Chemotherapy Alters Monocyte Differentiation to Favor Generation of Cancer-Supporting M2 Macrophages in the Tumor Microenvironment

Current treatment of advanced cervical and epithelial ovarian cancer includes platinum-based multimodality therapy (1, 2). Many patients with gynecologic cancer develop resistance to platinum drugs, thereby limiting further treatment options and decreasing overall survival rates (1, 3). Gynecologic cancers are generally known as immunogenic, and prominent correlations exist between the infiltration of tumors by immune cells and clinical outcome (4, 5). Macrophages are the most abundant immune cells present in the tumor microenvironment. Macrophages originate from monocytic precursors in the blood and undergo specific differentiation depending on cues in the local tissue. Two extreme polarization states of macrophages are known, M1 and M2, of which the latter has poor antigen-presenting capacity, prevents T-cell activation, contributes to suppressing dendritic cell (DC) functions, as well as enhances angiogenesis and metastasis (6). The presence of M2 macrophages in tumors is correlated to poor prognosis in several human cancers (7, 8). Previously, we and others showed the influence of cervical (9) and ovarian cancer cells (10, 11) on differentiation of monocytes into DC or macrophages. The majority of cancer cells either hampered monocyte to DC differentiation or skewed their differentiation toward M2-like macrophages, depending on their ability to produce prostaglandin E2 (PGE₂) and/or interleukin (IL)-6. Blocking these cytokines completely restored their differentiation toward DC (9–11). Interestingly, in gynecologic malignancies, upregulation of the COX enzymes has been associated with platinum drug resistance (12, 13). In addition, high levels of IL-6 in sera and ascites of patients with these gynecologic malignancies have also been related to chemoresistance and poor clinical outcome (14, 15). IL-6 is one of the major immunoregulatory cytokines present in the tumor microenvironment and induces several pathways leading to tumor proliferation, angiogenesis, and chemoresistance (16, 17). One important pathway stimulated by IL-6 is activation of STAT3 by phosphorylation. High levels of phosphorylated STAT3 are found in tumors (18) and tolerogenic antigen-presenting cells (APC; ref. 19). Consequently, STAT3 signaling in APC is linked to the induction of T-cell tolerance (20).

In view of the observations that PGE₂ and IL-6 are associated with chemoresistance and with the tumor-induced differentiation of tumor-promoting M2 macrophages as well as recent literature indicating that chemotherapy may bear impact on the function and thereby the efficacy of tumor-infiltrating immune cells (21), we investigated the potential effects of platinum-based chemotherapeutics on the differentiation and function of APC under the influence of cervical and epithelial ovarian cancer cells in vitro.

We found that the treatment of tumor cells with cisplatin or carboplatin increased the potency of some tumor cell lines to skew monocytes to M2-like macrophages. These M2-like macrophages displayed IL-6–mediated increased levels of activated STAT3 and PGE₂-mediated decreased levels of activated STAT1 and STAT6, which are associated with immune potentiating pathways. The underlying mechanism was a cisplatin or carboplatin-induced enhanced activation of the NF-κB pathway likely through the chemotherapy-induced DNA damage response (DDR). This resulted in an increased production of PGE₂ and IL-6 by cancer cells, but only if they already produced these factors, and an enhanced skewing of monocytes toward M2-like macrophages. Increased numbers of tumor-promoting M2-like macrophages may form an indirect mechanism for chemoresistance suggesting that concomitant therapy with COX inhibitors and/or blocking of IL-6 receptor (IL-6R) might increase the antitumor effect of chemotherapy.

