Nivolumab Enhances In Vitro Effector Functions of PD-1+ T-Lymphocytes and Leishmania-Infected Human Myeloid Cells in a Host Cell-Dependent Manner

Functional impairment of T-cells and a concomitant augmented expression of programmed death-1 (PD-1) have been observed in visceral leishmaniasis patients, as well as in experimental models for visceral and cutaneous leishmaniasis. The PD-1/PD-1-ligand (PD-1/PD-L) interaction negatively regulates T-cell effector functions, which are required for parasite control during leishmaniasis. The aim of this study was to elucidate the impact of the PD-1/PD-L axis in a human primary in vitro infection model of Leishmania major (Lm). Blocking the PD-1/PD-L interaction with nivolumab increased T-cell proliferation and release of the proinflammatory cytokines TNFα and IFNγ during the cocultivation of Lm-infected human monocyte-derived macrophages (hMDMs) or dendritic cells (hMDDC) with autologous PD-1+-lymphocytes. As a consequence Lm infection decreased, being the most pronounced in hMDDC, compared to proinflammatory hMDM1 and anti-inflammatory hMDM2. Focusing on hMDDC, we could partially reverse effects mediated by PD-1 blockade by neutralizing TNFα but not by neutralizing IFNγ. Furthermore, PD-1 blockade increased intracellular expression of perforin, granulysin, and granzymes in proliferating CD4+-T-cells, which might be implicated in reduction of Lm-infected cells. In all, our data describe an important role for the PD-1/PD-L axis upon Lm infection using a human primary cell system. These data contribute to a better understanding of the PD-1-induced T-cell impairment during disease and its influence on immune effector mechanisms to combat Lm infection.


inTrODUcTiOn
The parasitic disease leishmaniasis is still endemic in 97 countries, causing up to 30,000 deaths annually, a number potentially increasing due to climate changes and global warming (1). A pre requisite for controlling Leishmania infection is a strong adaptive immune response. Based on experimental mouse models, it is widely accepted that disease susceptibility is associated with IL10 and IL4 producing Thelper 2 (TH2) cells, whereas a strong Thelper 1(TH1)mediated IFNγ production promotes healing by inducing leishmanicidal nitric oxide in the Leishmaniaharboring cells (2). In human leishmaniasis this TH1/TH2 dichotomy does not always hold true and the resulting Tcell response strongly depends on the Leishmania strain and the immune status of the host (3)(4)(5)(6). In addition, in vitro data from cutaneous Leishmaniasis patients show parasite control to be mediated rather by IFNγinduced reactive oxygen species (ROS) then by nitric oxide (7,8). Macrophages and dendritic cells, the final host cells of Leishmania parasites, play an important role in the initiation of the adaptive immune response. Several in vitro studies demonstrated Leishmanianaive healthy human donors to possess a natural Tcell response against live parasites, antigen extracts or specific components of different Leishmania strains (9)(10)(11)(12)(13)(14)(15)(16). This early MHC class II dependent Tcell response was shown to dampen Leishmania parasite burden in autologous human macrophage/Tcell cocultures (11). The activation of CD8 + and CD4 + Tcells is regulated by various signals such as costimulatory molecules, which can either positively or nega tively influence Tcell priming.
The coinhibitory receptor programmed death1 (PD1, CD279), which is a member of the B7CD28 family, is expressed on activated Tcells and Bcells. Upon association with its ligands PDL1 (CD274) or PDL2 (CD273), which are expressed on, e.g., macrophages and dendritic cells, Tcell activation is sup pressed by inhibition of CD28 signaling (17). The role of the PD1/PDL axis in Tcell exhaustion, a functional impairment of Tcells, is very well studied in the field of cancer and in chronic infections such as HIV, HCV, or lymphocytic choriomeningitis virus (LCMV) (18)(19)(20). Recent publications indicate that the PD1/PDL pathway may play a similar role in Leishmania infection (21)(22)(23)(24). In the canine and mouse model of visceral leishmaniasis, PD1/PDLmediated Tcell exhaustion together with an impaired phagocyte function was observed. Blocking the PD1/PDL interaction in these models partially rescued effector functions of exhausted Tcells, which resulted in a lower parasite burden (21,23). In splenic aspirates of visceral leishmaniasis patients an anergic/exhausted CD8 + Tcell phenotype plus an augmented expression of PD1 was found (24). Nevertheless, functional data regarding the involvement of the PD1/PDL axis in human leishmaniasis is scarce.
