Tumor location determines tissue-specific recruitment of tumor-associated macrophages and antibody-dependent immunotherapy response

See allHide authors and affiliations

Science Immunology  06 Jan 2017:
Vol. 2, Issue 7, eaah6413
DOI: 10.1126/sciimmunol.aah6413

TAMpering with tumors

Immunotherapeutic antibodies are a promising cancer therapy, but little is known about the nontargeted effects of these antibodies on immune cells through Fc receptor binding. Tumor-associated macrophages (TAMs) and tumor-associated neutrophils (TANs), which have been implicated in both promoting and inhibiting tumor growth, express abundant Fcγ receptors. Lehmann et al. examined these cells in tumors growing in different sites—skin and lung. They found that the organ environment in which a tumor grows determined which TAM and TAN subpopulations contributed to antibody-dependent tumor immunotherapy. Different cellular pathways underlay the antibody response in different tissues. These data may help fine-tune therapeutic strategies to target cells that promote tumors and not those that fight them.


Despite recent advances in activating immune cells to target tumors, the presence of some immune cells, such as tumor-associated macrophages (TAMs) or tumor-associated neutrophils (TANs), may promote rather than inhibit tumor growth. However, it remains unclear how antibody-dependent tumor immunotherapies, such as cytotoxic or checkpoint control antibodies, affect different TAM or TAN populations, which abundantly express activating Fcγ receptors. In this study, we show that the tissue environment determines which cellular effector pathways are responsible for antibody-dependent tumor immunotherapy. Although TAMs derived from Ly6Chigh monocytes recruited by the CCL2-CCR2 axis were critical for tumor immunotherapy of skin tumors, the destruction of lung tumors was CCL2-independent and required the presence of colony-stimulating factor 2–dependent tissue-resident macrophages. Our findings suggest that TAMs may have a dual role not only in promoting tumor growth in certain tissue environments on the one hand but also in contributing to tumor cell destruction during antibody-mediated immunotherapy on the other hand.


In many solid tumors, such as malignant melanoma, an infiltration with T cells correlates with a good prognosis and is a predictive factor for responding to tumor immunotherapy (1, 2). In contrast, the abundant presence of tumor-associated macrophages (TAMs) or tumor-associated neutrophils (TANs) is mostly indicative of a poor prognosis in breast, cervical, and bladder cancer, for example (35). Many tumor cells—including ovarian, squamous cell, and cervical carcinoma—were shown to produce the chemokine CCL2 [also called monocyte chemoattractant molecule–1 (MCP-1)], which recruits CCR2-positive inflammatory (or classical) monocytes to the tumor, where they can differentiate into TAMs and dendritic cells (3, 611). Apart from CCL2, high levels of colony-stimulating factor 1 (CSF1) were shown to correlate with a high level of TAM and poor prognosis in breast and ovarian cancer (3, 12). TAM can support tumor growth by the production of proangiogenic factors, which are essential for connecting the tumor to the vasculature (9, 13). In addition, TAM can suppress T cell responses by triggering inhibitory receptors on T cells including CTLA4 and PD-1, by secretion of interleukin-10 (IL-10) and transforming growth factor–β, or by recruiting regulatory T cells (9, 14, 15). Moreover, in a model of virus-induced carcinogenesis, proinflammatory signals mediated through activating Fcγ receptors (FcγRs), which are broadly expressed on myeloid cells, were shown to further enhance tumorigenesis, suggesting that TAMs and the activating FcγR system are potent drivers of tumorigenesis (16). On the basis of this concept, strategies that aim to prevent the recruitment of inflammatory monocytes to the tumor by targeting CSF1 receptor signaling or the CCL2-CCR2 axis, for example, were developed (5, 9, 1719).

On the other hand, myeloid cells, such as the subset of patrolling monocytes, were shown to control tumor metastasis (20), and the presence of TAM correlated with a good prognosis in stomach cancer and some but not all patient cohorts with prostate and lung cancer (3, 21, 22). Furthermore, tumor-specific antibodies, such as Herceptin, which recognizes the Her2/Neu receptor on a variety of solid tumors, are used successfully in human therapy of cancer (2325) and result from preclinical animal model systems, including breast cancer, lymphoma, and melanoma, suggest that cellular FcγRs are central mediators of therapeutic antibody activity in vivo (2629). The set of activating FcγRs involved in the activity of most therapeutic antibodies suggests that myeloid cells, and not natural killer (NK) cells, are responsible for tumor cell depletion (3037). With regard to antibody-dependent depletion of B cells, there is convincing evidence that B cells are removed in the liver through monocyte and tissue-resident macrophage populations (32, 38, 39). Moreover, it was demonstrated that one underlying mechanism of the immunomodulatory activity of CTLA4-specific antibodies is through intratumoral depletion of regulatory T cells via myeloid effector cells (40). In summary, these studies provide evidence that, in a solid tumor environment, myeloid cells may be critical for therapeutic and immunomodulatory antibody activity, although it is largely unclear whether these already reside in the tumor or become recruited to the tumor via the blood or from neighboring tissues. Considering the vast heterogeneity of tissue-resident macrophages and blood-derived monocyte populations, a quite complex array of possible effector cell populations expressing overlapping sets of activating FcγRs emerges (4145). Distinguishing between the contributions of these different myeloid cell subsets to therapeutic antibody activity will be critical to guide future strategies that aim to optimize antitumor antibody activity.