APC and cancer cell lines were cultured as described earlier (9) in RPMI-1640 medium (Invitrogen) supplemented with 10% fetal calf serum (FCS; Greiner Bio-one), 2 mmol/L l-glutamine (Cambrex), 100 U/mL penicillin, 100 μg/mL streptomycin, and 50 μmol/L β-mercaptoethanol (Invitrogen), also referred to as complete or control medium. Adherent cells were harvested using trypsin/EDTA (Invitrogen). The following factors were used to culture APC: 500 U/mL IL-4 (Invitrogen), 800 U/mL granulocyte macrophage colony-stimulating factor (GM-CSF; Immunotools), 25 ng/mL macrophage colony-stimulating factor (M-CSF; R&D), 1 to 10 ng/mL PGE₂ (Sigma-Aldrich), and 1 to 50 ng/mL IL-6 (Immunotools). The toll-like receptor (TLR) ligand 0.25 μg/mL lipopolysaccharide (LPS; Sigma-Aldrich) was used to activate APC; to mimic T-cell interaction, APCs were stimulated with irradiated CD40 ligand (CD40L)-expressing mouse fibroblasts. The following cytokines were used to induce STAT phosphorylation: 500 U/mL IFN-γ (Bender Medsystems), 10 ng/mL IL-10 (Peprotech), or 500 U/mL IL-4.

Cells were treated with 0.2 to 50 μg/mL cisplatin or 2 to 500 μg/mL carboplatin (Pharmachemie) and/or 25 μmol/L indomethacin (Cayman Chemical) dissolved in dimethyl sulfoxide (DMSO) or as a control only with the corresponding concentration of DMSO. The monoclonal antibody (mAb) used to block the (s)IL-6R was tocilizumab, 5 to 50 μg/mL (RoActemra, Roche BV). We used 0.02 to 200 μg/mL of the selective inhibitor of κB-α (Iκ-Bα) inhibitor Bay 11-7082 (Sigma-Aldrich).

The human cervical cancer cell lines HELA, CASKI, CSCC1, CSCC7, and CC8 were typed and cultured as described earlier (9). Ovarian cancer cell lines SKOV3 and A2780 were purchased from the European Collection of Animal Cell Cultures. CAOV3 was obtained from the American Type Culture Collection (ATCC), and OVCAR3 and COV413B were kindly provided by the department of Clinical Pathology of the Leiden University Medical Center (Leiden, the Netherlands). All human ovarian cancer cell lines were of epithelial origin. Cell lines were authenticated every half year by short tandem repeat (STR) DNA markers as described previously (22). In brief, PCR amplification of 8 highly polymorphic microsatellite STR loci and gender determination were measured, and the uniqueness of DNA profiles was compared for identity control within the STR database of ATCC (23). Stock vials were thawed and cultured for 10 passages and routinely tested for the presence of mycoplasma.

Cell lines were grown in culture flasks at 80% to 90% confluence and harvested with trypsin/EDTA and cultured in 6- or 12-well plates (Corning) for 24 hours, treated with chemotherapeutics and/or COX inhibitors and/or tocilizumab as indicated. After 24 hours of treatment, cells were washed carefully and medium was refreshed. Tumor supernatants (TSN) were harvested after an additional 24 hours of culture and stored at −20°C. Cancer cell lines were cultured in the presence of 2 μg/mL cisplatin or 20 μg/mL carboplatin for 24 hours. These doses are estimated levels of chemotherapy in the tumor tissue because in patients, the maximum doses for cisplatin or carboplatin as measured in the blood are 5 to 6 μg/mL and 40 to 80 μg/mL, respectively. Assuming that the levels in (poorly vascularized) tumor tissue are somewhat lower, the doses used in vitro are representative for the in vivo situation (24–29).

APC were differentiated as described earlier (9). A brief explanation is given in Supplementary Material and Methods. To address the direct cytotoxic effect of chemotherapy on APC, the cultures were supplemented with 20% TSN or control medium (monocyte-derived DC; mo-DC) and titrated doses of chemotherapeutics, as indicated. To investigate the capacity of chemotherapy to alter the differentiation of APC by acting through tumor cell-mediated mechanisms, the cultures were supplemented with 20% TSN of untreated and treated cancer cell lines or control medium (mo-DC).

