The influence of chemotherapy on adenosine-producing B cells in patients with head and neck squamous cell carcinoma

Introduction Head and neck squamous cell carcinoma (HNSCC) strongly suppresses the immune system, resulting in increased metastasis and recurrent disease. Chemotherapy is part of the multimodal treatment but may further immunosuppression. Recently, we demonstrated that regulatory B cells (Breg), defined as CD19+CD39+CD73+ B cells, play a significant role in the production of immunosuppressive, extracellular adenosine (ADO). Here, we tested the influence of chemotherapy on Breg function. Results In HNSCC patients, Breg were diminished in absolute number and frequency after chemotherapy (paired samples). Chemotherapeutic drugs had variable effects; while platinum-based chemotherapy decreased the expression of CD39, methotrexate led to a functional increase in CD39 expression and increased production of immunosuppressive ADO. These findings were confirmed in a second patient cohort. Surface expression of CD39 correlated strongly with the production of ADO as measured by mass spectrometry. Conclusions Platinum-based anti-tumor-therapy reduces the number of adenosine-producing B cells and, consequently, potential immunosuppression within the tumor environment. Breg function in terms of ADO production and their potential capacity to suppress CD4+ T cells are promoted by methotrexate treatment amplifying anti-inflammatory therapeutic effects. Our results add to the understanding of how chemotherapeutic drugs can influence the human immune system and may therefore help to orchestrate standard oncologic therapy with new immune modulating approaches. Methods Mononuclear cells were collected prospectively from HNSCC patients before and after chemotherapy (n = 18), from healthy donors (n = 20), and an additional cohort sampled several months after chemotherapy (n = 14). Frequency, phenotype, and function of Breg were determined by multicolor flow cytometry, ATP luminescence assay as well as mass spectrometry measuring 5′-AMP, ADO, and inosine. Isolated B cells were incubated with chemotherapeutic drugs (cisplatin, methotrexate, paclitaxel, 5-fluorouracil) in vitro for functional studies.


INTRODUCTION B cells in cancer
The role of tumor-infiltrating B lymphocytes (TIL-B) in solid cancers has long been strongly underestimated [1]. It has recently become clear, that the presence of TIL-B frequently correlates with a good prognosis in a variety of solid cancers, including breast, colon, and lung cancer as well as head and neck squamous cell carcinoma (HNSCC) [2][3][4]. Furthermore, it has been suggested that the relevance of TIL-B as a prognostic marker may exceed the role of CD8 + T lymphocytes, given that the interaction of both cell types is mandatory for a sufficient anti-tumor response [5]. Additionally, their controversial function includes a tumorpromoting role [6] and moreover, several murine tumor models implement a pro-tumor effect of TIL-B [7,8]. In addition, B cells inhibit the anti-tumor effect of vaccination against murine malignant melanoma and they induce radioresistance in murine prostate cancer [9,10]. This effect may be partly explained by conversion and attraction of regulatory T cells (Treg) into the tumor micro-environment [1]. However, it also reiterates the difficulty in correlating murine models with the human system.
Nevertheless, it may be worthwhile to further subdivide the B cell population into immune competent B cells and regulatory B cells (Breg). Breg are commonly characterized by their potential to produce immune suppressive mediators, e.g. IL-10 and granzyme B [11,12]. In humans, the presence of IL-10 producing regulatory TIL-B (TIL-Breg) is associated with a worse prognosis in HNSCC and gastric cancer [13,14]. Similarly, the increased frequency of circulating Breg is correlated with advanced tumor stages in human hepatocellular carcinoma [15]. It has further been shown that TIL-Breg suppress the proliferation of CD4 + and CD8 + T cells as well as NK cells requiring a direct cell contact [16]. Additionally, IL-10 secreting TIL-Breg induce immunosuppressive regulatory T cells in breast cancer patients [17].

