Co-administration of RANKL and CTLA4 Antibodies Enhances Lymphocyte-Mediated Antitumor Immunity in Mice

Immunotherapy has shown great promise in the treatment of certain advanced solid organ malignancies, most prominently in melanoma. Breakthrough trials using ipilimumab (a monoclonal antibody targeting CTLA4, the archetypal T-cell immune checkpoint) demonstrated, for the first time, improved survival in advanced melanoma; recent pooled analysis of long-term survival data for ipilimumab monotherapy showed a plateau of 21% in the overall survival (OS) curve commencing at about 3 years (1). More recently, adjuvant ipilimumab improved survival in stage 3 resected melanoma (2). Combining anti-CTLA4 with an mAb-targeting another checkpoint, PD-1, produced superior tumor responses and survival benefit in advanced melanoma, demonstrating the importance of combination immunotherapy targeting non-redundant mechanisms of immune evasion by tumors (3–5). However, one challenge in the immunotherapy of solid cancers is the discovery of new targets for patients who display primary resistance to current immunotherapy combinations.

The possible utility of receptor activator of NF-κB ligand (RANKL) pathway blockade as a feasible therapeutic partner for anti-CTLA4 mAbs was suggested by a case report of a patient with rapidly progressive metastatic melanoma with skeletal and extra-skeletal metastasis, who received denosumab (anti-RANKL mAb) for palliation of bony pain, but then experienced an unexpected and rapid near-complete response to single agent ipilimumab (6). This report was accompanied by a preclinical experiment that showed a combination of anti-RANKL and anti-CTLA4 mAbs significantly reduced experimental mouse melanoma lung metastases via natural killer (NK) cells (6). A second case report of complete response following ipilimumab and denosumab in metastatic melanoma followed (7). However, to date the general utility and precise mechanisms of action of anti-RANKL and anti-CTLA4 have remained unclear.

RANK and RANKL are members of the tumor necrosis factor receptor and ligand superfamilies, respectively, with closest homology to CD40 and CD40L. RANK (TNFRSF11a) and RANKL (TNFSF11) are currently best known in clinical practice for their role in bone homeostasis, as the differentiation of osteoclasts from the monocyte–macrophage lineage requires RANKL interaction with RANK expressed on the myeloid osteoclast precursors (8, 9). However, RANKL was initially identified as a dendritic cell (DC)–specific survival factor, which was upregulated by activated T cells and interacted with RANK on the surface of mature DCs to prevent apoptosis (10, 11). The fully human IgG2 anti-RANKL antibody (denosumab) is widely used in clinical practice as a potent and well-tolerated antiresorptive agent for the prevention of skeletal-related events arising from bone metastases, and the management of giant cell tumor of bone and osteoporosis (12, 13). Intriguingly, denosumab increased OS in a post-hoc exploratory analysis of a phase III trial in patients with non–small cell lung cancer and bone metastases, compared with zoledronic acid (14).

Taken together, the investigation of RANK and RANKL as possible novel immunotherapy targets in cancer is a rational approach. Here we have defined the mechanism of action of RANKL–RANK blockade in combination with anti-CTLA4, and provide insight into the combination efficacy observed in the case reports.

Mouse melanoma cell lines B16F10 (ATCC) and LWT1, colon cancer cell line CT26, and prostate carcinoma cell line RM-1 were maintained, injected, and monitored as previously described (15–18). The mouse prostate cancer cell line Tramp-C1 was maintained as described previously (19) but without dehydrosolandosterone. The mouse fibrosarcoma cell line MCA1956 (derived from a MCA-inoculated C57BL/6 mouse) was kindly provided by Robert Schreiber (Washington University School of Medicine, St Louis, MO). All cell lines were routinely tested negative for Mycoplasma, but cell line authentication was not routinely performed.

C57BL/6 wild-type (WT) mice were bred-in-house or purchased from the Walter and Eliza Hall Institute for Medical Research. C57BL/6 perforin-deficient (pfp^−/−), interferon-deficient (IFNγ^−/−), Fc receptor deficient (FcγRIII, FcγRIV, and FcϵRγ), Batf3 transcription factor–deficient (Batf3^−/−; ref. 20) and FoxP3-DTR (21) mice were bred in-house at the QIMR Berghofer Medical Research Institute (QIMRB). All mice were used between the ages of 6 to 16 weeks. Groups of 5 to 13 mice per experiment were used for experimental tumor metastasis assays and subcutaneous tumor growth. All experiments were approved by the QIMRB Animal Ethics Committee.

