FcγR interaction is not required for effective anti‐PD‐L1 immunotherapy but can add additional benefit depending on the tumor model

Immunomodulatory antibodies blocking interactions of coinhibitory receptors to their ligands such as CTLA‐4, PD1 and PD‐L1 on immune cells have shown impressive therapeutic efficacy in clinical studies. The therapeutic effect of these antibodies is mainly mediated by reactivating antitumor T cell immune responses. Detailed analysis of anti‐CTLA4 antibody therapy revealed that an optimal therapeutic efficacy also requires binding to Fc receptors for IgG, FcγR, mediating depletion of intratumoral regulatory T cells. Here, we investigated the role of Fc binding in anti‐PD‐L1 antibody therapy in the MC38 C57BL/6 and CT26 BALB/c colon adenocarcinoma tumor models. In the MC38 tumor model, all IgG subclasses anti‐PD‐L1 showed similar therapeutic efficacy when compared to each other in either wild‐type mice or in mice deficient for all FcγR. In contrast, in the CT26 tumor model, anti‐PD‐L1 mIgG2a, the IgG subclass with the highest affinity for activating FcγR, showed stronger therapeutic efficacy than other IgG subclasses. This was associated with a reduction of a myeloid cell subset with high expression of PD‐L1 in the tumor microenvironment. This subclass preference for mIgG2a was lost in C57BL/6 × BALB/c F1 mice, indicating that the genetic background of the host may determine the additional clinical benefit of the high affinity antibody subclasses. Based on these data, we conclude that FcγR are not crucial for anti‐PD‐L1 antibody therapy but might play a role in some tumor models.


Introduction
Expression of programmed death ligand 1 (PD-L1), the ligand for programmed death 1 (PD-1), plays a paramount role in suppressing antitumor responses in human cancer patients. Because of the clinical success of PD-L1 blocking antibodies, [1][2][3][4][5][6] Atezolizumab, Avelumab and Durvalumab have been approved by the U.S. Food and Drug administration (FDA) for the treatment of various advanced cancers.
Several lines of evidence suggest that the main effect of anti-PD-L1 mAb is abolishing suppression of on-going antitumor T cell responses through blocking PD-L1 on the tumor or tumor infiltrating myeloid cells from interacting with PD-1 on T cells. 5,[7][8][9] Theoretically, this mechanism should be FcγR independent. Nevertheless, it has been suggested that antibody subclass, and subsequent FcγR binding might contribute to the therapeutic efficacy. 10 For example, many studies have demonstrated an essential role for activating FcγR in the antitumor activity of mAbs targeting the coinhibitory receptor C TLA-4. [11][12][13] Binding of the Fc part of anti-CTLA4 mAb to FcγR resulted in the depletion of regulatory T cells by antibody-dependent effector mechanisms within the tumor.
In the clinic, Atezolizumab and Durvalumab are engineered human IgG1 anti-PD-L1 mAbs, with a modification in their Fc-domain to prevent antibody-dependent cellular cytotoxicity (ADCC) and complement-dependent cytotoxicity (CDC). 14,15 On the other hand, Avelumab has a wild-type human IgG1 Fc region which has been shown in vitro to engage FcγR expressing cells to induce ADCC. Since PD-L1 is expressed on tumor cells and tumor infiltrating immune cells, it might be that anti-PD-L1 mAb with the capability to engage FcγR influences the antitumor responses not only by blocking PD1-PD-L1 signaling pathway but also by depleting PD-L1 expressing tumor cells and/or tumor associated myeloid cells. Conversely, Fc-mediated effector mechanisms might be detrimental to the immune response due to the depletion of PD-L1 + immune effector cells.
In our study, we investigated the effect of the use of different mAb IgG subclasses on the efficacy of anti-PD-L1 in two mouse models of colorectal cancer (MC38 on C57Bl/6 and CT26 on BALB/c background). For this, we replaced the constant regions of anti-mouse PD-L1 mAb (rat IgG2a; clone MIH5) with different mouse constant regions to vary the binding to mouse FcγR. Our data suggest that FcγR are not critical for anti-PD-L1 mAb therapy but depending on the genetic background, the therapeutic efficacy of anti-PD-L1 could be enhanced by using the IgG subclass with the highest affinity for activating FcγR.
