymphocytes, dendritic cells, macrophages, neutrophils and natural killers. These cells have the ability to establish contacts as well as distant interactions through the release of soluble factors. In turn, this will either favor or impair tumor growth, invasion and metastasis. One must emphasize that tumor cells are able to shape the stromal compartment though their physical or distant interactions with stromal cells, leading for instance to the acquisition of cancer associated fibroblast phenotype for fibroblasts or mesenchymal stem cells [
3,
4]. In a similar manner, recruitment of immune cells will also be affected by tumor cells and these cells may display pro- or anti-tumor properties. This has led to the identification of different types of tumor associated macrophages [
5], as well as of tumor associated neutrophils [
6,
7].
Among the soluble factors secreted both by epithelial tumor cells and stromal cells, chemokines have been identified as key players. The superfamily of chemokines can modulate basically all events of tumor progression including growth, survival, angiogenesis, invasion, extravasation and metastasis [
8,
9]. Chemokines are chemotactic cytokines and have been divided in four subtypes (CXC, CC, C or CX3C), based on the location of cysteines in N-terminus of the protein [
10]. They act through Gαi protein-coupled receptors leading to the activation of phosphatidylinositol-3 kinase/Akt, phospholipase C/protein kinase C and mitogen-activated protein kinase/p38, Ras/Erk and Janus kinase 2/signal transducer and activator of transcription (STAT3) and NF-κB pathways [
11].
We have previously shown that the ligands of human chemokine receptor CXCR2 (CXCL1, CXCl2, CXCL3, CXCL5, CXCL6, CXCL7 and CXCL8) were coregulated in breast cancers, presumably because of their common location in a narrow region of chromosome 4q [
12,
13]. CXCR2 ligands are expressed at higher levels in triple negative breast cancer (TNBC), compared to luminal and Her2 breast tumors or cells lines [
12,
13,
14]. Coexpression of CXCR2 ligands involves in particular nuclear factor-kappa B (NF-KB) and activator protein 1 (AP-1) pathways [
12,
15,
16].
CXCR2, which is expressed at high levels in neutrophils and at lower levels in endothelial cells, appears essential in the control of angiogenesis, through the binding of tripeptide glutamic acid-leucine-arginine (ELR)-motif containing chemokines, present in the N-terminal part of the protein [
17]. We hypothesized that CXCR2 expression could be also altered in breast cancer and could account for a differential recruitment in the tumor microenvironment.
The goal of this study was now to determine the expression and prognosis value of CXCR2 in breast cancer, which has been poorly studied.
3. Discussion
In this study, we aimed to understand the potential involvement of the chemokine receptor CXCR2 in the breast tumor microenvironment. We and others have previously shown that human triple-negative breast cancers were expressing higher levels of most CXCR2 ligands (CXCL1, 2, 3, 5, 6, 7 and 8) than luminal or Her2-positive tumors [
8,
9,
12,
13,
14,
19]. This is presumably due to a coregulation of these chemokines, which are present in a very narrow region of chromosome 4 (4q21) [
12]. Some of these chemokines such as CXCL1, CXCL5, CXCL6 and CXCL8 are produced at high levels by tumors cells, whereas some are also produced by blood cells, cancer associated fibroblasts or endothelial cells [
12,
14,
16]. Based on the fact that aggressive TNBC tumors express high levels of CXCR2 ligands, it was important to study the expression of CXCR2 in breast tumors.
Using a cohort of breast cancer patients for which we had obtained data on multiple cytokines, including CXCL8 as well as immune infiltration, we have first performed an extensive review of available CXCR2 Ab, as CXCR2 immunostaining remains controversial. We selected seven Ab recommended for IHC by the manufacturers. Not surprisingly, the different Ab tested gave various types of staining. However, we managed to identify three different Ab that could provide a similar pattern of staining and which were also in agreement with ISH experiments to detect CXCR2 mRNA. These selected Ab and ISH were all showing CXCR2-positive cells only in the stromal compartment or within vessels but never on tumor cells. Moreover, the immune-reactive cells had a granulocytic shape, looking as possible neutrophils. We also observed in IF experiments of costaining with CXCR2 and the neutrophil marker CD66b that most of the CXCR2-positive cells were neutrophils. This is in agreement with earlier findings showing that neutrophils are expressing the highest levels of CXCR2 [
20,
21]. However, we cannot exclude that other types of stromal cells could express CXCR2. Answering this question will require a complete study by itself. CXCR2 might be also expressed at lower levels on endothelial cells, but this remains controversial [
22,
23] and was not objectified in our series. We also previously showed at the mRNA level that breast tumor cells did not express significant levels of CXCR2 mRNA [
12]. Moreover, to confirm the specificity of the CXCR2 Ab (clone E-2) that we selected to perform the complete study, we tested this Ab on HEK-293 cells transfected with CXCR2 cDNA, further reinforcing that this Ab was adequate.