In this study, we set out to identify effector cell populations responsible for killing tumor cells that grow in different anatomical sites, namely, skin and lung. We show that, although antibody-dependent tumor cell killing in the skin was essentially dependent on the CCL2-dependent recruitment of Ly6Chigh inflammatory monocytes from the blood, depletion of melanoma metastasis in the lung was independent of the CCL2-CCR2 axis and rather required the presence of CSF2-dependent tissue-resident macrophage populations.


Skin and lung tumor tissues are characterized by distinct TAM populations

To study how the organ environment in which a tumor grows affects the cellular and molecular pathways underlying tumor-specific antibody activity, we chose a syngeneic orthotopic mouse melanoma model in which tumor growth can be induced either subcutaneously or in the lung, depending on the route of tumor cell injection. Because genetically identical tumor cells are used, this approach allows one to focus on host factors, such as different immune cell subsets present in the skin and the lung, which may contribute to therapeutic antibody activity. Consistent with previous studies in different in vivo model systems, the capacity of the gp75-specific mouse immunoglobulin G2c (IgG2c) antibody TA99 was abrogated in mice deficient in all FcγRs or in mice selectively lacking FcγRI and FcγRIV expression (Fig. 1, A to C) (36, 37, 46, 47). The prominent role of these activating Fc receptors (FcRs) would suggest that either Ly6Clow (nonclassical or resident/patrolling) monocytes or tissue-resident macrophages, which are the only two cell types expressing this combination of activating FcγRs (32), and not NK cells (FcγRIII-positive) or neutrophils (FcγRIII/FcγRIV-positive), may be key effector cells for antibody-dependent tumor cell killing. Consistent with this interpretation, reduction of neutrophil numbers with a Ly6G-specific antibody did not affect antitumor antibody activity (Fig. 1, C to F). Moreover, TA99 could limit tumor growth in growth factor independence–1 (Gfi-1) knockout mice, which lack mature neutrophils (Fig. 1, G to J) (48). This suggests that TANs may at least not be the dominant effector cell responsible for antibody-mediated tumor destruction in the skin or lung. To investigate which other innate immune subsets were present in malignant lesions, we analyzed the immune cell infiltrate in more detail. As shown in fig. S1 (A and B) and Fig. 2, a heterogeneous mixture of CD68+ macrophages expressing all or select combinations of activating FcγRs was present within or around melanoma tumors growing in the skin. With respect to FcγRIV expression, F4/80-positive macrophages showed the highest level of expression, whereas Gr1-positive neutrophils appeared to express barely detectable amounts of FcγRIV, consistent with previous results (fig. S2 and Fig. 2E) (32). CD68+ macrophages with ingested tumor material were abundantly present within tumors (Fig. 2B). A more detailed analysis demonstrated that CD68/CD11c/major histocompatibility complex II (MHCII)–positive TAMs and CD68-negative, CD11c/MHCII-positive dendritic cells could be abundantly detected especially in the case of subcutaneous tumors (fig. S1A and Fig. 2, C to E). In lung tumors, the very bright expression of CD11c and CD68 and the lower expression of MHCII and CD11b on TAM suggest that these cells are predominantly alveolar macrophages (fig. S1B) (49). A similar predominance of CD11c-positive alveolar macrophages became evident in a human lung cancer tumor sample, suggesting that the findings obtained from this mouse lung tumor model may be relevant at least for select human lung tumors (fig. S1C).

Fig. 1 Involvement of activating FcγR and neutrophils for antibody-mediated protection against melanoma growth in the skin and lung.

C56BL/6 (BL/6) mice and mice lacking FcγRI and FcγRIV (FcγRI/FcγRIV−/−) or all activating FcγRs (FcRγ-/-) were injected subcutaneously (A) or intravenously (B) with B16F10 cells and left untreated or treated with tumor-specific antibody TA99. (A) Relative skin tumor weight (in percentage) of TA99-treated compared with PBS-treated groups 11 days after tumor cell injection (n ≥ 7 mice per group from at least four experiments; Mann-Whitney test, ***P < 0.001, *P < 0.05). (B) Relative numbers of lung tumor colonies of TA99-treated compared with PBS-treated groups 11 days after intravenous injection of B16F10 cells (n ≥ 7 mice per group from at least three independent experiments; t test, ***P < 0.001, *P < 0.05). (C) Representative images of tumor-infiltrating neutrophils in subcutaneous and lung tumors as detected by immunofluorescence analysis with the indicated antibodies. Scale bars, 100 μm. (D) Numbers of neutrophils in blood of C57BL/6 mice treated with the neutrophil-specific antibody 1A8 1 day before and daily during the first week of tumor immunotherapy with the TA99 antibody at days 0 and 3 (n ≥ 6 from one representative experiment; t test, ***P < 0.001) and in lungs at day 1 (n ≥ 6; t test, **P < 0.01). (E and F) Mean tumor weight of subcutaneous tumors (n ≥ 9 from three independent experiments; Mann-Whitney test) (E) and the number of lung tumor colonies (n = 5; t test) (F) in C57BL/6 mice after TA99 or PBS therapy without and with pretreatment of the neutrophil-depleting antibody 1A8 (***P < 0.001). (G and H) Mean tumor weight of subcutaneous tumors (G) (n = 8 from two experiments; t test, *P < 0.05) and the number of lung tumor colonies (H) (n ≥ 8 from three experiments; t test, ***P < 0.001) in GFI1-deficient mice (GFI−/−) lacking neutrophils. (I and J) Detection of neutrophils (Gr1- or Ly6G-positive) in subcutaneous and lung tumors (I) and blood (J) (blue) of C57BL/6 and GFI−/− mice. Scale bars, 100 μm. SSC and FSC, side scatter and forward scatter, respectively.