Monocytes were cultured in a 24-well plate in complete medium with or without 2 μg/mL cisplatin or 20 μg/mL carboplatin. Cancer cells were cultured in the top compartment of a Transwell 0.4 μm pore insert (Corning). After 3 days, complete medium with cytokines was added. At day 6, cells were analyzed by flow cytometry.

To determine the survival of the cancer cells upon chemotherapeutic treatment, an MTT assay (Trevigen) was conducted, according to the manufacturer's instructions. Cell survival was calculated as follows: (OD_{570–655 nm} for treated/OD_{570–655 nm} for untreated) × 100%. To determine APC survival upon chemotherapeutic treatment, cells were analyzed by flow cytometry. Cell survival was calculated as follows: (% of cells in live gate for treated/% of cells in live gate for untreated) × 100%.

Phosphorylated STAT (pSTAT) analysis was conducted according to Krutzik and colleagues (30). Cells were fixed in 1.5% paraformaldehyde (Sigma-Aldrich) for 10 minutes at room temperature, harvested, and washed twice in PBS containing 2% FCS (PAA) and 0.02% sodium azide (AZL Pharmacy). Then, cells were permeabilized in 90% methanol (Sigma-Aldrich) for 10 minutes on ice, washed, and stained for pSTAT1 (pY701), pSTAT3 (pY705), or pSTAT6 (pY641; all PE; all BD Biosciences). Expression was calculated as follows: geometric mean of fluorescence intensity of condition of interest − geometric mean of fluorescence intensity of the corresponding unstained control. Relative expression (ratio) was calculated as follows: expression of condition of interest/expression of control condition (mo-DC).

IL-12p70 and IL-10 were analyzed using ELISA Kits from BD Biosciences or by inflammatory cytometric bead array (CBA) according to the manufacturer's instructions. To evaluate the cytokines present in supernatant of cancer cells, VEGF, IL-1β, IL-6, and IL-8 were determined by ELISA or CBA and M-CSF by Bioplex (Bio-Rad). PGE₂ levels were measured using the competitive PGE₂ Immunoassay Kit (Enzo Life Sciences).

Lentivirus expressing short hairpin RNA (shRNA) against peroxiredoxin 1 (PRDX1) or TurboGFP (sh control) were produced and used to infect the CC8 cells at multiplicity of infection 5. Further details can be read in the Supplementary Material and Methods.

Previously, we showed that when the TSN of cervical cancer cells was added to monocyte cultures, this could skew the differentiation of monocytes to cells resembling M2 macrophages (M2-like macrophages; ref. 9). We investigated whether also ovarian cancer cells could have this effect. Therefore, not only 5 cervical cancer cell lines but also 5 ovarian cancer cell lines were tested. First, the phenotype of in vitro differentiated monocytes cultured with TSN of cervical cancer and ovarian cancer cell lines was assessed by the use of a panel of APC markers at day 6 of differentiation. We found that TSN from the 2 ovarian cancer cell lines, COV413B and CAOV3, skewed differentiation of monocytes toward a M2-like macrophage phenotype, and we confirmed the M2-like macrophage skewing capacity of HELA, CC8, and CSCC7. We will refer to these as M2-like macrophages and the TSN inducing these cells as M2-TSN (HELA, CC8, CSCC7, COV413B, and CAOV3). Monocytes cultured in the presence of cancer cell lines that hampered monocyte differentiation, but did not induce M2-like macrophages, are referred to as tumor APC and the supernatant inducing this phenotype as APC-TSN (CSCC1, CASKI, SKOV3, OVCAR3, and A2780; Supplementary Table S1).