The pattern of CXCR2 expression in breast cancer remains a controversial question, presumably due to the use of a variety of Ab, with distinct specificities. A previous small study has shown that another CXCR2 Ab (no more commercially available) stained tumor cells in a series of only 37 patients [
24], in which no validation of the Ab used was presented. In the same line, Xu et al. [
25] used an Ab from Sigma Aldrich (c-6348), which stained also breast tumor cells. However, this Ab is not intended for use in IHC experiments but for flow cytometry and neutralizing studies. Thus, we did not select this Ab to be tested in the present study. Moreover, this study never shows any specificity controls such as positive or negative controls of IHC. In addition, IF experiments presented in the study from Xu et al. shows a clear cytoplasmic staining, which is in disagreement with the fact that CXCR2 is a transmembrane receptor. So, together, these data could explain the difference in terms of results with our study. Finally, a third study from Romero-Moreno et al. [
26] used another Ab (ab14935) mainly to stain mouse tissue. They report also a staining of bone and marrow cells from a human femoral head, but the nature of the cells expressing CXCR2 was not precise. One could speculate that it could be neutrophils as these cells are highly present among marrow cells.
We next compared the expression of CXCR2, CD11b and CD66b between normal breast tissue and breast cancers. We observed that CXCR2 was more expressed in breast tumors than in a normal breast, whereas CD11b was present at lower levels in breast tumors and CD66b was not different between normal and cancer tissues. Interestingly, if CXCR2 and CD66b should label the neutrophils, our results suggest that CXCR2 is an independent marker of neutrophils, with some neutrophils that do not express CXCR2. Higher expression of CXCR2 in tumor tissues compared to normal tissue has also been reported in the pancreas, lung or colon cancers [
27,
28,
29]. It is also interesting to note that CXCR4, another chemokine receptor, shows a higher expression in breast cancer tissue, whereas its ligand CXCL12 is downregulated in the same tissue, which is in contrast to the situation observed for CXCR2 [
30].
When looking in more details to breast cancer tissues, we show that CXCR2, CD11b and CD66b high levels were correlated to a high tumor grade and to TNBC tumors. To our knowledge, CD11b and CD66b have not been evaluated so far in primary breast tumors in terms of association with clinical parameters. It is interesting to note that CXCR2 expression parallels one of its ligands, which is also expressed at higher levels in TNBC tumors [
12,
13,
14], even if we did not observe a strict correlation between CXCL8 and CXCR2 in this study (data not shown). Interestingly, other chemokines or their receptors can be associated with ER. This is the case of CXCL12, which its expression is regulated by estrogens but can also activate ER activity [
31]. Moreover, CXCR4, one of the receptor of CXCL12, is a poor prognosis factor in breast cancer patients, independently of ER status, and its high expression has been shown to promote estrogen-independent tumor growth in vivo [
32].
We also reported that high staining for CD11b and CXCR2, but not CD66b were correlated to a higher infiltration of B and T lymphocytes. Whether this is a consequence of the mutual attraction of each type of cells, or that all these cells are infiltrating the tumor because of the presence of common chemoattractants remains to be addressed. We have previously reported that breast tumors with a high content of CCL2 or CCL4 had a higher infiltration of B and T lymphocytes [
13]. Moreover, one hypothesis for the increase in CXCR2 positive cells could be the higher release of CXCR2 ligands by other cells. In particular, we and others have shown that both CXCL1 and CXCL8, ligands of CXCR2, could be produced at higher levels by mesenchymal stem cells or fibroblasts, which have been “educated” by TNBC cells [
16,
33,
34]. Similar overexpression of CXCR2 ligands occurs also for breast cancer cells and could represent another source of chemokines production [
9,
12,
13,
14,
19].
In the search of potential molecular mechanisms accounting for a high expression of CXCR2, we were able to show that it was correlated to high expression of AP-1 transcription factors but not NF-KB. This is reminiscent to CXCL8, one of the ligand of CXCR2, which its expression is also correlated to AP-1 activity [
12,
13].