B cells and adenosine
Our research team recently described a new group of Breg defined by their ability to produce immune suppressive adenosine (ADO) from exogenous ATP using ectonucleotidases CD39 and CD73 [18]. The coexpression of both enzymes is found in the majority (>65%) of all peripheral B cells in healthy donors. As we have previously shown, a variety of other immune cell populations also carry CD39 and/or CD73 on their surface, including regulatory T cells (Treg) [19], mesenchymal stem cells (MSC) [20], T helper cells [21] as well as noncellular exosomes [22]. However, among the immune cells B lymphocytes display the highest CD39/CD73 expression and the largest potential for production of ADO [18]. In addition, several tumor cell lines have also been shown to express ectonucleotidases, contributing to ADO metabolism [23].

Adenosine in cancer
Recently, the role of adenosine in the progression of solid cancers has been discovered and investigated in detail [24]. In the tumor micro-environment of HNSCC, Treg upregulate the expression of ectonucleotidases [25]. In contrast, MSC display a decreased production of ADO when isolated from HNSCC tissue and compared to MSC from autologous healthy tissue [20]. Tumor cells of HNSCC and prostate cancer are stimulated by ADO via ADO receptor (ADOR A2 ) in vitro, indicating a protumorigenic effect of exogenous ADO [26,27]. In humans, the ectonucleotidase CD73 is a strong and independent prognostic marker in a large number of ovarian cancer and prostate cancer patients and is correlated with poor outcome [28,29]. Both the immune suppressive effect and tumor-promoting effect make ADO an interesting therapeutic target. The ADO pathway can be inhibited either by inhibition of the enzymes CD39 and CD73 or by the blockade of ADO-specific receptors. While the modulation of CD39 is still in preclinical testing, an antibody against CD73 has recently entered the clinical trials stage for solid tumors (phase I, NCT02503774). Interestingly, caffeine acts as a natural inhibitor of ADOR A2a and reduced tumor growth in a murine 3-MCA and melanoma model [30]. Similarly, the blockade of ADOR A2a increases the efficacy of an anti-PD1 therapy in a murine breast carcinoma model [31,32]. Based on these promising findings, a selective ADOR A2a inhibitor is currently being tested in a clinical trial for advanced non-small cell lung cancer (phase I, NCT02403193). In summary, inhibition of ADO receptors can have two beneficial effects for cancer patients: (I) improved functionality of immune cells, and (II) decreased proliferation of cancer cells, making this a promising treatment option.
Despite the importance of ADO in cancer progression, very little is known about the influence of cancer therapy on ADO-producing immune cell populations. It is now clear that the frequency of Treg is slightly increased in patients with HNSCC. However, after chemoradiotherapy (CRT) the frequency of peripheral CD39 + Treg increases dramatically due to their relative resistance to CRT as compared to CD4 + T helper cells [33]. This may contribute to increased ADO levels and further immune suppression resulting in recurrent disease after CRT. Similarly, treatment with the EGF inhibitor cetuximab increased the frequency of CD39 + Treg in the micro-environment of HNSCC [34]. In contrast, a decrease in frequency and number of CD39 + Treg was observed after vaccination with p53-specific dendritic cells with a possible immune enhancing effect due to decreased ADO concentrations [35]. www.impactjournals.com/oncotarget In the present study we describe for the first time the influence of chemotherapy on adenosine-producing Breg in two independent patient cohorts with HNSCC.