Fig. 2 The tissue determines the tumor immune cell infiltrate.

(A) Overview of a representative hematoxylin and eosin–stained subcutaneous B16F10 tumor with corresponding immunofluorescent staining showing the influx of CD68+FcγRIV+ macrophages into the tumor capsule, the central and necrotic area (N) within the tumor. (B) Overlay of bright-light and immunofluorescent staining demonstrating that intratumoral CD68+FcγRIV+ macrophages engulf melanoma cells (arrows) in subcutaneous (left) and lung (right) tumors. (C) Immunofluorescent staining of subcutaneous or lung melanoma tumors showing the infiltration of CD11c+MHCII+CD68 (dendritic cells) and CD11c+MHCII+CD68+ macrophages. (D and E) Expression of FcγRI and FcγRIII on CD68+ (D) and FcγRIV on F4/80+ and/or CD11b+ macrophages (E) in subcutaneous and lung melanoma tumors. One representative of four independent experiments is shown. Scale bars, 500 μm [overview in (A)] and 100 μm (all other images).

Blood-derived TAMs are responsible for antibody-mediated control of skin tumors

To test whether macrophages found within skin tumors were recruited from the blood or were tissue-derived, we generated bone marrow chimeric mice with a titrated irradiation approach, in which skin-resident macrophages remain largely host-derived (CD45.2), whereas hematopoietic cells originated from the bone marrow donor (CD45.1) (50). In contrast to the surrounding skin, the vast majority of macrophages localized within skin tumors were donor-derived, suggesting that they were recruited from the blood (Fig. 3, A and B). To provide functional evidence for a contribution of blood-derived monocyte subsets to therapeutic antibody–mediated destruction of skin tumors, we generated bone marrow chimeric mice with FcγR-sufficient hosts engrafted with FcγR-deficient bone marrow and vice versa. As shown in Fig. 3C, only those chimeric animals in which FcγRs were present on hematopoietic cells responded to tumor immunotherapy. Previous studies have demonstrated that TAMs may derive at least in part from Ly6Chigh monocytes (also referred to as classical or inflammatory monocytes) (9, 51), prompting us to investigate whether this monocyte subset may play an active role during tumor immunotherapy. Ly6Chigh monocytes express activating FcγRI and FcγRIII but lack FcγRIV expression during the steady state (32). Upon stimulation with proinflammatory cytokines or during differentiation into macrophages, such as osteoclasts, they can up-regulate FcγRIV expression (fig. S3) (52), demonstrating that they express all relevant activating FcγRs shown to be essential for antibody-dependent tumor immunotherapy. Incubating monocytes with tumor material also resulted in an up-regulation of FcγRIV on monocytes, suggesting that FcγRIV up-regulation may occur once inflammatory monocytes enter the tumor environment (Fig. 4A). Ex vivo, all monocyte subsets could ingest tumor cells in an antibody-dependent manner, providing evidence for a potential role of inflammatory monocyte–derived TAMs in antitumor antibody activity (Fig. 4B). Because therapeutic antibody activity was normal in Nr4a1 knockout mice lacking Ly6Clow monocytes (53), we concentrated on the Ly6Chigh monocyte subset (Fig. 4, C and D).

Fig. 3 Antibody-mediated destruction of subcutaneous tumors requires FcγR expressing blood-derived macrophages.

(A) Infiltration of host (CD45.2)– and donor (CD45.1)–derived macrophages into subcutaneous tumors in C57BL/6 (CD45.2) mice 4 weeks after irradiation with 6 Gy and reconstitution with bone marrow of CD45.1 animals by immunohistochemistry. Although most of the tumor-infiltrating macrophages are derived from the bone marrow donor (CD45.1), skin- and liver-resident macrophages are predominantly recipient (CD45.2)–derived. (B) Quantification of the chimerism in the blood (n = 8), liver (n = 7), tumor (n = 5), and skin (n = 3) of C57BL/6 mice (CD45.2) reconstituted with bone marrow from CD45.1 donors. (C) Tumor weight in the indicated mouse strains 13 days after subcutaneous injection of melanoma cells in the presence (TA99) or absence (Untreated) of antibody immunotherapy. To study the role of FcγR expression on hematopoietic or irradiation-resistant tissue-resident cells, we reconstituted FcγR-deficient mice with bone marrow from C57BL/6 (FcRγ−/− rec. BL/6), and we reconstituted C57BL/6 mice with bone marrow from FcγR-deficient animals (BL/6 rec. FcRγ−/−). One representative of two independent experiments with at least four mice per group is shown. n.d., not detected. Scale bars, 100 μm.

Fig. 4 Expression of FcγRIV on inflammatory monocytes and antibody-dependent tumor cell uptake by innate immune effector cells.