Purified anti-mouse anti-RANKL (IK22-5; rat IgG2a; ref. 22), anti-CTLA4 (UC10-4F10; hamster IgG) and control antibodies (1-1 or 2A3; rat IgG2a) were produced in house or purchased from BioXcell (West Lebanon, NH). Anti-CTLA4 clone 9D9 (various isotypes as indicated), and control antibody (1D12; mouse IgG2a), were supplied by Bristol–Myers Squibb. Recombinant human OPG-Fc (consisting of the ligand binding residues of OPG (residues 22–194) fused to the Fc portion of human IgG1; ref. 23) was supplied by Amgen Inc. Antibodies to deplete NK cells (anti-asialoGM1, Wako) or anti-CD8β (53.5.8, BioXcell) were administered as indicated.

For B16F10 (1 × 10⁵), RM-1 (5 × 10⁴), CT26 (1 × 10⁵), MCA1956 (1 × 10⁶) and Tramp-C1 (1 × 10⁶) tumor formation, cells were inoculated subcutaneously into the abdominal flank of female (B16F10, MCA1956) or male (RM-1, CT26, Tramp-C1) mice. Therapeutic antibody treatment commenced as indicated on day 3 to 12 after tumor inoculation and was given every 2 to 4 days up to a maximum of 4 doses. Digital callipers were used to measure the perpendicular diameters of the tumors. The tumor size was calculated as the product of the two measurements and is presented as mean ± SEM.

Single-cell suspensions of B16F10 (2 × 10⁵), RM-1 (1 × 10⁵), or LWT1 (7.5 × 10⁵) were injected intravenously into the tail vein of the indicated strains of mice. Lungs were harvested on day 14, and surface tumor nodules were counted under a dissection microscope. Antibody treatments were as indicated, with anti-CTLA4 and/or anti-RANKL mAbs administered on days −1, 0, and 2 relative to tumor inoculation. Antibodies to deplete CD8⁺ T cells or NK cells were administered where indicated on days −1, 0, and 7 relative to tumor inoculation.

Tumor-bearing mice were sacrificed at two time points: day 9, or at end-point (when the experiment was terminated due to tumors reaching ethical endpoint size). Tumor and spleen were collected and wet weight was recorded. Single-cell suspensions were generated from indicated organs as previously described (21).

The following antibodies (from Biolegend, eBioscience, BD Biosciences) were used: CD4-BV605 (RM4-5), CD8-BV711 (53-6.7), CD11b-BV650 (M1/70), CD11b-PE (M1/70), CD11c-PE (N418), purified CD16.2 (9E9) followed by goat-anti-hamster FITC, CD206-AF647 and CD206-PECy7 (C068C2), and Zombie Aqua live/dead dye; TCRβ-PerCP-Cy5.5 (H57-597), CD45.2-A780 (104), Ly6C/Ly6G (GR-1)-EF450 (RB6-8C5), MHCII-APC (M5/114.15.2), CD265 (RANK)-PE (R12-31); RANKL-AF647 (IK22-5). For intracellular cytokine staining (ICS), cells were stimulated for 4 hours with Cell Stimulation Cocktail (1/1,000; eBioscience). Cells were then surface stained as described above before being fixed/permeabilized with Cytofix/Cytoperm (BD Biosciences) and stained with IFNγ-AF488 (Biolegend), TNFα-PE (BD Biosciences), and IL-2-Pacific Blue (Biolegend).

For intracellular transcription factor staining, cells were surface stained as described above before being fixed and permeabilized using the Foxp3/Transcription Factor Staining Buffer Set (eBioscience), according to the manufacturer's protocol and stained with FoxP3-EF450 or FoxP3-AF488 (FJK-16s) and Ki67-EF450 (Sol185; eBioscience). All immune cell analysis was first gated on live, single CD45.2⁺. T cells were defined as TCRβ⁺NK1.1⁻. NK cells were defined as TCRβ⁻NK1.1⁺. DCs were defined as CD11c⁺MHCII^highcells. Tumor-associated macrophages (TAM) were defined as CD11b⁺F4/80⁺ non-DC cells. MDSCs were defined as CD11b⁺, Ly6C/Ly6G (GR-1)^hi, non-TAM, non-DC cells. To determine absolute counts in samples, liquid-counting beads (BD Biosciences) were added immediately before samples were ran on a flow cytometer. All data were collected on a Fortessa 4 (BD Biosciences) flow cytometer and analyzed with FlowJo v10 software (Tree Star, Inc.).