In this study, we aimed to define a role for the PD1/PDL axis during Leishmania infection of human primary myeloid and lymphoid cells. By using a newly established autologous in vitro model consisting of functionally impaired PD1 + T lymphocytes, three potential Leishmania major (Lm) host cell types and the cancer therapeutic antiPD1 antibody nivolumab, we could demonstrate that PD1 blockade reinvigorated Tcell effector functions. Depending on the type of parasitized human primary myeloid cell, the magnitude of Tcellmediated parasite elimination varied. Focusing on dendritic cells, we found PD1 blockademediated effects to be partly TNFα dependent. Furthermore, PD1 blockade enhanced almost exclusively Lm induced proliferation of CD4 + and not CD8 + Tcells. Moreover, an increased expression of cytolytic Tcell effector molecules was detected, which are likely to be implicated in reduced parasite survival.
In all, our study gives insight into the role of the PD1/PDL axis during Leishmania infection of primary human cells. This information may be useful for the development of immunothera peutic strategies targeting leishmaniasis.

infection of human Primary Macrophages or Dendritic cells
Human monocytederived macrophages or dendritic cells were detached, counted and seeded in 1.5 or 2 mL microcentrifuge tubes. For infection, stationaryphase Lm promastigotes (wild type or dsRED parasites) were added at a multiplicity of infection (MOI) of 10. After 24 h of incubation at 37°C, 5% CO2, extracellu lar parasites were removed by centrifugation of microcentrifuge tubes and washing steps with CM. (Non) infected hMDM/ hMDDC were analyzed by flow cytometry or used in the CFSE based proliferation assays (see below).

Flow cytometry
For flow cytometric analysis of human primary cells, at least 0.3 × 10 6 cells were labeled with fluorescently labeled antibodies (Table S1 in Supplementary Material) and corresponding isotype controls as defined by the manufacturer. Intracellular proteins were labeled by prior fixation (4% paraformaldehyde) and per meabilization (0.5% saponin) of cells. Intranuclear transcription factors Tbet and GATA3 were labeled by using eBioscience™ Foxp3/Transcription Factor Staining Buffer Set. To label specifi cally nivolumab, a goat antihuman IgGFc polyclonal Fab2 RPE (Dianova; 109116098) was used. Upon analyzing, at least 10,000 events (human cells) were recorded using a BD LSR II flow cytometer (BD Bioscience, Heidelberg). Data were analyzed by FlowJo software (Treestar). elisa TNFα and IFNγ levels were analyzed in the supernatants by using Human TNFalpha DuoSet ELISA (DY008) or Human IFNgamma DuoSet ELISA (DY285) from R&D systems accord ing to the manufacturer's protocol plus a TECAN ® Infinite F50 ® microplate reader.

statistical analysis
Samples were tested for normal Gaussian distribution using D' AgostinoPearson omnibus normality test. In case of nor mally distributed paired samples a parametric paired ttest was performed. Otherwise, the Wilcoxon signedrank test was used. All calculations were done using GraphPad Prism version 7. A value of P < 0.05 was considered statistically significant. Pearson's r correlations were also calculated by GraphPad Prism version 7. r ≥ 0.7 indicates a positive correlation; r ≤ −0.7 indi cates an inverse correlation.