Then, the functionality of these cultured APC was tested by a subsequent stimulation with LPS (a TLR4 agonist), most often used to stimulate tumor resident DC in vitro or with CD40L-expressing fibroblasts (CD40L) to mimic APC-T–cell interaction. The production of IL-12 and IL-10 was measured after 48 hours. In accordance with our previous results, monocytes skewed by the TSN of ovarian cell lines COV413B and CAOV3 toward an M2-like phenotype produced almost no IL-12, but instead high levels of IL-10 when compared with mo-DC. APC-TSN displayed a similar balance in IL-12/IL-10 production as mo-DC.

To understand which soluble products in the TSN of ovarian cancer cells drove the differentiation of M2 macrophages, we measured the production of TGFβ, IL-1β, IL-6, IL-8, VEGF, PGE₂, and M-CSF (Supplementary Table S2) and found that COV413B produced large amounts of IL-6 (25 ng/mL) but no PGE₂, whereas CAOV3 produced both IL-6 (14 ng/mL) and PGE₂ (2 ng/mL). Blocking of both these cytokines revealed that they are responsible for induction of phenotypically and functionally M2-like macrophages not only by cervical cancer cell lines (9) but also by ovarian cancer cell lines (data not shown).

First, to address the cytotoxic effect of cisplatin and carboplatin on our panel of cancer cell lines, we applied an MTT assay (Supplementary Fig. S1). We observed that the PGE₂ and/or IL-6–producing cell lines, HELA and CC8, were most sensitive for chemotherapy, whereas COV413B, CAOV3, and CSCC7 producing PGE₂ and/or IL-6, were most resistant to chemotherapy. In addition, SKOV3 was chemoresistant, whereas the cells did not produce PGE₂ and/or IL-6 (Supplementary Fig. S1). This showed that there was no direct relation between the production of IL-6 and/or PGE₂ and chemoresistance of cancer cells.

Furthermore, the direct cytotoxic effect of cisplatin and carboplatin on the different types of APC (mo-DC, APC-TSN, and M2-TSN) was addressed. Monocytes were cultured in the presence of TSN and increasing doses of chemotherapy. Following treatment, cells were analyzed by flow cytometry. Interestingly, M2-like macrophages were more vulnerable to chemotherapy than tumor APC or mo-DC (Fig. 1A). Subsequently, the sensitivity of classical M1 macrophages (monocytes cultured in the presence of GM-CSF) and classical M2 macrophages (monocytes cultured in the presence of M-CSF) to cisplatin and carboplatin was tested (31). The classical M2 macrophages were also more sensitive to chemotherapy than M1 macrophages and DC (Fig. 1B).

We previously showed that IL-6 and PGE₂ promoted the differentiation of M2 macrophages (9), cells, which are associated with a worse response to therapy (7, 8). Therefore, we studied the effect of chemotherapy treatment on tumor cells focusing on the tumor-induced differentiation of APC. Tumor cells were incubated with a dose of chemotherapy representative for the level within the tumor microenvironment. Subsequently, monocytes were cultured in the presence of TSN isolated from untreated, cisplatin, or carboplatin-treated cancer cells. A clear increase in the percentage of CD1a-CD14+CD206+CD163+ M2 macrophages was observed when M2-TSN from treated cancer cells was used compared with M2-TSN from untreated cells. This effect was not observed for tumor cells producing APC-TSN (Fig. 2), excluding the possibility that cell debris would have been the cause for the effects observed with the M2-TSN of chemotherapy-treated tumor cells. To mimic the natural situation, we used a Transwell system to culture monocytes and cancer cells together in the presence of cisplatin or carboplatin. This set-up confirmed that platinum-containing chemotherapeutics did not influence the differentiation of the APC directly as the phenotype of mo-DC remained the same upon treatment. However, addition of platinum chemotherapy to the culture system resulted in an increased percentage of M2-like macrophages (Fig. 2A).