GraphPad Prism and R software was used for statistical analysis. For column analyses, the Brown–Forsythe test was used to assess equal variances. If non-significant, one-way ANOVA with multiple comparisons was used. In the event of unequal variances between groups, Kruskal–Wallace analysis with Sidak's or Dunnett's multiple comparisons were employed as appropriate. For longitudinal tumor growth analysis, treatment group random effects models were used, for within-experiment mice only. Data were considered to be statistically significant where the P value was less than 0.05.

In mice bearing experimental B16F10 melanoma lung metastases, wild-type (WT) mice treated with the combination of hamster anti-CTLA4 (UC10-4F10) and rat anti-RANKL (IK22-5) mAbs showed superior resistance to metastases compared with mice treated with either antibody alone or control immunoglobulin (cIg; Fig. 1A). The mechanism of action of anti-CTLA4 and anti-RANKL combination therapy was determined in WT mice depleted of CD8⁺ or NK cells or mice deficient for perforin or IFNγ. As shown in Fig. 1B, the efficacy of the combination relied on the presence of NK cells but not CD8⁺ T cells; and IFNγ was critical and to a lesser extent perforin (Fig. 1C). A similar dependence on NK cells was demonstrated for the effective control of prostate carcinoma RM-1 experimental lung metastases following treatment the same with anti-CTLA4 and anti-RANKL combination therapy (Supplementary Fig. S1A).

Given that the immunoglobulin constant region of anti-CTLA4 has been reported to influence antitumor activity (24), we next assessed the impact of how different anti-CTLA4 antibody isotypes synergized with anti-RANKL in suppressing experimental B16F10 lung metastases (Fig. 2). The anti-CTLA4 clone 9D9 has been produced as a number of isotypes, including mouse IgG1, IgG2a and IgG2b; whereas another isotype (IgG1-D265A) contained a mutation that eliminated binding to all Fcγ receptors (FcγR; ref. 24). As shown in Fig. 2A, the IgG2a isotype of the anti-CTLA4 (anti-CTLA4; mIgG2a) alone resulted in greater suppression of lung metastases compared with the hamster clone of anti-CTLA4 (anti-CTLA4; h), and this suppression was further increased with the addition of anti-RANKL to either anti-CTLA4 clones. Similarly, significant suppression of RM-1 and LWT1 lung metastases were also seen with anti-CTLA4 (mIgG2a) and anti-RANKL combination therapy (Supplementary Fig. S1B and S1C).

Interestingly, the other three anti-CTLA4 isotypes (IgG2b, IgG1, or IgG1-D265A) were not as effective in suppressing lung metastases as monotherapy compared with the anti-CTLA4 (mIgG2a) isotype as they did not result in significant suppression of metastases compared with the cIg treated group (Fig. 2B). However, the addition of anti-RANKL to anti-CTLA4 (h) or anti-CTLA4 (mIgG2a and mIgG2b) resulted in significant suppression of lung metastases compared with cIg. Nevertheless, the group treated with anti-RANKL and the anti-CTLA4 (mIgG2a) clone was significantly superior to the combination anti-RANKL with anti-CTLA4 (mIgG2b; Fig. 2B). Overall, anti-RANKL treatment alone did not significantly suppress metastasis, although it significantly improved the control of metastases when used in combination with specific anti-CTLA4 isotypes, most notably mIgG2a (Fig. 2A and B).