PD-1 ligand cell surface expression on Macrophages and Dendritic cells is Differently Modulated by Leishmania infection
To assess whether PD1/PDLigand (PD1/PDL) interactions can influence Lm infection of human myeloid cells, we first determined cell surface expression of PDL1 and PDL2 after Lm infection. For this purpose, we generated CD14 + CD163 − proinflammatory macrophages (hMDM1), CD14 + CD163 + antiinflammatory macrophages (hMDM2), or CD1a + CD14 − dendritic cells (hMDDC) as described previously (27). We ana lyzed all three host cell types, as their individual roles in human Lm disease are insufficiently clarified. The cell surface density of both PDL1 and PDL2 was quantified by flow cytometry. Therefore, the relative fluorescence intensity (RFI) was measured as the ratio of the mean fluorescence intensity of specific markers to the mean fluorescence intensity of isotype controls. Low levels of PDL1 were detected on hMDM1 (RFI: 1.42 ± 0.28) and hMDM2 (RFI: 1.21 ± 0.23). After Lm infection, expression of both ligands significantly increased (hMDM1: RFI: 2.01 ± 1.09 and hMDM2: RFI: 1.78 ± 0.41, respectively) ( Figure 1A). Interestingly, basal surface expression of PDL1 on hMDDC (RFI: 2.13 ± 0.37) was higher compared to hMDM1 or hMDM2, which however did not increase upon Lm infection (RFI: 2.33 ± 0.36) ( Figure 1A). Focusing on PDL2 expression, also a low basal surface expres sion was observed on hMDM1 (RFI: 1.91 ± 0.88), hMDM2 (RFI: 1.90 ± 0.74), and hMDDC (RFI: 3.27 ± 2.36), respectively. During Lm infection, PDL2 surface expression significantly increased on hMDM1 (RFI: 2.78 ± 1.44) ( Figure 1A). However, surface expression levels of PDL2 on hMDM2 and hMDDC did not differ in presence or absence of Lm infection. Taken together, we show that PDL1 and PDL2 are expressed on all three host cell types and their expression is partially upregulated by Lm infec tion, which is a prerequisite to modulate the PD1/PDL axis by antiPD1 blockade.

Pha Treatment Mimics T-cell exhaustion as Determined by surface expression of Various Marker Proteins
In the chronic LCMV mouse model, where Tcell exhaustion was initially defined, persistent antigen stimulation leads to Tcell exhaustion, which is characterized by a stepwise upregulation of several inhibitory molecules like PD1, PDL1, 2B4, lymphocyte activation gene 3 (LAG3), Tcell immunoglobulin and mucin domain containing3 (TIM3), and CD160 (28,29). To mimic an exhausted phenotype, we stimulated PBLs with PHA for 6 days. In contrast to CD3/CD28 stimulation, PHA stimulation has been reported to lead to an expansion of functionally impaired Tcells (30). After 6 days of PHA stimulation, we observed that most of the Bcells and NKcells were overgrown by Tcells shown by the significant increase of CD3 positivity (96.3 ± 1.58%), compared to the unstimulated control (80.9 ± 5.59%) ( Figure 1B). To assure that the Tcells were not functionally impaired by senescence, characterized by downregulation or loss of CD28, we addition ally analyzed expression of this costimulatory molecule. CD28 expression tended to be higher in PHAstimulated Tcells (RFI: 15.65 ± 10.06) compared to the untreated control (RFI: 12.7 ± 5.81) ( Figure 1B). Therefore, we excluded Tcell senescence. Next, we investigated surface expression of several Tcell exhaustion mark ers on Tcells of PHAprestimulated PBLs (PBLs PHA ). Compared to the unstimulated control, PHAprestimulated Tcells expressed the coinhibitory molecules PD1   (MOI 10) for 24 h and expression levels of PD-L1 or PD-L2 were determined by flow cytometry, respectively. hMDM and hMDDCs were gated using their FSC/SSC properties. Data are presented as mean ± SD RFI (ratio of the mean fluorescence intensity of specific markers to the mean fluorescence intensity of isotype controls). Representative histograms