(A) Up-regulation of FcγRIV on monocytes cultured in the presence or absence of B16F10 tumor suspensions for 4 hours ex vivo. FACS plots of Ly6Cneg and Ly6Chigh inflammatory monocytes and the quantification of FcγRIV up-regulation on Ly6Chigh monocytes are shown. Because of the progressive down-modulation of CD62L expression during in vitro culture, FcγRIV expression was quantified on the CD62Llow subset (n = 5 biological replicates; t test, *P < 0.05). (B) Ex vivo tumor uptake by innate immune effector cells. Blood of C57BL/6 and FcRγ-deficient (Fcγ−/−) mice (n = 4 each) was cultured with PKH26-labeled B16F10 cells in the presence or absence of TA99, and uptake of B16F10 material by blood cells was determined by acquiring PKH26-dependent fluorescence. As a control, blood cells were cultured in the absence of tumor cells. NK cells, Ly6Chigh monocytes (Ly6C+ mono), Ly6Clow monocytes (Ly6C mono), neutrophils, and eosinophils are shown (Mann-Whitney test, *P < 0.05). (C and D) Weight of melanoma tumors 15 days after subcutaneous injection (C) and the number of metastases 11 days after intravenous injection (D) of B16F10 cells in C57BL/6 or Nr4a1-deficient mice in the presence or absence of tumor immunotherapy with the tumor-specific antibody TA99. (C) n ≥ 9 mice from three independent experiments; Mann-Whitney test, **P < 0.01. (D) n ≥ 10 mice from three independent experiments; t test, **P < 0.01, ***P < 0.005.

The CCL2-CCR2 axis is critical for antibody-mediated killing of skin tumors

The chemokine CCL2 and its receptor CCR2 were demonstrated to be essential for recruitment of inflammatory monocytes into peripheral organs and solid tumors, resulting in their development into TAMs (10, 11, 19, 54). To test whether the CCL2-CCR2 axis is essential for effective tumor immunotherapy, we studied antibody activity in CCR2-deficient mice. In these animals, the number of Ly6Chigh monocytes in the blood is diminished (fig. S4A) (54). Consistent with a role of Ly6Chigh monocytes, the antitumor antibody activity of the TA99 antibody was severely impaired in mice lacking CCR2 (Fig. 5, A to C). Furthermore, CCL2 production was detectable in situ in TAMs present in subcutaneous tumors and in melanoma tumor supernatants from subcutaneous but not from lung tumors (Fig. 5, D and E). FcγRIV expression on intratumoral TAM and TAN populations did not differ between solid tumors growing in C57BL/6 versus CCR2-deficient mice, excluding the idea that alterations in receptor expression accounted for the phenotype (fig. S2). Additional evidence that blood-derived CCR2-expressing cells were essential for successful tumor immunotherapy was provided by bone marrow chimeric mice in which CCR2 expression was absent selectively on hematopoietic cells (Fig. 5F). Despite the normal expression of activating FcγRs, the absence of CCR2 on hematopoietic cells resulted in a loss of antibody activity, strongly suggesting that the recruitment of inflammatory monocytes to the tumor is essential for antibody-dependent tumor immunotherapy. Because B16F10 tumor cells did not produce CCL2 (fig. S5A), this suggested that the interaction of tumor cells with tumor-infiltrating hematopoietic cells triggered a CCL2 production. Injection of tumor cells into mice resulted in an immediate but transient CCL2 production, which was blocked if the antitumor antibody was co-injected for immediate depletion of tumor cells (fig. S5, B and C). In a similar manner, the injection of a human melanoma cell line into C57BL/6 mice triggered a transient CCL2 production (fig. S5D). Potential reservoirs from which Ly6Chigh monocytes may become recruited to the tumor include the spleen, lymph nodes, and the bone marrow (55). To test this, we analyzed antitumor antibody activity in splenectomized and tumor necrosis factor–α (TNFα)/lymphotoxin-α (LTα)–deficient mice, lacking lymph nodes and an intact splenic architecture. However, in both instances, antibody-dependent inhibition of tumor growth remained intact (Fig. 5, G and H). Moreover, other cell types such as mast cells or basophils were also dispensable for antibody activity (fig. S6), suggesting that inflammatory monocytes recruited from the bone marrow via CCL2 were essential for antibody-mediated tumor immunotherapy of melanoma cells growing subcutaneously.

Fig. 5 Importance of CCL2-CCR2 axis for therapeutic antibody–mediated protection against subcutaneous melanoma.

(A to C) Loss of antibody-mediated tumor killing in CCR2−/− mice. The mean tumor weight (A) (n ≥ 12 from four experiments; t test) or total photon flux (B and C) in mice that received B16F10 (A) or luciferase-expressing B16F10-luc (B and C) (n = 3) melanoma cells subcutaneously followed by treatment with the tumor-specific antibody TA99 or left untreated is shown. (D) Cytokines in supernatant of B16F10 melanomas isolated from the skin [subcutaneously (sc)] or lung (n ≥ 6; t test). (E) Detection of CCL2 production by immunofluorescent analysis of subcutaneous and lung tumor sections. CD68- and Gr1-specific antibodies were used to identify TAMs and TANs, respectively. Arrows mark CD68/CCL2 double-positive cells. (F) Weight of melanomas 15 days after subcutaneous injection in C57BL/6 or mice in which CCR2 expression is restricted to irradiation-resistant tissue-resident cells (BL/6 rec. CCR2−/−) in the presence or absence of tumor immunotherapy with the tumor-specific antibody TA99 (n ≥ 5; t test). (G and H) Secondary lymphoid organs are dispensable for TA99-mediated protection from melanoma growth. The weight of subcutaneous melanomas in splenectomized or sham-operated C57BL/6 animals (G) (n = 4 per group; Mann-Whitney test) and in TNFα/LTα-deficient mice (H) (n ≥ 5 per group, t test) in the presence or absence of antibody immunotherapy is depicted. One of two representative experiments is shown. Scale bars, 100 μm. *P < 0.05, **P < 0.01.