The resulting single PCR fragments of the variable regions of both the heavy and light chain were analyzed by DNA sequencing. This sequence was codon optimized for expression in human cells. Based on this information, the heavy and light chain variable region encoding DNA fragments were assembled from GBLOCKS (Integrated DNA Technologies, Leuven, Belgium). The anti-mouse PD-L1 heavy chain variable region encoding DNA fragment was cloned in the expression vectors pFUSE_CHIg_mG1, pFUSE_CHIg_mG2a and pFUSE_CHIg_mG3 to express the anti-mouse PDLI mIgG1, mIgG2a and mIgG3 heavy chain respectively. The light chain variable region encoding DNA fragment was cloned in pFUSECLIg mk to express the anti-mouse PD-L1 Kappa Light chain. Unfortunately, these vectors showed low expression of the proteins when transfected into HEK293 cells and therefore the antibody encoding sequences were cloned into the expression vector pCDNA3.1(A) which showed good transient expression in these human cells. In addition, an anti-PD-L1 mIgG2a antibody was generated with a D256A mutation in the Fc domain of mIgG2a engineered according to the protocol as described in Liu et al. 16 Mice C57BL/6 and BALB/c mice were purchased from Charles River (L'Arbresle, France) and housed in the animal facility of the Leiden University Medical Center (LUMC). FcγRI/II/III/ IV deficient C57Bl/6 mice were generated in the transgenic mouse facility of the LUMC. 17 These transgenic mice were bred in house and routinely checked for their genotype by PCR. All mice were housed in individually-ventilated-cage (IVC) systems under specific pathogen-free conditions and used at 8-12 weeks of age.
The health status of the animals was monitored over time. Animals tested negative for all agents listed in the FELASA (Federation of European Laboratory Animal Science Associations) guidelines for SPF mouse colonies. 18 All mouse studies were approved by the animal ethics committee of the LUMC. Experiments were performed in accordance with the Dutch Act on Animal Experimentation and EU Directive 2010/63/EU ('On the protection of animals used for scientific purposes').
Syngeneic tumor models and tumor therapy CT26 (kindly provided by Mario Colombo, Milano) and MC38 tumor cells were cultured in Iscove's modified Dulbecco's medium (IMDM) (Lonza) supplemented with 8% Fetal Calf Serum (Greiner), 25 μM 2-mercaptoethanol and 100 IU/mL penicillin/streptomycin (Gibco). Cell lines were mycoplasma and MAP-tested before the start of experiments. The tumor cells (2.5 × 10 5 for MC38/MC38 PD-L1 −/− and 1 × 10 5 for CT26/ CT26 PD-L1 −/− ) were injected subcutaneously into 8-12 week-old mice in 100 μL of PBS. PD-L1 deficient MC38 and CT26 cell lines were generated using What's new? Monoclonal antibodies against T cell "exhaustion"-inducing molecules show striking success in reactivating the antitumor immune response, but it is unclear if this success involves the so-called Fc region of the antibody that activates the immune system. Here the authors systematically swapped constant regions of a known mouse antibody against programmed deathligand-1 (PD-L1) to alter interactions with the Fc receptor. The outcome was nuanced with different results obtained in different mouse tumor models leading the authors to conclude that in some tumors anti-PD-L1 immunotherapies might be enhanced with proper Fc receptor interactions.
CRISPR/Cas9 technology as described in Kleinovink et al. 2017. 9 Two hundred micrograms of therapeutic antibody was injected intraperitoneally at day 6, 9 and 12 after tumor inoculation. Once palpable tumors were present, tumor size was measured twice per week, using a caliper. Mice were sacrificed when tumors reached a size of 100 mm 2 because of ethical reasons.

Rechallenge tumor studies
After an 80 days tumor-free period, mice with complete regression of MC38 and CT26 tumors, as well as naïve control mice, were reinoculated subcutaneously in the left flank with either 2,5 × 10 5 MC38 or 1 × 10 5 CT26 tumor cells. Once palpable tumors were present, tumor size was measured twice per week, using a caliper. Mice were sacrificed when tumors reached a size of 100mm 2 because of ethical reasons.
For assessing the binding of murinized anti-PD-L1 mAb to mouse PD-L1, PD-L1 expressing MC38 or B16F10 tumor cells (pretreated with 20 U/mL mIFNγ for 24 h) were preincubated for 15 min with titrated murinized anti-PD-L1 mAb or serum collected from anti-PD-L1 mAb treated mice prior to staining with PE-labeled anti-PD-L1 mAb (clone MIH5). Cells were analyzed by flow cytometry for PE labeling, which was blocked by the presence of unlabelled murinized anti-PD-L1 mAb. Analysis were performed using LSRII cytometer (BD) using FacsDIVA software (BD) and FlowJo Software (Tree Star).

Statistical analyses
Data was analyzed using Prism 7.0 (GraphPad Software). Statistical significance was calculated using the two-way ANOVA and Mann Whitney nonparametric test. Statistical significance was defined as p < 0.05. Tumor survival data was analyzed with the Kaplan-Meier method and the log-rank (Mantel-Cox) test.