As expected (9), stimulation of differentiated monocytes with LPS or CD40L resulted in a strong cell-surface expression of the markers CD80, CD83, CD86, HLA-DR, and PD-L1 (data not shown). Furthermore, chemotherapeutic treatment resulted in a significant (P < 0.01) functional enhancement of M2-like macrophage function (measured by indirect as well as Transwell assay). These M2-like macrophages produced significantly less IL-12 and more IL-10 upon activation with LPS (data not shown) and CD40L after chemotherapy compared with their untreated controls (Fig. 2C).

The capacity of cervical and ovarian cancer cells to skew monocytes to M2-like macrophages depends mainly on their ability to produce PGE₂ and/or IL-6 (9–11). Treatment of tumor cells with cisplatin resulted in an increased production of PGE₂ and IL-6 in cell lines that already produced these cytokines: CASKI and COV413B (IL-6), CSCC7 (PGE₂), CC8, HELA, and CAOV3 (both PGE₂ and IL-6). Carboplatin displayed a significant effect on PGE₂ levels in HELA and CAOV3 as well as enhanced the production of IL-6 in all IL-6–producing cell lines (Fig. 3). As expected by the increased cytokine levels, the expression of both COX-2 and/or IL-6, but not COX-1 mRNA levels were increased upon platinum-based treatment in accordance with the cytokine profile of the cell lines (Supplementary Fig. S2). Notably, cisplatin or carboplatin treatment of the cancer cells did not induce or enhance the production of TGFβ, IL-1β, IL-8, VEGF, and M-CSF (Supplementary Table S2).

To elucidate the mechanisms underlying the effect of chemotherapy on APC differentiation and function, we studied the involvement of STAT. Therefore, the intracellular pSTAT levels of APC were determined (Fig. 4A). Compared with mo-DC, APC cultured in the presence of all IL-6–producing cancer cell lines (CASKI, CC8, HELA, COV413B, and CAOV3), had significantly increased levels of pSTAT3 (Fig. 4B and data not shown). A number of APC cultures displayed a decrease in the levels of pSTAT1 and pSTAT6. This was associated with the presence of PGE₂ in the TSN (Fig. 4B and data not shown). Monocytes cultured in the presence of cancer cell lines that did not produce PGE₂ or IL-6 (CCSC1, SKOV3, OVCAR3, and A2780) did not show alteration in pSTAT levels. To test whether these factors were responsible for the altered levels of phophorylated STATs, we differentiated monocytes in the presence of increasing doses of PGE₂ and IL-6. Indeed, the levels of pSTAT1 and pSTAT6 were dose dependently decreased in response to PGE₂, whereas the levels of pSTAT3 increased upon increasing amounts of IL-6 (Fig. 4C). We hypothesized that the observed increase in M2-skewing capacity of tumor cells treated with cisplatin or carboplatin would correlate with further alteration of pSTAT levels. Indeed, the use of M2-TSN isolated from cancer cell lines treated with cisplatin and carboplatin resulted in M2-like macrophages displaying significantly higher pSTAT3 and decreased pSTAT1 and pSTAT6 levels than M2-like macrophages cultured with M2-TSN from untreated cancer cells in almost all cases (Fig. 4C). Notably, M2-like macrophages cultured with TSN of CAOV3 already displayed low levels of pSTAT1 and pSTAT6 compared with mo-DC, and when the TSN of CAOV3 treated with carboplatin was used, no significant further decrease in pSTAT1/6 levels was found, despite the increase in PGE₂ production.

We blocked the effect of IL-6 with tocilizumab, a human monocloncal antibody against (s)IL-6R that is clinically successful in the treatment of rheumatoid arthritis and Castleman disease (32, 33). The effect of PGE₂ was blocked by treating the cancer cells with the COX (1 and 2) inhibitor indomethacin to inhibit PGE₂ production via COX-2 (Fig. 5A). Restoration to the full phenotype of mo-DC (Fig. 5B) as well as function, measured by the production of IL-12 and IL-10, could be obtained when the cancer cells were treated with indomethacin and/or tocilizumab (Fig. 5C). Targeting COX-2 and (s)IL-6R also had a clear effect on the levels of pSTAT. As expected from our previous experiments, treatment with tocilizumab decreased the levels of pSTAT3, whereas indomethacin treatment restored pSTAT1 and pSTAT6 levels similar to that observed in mo-DC (Fig. 5C).