To obtain further evidence that the CCL2-CCR2 axis is of general importance for antibody-mediated destruction of a subcutaneous tumor, we turned to another tumor model system in which subcutaneously growing EL4 lymphoma cells are depleted with a Thy1.2-specific antibody 30H12, which is a rat IgG2b subclass (fig. S7). Consistent with our previous study, the EL4 tumors were infiltrated by a heterogeneous mixture of TAMs and dendritic cells, and the therapeutic activity of the 30H12 antibody was abrogated in CCR2-deficient mice, in which antibody therapy rather showed a tumor-promoting activity.

CSF2-dependent tumor-infiltrating macrophage populations are critical for antibody-mediated tumor cell killing in the lung

However, in contrast to subcutaneously growing melanoma tumors, antibody-dependent depletion of melanoma colonies in the lungs seemed to require different effector cells. Consistent with the lack of intratumoral CCL2 production (Fig. 5, D and E), CCR2-deficient animals responded normally to antibody-dependent immunotherapy of lung tumors (Fig. 6, A to C). Furthermore, blood monocyte depletion via clodronate liposomes did not affect antibody activity (Fig. 6, D to F). Clodronate treatment did not deplete lung-resident macrophages, suggesting that they may contribute to antibody activity (Fig. 6, E and F). To provide direct evidence that lung-resident macrophages were responsible for antitumor antibody activity, we analyzed Csf2-deficient mice, which have a diminished number of functionally impaired alveolar macrophages but normal levels of monocyte subpopulations in the blood (fig. S5C) (56). CD11chighCD68+ macrophages were largely absent from tumors of Csf2-deficient mice (Fig. 6, G and H). Removal of lung tumors via TA99 was strongly impaired in Csf2-deficient mice, suggesting that lung-resident macrophages and not the recruitment of blood-derived inflammatory monocytes via CCL2 were critical for therapeutic antibody activity in the lung (Fig. 6, G and I). In an effort to further distinguish between the contribution of blood- and lung tissue–derived myeloid cell compartments to tumor immunotherapy (as shown in Fig. 3 for the skin), we again generated bone marrow chimeric animals in which either blood-derived or tissue-derived immune cells expressed activating FcγRs. However, in contrast to the skin and liver, lung-resident macrophages were almost fully replaced by blood-derived progenitors in the course of the experiment, not allowing us to address which role each cell population plays for antibody-mediated tumor destruction with this method (fig. S8).

Fig. 6 Involvement of Csf2-dependent lung-resident macrophages in antibody-mediated protection against lung melanoma.

(A to C) The CCR2-CCL2 axis is not required for TA99-mediated depletion of lung tumors. (A) The number of melanoma colonies in the lung of C57BL/6 and CCR2−/− mice that either received immunotherapy (+TA99) or were left untreated (n ≥ 18 from six experiments; t test). Representative image (B) and quantification of the total photon flux (C) in CCR2−/− mice that were injected with luciferase-expressing B16F10-luc cells and either received the melanoma-specific antibody TA99 or were left untreated (n = 4; Mann-Whitney test). (D to F) Blood-derived monocytes are dispensable for lung tumor immunotherapy. The amount of inflammatory (Ly6C+) and resident monocytes (Ly6C-) (D) and lung-resident CD68+ macrophages (E) in mice after injection of clodronate liposomes or control liposomes is shown. (F) The number of lung melanoma colonies in C57BL/6 mice with or without pretreatment with clodronate liposomes (Clodronate) in the presence (+TA99) or absence of immunotherapy (untreated) is shown (n ≥ 4 mice per group; Mann-Whitney test). f.o.v., field of view. (G to I) Csf2-dependent lung macrophages are critical for antitumor antibody activity. Representative immunofluorescent staining (G) and quantification (H) of CD11c+CD68+ alveolar macrophages and the expression of CD11c on CD68+ macrophages in lung tumors of C57BL/6 and Csf2−/− mice with or without antibody immunotherapy (±TA99; n = 4; Mann-Whitney test). (I) The number of lung melanoma colonies in C57BL/6 and Csf2−/− mice (n = 4 mice per group; Mann-Whitney test) in the presence or absence of immunotherapy. Scale bars, 100 μm. *P < 0.05, **P < 0.01, ***P < 0.001.


In this study, we investigated how the tissue environment in which a solid tumor grows determines the effector cells contributing to therapeutic antibody–mediated control of tumor growth. Although it has become clear that myeloid cells play a key role for cytotoxic antibody–mediated depletion of normal and malignant B cells, for example, it is largely unknown how the environment of a solid tumor growing in different tissues affects the molecular and cellular pathways underlying antibody activity (30, 38, 39, 57). By using a syngeneic melanoma model that allows us to induce tumor growth in the skin and the lung, we now show that different TAM subsets reside in melanoma tumors growing in different organ environments and that only select TAM subpopulations contribute to therapeutic antibody–dependent tumor cell killing.

In line with previous studies, the FcγR system was critical for therapeutic antibody activity irrespective of the organ in which the tumor grew (3639, 46, 47, 5760). However, the cell types responsible for antibody-dependent control of melanoma growth differed depending on the organ. Thus, the recruitment of Ly6Chigh monocytes via the CCL2-CCR2 axis was critical for antibody-dependent depletion of melanoma and lymphoma cells in the skin, whereas the absence of CCR2 did not impair antibody-dependent depletion of melanoma cells in the lung. Consistent with previous studies showing that recruitment of Ly6Chigh monocytes via CCR2 supports breast cancer metastasis in the lung (10, 11), we noted a reduced number of melanoma metastasis in the lung but not in the subcutaneous tumors, where a trend toward slightly larger tumor size became evident (Figs. 5 and 6). Thus, our results support the notion that, depending on the tissue environment, the recruitment of inflammatory monocytes via the CCL2-CCR2 axis can either promote tumor growth without contributing to antibody immunotherapy (in lung tumors) or may be a central component of antibody-mediated tumor rejection (in skin).