Characterization of murinized anti-mouse PD-L1 IgG monoclonal antibody
To determine the role of FcγR in anti-PD-L1 mAb therapy, the variable region of the rat anti-mouse PD-L1 mAb (clone MIH5, rat IgG2a) was fused to the different mouse immunoglobulin heavy chain constant regions by molecular cloning and protein was purified from HEK293 cells transfected with the expression vectors encoding the recombinant murinized immunoglobulins. These included mIgG1, mIgG2a, mIgG3 and a mIgG2a containing the D265A mutation (IgG2a D265A) which has been reported to abrogate binding to FcγR and strongly reduces complement activity. 20, 21 We also assessed whether the murinized mAbs bound to the same epitope on PD-L1 as the parental anti-PDL1 rat IgG2a mAb. PD-L1 expressing MC38 tumor cells were preincubated with respective concentrations of purified murinized anti-PD-L1 mAb and then stained with fluorescently labeled anti-PD-L1 rat IgG2a. Each of these murinized mAbs showed equivalent binding to PD-L1 expressing MC38 tumor cells (Fig. 1a), suggesting that the murinized anti-PD-L1 and parental anti-PD-L1 mAbs recognize the same epitope on PD-L1 and that the different IgG subclasses bind with the same affinity. In addition, the binding affinity of the Fc portion of each mAb to mouse FcγRI, II, III, and IV was measured using surface plasmon resonance (Fig. 1b). Consistent with literature, mIgG1 bound with low affinity to FcγRIII and FcγRII, mIgG2a bound with high affinity to FcγRI, intermediate affinity to FcγRIV and low affinity to FcγRIII, mIgG3 did not bind to any FcγR. The mIgG2a D265A showed only residual binding to FcγRI (Fig. 1b). The calculated K D values are shown in Figure 1c.
Therapeutic efficacy of murinized anti-PD-L1 IgG monoclonal ab is FcγR independent in MC38 tumor model The MC38 colon adenocarcinoma syngeneic model on C57BL/6 background is highly immunogenic and it has been demonstrated to be sensitive to anti-PD-L1 mAb monotherapy. 9,22 The therapeutic efficacy of the murinized anti-PD-L1 mAbs was analyzed in this tumor model. Strong inhibition of tumor growth (Fig. 2a), increased survival rates (Fig. 2b) as well as long-term immune memory (Fig.S1A, Supporting Information) were observed after i.p. administration of each of the murinized anti-PD-L1 mAbs (Figs. 2a and 2b). We did not observe significant differences in delay of tumor outgrow and long-term survival between mice treated with anti-PD-L1 mAb of different IgG subclasses including the D265A mIgG2a, indicating that FcγR engagement was not required for the optimal therapeutic efficacy of anti-PD-L1 mAb. To confirm this, we studied the therapeutic efficacy of IgG1 and IgG2a anti-PD-L1 in MC38 tumor bearing C57BL/6 mice deficient for the ligand binding chains of all four FcγR (FcγRI/II/II/IV −/− mice) which show normal innate and adaptive immunity. 17   Information). Collectively, our data demonstrate that interactions with FcγR are dispensable for the therapeutic efficacy of anti-PD-L1 IgG in the MC38 tumor model.