for a surface marker staining of PD-L1 and PD-L2 on infected (black filled histogram) or non-infected (black non-filled histogram) hMDMs and hMDDCs, compared to the respective isotype (black dashed line) control. Statistics were calculated by Wilcoxon matched-pairs signed rank test, P < 0.05 is considered statistically significant (*P < 0.05; **P < 0.01). At least three independent experiments were performed (n = 7-9). Expression of exhaustion marker after PHA-stimulation of PBLs. (B) PBLs were incubated for 6 days with/without PHA (0.5 µg/mL) and expression levels of the indicated surface molecules were determined via flow cytometry. T-cells were gated via their FSC/SSC properties and CD3. Data are presented as RFI (ratio of the mean fluorescence intensity of specific markers to the mean fluorescence intensity of isotype controls) or percentage of CD3 + cells. Statistics were calculated using the parametric paired t-test (CD3, CD28) or Wilcoxon matched-pairs signed rank test; P < 0.05 is considered statistically significant (*P < 0.05; **P < 0.01). At least three independent experiments were performed (N = 6-9). hMDM1, proinflammatory human monocyte-derived macrophages type 1; hMDM2, anti-inflammatory human monocyte-derived macrophages type 2; hMDDC, human monocyte-derived dendritic cells; Lm, Leishmania major; PD-L1, programmed death-1 ligand 1; PD-L2, programmed death-1 ligand 2; MFI, mean fluorescence intensity; PHA, phytohemagglutinin; PBL, peripheral blood lymphocytes; LAG-3, lymphocyte-activation gene 3; TIM-3, T-cell immunoglobulin and mucin-domain containing-3. First, we addressed, whether Tcell effector functions of PBLs PHA can be reinvigorated by the therapeutic antiPD1 antibody nivolumab and whether this is influenced by the type of antigen presenting cell. Therefore, we compared hMDM1, hMDM2 and hMDDC, respectively, as Lm antigen presenting cells in cocultures with autologous PBLs PHA . Cocultivation of Lminfected hMDM1 with PBLs PHA significantly increased Tcell proliferation (12.32 ± 9.49%), compared to the unin fected control (4.63 ± 3.33%) (Figure 2A). This indicates that PHAprestimulated Tcells have residual effector func tions. Furthermore, by blocking PD1, Tcell proliferation (25.96 ± 15.07%) was significantly enhanced (Figure 2A). Concomitantly, we observed a significant reduction in Lm infec tion in the sample cocultivated with PBL PHA (44.62 ± 6.87%) compared to infected hMDM1 only (69.66% ± 5.07%), which consequently was reduced to a stronger extent upon PD1 block ade (39.47 ± 9.48%) (Figure 2A). In cocultures of Lminfected hMDM2 with autologous PBL PHA , no significant difference in Tcell proliferation was observed (3.84 ± 1.74%) compared to the uninfected control (3.21 ± 2.05%) ( Figure 2B). However, PD1 blockade enhanced Tcell proliferation significantly in the infected sample (9.71 ± 5.13%) ( Figure 2B) Figure 2B). Compared to hMDM1 and hMDM2 Lminfected hMDDC induced the highest Tcell proliferation in the PBLs PHA coculture (17.71 ± 10.73%) and, after PD1 blockade, enhanced Tcell proliferation the most (49.30 ± 13.81%) ( Figure 2C). Consequently, less Lminfected hMDDCs were observed in the presence of PBLs PHA (42.02 ± 9.17%) and PD1 blockade damp ened the infection rate of hMDDCs the most (24.68 ± 8.15%), compared to hMDM1 and hMDM2 ( Figure 2C). PD1 blockade in the absence of PBLs PHA did not influence Lm infection rate in hMDM1, hMDM2 or hMDDC ( Figure S1 in Supplementary Material). Furthermore, we showed that PD1 is blocked on Tcells throughout the whole coculture experiment, indicating that the used amount of nivolumab is sufficient ( Figure S2 in Supplementary Material). Our data demonstrate that Tcell func tion is enhanced by PD1 blockade, however, the extent of the effects strongly depends on the Lm host cell phenotype.