Consistent with the genetic data in CCR2-deficient animals, one distinguishing feature between the two tumor entities growing in a skin or lung environment was the much higher (or more sustained) level of CCL2 production in skin tumors and the predominant recruitment of TAM from the bone marrow (Figs. 3 and 5E). Although many tumor cells have been described to produce CCL2, B16F10 melanoma cells showed no autonomous production of this chemokine (3, 8, 61), which was rather produced by TAM. Functional evidence that bone marrow–derived cells are critical for controlling melanoma growth in the skin was provided by bone marrow chimeric mice in which the irradiation dose was titrated to predominantly exchange the hematopoietic system but to maintain tissue-resident macrophages in the skin. These studies demonstrated that activating FcγRs expressed on bone marrow–derived but not skin-resident macrophages were critical for therapeutic antibody activity.

Because the absence of neutrophils, mast cells, or basophils did not impair antibody activity and because the source of CCL2 was bone marrow–derived cells (Fig. 5F and fig. S2), a model in which inflammatory monocytes are recruited to the tumor where they differentiate into CCL2 producing TAM, which in turn leads to a positive feedback loop, seems likely. With respect to the lung, further studies will be necessary to investigate why CCL2 production by TAM is not maintained and why predominantly CD11chigh/CD68+/CD11bneg/MHCIIdim cells, most likely representing alveolar macrophages, can be found within the tumor. The titrated irradiation experiment to distinguish between bone marrow–derived and lung-resident macrophages did not yield clear results in the lung model, because alveolar macrophages (in contrast to skin-resident macrophages) became replaced by bone marrow–derived cells during the course of the experiment, consistent with previous results (44, 50, 56). Arguing against an involvement of blood-derived monocytes, treatment with clodronate liposomes, which diminished blood monocyte numbers but retained alveolar macrophages, did not impair antibody activity (Fig. 6, D to F). To provide more direct evidence that alveolar macrophages contribute to antibody-dependent tumor immunotherapy, we further used Csfr2−/− mice deficient in granulocyte-monocyte CSF (GM-CSF) in which the development and function of alveolar macrophages are impaired (56). Supporting a model in which alveolar macrophages are responsible for controlling tumor growth in the lung, antibody-mediated control of lung tumors was impaired in GM-CSF–deficient mice, consistent with a reduced level of tumor-associated CD11chighCD68+ macrophages (Fig. 6, G to I).

However, caution is needed when interpreting results from inbred and knockout animal model systems, in which a general deletion of target genes occurs. Csfr2−/− mice, for example, not only show reduced lung-resident macrophage numbers but also develop a pronounced proteinosis during age, which may affect immigration of other immune cell subsets. Moreover, deletion of the GM-CSF receptor on Ly6Chigh monocytes has been shown to affect central nervous system inflammation in a model of experimental autoimmune encephalitis, suggesting that Ly6Chigh monocyte function may be altered (62). In a similar manner, Gfi−/− mice show other alterations in immune system function apart from lacking mature neutrophils, such as altered B cell, T cell, and myeloid development (63). This may at least in part explain why an increase in tumor load was detectable in the lung model and why tumor eradication was not complete. Alternatively, we cannot completely rule out that neutrophils contribute at least in part in inhibiting tumor growth, as described previously (35). Despite the fact that we have noted a similar CCL2-dependent pathway underlying therapeutic antibody activity in two independent syngeneic skin tumor models, future studies in spontaneous tumor models growing in different locations will be necessary to identify which TAM subsets contribute to antibody-dependent tumor immunotherapy throughout the body. A previous study identified Kupffer cells as a possible effector cell population responsible for TA99-mediated depletion of melanoma metastasis in the liver (37). Thus, it seems plausible that, depending on the organ, either tissue-resident (lung and liver) or blood-derived (skin) TAM populations contribute to antitumor antibody activity. Last, future studies with samples of human tumors growing in the skin and the lung will need to show whether similar rules apply in the human system because immune cell subsets differ, especially those in the skin.

Together, our study shows that, despite a critical requirement of the same set of activating FcγRs for therapeutic antibody–dependent tumor cell killing, different cellular pathways were underlying tumor-specific antibody activity. These results may suggest that the local tissue environment is a strong determinant of which immune effector cells contribute to tumor rejection by therapeutic antibodies and show that approaches that aim to deplete TAM subsets by targeting the CCL2-CCR2 pathway may also target an important effector cell population essential for therapeutic antibody activity at least if the tumor resides in the skin. Our results may be helpful for optimizing clinical strategies that aim to limit the infiltration of tumor-supporting macrophages while maintaining TAM subsets that contribute to therapeutic antibody activity.


Study design

In the in vivo experiments, animals were randomly distributed into the different experimental groups. Upon injection of tumor cells, animals were split into treatment or control groups without blinding the investigator. The decision of how many animals needed to be included within one experimental group to allow statistical evaluation was based on previous studies in which the B16F10 melanoma model was used. A more detailed description of the experimental parameters and statistical calculations can be found in the following paragraphs and in the respective figure legends.