Changes in frequency, absolute numbers and phenotype of B cells before and after CRT
For each HNSCC patient of cohort #1, the frequency and absolute number of CD19 + B cells in the peripheral blood were compared before and after CRT. Both the frequency and the absolute number of CD19 + B cells were decreased (each p < 0.05) after CRT ( Figure 1A). Representative density plots are shown in Figure 1B. In cohort #1, the frequency of CD4 + T cells also decreased significantly (Supplementary Figure 1A), while the frequency of CD8 + T cells was not significantly affected, confirming the data from previous publications [33]. While these changes applied to patients treated with a platinumbased chemotherapy, patients treated with methotrexate showed no alterations (Supplementary Figure 1B).
Furthermore, B cells in patient cohort #1 were tested by flow cytometry for expression of various immunologic surface markers. IgM surface expression, as well as the IgM + B cell subset, were significantly increased after CRT (Supplementary Figure 1C). In addition, there was an increase in the CD19 + CD5 + B cell compartment after CRT, which is considered critical regarding the promotion of further tumor growth ( Figure 1C, 1D) [37]. Both surface markers, IgM and CD5, were found to be unchanged after methotrexate therapy. B cells were negative for CD26 and no expression was induced by CRT. Expression rates and percentages of CD25 + , PD1 + , CCR7 + , IgA + , and CD40 + B cells also showed no significant alteration after treatment (Supplementary Figure 1E and 1F).

Phenotypic characterization of ADO-producing B cells
In patient cohort #1, flow cytometry analysis showed that up to 82% of B cells co-expressed CD39 and CD73 on their cell surface. As previously reported, these cells demonstrate an immunosuppressive potential by hydrolyzing exogenous ATP to ADP, 5′-AMP, and ADO [18]. Therefore, we were especially interested in therapyinduced changes in this Breg subset. Within the CD19 + B cell compartment, the frequency and the absolute number of these CD39 + CD73 + Breg was significantly decreased after CRT (p < 0.005) (Figure 2A, 2B). Consequently, the subsets of CD39 + CD73 neg as well as CD39 neg CD73 + B cells were increased (p < 0.01, data not shown). As shown in Figure 2C, the mean fluorescence intensity (MFI) of both ectonucleotidases, CD39 and CD73, was significantly reduced in the CD19 + B cell compartment after platinumbased chemotherapy (p < 0.001). Interestingly, MTX treatment showed no reduction in the ectonucleotidases ( Figure 2D) and also no decrease in co-expressing cells (Supplementary Figure 1D).

In vitro changes by cytostatic drugs
To test the different effects of cytostatic drugs on ADO-producing B cells, isolated B cells of healthy donors were treated with chemotherapy in vitro for 7 days as described above. Cytostatic drugs were chosen from regimens used in standard therapy for HNSCC patients in clinically administered concentrations (cisplatin, paclitaxel, 5-FU and MTX). Cisplatin, paclitaxel, and 5-FU induced a dose-dependent decrease in CD39 surface expression (p < 0.05). However, methotrexate induced a significant increase in CD39 expression on B cells (p < 0.05, Figure 3A), while the mean fluorescence intensity of CD73 was significantly increased after incubation with cisplatin and paclitaxel (each p < 0.05). 5-FU and methotrexate did not influence surface expression of CD73 ( Figure 3B). Cetuximab did not show either of these significant influences regarding the expression of CD39 and CD73 (data not shown).

ATP hydrolysis by luminescence
ATP is efficiently hydrolyzed to ADP and AMP by the enzymatic activity of CD39. To quantify the function of this enzyme and further evaluate the influence of chemotherapy, precultured B cells were separated and incubated with one of the cytostatic drugs. The same number of B cells (3 × 10 4 ) separated from the cultures were then analyzed in terms of ATP consumption by luminescence assay. Trypan blue dye was used for cell count and to determine viable cells. B cells cultured in the presence of high dose of cisplatin (p < 0.05), paclitaxel (p < 0.01), and 5-FU (p < 0.01) all showed lower ATP hydrolysis than B cells of the same healthy donors cultured in the absence of any cytostatic drug. Furthermore, 5-FU and paclitaxel each induced a significant reduction in ATP consumption, even in low or moderate dosages. In contrast, MTX showed a higher ATP hydrolyzation (p = 0.1; Figure 4A). There was a significant correlation between ATP utilization and CD39 expression for all samples (p < 0.0001), indicating a direct relationship ( Figure 4B). This correlation was found to be particularly significant and dose-dependent in the presence of paclitaxel (p < 0.001; Supplementary Figure 1G).