Given these results in experimental metastasis models, the efficacy of co-administration of anti-RANKL and anti-CTLA4 mAbs in mice bearing subcutaneous tumors was next assessed (Fig. 3). Similar to the lung metastasis models, optimal B16F10 subcutaneous growth suppression by combination anti-CTLA4 and anti-RANKL was again dependent on antibody isotype, with significant suppression of growth observed with anti-CTLA4 (mIgG2a; Fig. 3A), rather than the hamster isotype (Supplementary Fig. S2A). Antitumor efficacy was also confirmed in additional subcutaneous tumor models, mouse colon cancer CT26 (Fig. 3B) and mouse prostate carcinoma Tramp-C1 (Supplementary Fig. S2B). We next performed a longitudinal analysis of seven independent pooled experiments using B16F10 comparing anti-CTLA4 (mIgG2a) or the FcγR–non-engaging clone of anti-CTLA4 (IgG1-D265A) and/or anti-RANKL with control Ig (Supplementary Fig. S2C). Overall our data demonstrated the combination of anti-CTLA4 (mIgG2a) with anti-RANKL significantly suppressed tumor growth and resulted in superior efficacy compared with either as monotherapy or with cIg (Supplementary Fig. S2C). By contrast, the combination of anti-CTLA4 (IgG1-D265A) and anti-RANKL was superior to cIg-treated groups, but combinatorial therapy was not superior to either treatment as a monotherapy (Supplementary Fig. S2C). Similarly, although B16F10 tumor mass at endpoint of mice treated with combination anti-RANKL with anti-CTLA4 (mIgG2a) was also significantly decreased compared with the respective monotherapy-treated groups, there was no additive benefit of adding anti-RANKL to anti-CTLA4 (IgG1-D265A; Fig. 3C).

To confirm these findings, anti-CTLA4 mIgG2a and IgG1-D265A isotypes were assessed in the treatment of subcutaneous CT26 tumors in combination with two different therapies targeting RANKL: anti-RANKL mAb (IK22-5) or the well-validated RANKL blocker OPG-Fc (23). Anti-CTLA4 (mIgG2a) was the significantly more effective isotype in suppressing tumor growth in combination with either RANKL-directed therapy (Fig. 3D), suggesting that FcR engagement by the anti-CTLA4 mAb was required for optimal combination therapy. Similar results were found in the treatment of fibrosarcoma MCA1956 (Supplementary Fig. S2D). OPG-Fc demonstrated similar efficacy as anti-RANKL mAb when used in combination with anti-CTLA4 (mIgG2a) in CT26 (Fig. 3D) as well as B16F10 subcutaneous tumors (Fig. 3E), confirming efficacy of different RANKL-directed therapies. When treatment was commenced at a later timepoint (with established B16F10 tumors), the combinatorial efficacy of anti-CTLA4 (mIgG2a) with anti-RANKL or OPG-Fc was still observed (Fig. 3F). Despite the optimal effect of anti-CTLA-4 (mIgG2a) and anti-RANKL, in the more sensitive CT26 and MCA1956 tumor models, it was obvious that co-blockade using anti-CTLA-4 (IgG1-D265A) and anti-RANKL was also more effective than either monotherapy (Fig. 3D, Supplementary Fig. S2D). These data suggested that anti-RANKL provides additional antitumor benefit to either Treg-depleting anti–CTLA-4 (IgG2a) or blocking only anti–CTLA-4 (IgG1-D265A) in the CT26 and MCA1956 models.

Expression of RANKL and RANK in the B16F10 tumor microenvironment (TME) was next defined (Fig. 4A and B). The majority of intratumor RANKL was expressed by a small fraction of T cells, with a higher proportion expressing RANKL at an earlier timepoint (day 9) and higher in tumor than in spleen, with more CD8⁺ T cells compared to CD4⁺ T cells expressing RANKL in the tumor (Fig. 4A, Supplementary Fig. S3A). Overall, about 20% of tumor-infiltrating leukocytes (TIL) expressed RANK with greater than 90% of those also staining positive for CD11b, suggesting intratumor RANK was expressed almost exclusively by tumor-infiltrating myeloid cells (Supplementary Fig. S3B). Approximately 40% of tumor-infiltrating macrophages (TAM), 60% of MDSCs, and a low proportion of DCs (7%), expressed RANK (Fig. 4B, Supplementary Fig. S3C). Treatment with anti-RANKL did not significantly alter myeloid RANK expression on these cell types (Fig. 4B). Recently, it was reported in a B16 melanoma model that Ly6C^lowMHCII^high intratumor macrophages had an RNA expression profile consistent with an inflammatory M1 subtype, whereas those with MHCII^low/negative expression were more consistent with an immunosuppressive M2 phenotype (25). Interestingly, in our B16F10 model, a higher proportion of Ly6C/Ly6G (GR-1)^low TAMs expressing RANK had negative or low MHCII expression compared with those not expressing RANK, suggesting that the RANK-expressing TAM population may be more suppressive than those TAMs not expressing RANK (Supplementary Fig. S3D and S3E. Less than 1% of RANKL- or RANK-expressing cells are CD45.2 negative (indicating a negligible level of intratumor expression of either in vivo), and additionally all tumor cell lines used in this study when assessed by flow cytometry were negative for RANKL or RANK expression. Thus, RANKL and RANK expression was largely restricted to tumor-infiltrating lymphocytes and myeloid cells in the TME.