C57BL/6J mice were purchased from Janvier. PepBoy (CD45.1), Nr4a1−/−, and CCR2−/− mice were from the Jackson Laboratory. FcRγ chain–deficient (Fcγ−/−) and FcγRIV−/− mice on the C57BL/6 background were provided by J. Ravetch. FcγRI−/− mice were provided by M. Hogarth (64). FcγRI−/− and FcγRIV−/− mice were crossed to obtain FcγRI/FcγRIV−/− double-deficient mice. Cpa3Cre/+, Csf2−/−, Mcpt8-cre, and TNFα/LTα mice were provided by H.-R. Rodewald, G. Weber, D. Vöhringer, and T. Winkler, respectively. Mice between 6 and 14 weeks of age were used for individual experiments. All mice were maintained under specific pathogen–free conditions and according to the guidelines of the National Institutes of Health and the legal requirements in Germany.

B16F10 melanoma model

Experiments were performed as described previously (47). Briefly, pulmonary metastases were induced by intravenous injection of 5 × 105 B16F10 cells, whereas skin tumors were induced by subcutaneous injection of 5 × 104 B16F10 cells (American Type Culture Collection). Antibody treatment for pulmonary and subcutaneous tumors consisted of an intraperitoneal application of 100 μg (days 0, 2, 4, and 7) and 150 μg (days 0, 2, 4, 7, 9, and 11) of tumor-specific TA99-IgG2c (anti-gp75, BioXCell), respectively. This antibody was also referred to as an IgG2a subclass in previous publications. Because it originates from C57BL/6 mice, we referred to it as IgG2c in this study. Mice receiving treatment were chosen at random. The investigators were not blinded regarding allocation of animal groups. Mice were sacrificed at day 11 (lung model) or after formation of visible solid tumors between days 12 and 15. For in vivo imaging, a luciferase-expressing B16F10 clone (B16F10-luc) that was generated in our laboratory was used. To detect tumors in situ, mice were shaved and received an intravenous injection of 3 μg of d-luciferin. Bioluminescence was measured using an IVIS Imaging Systems (IVIS Lumina, Caliper Life Science).

EL4 tumor model

C57BL/6 and CCR2−/− mice were subcutaneously injected with 5 × 104 EL4 cells alone or in addition with 100 μg of anti-Thy1.2 (clone 30H12, BioXCell) in equal volumes of phosphate-buffered saline (PBS). Tumor growth was monitored over the next 2 weeks without blinding of the two experimental groups.

Generation of bone marrow chimeras

Recipient mice were irradiated with 6 Gy (gray), leaving tissue-resident macrophages largely unaffected while simultaneously exchanging the majority of the hematopoietic system after bone marrow transfer. Chimerism was evaluated 3 to 4 weeks later by flow cytometry of peripheral blood, and mice with a chimerism greater than 90% were used for further experiments.

Depletion of leukocyte populations

For depletion of neutrophils, 0.5 mg of Ly6G-specific antibody was injected intraperitoneally daily during the first week of the experiment (clone 1A8, BioXCell). Monocytes were depleted by intraperitoneal application of 100 to 200 μl of dichloromethylene bisphosphonate containing liposomes (clodronate liposomes, purchased from every other day in the first week of the experiment. The amount of clodronate liposomes was titrated before experiments for every batch. Depletion efficiency was verified by flow cytometry.

Flow cytometry

Single-cell suspensions of the different organs were incubated for 15 min with Fc block (anti-FcγRII/FcγRIII, clone 2.4G2; anti-FcγRIV, clone 9E9) on ice to prevent unspecific binding (omitted when respective receptors were stained) followed by staining with fluorochrome-coupled antibodies for 15 min. Antibodies against the following antigens were used: Ly6G (clone 1A8), NK1.1 (clone PK136), and CD45.2 (clone 104) conjugated to fluorescein isothiocyanate (FITC). CD45.1 (clone A20), CD11c (clone HL3), Gr1 (clone RB6-8C5), and SiglecF (clone E50-2440) were labeled with phycoerythrin (PE). Anti-CD11b–PerCp-Cy5.5 (clone M1/70), anti-CD45–allophycocyanin (APC)–Cy7 (clone 30-F11), anti-CD62L–PE-Cy7 (clone MEL-14), anti-F4/80 (clone BM8), and anti-CD68 (clone FA-11) were conjugated to Alexa Fluor 488, anti-FcγRIV–Alexa Fluor 647 (clone 9E9), and anti-MHCII–APC (clone M5/114.15.2). All antibodies were from BD, BioLegend, or AbD Serotec. Dead cells were excluded from analysis using 4′,6-diamidino-2-phenylindole. Measurements were carried out on a FACSCanto II flow cytometer (BD Bioscience). Briefly, single viable CD45+ leukocytes were divided into SSChigh granulocytes and further distinguished into neutrophils and eosinophils by Ly6G. SSClowCD11b+ cells negative for NK1.1 (NK cells) were classified as monocytes that were further characterized by expression of CD62L and Gr1 (inflammatory monocytes) or their lack of it (resident monocytes as well as FcγRIV+). For antibody staining, the fluorescence minus one control was used to deduct the cell subset–specific levels of autofluorescence in each channel. These values are indicated as the Δ in mean fluorescence intensity (ΔMFI). A representative isotype control staining with the different mouse and rat IgG subclasses used in fluorescence-activated cell sorting (FACS) analysis is shown in fig. S9.