We also investigated the potential of CXCR2, CD11b and CD66b as prognostic factors. We showed in a univariate and multivariate analysis that high CXCR2 expression was associated with a better prognosis in terms of TTR, whereas CD11b and CD66b were not. This further reinforces our data showing that CXCR2 and CD66b are not equivalent markers of neutrophils. This could suggest potentially different functions of CXCR2 negative and positive neutrophils. In line with this, it was shown on neutrophils isolated from the blood of patients that the percentage of CXCR2 positive neutrophils was lower in breast cancer patients compared to healthy patients [
35]. Another study has also shown that CD66b was an independent poor prognosis factor for disease-free survival of breast cancer patients. CD66b marker, as a marker of neutrophils will require additional studies to decide if it can be used as a prognosis factor. The survival analysis was performed on the whole cohort of breast cancer patients. Due to the number of patients included in the cohort, it was not possible to analyze TTR for each group and in particular TNBC separately, as it would not have provided enough statistical power. However, for the group ER+/PR+ patients, which was large enough to be analyzed, we observed the same favorable TTR with high CXCR2 patients (data not shown). It would be interesting in the future to determine the prognostic value of CXCR2 for each subgroup of breast cancer patients. In a variety of cancers, it has been shown that CXCR2 was expressed in cancer cells and frequently associated with a poor prognosis [
9,
36]. It is interesting to mention that the prognosis value of CXCR2 has remained so far poorly studied in breast cancer, whereas it has been analyzed more thoroughly in other cancers. Xu et al. have shown that a high level of CXCR2 was a poor prognosis factor of disease free survival in breast cancer [
25]. This is in disagreement with our results, but could be explained by the fact the Ab that they used stained mainly cancer cells and not stromal cells, as mentioned earlier. Interestingly, CXCR2 could be also a potential cancer stem-like cell marker, as a costaining of some breast cancer cells with NANOG and SOX2 has been reported and also that CXCR1/2 inhibition decreases mammosphere formation [
24,
37]. This is reminiscent of the role of another chemokine receptor, CXCR1, which binds some ELR chemokines and is critical for the functionality of cancer stem cells [
38]. Overall, the discrepancy observed for the prognosis value of CXCR2 in breast cancer might be related to the lack of specificity of the Ab used. As we conducted various validation tests to insure the specificity of CXCR2 staining reported here, we believe that CXCR2 is not expressed in breast epithelial cells. Staining of cancer cells would definitely alter the value of CXCR2 for outcome. Altogether, our results further reinforce the unique role of stromal compartment in cancer development.
4. Materials and Methods
4.1. Patients and Tumor Samples
One hundred and five formalin-fixed paraffin-embedded tumors samples, corresponding to the frozen samples previously used in our first study [
13], were selected from the Montpellier Cancer Institute (ICM) biological resources center for IHC analyses. The tumors were graded according to the modified Nottingham SBR system and categorized according to the sixth edition of the AJCC (American Joint Committee on Cancer) Cancer Staging Manual for pTNM staging. The ER, PR and Her2 status were determined at the protein level by immunohistochemistry. When equivocal, Her2 results were confirmed by fluorescence ISH. The same samples as those previously used in our first study were selected for IHC analyses [
13], as previously published. A follow up was updated on September 2018 by review of the medical files. The clinical and pathological characteristics of the patients as well the measure of B lymphocytes, T lymphocytes, macrophages infiltration and the levels of AP-1 and NF-KB transcription factors are described in
Table 1.
4.2. Ethics Approval and Consent to Participate
Breast tumor surgical specimens were provided by the biological resource center (Biobank number BB-0033-00059) after approval of the Montpellier Cancer Institute Institutional Review Board (ICM-CORT-2018-06, 11 March 2018), following the Ethics and Legal national French dispositions for the patients’ information and consent. All patients were informed before surgery that their surgical specimens might be used for research purposes.
4.3. Immunohistochemistry
IHC analyses were performed using 3-μm-thin sections from formalin-fixed paraffin-embedded tissue blocks on the Dako Autostainer Link 48 platform (Dako-Agilent, Glostrup, Denmark). Ab clones, suppliers, antigen retrieval procedures, dilutions and staining protocols are listed in
Table S1. The PTLink system (Dako-Agilent) was used for pretreatment, allowing simultaneous deparaffinization, rehydration and antigen retrieval. Heat-induced antigen retrieval was executed for 15 min at 95 °C in high (pH 9) or low (pH 6.1) pH buffer (Dako-Agilent) according to the Ab. Endogenous peroxidase was quenched using Flex Peroxidase Block (Dako-Agilent) for 5 min at room temperature. Slides were then incubated with the selected Ab (anti-CXCR2 mouse monoclonal Ab Sc-7304 at 1/500, Santa Cruz Technology, Dallas, TX, USA; anti-CD11b rabbit monoclonal Ab EP45 at 1/400, BioSB, Santa Barbara, CA, USA or anti-CD66b mouse monoclonal Ab 80H3 at 1/200, BioRad, Marnes-la-Coquette, France) for 30 min at room temperature. After 2 rinses in buffer, the slides were incubated with a horseradish peroxidase-labeled polymer coupled to secondary anti-mouse and anti-rabbit Ab for 20 min, followed by appliance of 3,3′-diaminobenzidine for 10 min as a substrate. Counterstaining was performed using Flex Hematoxylin (Dako-Agilent) followed by washing the slides under tap water for 5 min. Finally, slides were mounted with a coverslip after dehydration. A set of samples were also processed with irrelevant IgG isotypic control at the same concentration as CXCR2 Ab, and used as negative control.