Anti-PD-L1 IgG2a is more effective than anti-PD-L1 IgG1 and D265A mIgG2a in CT26 tumor model As FcγR were dispensable for the therapeutic efficacy of anti-PD-L1 antibody in MC38 tumor model, we aimed to test if this is a general feature of this treatment modality, by verifying this in another tumor model. For this, we selected the CT26 colon adenocarcinoma syngeneic model on BALB/c background which is also known to be immunogenic and hence highly responsive to anti-PD-L1 mAb monotherapy. 9,23 When CT26 tumor bearing BALB/c mice were treated with murinized anti-PD-L1 mAb, all mIgG subclasses including D265A mIgG2a again effectively delayed tumor outgrowth resulting in prolonged survival (Figs. 3a and 3b) and formation of antitumor memory response (Fig. S1C, Supporting Information). However, in contrast to the MC38 tumor model, overall survival rate of mice treated with anti-PD-L1 mIgG2a was significantly higher compared to mice treated with anti-PD-L1 mIgG1 or D265A IgG2a (Fig. 3b). To establish that this was not caused by differences in antibody clearance we measured the serum levels of circulating murinized anti-PD-L1 mAb in CT26 tumor bearing BALB/c mice using flow cytometry. We found that serum levels of the four murinized anti-PD-L1 mAb were similar. Thus, the higher efficacy of the murinized anti-PD-L1 mIgG2a in the CT26 tumor model cannot be explained by differences in drug exposure (Fig. S2, Supporting Information).  (Fig. 4a), improved long-term survival was observed only in mice treated with anti-PD-L1 mIgG2a but not mIgG1 (Fig. 4b). Our result suggests that the treatment of PD-L1 −/− tumor can also benefit from the use of  Anti-PD-L1 mIgG2a but not mIgG1 modulates the tumor infiltrating myeloid cell subsets in the CT26 tumor microenvironment Because PD-L1 is not only expressed on tumor cells but also on tumor infiltrating immune cells, we evaluated the relative expression of PD-L1 on multiple immune cell populations within the MC38 and CT26 tumor microenvironment. High levels of PD-L1 expression on F480 hi Ly6C lo myeloid cells was observed as compared to F4/80 int Ly6C hi , F4/80 lo Ly6C lo myeloid cells, CD3 T lymphocytes, granulocytes (Ly6G + ) and CD45-negative cells (Fig. 5a). Based on the differences in PD-L1 density, we hypothesized that anti-PD-L1 mAb therapy could result in selective modulation of myeloid cell subsets. Therefore, we compared the impact of murinized anti-PD-L1  mIgG1 and mIgG2a on the frequency of immune cells in the tumor microenvironment of both tumor models. Administration of anti-PD-L1 mAb resulted in an increased frequency of CD4 + and CD8 + T cells in both CT26 and MC38 tumor (Fig. 5b). In MC38 tumors, the frequency of myeloid cell subsets between mice treated with murinized IgG1 or IgG2a anti-PD-L1 mAb and untreated mice was similar. In contrast, in CT26 tumors,anti-PD-L1 mIgG2a treatment resulted in reduction of F480 hi Ly6C lo cells, whereas the mIgG1 subclass did not affect this myeloid subset (Fig. 5c). Therefore, we hypothesized that the additional antitumor effect of anti-PD-L1 mIgG2a in the CT26 tumor model is due to its potential capability to directly target tumor-associated myeloid cells.
Similar therapeutic efficacy of IgG1 and IgG2a murinized anti-PD-L1 antibodies in C57BL/6 x BALB/c F1 mice We next analyzed whether the genetic background (BALB/c versus C57BL/6) of the mice or tumor cell intrinsic differences (MC38 versus CT26) determined the difference in therapeutic efficacy of anti-PD-L1 mIgG1 and mIgG2a. For this, C57Bl/6 x BALB/c F1 mice (strain CB6F1/J) were inoculated with either MC38 or CT26 tumor cells and treated with anti-PD-L1 mIgG1 or mIgG2a. In both tumor models, no difference in survival was observed between mice treated with anti-PD-L1 mIgG2a or mIgG1 (Figs. 6a and 6b) indicating that the genetic background of the mice but not tumor cell intrinsic differences played a paramount role in the therapeutic efficacy of these antibodies.

Discussion
It has been proposed that the underlying mechanism of tumor rejection in anti-PD-L1 antibody therapy is reactivation and increase of intratumoral T cells as the consequence of blocking PD1-PD-L1 inhibitory activity. It has recently become apparent, however, that the therapeutic efficacy of several other immunomodulatory antibodies such as anti-OX40, GITR, and CTLA4 mAb can also be attributed to the activation of Fc mediated pathways. 11,13,24,25 In contrast, two Statistical significance of the mean tumor size was determined by two-way ANOVA (*p < 0.05; ** p < 0.01; *** p < 0.001). Logrank test was used to determine the statistical significance of the survival. Data from one experiment, eight mice per group.
preclinical studies have revealed that interaction with its Fc part is detrimental to the therapeutic efficacy of anti-PD-1 mAb, by facilitating macrophages to deplete PD1 + effector T cells 10 or to remove this mAb from T cells. 26 This illustrates that it is important to understand the impact of the interaction with FcγR on the therapeutic efficacy of any immunomodulatory antibody. Here, we investigated the role of FcγR in the antitumor effect of PD-L1 blocking antibody. From our studies with the MC38 and CT26 syngeneic colon adenocarcinoma models, we conclude that the therapeutic effect of anti-PD-L1 mAb is predominantly based on FcγR independent blocking of PD-L1. This is conflicting with a recent study showing a minor enhancement of therapeutic efficacy of anti-PD-L1 mIgG2a over other IgG subclasses in MC38 tumor model. 10 This might be caused by differences in experimental conditions as some of the genetically modified mouse strains and anti-PD-L1 mAb used in their study are different from ours.