Acetone-fixed sections of frozen tissues were incubated with 5% goat serum for 45 min to block unspecific binding followed by staining with the respective antibodies for 45 min at room temperature. The following antibodies were used: anti-CCL2–PE (clone 2H5, BioLegend), anti-CD11b–PE (clone M1/70, BD), anti-CD11c–PE (clone HL3, BD), biotinylated anti-CD45.1 (clone A20, BioLegend), anti-CD45–PE (clone A20, BD), biotinylated anti-CD45.2 (clone 104, BioLegend), anti-CD45.2–PE (clone 104, BD), anti-CD68–FITC (clone FA-11, AbD Serotec), anti-F4/80–Alexa Fluor 647 (clone MCA497, AbD Serotec), anti-F4/80–FITC (clone CI:A3-1, BioLegend), anti-FcγRI–Alexa Fluor 647 or anti-FcγRI–PE (clone X54-5/7.1, BD), anti-FcγRIII–PE (R&D Systems), anti-FcγRIV–Alexa Fluor 647 (clone 9E9), anti-Gr1–APC (clone RB6-8C5, BD), and anti-MHCII–Alexa Fluor 647 (clone M5/114.15.2, eBioscience). For staining of human lung sections, anti-CD11c–PE (clone 3.9, eBioscience), anti-CD68–FITC (cloneY1/82A, BD), and anti–HLA-DR–APC (clone L243, BioLegend) were used. When biotinylated antibodies were used, blocking of endogenous biotin was performed with an Avidin-Biotin Blocking Kit (Invitrogen) according to the manufacturer’s instructions. Sections were analyzed using an Axiovert 200M microscope and AxioVision software (Carl Zeiss AG).

Cytokine analysis

The measurement of cytokines from sera and tumor supernatants was performed with cytometric bead arrays (BD) according to the manufacturer’s instructions. For cytokine extraction from subcutaneous or lung tumors, tumor tissue was excised and passed through a 70-μm nylon mesh in 1 ml of PBS. After centrifugation, the supernatants were collected for further analysis. The amount of cytokines was normalized to the weight of the tumor tissue.

Ex vivo stimulation of blood leukocytes and B16F10 uptake

For ex vivo experiments, blood samples of mice were depleted of erythrocytes by hypotonic water lysis and split into two or three wells of a 96-well plate in 200-μl RPMI medium supplemented with 5% fetal calf serum, 1% penicillin/streptavidin, and 1% glutamine. For interferon-γ (IFN-γ) and CCL2 stimulation, 20 ng/ml of the respective cytokine (PeproTech) was added. For tumor cell uptake, B16F10 cells were labeled with PKH26 using the PKH26 Red Fluorescent Cell Linker Kit (Sigma-Aldrich) and cocultured with blood leukocytes in the presence or absence of the tumor-specific antibody TA99 (20 ng/ml). Cells were analyzed by flow cytometry after 4 hours of incubation at 37°C in a 5% CO2 atmosphere.


All data are means ± SEM. Statistical analysis was performed using GraphPad Prism software (GraphPad Software Inc.). Samples were tested for Gaussian distribution. When tested positive, t test was used for comparison of two groups, and analysis of variance (ANOVA) was used for comparison of multiple groups. If samples were not normally distributed, the Mann-Whitney test was used for comparing two groups of samples or the Kruskal-Wallis test was used for comparing multiple groups. Statistical analysis of tumor weight differences between treated and untreated groups could not be carried out when no tumor growth is detected in the treated group because of lack of variance. Significant differences were indicated by the exact P values.


Fig. S1. Leukocyte subsets in murine and human lung tumor tissue.

Fig. S2. FcγRIV-expressing immune cell subsets in subcutaneous tumors.

Fig. S3. FcγRIV expression on blood leukocytes after cytokine stimulation.

Fig. S4. Leukocyte distribution in CCR2- and CSF2-deficient mice.

Fig. S5. Cytokines in B16F10 cell culture supernatant and serum of tumor-bearing mice.

Fig. S6. Role of mast cells and basophils in antibody-dependent tumor immunotherapy.

Fig. S7. Role of the CCL2-CCR2 axis in antibody-mediated protection against EL4 subcutaneous tumors.

Fig. S8. Impact of lung tissue–resident cells on therapeutic antibody activity.

Fig. S9. Isotype control staining.

Table S1. Primary source data.


Acknowledgments: We thank H. Danzer, M. Kießling, and H. Albert for expert technical assistance. We are grateful to J. Ravetch (Rockefeller University), M. Hogarth (Burnet Institute), D. Vöhringer, G. Weber (University Hospital Erlangen), T. Winkler (University of Erlangen), and H.-R. Rodewald (Deutsches Krebsforschungszentrum Heidelberg) for providing mice. We thank B. Weigmann (University Hospital Erlangen) for access to the IVIS platform. B16F10-OVA and EL4 cells were provided by D. Dudziak. H. Sirbu (University Hospital Erlangen) provided the human lung tumor sample. S. Gross and N. Schaft provided the human melanoma cell line. Funding: This study was supported by a grant from the German Research Foundation (SFB643 to F.N. and D.D., WE4892/3-1 and 4/1 to G.F.W., and Vo944/7-1 to D.V.). Author contributions: B.L., M.B., C.B., A.I.-E., C.L., and S.G. performed experiments and analyzed the data. D.V., T.W., D.D., G.F.W., H.S., S.G., and N.S. provided essential reagents and helped with data analysis and interpretation. B.L. and F.N. wrote the manuscript. Competing interests: The authors declare that they have no competing interests.
View Abstract

Navigate This Article