TnFα and iFnγ release is strongly enhanced after PD-1 Blockade in the Lm-infected hMDDc coculture The proinflammatory cytokines TNFα and IFNγ play a critical role in shaping the immune response against Leishmania infection (31). Thus, we analyzed TNFα and IFNγ levels in the coculture supernatants by ELISA. PD1 blockade significantly increased TNFα levels in the infected hMDM1 samples (216.60 ± 114.90 pg/mL; 88.57 ± 52.52 pg/mL) ( Figure 2D). Also IFNγ release was augmented by PD1 blockade in the infected hMDM1 sam ples (80.61 ± 59.73 pg/mL; 16.95 ± 25.68 pg/mL) ( Figure 2D). TNFα levels in the hMDM2 samples were similar to the hMDM1 samples. In contrast to hMDM1, IFNγ was barely detectable in the PD1blocked infected hMDM2 sample (17.62 ± 24.18 pg/mL) ( Figure 2E) (Figure 2F). This suggests that effects induced by PD1 blockade may be mediated by TNFα and IFNγ. To investigate this idea, we performed a correlative analysis and neutralized both cytokines.

TnFα but not iFnγ neutralization Partially reversed effects of nivolumab Treatment
To assess whether cytokine release correlates with Tcell prolif eration or Lm infection rate, we performed a correlative analysis.
Remarkably, effector functions of exhausted Tcells can be rein vigorated by blocking the PD1/PDL interaction, thereby reduc ing LCMV loads in the chronic LCMV mouse model (18). This principle has been extended to other applications, like human cancers (42) and animal infection models (21,23). Experimental animal models of leishmaniasis indicate that the use of PD1/ PDL blocking antibodies might be beneficial for treatment of chronic forms of leishmaniasis (21)(22)(23).
Previously we reported that Lm-infected macrophages gener ated from monocytes of Leishmanianaive German blood donors induce a MHC class IIdependent Tcell response in autologous in vitro cocultures. This early Tcell response reduced parasite load in the infected macrophages (11,27). PD1/PDL interac tions are not prominent during the early Tcell priming phase, but they regulate the Tcell response during the effector phase (43). During chronic infections and cancer, PD1/PDL interactions play an important role in induction of Tcell exhaustion (44).
To investigate PD1/PDL interactions in Lm infection of human myeloid and lymphoid cells, we used a newly established  PHAstimulated Tcells express high levels of PD1, PDL1, LAG 3, TIM3, and 2B4, and thus resemble exhausted Tcells.
In the current study, we found the antiPD1 antibody nivolumab to enhance Tcell effector functions and reduce parasite survival differently depending on the Lm host cell phenotype that was initially infected. Our data suggest Lminfected hMDM2 not to be target of PD1/PDLmediated inhibition, whereas PD1/ PDL interactions strongly inhibit Tcell effector functions on Lminfected hMDDC. A reason for this might be that both PD1 ligands are higher expressed on hMDDC compared to hMDM2. Furthermore, antigenpresenting dendritic cells are much more potent in inducing Tcell responses compared to macrophages. Arnold et al. compared differently polarized monocytederived macrophages and dendritic cells in their ability to induce autolo gous Tcells by using mycobacterial purified protein derivate as recall antigen or keyhole limpet hemocyanin as a primary antigen for naive Tcell responses. In this context dendritic cells (hMDDC) induced the highest levels of Tcell proliferation to both antigens, whereas LPS + IFNγtreated macrophages (comparable to hMDM1) were less effective in inducing antigenspecific Tcell responses followed by IL4treated macrophages (comparable to hMDM2) (46).
Focusing on hMDDC, we could partially reverse effects of PD1 blockade by neutralization of soluble TNFα. Chiku et al. demonstrated elevated levels of TNFα and a concomitant decrease of parasite load after PD1/PDL blockade in the canine model of visceral leishmaniasis (47).
Furthermore, we found that neutralization of IFNγ, which was induced after PD1 blockade in the hMDDC samples, had no significant effect on Tcell proliferation or parasite load. IFNγ is reported to mediate resistance to Lm in mice by inducing iNOS expression (2). In human myeloid cells iNOS (or NOS2) expression and function is controversially debated (48,49). In vitro generated human myeloid cells are unable to express func tional iNOS (NOS2) because the essential iNOS cofactor BH4 is missing (48). In contrast to mice, ROS play an important role in parasite control during human Leishmaniasis. IFNγ was shown to induce ROS in Leishmaniainfected human monocytes, which dampened overall parasite load (8). Furthermore, ROS dependent parasite control was rather evident in monocytes from CL patients than in monocytes from healthy individuals (7). Thus, it could be that IFNγinduced ROS production in Lminfected hMDDC was to low to see significant effects of IFNγ neutralization on infection rate in our in vitro model. One reason for why we did not observe significant differences in Tcell proliferation might be the general increased survival of Tcells upon IFNγ. However, IFNγ does not increase the number of antigenspecific Tcell divisions. This was shown, e.g., for murine OVAspecific CD4 + Tcells (50).