Adenosine production measured by mass spectrometry
In order to confirm results of ATP hydrolysis measured by luminescence, and therefore, functional characterization of the ectoenzymes CD39 and CD73, ADO production was measured by mass spectrometry. As expected, B cells showing downregulation of CD39 after treatment with cisplatin were less capable of ATP hydrolyzation and produced lower levels of 5′AMP. Due to the lack of substrate, production of ADO and inosine was also reduced. Despite the small number of samples, the data showed a trend towards significance in terms of ADO concentrations (no chemo vs. cisplatin; p = 0.07). Figure 5 displays measurements of one representative patient including (A) expression of CD39 and CD73 by flow cytometry, (B) ATP consumption by luminescence, and (C) mass spectrometry results. In accordance with our previous results, levels of 5′-AMP, ADO, and inosine were clearly higher in supernatants of B cell cultures incubated with MTX. While altered CD39 expression influenced consequent ADO production, CD73 varied less and seemed to play a minor role in ADO pathways.

Drug-dependent suppression of B cell proliferation.
B cells of healthy donors were incubated with cytostatic drugs, and proliferation was measured in CFSE-based proliferation assays as described above. Drugs in concentrations ranging from 1 to 50 µg/mL were added on day 0. All drugs significantly inhibited B cell proliferation except for MTX ( Figure 4C, 4D). Cisplatin almost completely blocked B cell proliferation (p < 0.001). Paclitaxel decreased B cell proliferation in a dose-dependent manner (Supplementary Figure 2).

Long-term effects of CRT
Blood samples of patient cohort #2 were collected 14.2 ± 7 months after CRT allowing for interpretation of longterm effects of CRT. The expression of the ectonucleotidases CD39 and CD73 were found to be decreased (p = 0.01); Figure 6A). However, the B cell frequency was significantly increased after CRT. In addition, there was a significant correlation between B cell frequency and time-after-treatment (p < 0.001) suggesting B cell population recovery after exposure to CRT ( Figure 6B).

Role of B cells in cancer
Effects of systemic cancer treatment on the host immune system are of considerable relevance. The interaction and function of immune cells is affected by any kind of therapy with a consequent reduction of antineoplastic activity. Recently, the role of tumorinfiltrating B cells has been a focus of various research groups and as described above, their impact on tumor growth has been underestimated in the past and is a matter of dispute. In particular, the role of tumor-infiltrating regulatory B cells (TIL-Breg) is not well understood [1]. However, it is reasonable to expect that adenosineproducing B cells have an immunosuppressive effect in the tumor micro-environment.

Influence of cisplatin on B cells
Here, we show for the first time the effect of systemic chemotherapy on Breg in cancer patients. Platinum-based therapy not only reduces the frequency of Breg but also their ability to produce immunosuppressive adenosine. This effect was demonstrated in vitro as well as in two independent cohorts of HNSCC patients. Our previous work has shown opposing effects on regulatory T cells, which are strongly enhanced in frequency and function by platinum-based chemotherapy [33]. It therefore becomes evident that oncologic therapy has different effects on regulatory cell populations in the human immune system. Based on the results of other research groups, tumor escape mechanisms apparently benefit from the presence of immunosuppressive ADO; the dysfunction of the ectonucleotidases CD39 and CD73 is associated with reduced tumor growth [38][39][40], and upregulation of CD73 on tumor cells is linked to increased risk for metastasis and increased chemo resistance [41][42][43]. However, other studies report antitumor effects of ADO [44][45][46], describing intracellular and extracellular mechanisms leading to apoptosis of cancer cells through ADO.