The NanoZoomer slide scanner system (Hamamatsu Photonics, Massy, France) was used to digitalize glass slides at the 20× objective. The invasive component of each sample was manually surrounded on the virtual slide by a pathologist and identified as a region of interest (ROI). Necrotic areas, in situ carcinoma and normal breast were excluded from the ROI. Following the calibration of size, shape and color parameters, immune-reactive cells were automatically quantified with the HistoLab
® Image Analysis Software (Microvision, Evry, France) as previously described [
39]. Density of immune-reactive cells in the ROI (
Figure 5), recorded as the number of positive cells per cm
2 of tumor surface, was finally matched to clinicopathological data.
4.4. In Situ Hybridization
ISH was performed using specific HsCXCR2 probes and an RNAscope® 2.5 HD detection kit (Bio-Techne, Abingdon, UK) according to the manufacturer’s instruction. Serial sections were used for CXCR2 IHC and ISH procedures. Probes that target the housekeeping Homo sapiens PPIB gene and the bacterial DapB gene were used as a positive and negative control, respectively.
4.5. Immunofluorescence
Double IF was performed using the CXCR2 rabbit polyclonal Ab HPA031999 combined with the CD66b mouse monoclonal Ab (clone 80H3) on three selected samples showing various CXCR2 and CD66b expression levels following IHC. Briefly, 3-μm-thin sections were submitted to the antigen retrieval procedure using the low pH buffer (Dako-Agilent). Sections were incubated with donkey serum for 30 min at room temperature to block non-specific binding of immunoglobulins and then, with a combination of CXCR2 (1/100) and CD66b (1/200) for 16 h at +4 °C. After washing with PBS Tween, the slides were incubated with a mix of DAPI and secondary donkey anti-rabbit and anti-mouse Ab fluorescently labeled with Alexa Fluor 647 (CXCR2) and 488 (CD66b) dye molecules, respectively (ThermoFisher Scientific, Illkirch, France). The slides were mounted with Mowiol (Sigma-Aldrich, Darmstadt, Germany) for fluorescence microscopy.
4.6. Transfection of Human CXCR2 in HEK-293 Cells
HEK-293 cells were plated in 10 cm dishes and transfected using JetPEI (Ozyme, St Quentin Yvelines, France) according to the manufacturer’s recommendations, using 10 µg of pcDNA3 (control vector) or cDNA3-hCXCR2 expression vector (Interchim, Montluçon, France). After 18 h incubation, the medium was removed and the cells were placed into a fresh medium. Twenty-four hours later, cells were harvested and either included as a pellet for IHC experiments or prepared for protein extraction.
4.7. Protein Extraction and Western Blot
For protein extraction, HEK-293 cells transfected or not with human CXCR2 cells were lyzed in R1 buffer (20 mM Hepes pH 7.3; 20 mM MgCl2, 150 mM NaCl, 0.75% NP-40), supplemented with protease inhibitor cocktail (Roche, Meylan, France). Fifteen micrograms of total protein extracts were separated by SDS-PAGE and subjected to Western blot analyses using CXCR2 Ab (Santa Cruz sc-7304, 1/1000). Immunoreactivity was detected with the Merck Millipore ECL system (Molsheim, France).
4.8. Statistics
Patients’ characteristics were summarized using descriptive statistics. Continuous variables were described using medians and ranges, and categorical variables with frequencies and percentages.
Association between biomarkers and tumor characteristics was tested using the Kruskal–Wallis test. A p value of less than 0.05 was considered statistically significant.
TTR was defined by the interval (calculated in years) between the date of surgery and the date of the first event between local and metastatic recurrence. Only untreated primary tumors were selected for survival analysis (N = 90). TTR was estimated using the Kaplan–Meier method, and median follow-up durations using the reverse Kaplan–Meier method. Survival curves with their log-rank tests were generated. Censored data were descriptively summarized for the different equal groups.
Variables with a p-value less than 0.25 from univariate analysis, including prognostic factors, were introduced in the multivariate model with a set of forced variables (age, SBR grade and immune type). Then, backward and stepwise selection strategies of the non-forced variables were performed. The selection procedure consists of removing variables having the largest p-value greater than 0.10 from the model.
The multivariate analysis was performed using a Cox model adjusted for prognostic factors of survival and potential cofounders. HdRs are presented on a descriptive basis with 95% CI. Statistical analyses were performed using STATA 13.0 (StataCorp, College Station, TX, USA).