To find the underlying mechanism of the enhanced PGE₂ and IL-6 production within the tumor cells upon treatment with cisplatin and carboplatin, we investigated 2 hypotheses. Previously, Wang and colleagues showed that cisplatin can selectively crosslink a complex of the chaperone protein PRDX1 and NF-κB at the NF-κB promotor site of the COX-2 promotor region, thereby promoting COX-2 expression (34). Therefore, the expression of PRDX1 was blocked by shRNA, to prevent this protein to form a complex and hence, to prevent the increased production of PGE₂. We successfully downregulated the expression of PRDX1, but this did not result in decreased COX-2 mRNA levels upon chemotherapy treatment (Supplementary Fig. S3).

The second mechanism that could play a role in the increased production of PGE₂ and IL-6 by cancer cells is the activation of DDR pathway as this can lead to NF-κB activation and subsequently to the production of PGE₂ and IL-6 (35, 36). To block the NF-κB pathway, Bay 11-7082, a selective inhibitor of I-κB phosphorylation and as such also of the canonical NF-κB pathway, was used. Both untreated and cisplatin-treated cancer cells were incubated with increased doses of Bay 11-7082. Indeed, the cisplatin-induced increase in PGE₂ and IL-6 production was blocked by Bay 11-7082 in a dose-dependent fashion (Fig. 6).

Here, we showed 2 potential effects of platinum-based chemotherapy of gynecologic cancers on the immune system. First, we showed that M-CSF–induced M2 macrophages and M2-like macrophages are most vulnerable for chemotherapy. Monocyte-derived DC and M1 macrophages were only affected at higher doses, suggesting a selective survival benefit of these cells. This can be considered beneficial for the patient. However, while it is clear that a sufficient dose of cisplatin or carboplatin to kill M2-like macrophages will be reached within the blood stream, one may question whether this also occurs within the tumor microenvironment. The second effect concerns tumors in which the NF-κB pathway, leading to the production of PGE₂ and/or IL-6, is already activated. Treatment of these tumors with cisplatin and carboplatin resulted in an increased production of these 2 inflammatory mediators and subsequently in a more pronounced skewing of monocyte differentiation toward the tumor promoting M2-like macrophages, reflected by their production of large amounts of IL-10, decreased production of IL-12, activation of tolerogenic STAT3 pathway, and a decrease in the immune potentiating STAT1 and STAT6 pathways. This effect should be considered detrimental to patients as it implies that, upon treatment with platinum-based regimens of PGE₂ and/or IL-6–producing tumors, the number of local tumor-promoting M2 macrophages may increase, helping the tumor to defy the chemotherapeutic treatment. This would fit well with the existing literature showing that chemoresistance of cervical and ovarian cancer is associated with increased levels of PGE₂ and IL-6 (12–15).

We studied 2 possible mechanisms responsible for this effect and showed that activation of the NF-κB pathway was required, which occurred most likely via the DDR pathway. Indeed platinum-containing chemotherapeutics act by binding to and causing crosslinking of DNA and thus may trigger the DDR pathway. Recently, several studies reported that upon DDR, NF-κB activation may result in the production of PGE₂ and IL-6 (35, 36). More recently, it was shown that excessive DNA damage-induced NF-κB activation promoted expression of IL-6 in HELA cells (37). Upon chemotherapy-induced DDR, the I-κB kinase (IKK) complex phosphorylates Iκ-Bα for ubiquination and proteasomal degradation, thus allowing the NF-κB complex to translocate to the nucleus (18, 38). As the DDR pathway-activated NF-κB–mediated production of PGE₂ and IL-6 only occurred in tumor cell lines originally producing sufficient amounts of IL-6 and/or PGE₂, an autocrine role for these cytokines can be envisaged ref. (39; Fig. 7).