In our in vitro model, preferentially CD4 + Tcells proliferated upon Lm infection and nivolumab treatment. In the classical mouse model of Lm, induction of TH1 response leads to resist ance whereas the induction of a TH2 response promotes disease (2). Human cutaneous leishmaniasis patients with moderate disease symptoms show a balanced TH1/TH2 response, whereas an imbalance of TH1/TH2 is associated with disease severity (51).
Experiments with blood of prostate and advanced melanoma cancer patients revealed that PD1 blockade augments TH1 responses und suppresses TH2 responses (52). By analyzing intranuclear expression of the TH1specific Tbet and TH2specific GATA3, we found PD1 blockade to shift Lminduced CD4 + Tcells more toward a TH1 or TH1/TH2 phenotype, respectively. Thus, the higher abundance of TH1 Tcells and their effector func tions may be implicated in reduction of Lm infection.
Specific killing of intracellular parasites in a concerted action of perforin, granzymes, and granulysin (expressed by Tcells) with minimal collateral damage to the host cell was demonstrated in transgenic mouse models (53). PD1 blockade is described to increase perforin, granzyme B, and granulysin expression in Tcells of tuberculosis and cancer patients (54). Focusing on Lminduced CD4 + Tcells, we detected increased intracellular levels of perforin, granzyme A and B, and granulysin after PD1 blockade. This suggests that cytolytic molecules might contribute to the reduction of Lm in hMDDC.
Altogether, our data suggest how the PD1/PDL axis could modulate Lm host cells and CD4 + Tcells in patients suffering from chronic forms of leishmaniasis. So far, the focus of research groups is mostly on CD8 + Tcell exhaustion, which was observed in diffuse cutaneous (40) and visceral human leishmaniasis (24). In case of the cutaneous forms of human leishmaniasis, CD8 + Tcells have a dual role (55) but their contribution in resolving primary cutaneous Leishmania infection might be negligible (56,57). CD4 + Tcells activate leishmanicidal functions of infected macrophages and dendritic cells. In our experiments infected dendritic cells benefit the most from PD1 blockade, as this strongly enhanced CD4 + Tcell effector functions and parasite killing. Additionally we found increased levels of matura tion markers on Lminfected dendritic cells after PD1 blockade ( Figure S3 in Supplementary Material). To induce a strong cell mediated immunity, e.g., after vaccination, adequate maturation of dendritic cells is important. In animal models, it was demon strated that vaccination using L. mexicana LPG induced PD1/ PDL2 expression on several immune cells in a dose dependent fashion (58). Blocking the PD1/PDL interaction could be a valuable approach to enhance efficacy of Leishmania vaccine candidates. Dendritic cellbased immunotherapy in combina tion with antimonials has been shown to significantly reduce parasite burden in experimental models of visceral leishmaniasis (59,60). This approach might also benefit from PD1 checkpoint inhibitors.
Collectively, by using a limited reductionist approach the pre sent work provides new insights regarding the PD1/PDL axis in Lm infection of primary human cells and its consequence for adaptive immunity. Further experiments using material obtained from chronic leishmaniasis patients can contribute to a better understanding of PD1 blockademediated effects.
aUThOr cOnTriBUTiOns CF, PC, ZW, GR, and GVZ contributed to conception and design of the study; CF, KA, and GANN performed and analyzed experiments; CF wrote the first draft of the manuscript; all authors contrib uted to manuscript revision, read, and approved the submitted version.

acKnOWleDgMenTs
We thank Wijnand Helfrich for carefully reading and comment ing on this manuscript. We thank Bianca Walber and Silvia Isabel Rösser for the technical support.

FUnDing
This Project was funded by PaulEhrlichInstitut internal funding.