On the other hand, in the subcutaneous CT26 tumor model, anti-PD-L1 mIgG2a elicited stronger therapeutic effect and survival compared to anti-PD-L1 mIgG1, mIgG3 and mIgG2a D265A. We observed that this enhanced antitumor efficacy correlated with the reduction of a tumor-infiltrated myeloid subset in CT26 tumor. For this tumor model, an association between an elevated number of myeloid cells and the increased magnitude of their immunosuppressive tumor microenvironment was reported while elimination of these cells can lead to strong antitumor responses. 27,28 Preferential reduction of this myeloid subset by anti-PD-L1 mIgG2a may be due to its high levels of surface PD-L1 expression compared to other immune and nonimmune cells, similar to that proposed for anti-CTLA4 mAb, which preferentially depletes high CTLA4 expressing regulatory T cells. 12 As mIgG2a (but not IgG1) can also efficiently activate complement, 29,30 we cannot exclude that the higher efficacy of IgG2a anti-PD-L1 in the CT26 tumor model is caused by the activation of complement pathways. The potential involvement of various Fc-mediated effector functions toward myeloid cell eradication and how the eradication of a myeloid subset in the CT26 tumor microenvironment augment subsequent adaptive immune responses will need further investigation.
We observed a comparable therapeutic efficacy of anti-PD-L1 mIgG1 and mIgG2a in CT26 tumor bearing CB6F1 mice, offspring of a cross between BALB/c and C57Bl/6 mice, suggesting that the stronger antitumor efficacy of anti-PD-L1 mIgG2a is strain-dependent. Although multiple in vitro and in vivo data indicate that human Fc polymorphisms can influence the therapeutic activity of anti-CD20 (rituximab), anti-CD52 (alemtuzumab), anti-Her2 (trastuzumab) and anti-EGFR (cetuximab), in mice two polymorphisms are reported, one in FcγRII and one in FcγRIII, but no difference in IgG binding was observed. 31 Nevertheless, polymorphism in the Fc region of IgG has been found in mice. Whereas most inbred mouse strains including BALB/c express IgG2a, C57Bl/6 mice express an allelic variant of that, named IgG2c. 32 To the best of our knowledge, it is not known whether there is any difference in functionality between IgG2c and IgG2a. However, we cannot exclude the possibility that IgG2a is immunogenic in C57Bl/6 mice and induces an immune response decreasing its efficacy over time.
The objective response rate with approved anti-PDL1 mAb as monotherapy is~20% in urothelial carcinomas, 6,33,34~1 5% in nonsmall-cell lung cancer (NSCLC) 35,36 and~30% in Merkel cell carcinoma. 2,37 Anti-PD-L1 mAb therapy is very effective but so far has only be tested in a restricted number of specific cancer types. However, in animal studies, antitumor activity of anti-PD-L1 treatment varies across a range of syngeneic tumor models. 38 Thus, evaluation of the role of different tumor types and genetic background in the additional effect of anti-PD-L1 mIgG2a may have clinical implications, because the results might guide design and development of more effective anti-PD-L1 mAb for cancer therapy. To date, antibodies targeting PD-L1, including Atezolizumab, Avelumab and Durvalumab have been approved by FDA for treating various cancers. Unlike Atezolizumab and Durvalumab which are engineered human IgG1 mAbs to avoid Fc interaction, Avelumab has a wild-type human IgG1 Fc region. Although results of in vitro studies have suggested that Avelumab is capable of inducing ADCC of tumor cells, [39][40][41] as far as we know, no evidence from clinical results has been published to support this. As previously reported, PD-L1 expression on immune but not tumor cells in the tumor microenvironment was significantly associated with higher response to PD-L1 blocking antibodies, 3,6 indicating that blocking of PD-L1 on immune cells with antibodies (Atezolizumab and Durvalumab) is critical for the effectiveness of this therapy. Hence, it remains to be evaluated whether Avelumab has an additional therapeutic benefit in patients whose tumors are infiltrated by PD-L1 + myeloid cells as our study in the CT26 model suggests. A study by Boyerinas et al. 41 showed very low levels of Avelumab mediated in vitro lysis of PD-L1 + PBMC subsets of patients receiving Avelumab, but this does not rule out the possibility that PD-L1 + immune effector cell subsets in the tumor microenvironment would be subject to some degree of depletion. The concerns related to toxicity versus clinical benefits of Avelumab have to be defined in larger clinical studies. Overall, our work suggests that anti-PD-L1 mAbs work primarily by blocking PD-L1 and binding to FcγR is no prerequisite for their antitumor activity. However, depending on the tumor model, interaction with FcγR can potentially enhance the therapeutic efficacy of anti-PD-L1 mAb.