STATs represent central regulators of cancer-associated inflammation and influence interactions between cancer cells and their immune microenvironment that determine whether the inflammation promotes or inhibits cancer. The majority of gynecologic tumors have high STAT3 activity and this was associated with poor survival and chemoresistance (40). Interestingly, it has been reported that COX-2 is a transcriptional target of STAT3 signaling (41). Furthermore, high COX-2 expression results in enhanced STAT3 phosphorylation in an IL-6–dependent manner (42). Together, this may form an autocrine mechanism. Notably, STAT3 and NF-κB interact at multiple levels, thereby promoting protumorigenic inflammatory conditions in the tumor microenvironment (increased PGE₂ and IL-6 production), increasing tumor cell proliferation and survival as well as chemoresistance, tumor angiogenesis, and metastasis (18). Our data suggest that the autocrine signaling loops of PGE₂ and IL-6 in the tumor cells are enhanced by the NF-kB–activating signals of the DDR pathway induced by platinum treatment.

Tumor-produced IL-6 induced the activation of the STAT3-signaling pathway in M2 macrophages, a signaling pathway that is known to be activated in tolerogenic APC (19, 20) In contrast, STAT1 and STAT6 can support antitumor immunity (18). We found a reduction of pSTAT1 and pSTAT6 in M2-like macrophages and showed that this was PGE₂ dependent. These effects were enhanced when cancer cells were treated with cisplatin and carboplatin and sustain the notion that platinum-based chemotherapy may indirectly skew the local immune environment to a more tolerogenic and tumor-promoting milieu in cancers actively producing PGE₂ and/or IL-6. In a mouse tumor model, platinum-based therapy caused immunogenic cell death, thereby activating local APC and enhancing antitumor T-cell responses. Cisplatin triggered the release of the TLR4-stimulating protein HMGB-1 (43, 44). We previously showed (9), and confirm here, that tumor-induced macrophages respond to the TLR4 agonist LPS as reflected by the production of IL-10. Therefore, it is highly likely that chemotherapy-mediated release of HMGB-1 may not only activate DC, but activate local tumor-promoting macrophages as well. Because tumor-associated macrophages can easily outnumber tumor-infiltrating DC, the overall effect may be less beneficial in case there is an excess of M2 macrophages.

Chemoresistance was thought for a long time to arise as a consequence of cell intrinsic genetic changes, including upregulation of drug efflux pumps, activation of detoxifying enzymes, or apoptotic defects. Recent evidence suggests that resistance to chemotherapy can also result from cell extrinsic factors such as cytokines and growth factors (45), implying an important role for the tumor microenvironment. Our data support this hypothesis and suggest an immunologic explanation for the correlation of high COX-2 expression and IL-6 levels with poor response to treatment. We found no correlation of PGE₂ and/or IL-6 production of the cancer cell lines with chemoresistance of the cancer cell lines, whereas several other studies show that high COX-2 expression and high levels of IL-6 in serum and ascites identified patients with a poor response to cisplatin and unfavorable prognosis (13, 17, 46). This effect was explained by enhanced apoptosis resistance of the cancer cells but this was not reflected by our in vitro tests where cancer cell lines were incubated with different doses of cisplatin and carboplatin. Our observation that platinum-containing chemotherapy of cell lines with high COX-2 and/or IL-6 expression promoted the differentiation of monocytes to tumor-promoting M2-like macrophages substantiates the notion that also cell extrinsic factors such as cytokines play a role in chemoresistance (48). In our opinion, chemoresistance of a tumor may not only be mediated by the resistance of the cancer cells themselves but can occur through a skewed tumor microenvironment that is geared to promote tumor cell growth for instance through the action of M2 macrophages. This fits well with the finding that in glioma, the COX-2 pathway promotes gliomagenesis by supporting the development of myeloid suppressor cells in the tumor microenvironment (47). If so, blocking the tumor-promoting effects of PGE₂ and IL-6 might even enhance the sensitivity of otherwise more resistant tumors to chemotherapy, but this will require new studies.