Recovery of B cells
In our second cohort of cancer patients, we could show that B cells have a fast recovery rate reaching normal levels about one year after termination of platinum-based chemotherapy. This is in contrast to our previous observations analyzing the Treg (CD4 + FoxP3 + ) population. There, we demonstrated that the frequency of Treg is increased for up to three years after chemotherapy [33]. Of note, the increased frequency of Treg after chemotherapy is caused by the decreased absolute number of CD4 + T cells. There was no increase observed   in the absolute number of Treg. These data suggest that although CRT adversely affects both lymphocyte subsets, it has exceptionally detrimental effects on the CD4 + T cell population. CD19 + B cells seem to be less sensitive in the long-term since they recover faster from CRT than CD4 + T cells. From an immunologic perspective, the altered lymphocyte homeostasis may partially explain, why HNSCC cells become resistant to an initially effective therapy.

Influence of MTX on B cells
When compared with platinum-based chemotherapy, MTX has an opposing effect on the human immune system. In our study population, Breg frequencies increased and showed enhanced ability to produce exogenous adenosine when patients were treated with MTX. These clinical data are supported by our in vitro experiments. MTX is not only used as an antineoplastic drug in cancer patients but also as an anti-inflammatory drug, e.g. in rheumatoid arthritis where it is used in a lower dose. Our results show an increased adenosine production in Breg induced by the presence of MTX, whose anti-inflammatory mechanism of action is not fully understood but is thought to be adenosine dependent [47]. It is further postulated by others that caffeine can suppress the therapeutic effect or unwanted side effects of MTX by targeting the adenosine receptors in an anti-inflammatory setting [47,48]. The importance of the MTX-ADO axis is also confirmed by Tsujimoto et al. who found an association between MTX-induced leukencephalopathy and the polymorphism of the ADO receptor A2a in pediatric leukemia patients [49]. Moreover, MTX was demonstrated to suppress the NFkappaB pathway by the release of ADO in Jurkat cells contributing to its antiinflammatory and antiproliferative effect [50]. Whether these observations can be translated into the antineoplastic effects of MTX in solid cancer patients is yet to be investigated.
Despite the analogy, caution is advised in extrapolating experimental results from arthritis patients to cancer patients as the diseases arise from entirely different conditions. Namely, it is not yet completely understood if immune suppression in the tumor micro-environment of cancer patients is beneficial or not. In addition, great differences may exist in the micro-environment of different kinds of cancer [22]. Interestingly, MTX may even have a promoting effect on adenosine metabolism in cancer cells as shown in glioblastoma cell lines due to the upregulation of the ectonucleotidase CD73 [51].

Immunotherapy relies on a functional immune system
Aside from the standard regimen of tumor therapy, e.g. surgery, radiation and chemotherapy, a variety of immunotherapeutic approaches have recently found their way into clinical routine. Special potential is seen in the checkpoint modulators, e.g. inhibitors of the programmed death receptor PD1, as they can lead to long-term survival Figure 5: Correlation of B cell phenotype, ATP hydrolysis, and production of its derivates 5′AMP, ADO, and inosine.
Juxtaposition of B cell treatment from one representative donor to confirm ATP hydrolysis and correlation with the respective phenotype. (A) Expression of ectonucleotidases after 7 days of cytostatic treatment as measured by flow cytometry. (B) ATP consumption as measured by luminescence. ATP concentrations measured without the influence of chemotherapeutic drugs were set to 100%. (C) Production of 5′AMP, adenosine, and derivatives as measured by mass spectrometry. Decreased ATP hydrolysis and adenosine production correlated well with less CD39 expression after treatment with cisplatin and paclitaxel while methotrexate showed opposite effects. www.impactjournals.com/oncotarget in advanced tumor disease [52]. All immunotherapeutic approaches rely on a functional immune system, and it is not yet clear how to optimally combine the different therapeutic regimens. Modified lymphocyte function and homeostasis induced by CRT could lead to therapy failure. This may be especially true for tumor cells arising from a chronic inflammatory setting, which is the case in HPV + HNSCC. Here, the alteration of extracellular adenosine production may be added to the mechanisms, which result in therapy resistance. On the other hand, these processes may have the potential to overcome CRT failure in an immunosuppressive tumor microenvironment [53]. As recently demonstrated by Maj et al., regulatory cell populations can induce resistance to PD1-mediated therapy in the tumor microenvironment (TME). Due to the oxidative stress in the TME, the preferred mechanism of suppression is adenosine (and not PD-L1, CTLA-4, TGF-b, IL-35, or IL-10) [54]. These findings demonstrate the importance of adenosine in the context of immune suppression and therapeutic failure.