Next, to continue dissecting mechanism of action of the combination, we assessed the reliance of the combinatorial efficacy of anti-RANKL with anti-CTLA4 (mIgG2a) on Fc receptors and the presence and function of effector lymphocytes in the subcutaneous B16F10 tumor model (Fig. 4C). Combination activity of anti-CTLA4 (mIgG2a) and anti-RANKL against B16F10 was abrogated in mice lacking FcγRIV or FcϵRγ, but not FcγRIII (Fig. 4C). This is consistent with the described mechanism of action of anti-CTLA4 (mIgG2a), which requires FcγRIV (24, 26). To determine whether Fc receptor engagement is necessary for anti-RANKL treatment alone, and because an Fc-mutant of anti-RANKL was not available, we assessed treatment effect of anti-RANKL in the immunogenic fibrosarcoma MCA1956, where anti-RANKL monotherapy resulted in significant subcutaneous growth suppression. This efficacy was preserved in mice lacking activating FcR (Supplementary Fig. S4A), demonstrating that FcR engagement was not essential for the antitumor activity of RANKL blockade. Next, the role of CD8⁺ T cells and NK cells in the control of B16F10 tumor growth by anti-CTLA4 (mIgG2a) and anti-RANKL was assessed by the selective depletion of each subset (Fig. 4D). When CD8⁺ T cells were depleted, the antitumor efficacy of the combination therapy was almost completely abrogated. By contrast NK-cell depletion was without effect, demonstrating the reliance of this combination therapy on CD8⁺ T cells (Fig. 4D). Similar to results observed in the metastatic setting, this combination therapy was IFNγ dependent, but does not require perforin (Fig. 4E). The essential role for cross-presenting CD8α⁺ conventional DCs in this combination therapy was also revealed by the use of mice deficient in the transcription factor Batf3; the efficacy of the combination therapy was abrogated in these mice compared with WT-treated mice (Fig. 4F). Similarly, combination efficacy against subcutaneous Tramp-C1 was abrogated in Batf3^−/− mice (Supplementary Fig. S4B).

To further understand the mechanism of the combination therapy and the role of CD8⁺ T cells, the composition of TILs was assessed in subcutaneous B16F10 tumors that had been treated with the optimal combination therapy of anti-CTLA4 (IgG2a) and anti-RANKL (Fig. 5). When assessed at tumor end-point, the proportion of CD45.2⁺ TILs was significantly increased in the combination therapy compared with cIg- or monotherapy-treated groups (Fig. 5A). The increase in CD45.2⁺ TILs in the combination therapy–treated group was largely accounted for by a marked increase in CD8⁺ T cells, both in proportion (Fig. 5B) and absolute numbers (Fig. 5C).

Given the superiority of anti-CTLA4 (mIgG2a) as a partner to RANKL blockade, we next assessed the level of Tregs in the TME, as anti-CTLA4 (mIgG2a) reportedly depletes tumor-infiltrating Tregs via ADCC (24). Interestingly, although the proportion of Tregs (CD4⁺Foxp3⁺) as a percentage of CD4⁺ T cells in the tumor was indeed reduced with anti-CTLA4 (mIgG2a) monotherapy, it was not further reduced with the addition of anti-RANKL antibody (Fig. 5D). In addition, the FcγR-IV expression on CD11b⁺ cells was not further increased with the combination therapy in the tumor (Supplementary Fig. S5A), suggesting that enhanced Treg depletion in the TME does not explain the mechanism of action of this combination. In the spleen, no significant changes in Treg proportion or number were detected between treatment groups, indicating that Treg levels were not altered systemically with the combination (Supplementary Fig. S5B).