Previously, we showed that monocytes differentiated in the presence of HELA, CC8, and CSCC7 or its supernatant induced an M2-like phenotype and that treatment with indomethacin (COX-inhibitor) and mAb against both IL-6 and IL-6R fully restored their phenotype and functionality toward that of DC. In recent clinical studies, where the effects of IL-6 were targeted with the mAb siltuximab against IL-6, the results were discordant (48–50). One explanation given was the fact that if tumors produce high levels of IL-6, it will be difficult to neutralize all the soluble IL-6 present. In our experiments, we used a mAb against both the cell-surface bound form and the soluble form of the IL-6R called tocilizumab. This antibody, which successfully blocks the detrimental effects of IL-6 in the clinic when treating rheumatoid arthritis and Castleman disease (31, 32), was capable to fully block the effects of IL-6 produced by untreated or chemotherapy-treated tumor cells. Pilot experiments revealed that this antibody to some extent also blocked IL-6 signaling in the tumor cells themselves (Supplemental Fig. S4).

In summary, chemoresistance of tumors have long been associated with the activation of COX-2 and the production of IL-6. Our data showed that tumor cells do not produce the same levels of IL-6 and/or PGE₂, and that there was no direct correlation between the efficacy of cisplatin and carboplatin to kill tumor cells and the levels of these cytokines_. In contrast, we found that these chemotherapeutic compounds elicited the NF-κB pathway leading to an increased production of PGE₂ and IL-6 in tumors that actively produce these inflammatory mediators. As IL-6 or PGE₂ by themselves or in combination can skew M2 macrophage differentiation, chemotherapeutic treatment of tumors will favor the differentiation of M2-like tolerogenic macrophages, despite the differences in levels produced by tumor cells. In the end, this will result in a stronger immune suppressive tumor-promoting tumor microenvironment known to be associated with therapy resistance (8). Therefore, our data suggest that a chemotherapy-induced increase in the number of intratumoral tumor-promoting M2 macrophages forms an indirect mechanism underlying chemoresistance. Successful blockers of PGE₂ and the IL-6/IL-6R pathway are already in the clinic. It will be of great interest to study the effects of a combined therapy of cytotoxic agents and these clinically available compounds in patients with apparently chemoresistant tumors.

No potential conflicts of interest were disclosed.

Conception and design: E.M. Dijkgraaf, J.W.R. Nortier, M.J.P. Welters, J.R. Kroep, S.H. van der Burg

Development of methodology: E.M. Dijkgraaf, M. Heusinkveld, B. Tummers, L.T.C. Vogelpoel, M.J.P. Welters

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E.M. Dijkgraaf, M. Heusinkveld, B. Tummers, L.T.C. Vogelpoel, R. Goedemans

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.M. Dijkgraaf, L.T.C. Vogelpoel, R. Goedemans, V. Jha, M.J.P. Welters, J.R. Kroep, S.H. van der Burg

Writing, review, and/or revision of the manuscript: E.M. Dijkgraaf, M. Heusinkveld, L.T.C. Vogelpoel, J.W.R. Nortier, M.J.P. Welters, J.R. Kroep, S.H. van der Burg

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B. Tummers, L.T.C. Vogelpoel, R. Goedemans

Study supervision: M.J.P. Welters, J.R. Kroep, S.H. van der Burg

The authors thank Drs. Thorbald van Hall and Bianca Querido for authenticating all cell lines.

M.J.P. Welters was financially supported by a grant from the Dutch Cancer Society 2009-4400 and B. Tummers and R. Goedemans are supported by NWO 40-00812-98-09012.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.