CONCLUSIONS
Our results add to the understanding of how chemotherapeutic drugs can influence the human immune system and may therefore help to orchestrate standard oncologic therapy with new immune modulating approaches. Giving consideration to the growing importance of tailored cancer treatment, further investigation of the interaction between antitumor immune response and systemic, oncologic therapy may be needed to adapt antineoplastic therapy regimes to individuals.

Patient cohort #1
Peripheral blood samples (n = 36) were obtained from 15 HNSCC patients before and after CRT and from 3 patients before and after palliative chemotherapy (11/2013-10/2016). Healthy volunteers (NC, n = 20) served as a control group for in vitro experiments.

Patient cohort #2
An additional cohort of HNSCC patients (n = 14) donated peripheral blood after CRT, which was terminated 14.2 ± 7 months prior to the respective blood draws. At this time, none of the patients showed evidence of recurrent disease (NED). Blood samples were compared to an additional healthy control group (NC, n = 24) in order to analyze long-term effects of CRT. All subjects signed

PBMC collection
Blood samples (50 mL) were drawn into citratebuffered tubes and centrifuged on Ficoll-Hypaque gradients (Greiner Bio-one, Kremsmuenster, Austria). Peripheral blood mononuclear cells (PBMC) were recovered, washed twice with RPMI medium (RPMI 1640, Gibco, Grand Island, NY) and phosphate-buffered saline (PBS, Gibco), and counted in a trypan blue dye. PBMC from study cohort #1 were cryo-conserved for subsequent analysis. PBMC from cohort #2 were immediately used for in vitro experiments.

FACS antibodies
The following anti-human monoclonal antibodies

Surface staining and flow cytometry
Briefly, cells were incubated with mAb specific for surface markers in 50 µl PBS for 30 minutes at room temperature in the dark and washed twice before acquisition for surface marker detection. Flow cytometry was performed using a Gallios TM 10-color flow cytometer equipped with Kaluza ® flow cytometry software (both Beckman Coulter, Brea, CA). The acquisition and analysis gates were restricted to the lymphocyte gate based on characteristic properties of the cells in forward and side scatter. At least 10 5 cells were acquired for analysis. Absolute numbers of CD19 + B cells and CD39 + CD73 + Breg were calculated by multiplying their frequency values by the absolute number of lymphocytes obtained from whole blood counts.  CRT: chemoradiotherapy; CIS: cisplatin (30-40mg/m² weekly for 7 weeks, or 20mg/m² daily in week 1 and 5); CARBO: carboplatin (AUC-area under the curve); MIT-C: mitomycin-C; 5-FU: 5-fluoruracil; MTX: methotrexate. Patients with adjuvant CRT underwent surgery before. Primary treated patients were treatment naive. n/a: not available. #16 was previously treated with CIS/5-FU/cetuximab; #17 and #18 were previously treated with taxane/CIS/5-FU, gemcitabine, cetuximab, and pembrolizumab. www.impactjournals.com/oncotarget

Statistics
Statistical analyses were performed using GraphPad Prism Software (GraphPad Software, Inc., La Jolla, CA). Error bars, where displayed, indicate the standard deviation of the mean or median data from replicate experiments. Significance of differences between samples within figures was confirmed using paired or unpaired t tests, depending on the experimental setting, with a significant level of α = 0.05. Correlations were calculated by the Spearman test considering R² > 0.5 to be significant.

Author contributions
AZ, UH, and SJ performed experiments; JD and SL analyzed results and prepared figures; EKJ performed mass spectrometry experiments, CB and PJS designed research; AZ and PJS wrote the paper; TKH, EKJ and CB edited the paper.

CONFLICTS OF INTEREST
None.