To further assess any essential role for Tregs in combinatorial efficacy of anti-RANKL in immunotherapy, the FoxP3-DTR mouse model was employed. In these mice, the diphtheria toxin receptor (DTR) is expressed under the control of the foxp3 locus, allowing for the conditional and near-complete depletion of Tregs through administration of diphtheria toxin (DT), resulting in enhanced anti-tumor immunity (21). A trend to greater B16F10 subcutaneous tumor growth suppression was seen (Supplementary Fig. S5C and S5D) and a higher proportion of mice were cured when anti-RANKL therapy was given in combination with DT compared with DT alone (Supplementary Fig. S5E). In addition, a similar trend of enhanced growth suppression of subcutaneous RM-1 was also seen in FoxP3-DTR mice treated with DT and anti-RANKL (Supplementary Fig. S5F). FACS analysis of RM-1 tumors at endpoint revealed near-complete Treg depletion by DT alone, with no additional depletion of Tregs noted when anti-RANKL was combined with DT (Supplementary Fig. S5G). Taken together, the mechanism of action of the combination therapies in these models does not appear to be due to more efficient Treg depletion, and is intact in settings of near-complete intratumoral Treg depletion.

Another potential mechanism of action of anti-RANKL could be the enhancement of T-cell proliferation. However, we did not observe any further increase in Ki-67 expressing CD8⁺ T cells in the combination treated groups compared to anti-CTLA4 monotherapy (Fig. 5E). This suggests that the additional CD8⁺ T cells observed in the tumor post combination treatment might be a result of selective CD8⁺ T cell recruitment. No significant changes were noted in proportions or numbers of myeloid subsets apart from a mild decrease in the proportion of CD11b⁺ cells as a fraction of total TILs with combination therapy, which likely is compensatory for the increased proportion of CD8⁺ T cells (Supplementary Fig. S6A–S6G). In addition, the proportion of TAMs expressing CD206 (an M2 marker) was not altered in either B16F10 or RM-1 subcutaneous tumors (Supplementary Fig. S6H and S6I). Influx of CD8⁺ T cells with the combination therapy, combined with a lack of increase in suppressive immune cells such as Tregs or myeloid cells, may change the TME to favor antitumor activity. Indeed, significant increase in CD8⁺-to-Treg ratio were noted when measured at an early time-point (day 9; Fig. 5F) or at tumor end-point (Fig. 5G). In addition, the ratio of CD8⁺ T cells to MDSCs was also significantly increased with the combination therapy (Fig. 5H). Importantly, these observed changes were specific for the TME as no significant changes in the proportions of leukocyte subsets were observed in the spleen of these tumor-bearing mice, such as T cells (Supplementary Fig. S6J and S6K).

Finally, given that the main changes in TME composition related to the T-cell compartment, we also assessed how this combination affected Th1 cytokine production (IFNγ, TNFα, and IL-2) from CD8⁺ and CD4⁺ T cells in B16F10 tumor at experimental endpoint (day 16; Fig. 6). TNFα was the most commonly produced cytokine ex vivo after stimulation, but significant differences were noted in the production of IFNγ by CD8⁺ T cells (Fig. 6A) following combination therapy compared to cIg or monotherapy alone. Furthermore, CD8⁺ T cells co-expressing IFNγ and IL-2 (Fig. 6B) or IFNγ, IL-2 and TNFα (Fig. 6C) was also increased with the combination therapy. Similar findings were also seen with CD4⁺ T cells, particularly in the proportion that produced IFNγ (Fig. 6D). The majority of CD8⁺ T-cell from the cIg-treated group produced no cytokines after stimulation, whereas the combination therapy generated T cells with the most polyfunctionality, with monotherapy treatment groups displaying intermediate phenotypes (Fig. 6E and Supplementary Fig. S7). The effect of the combination therapy on cytokine polyfunctionality was TME-specific, as these differences were not observed in the splenic T cells of tumor-bearing mice (Supplementary Fig. S8).

Longer-term OS data for key immunotherapy trials in advanced melanoma are now maturing, showing encouraging results but with significant scope for improvement. Subsequent trial experience and pre-clinical studies have underlined the importance of a combinatorial approach, with superior tumor responses and survival (at 2 years) reported with combination nivolumab and ipilimumab compared with ipilimumab alone which is likely to translate into longer-term survival benefit although this data is awaited (27). A priority now is to improve combination immunotherapy to overcome primary and acquired resistance to existing therapies; experimental strategies using novel partners in combination with either ipilimumab or mAbs targeting the PD-1/PD-L1 axis have reached the clinical trial stage (28–30). In this study, we have presented data indicating the combinatorial efficacy of anti-CTLA4 with molecules blocking RANKL–RANK interactions in metastatic and subcutaneous models, and demonstrated its mechanism of action is due to increased tumor-localized CD8⁺ T cells and enhanced effector function.

The full efficacy of CTLA4 blockade, at least in mice, relies on depletion of Tregs in addition to blocking the interaction of CTLA4 on effector T cells (Teff) with B7 ligands, which results in negative regulation of their function (31). We have shown that the optimal combinatorial efficacy of the anti-RANKL and anti-CTLA4 combination is achieved when using the anti-CTLA4 (mIgG2a) isotype. This isotype was previously demonstrated to selectively deplete intratumoral Tregs via antibody-dependent cell–mediated cytotoxicity (ADCC) involving upregulation of FcγRIV expression on CD11b⁺ TILs, thereby enhancing antitumor immunity (24, 26). The requirement for activating FcγRs has previously been established in mice for another anti-CTLA4 isotype (9D9 mIgG2b; ref. 32), as has the requirement of FcγRIIIA in the human setting in the ex-vivo lysis of Tregs by monocytes from peripheral blood when cultured with ipilimumab (33). However, in our models, neither increased depletion of Tregs, nor further upregulation of FcγRIV on myeloid cells was seen with the combination of this IgG2a mAb with anti-RANKL. Indeed, additional efficacy of combination anti-RANKL with DT-induced Treg depletion was seen in FoxP3-DTR mice compared with DT alone, where both DT-containing arms showed >95% Treg depletion compared with cIg at endpoint. Thus, the combinatorial efficacy with anti-RANKL remained intact even when more Tregs were depleted than could be achieved through the use of anti-CTLA4 (mIgG2a), indicating that the mechanism of action of anti-RANKL was not directly on Tregs. These findings are in contrast with the previously reported roles for RANKL-expressing Tregs in the promotion of metastasis in a mouse model of RANK-expressing mammary carcinoma (34), or the role of systemic Treg control via cutaneous RANKL–RANK interactions in the restraint of UV-induced cutaneous inflammation (35). Furthermore, we have shown that monotherapy efficacy of anti-RANKL is independent of activating FcR.

Overcoming T-cell exhaustion is a hallmark of current clinically available cancer immunotherapy, with restoration of effector T-cell function (such as IFNγ production; ref. 36). In our study, blockade of RANKL when added to anti-CTLA4 further improved effector function of intratumor T cells as measured by production of Th1-type cytokines such as IFNγ and IL-2, and increased the proportion of polyfunctional T cells. Such polyfunctional T cells have been shown to be important in the control of cancer and certain chronic viral infections and vaccination (37–40). Our data demonstrated the combination therapy led to marked increase of tumor-infiltrating CD8⁺ T cells most likely through recruitment rather than increased proliferation. In a study of human extra-mammary Paget's disease, the majority of RANK expression was associated with M2 macrophages, which produced the chemokine CCL17 when stimulated with RANKL, resulting in recruitment of T cells (41, 42). Although not addressed in this study, future analysis of chemokine production in the TME under the conditions of RANKL/RANK blockade may give further insights into this mechanism in the setting of cancer immunotherapy.

It is unclear whether the activity of anti-RANKL in combination with specific anti-CTLA4 isotypes resulted from the disruption of an immunosuppressive or tolerogenic axis between RANK-expressing myeloid cells and RANKL-expressing lymphocytes in the TME, although an essential role for DC-mediated cross-presentation was suggested through the use of Batf3-deficient mice. In our models, the most abundant RANK-expressing cells were TAMs, with expression also detected by DCs and MDSCs. The conditions under which myeloid cells express RANK in the TME and its downstream functions after RANK ligation remain unknown. Interactions between RANKL-expressing T cells and RANK-expressing myeloid cells can result in differential outcomes depending on factors such as anatomical site and pathologic setting. In one study, ligation of RANK on mucosal DCs by RANKL (but not splenic, mesenteric or peripheral lymph node DCs) resulted in increased IL-10 production, and oral RANKL administration at the time of antigen challenge increased oral tolerance, suggestive of a role for RANKL/RANK in tolerance induction (43). In a separate study, DCs that were co-cultured with genital squamous cell carcinoma cell lines showed an IL-10^high IL-12^low phenotype and functional deficits in a mixed lymphocyte reaction, but this phenotype could be reversed with RANKL blockade, which also resulted in increase in DC expression of co-stimulatory proteins (44). It is currently unknown whether the interaction between RANK-expressing DCs in the TME and RANKL-expressing lymphocytes results in a tolerogenic signal, although this supposition would be supported by our finding that treatment efficacy is abrogated in mice lacking Batf3.

RANK and RANKL are also implicated in the development of select lymphoid organs: gene-targeted mice deficient for RANK or RANKL lack peripheral lymph nodes, display thymic abnormalities, and are osteopetrotic with resultant reduction in B-cell development; in addition, a role for RANK and RANKL in the embryological development of Aire positive medullary thymic epithelial cells (mTEC) has been described (8, 9, 45–47). RANKL blockade as a mechanism of altering central tolerance has been demonstrated to increase tumor antigen-specific T-cell immunity in young (3–5-week-old) mice (48). The mouse thymus reaches peak stromal cellularity at 4 weeks, and thereafter into adulthood the turnover of certain thymic epithelial cells is 10 to 14 days (49). In the current study, all mice were aged at least 6 weeks, and time from treatment commencement to experimental completion was typically 12 to 13 (but up to 20) days, meaning that alteration of thymic negative selection through depletion of mTEC subsets is unlikely to account for the antitumor efficacy of the combination studied.

The combination of RANKL blockade with alternative immune checkpoint inhibitors, such as those targeting PD-1, should now be assessed. Although the published case reports suggest unusually deep and brisk antitumor responses with combination ipilimumab and denosumab, antibodies targeting PD-1 have demonstrated superior antitumor responses compared with ipilimumab as monotherapy with a more acceptable toxicity profile (50). These results suggest that exploration of the combination of anti-RANKL and anti-PD1/PD-L1 are warranted, to assess whether the improvement of effector T-cell responses seen with the combination studied here could be recapitulated. In humans, the currently available anti-RANKL mAb is well tolerated; in cancer trials, adverse events suggestive of either immune suppression or autoimmunity were not higher than with bisphosphonate therapy (13, 51).

Overall, these results suggest a potential role for combining antibodies targeting RANKL with immune checkpoint blockade. In clinical trials of such combinations, translational markers of interest would include those of T-cell function and proliferation, and functional markers of investigational RANK⁺ myeloid subsets. Identification of the essential RANK-expressing suppressive myeloid population(s) for combinatorial efficacy would be a key goal, as this may provide a biomarker to identify patients suitable for this approach. Furthermore, this could provide a new immunotherapeutic treatment option for patients with malignancies who currently show little response to immune checkpoint blockade, but which have a significant myeloid component in the TME, such as advanced prostate cancer (52–55).

W.C. Dougall reports receiving commercial research grants from Bristol-Myers Squibb. M.W.L. Teng reports receiving speakers bureau honoraria from Merck Sharp & Dohme. M.J. Smyth reports receiving commercial research grants from Aduro Biotech, Bristol-Myers Squibb, and Corvus Pharmaceuticals, and is a consultant/advisory board member for Arcus Biosciences, F-star, and Kymab. No potential conflicts of interest were disclosed by the other authors.

Conception and design: E. Ahern, D. Wyld, W.C. Dougall, M.W.L. Teng, M.J. Smyth

Development of methodology: E. Ahern, S. Allen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Ahern, H. Harjunpää, D. Barkauskas, K. Takeda, H. Yagita, M.J. Smyth

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Ahern, W.C. Dougall, M.W.L. Teng, M.J. Smyth

Writing, review, and/or revision of the manuscript: E. Ahern, D. Barkauskas, K. Takeda, H. Yagita, D. Wyld, W.C. Dougall, M.W.L. Teng, M.J. Smyth

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Ahern, S. Allen, M.W.L. Teng

Study supervision: D. Wyld, W.C. Dougall, M.W.L. Teng, M.J. Smyth

We thank Liam Town and Kate Elder for their mouse genotyping and maintenance. We also thank David Smith from the QIMR Berghofer Biostatistical Unit for statistical advice.

M.J. Smyth was supported by a National Health and Medical Research Council (NH&MRC) Senior Principal Research Fellowship (1078671) and The Cancer Council of Queensland (1102242). M.W.L. Teng was supported by a NHMRC Career Development Fellowship (1025552) and a QIMR-Berghofer RTCC-WEWC grant (15012). E. Ahern was supported by a University of Queensland (UQ) Australian Postgraduate Award (APA). H. Harjunpää was supported by a UQ International Postgraduate Research Scholarship, a UQ APA, and a QIMR Berghofer Top-Up award. H. Yagita was supported by a MEXT Japan grant (26290059).

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.