Impact of Selective CD28 Blockade on Virus-Specific Immunity to a Murine EBV Homolog

Abstract

CTLA-4Ig (belatacept) blocks the CD80/CD86 ligands for both CD28 and CTLA-4; thus, in addition to the intended effect of blocking CD28-mediated costimulation, belatacept also has the unintended effect of blocking CTLA-4-mediated coinhibition. Recently, anti-CD28 domain antibodies (dAb) that selectively target CD28 while leaving CTLA-4 intact were shown to more effectively inhibit alloimmune responses and prolong graft survival. However, the impact of selective CD28 blockade on protective immunity has not been extensively investigated. Here, we sought to compare the impact of CTLA-4Ig vs. anti-CD28dAb on CD8⁺ T cell immunity to a transplant-relevant pathogen, a murine homolog of Epstein-Barr virus. Mice were infected with murine gammaherpesvirus-68 (MHV) and treated with PBS, CTLA-4Ig, or anti-CD28dAb. While anti-CD28dAb resulted in a decrease in virus-specific CD8⁺ T cell numbers as compared to CTLA-4Ig, cytolytic function and the expression of markers of high-quality effectors was not different from CTLA-4Ig treated animals. Importantly, MHV-68 viral load was not different between the treatment groups. These results suggest that preserved CTLA-4 co-inhibition limits MHV-specific CD8⁺ T cell accumulation, but the population that remains retains cytolytic function and migratory capacity and is not inferior in its ability to control viral burden relative to T cell responses in CTLA-4Ig-treated animals.

INTRODUCTION

Limiting immune-mediated damage of the allograft while maintaining protective immunity following transplantation requires precise regulation of the immune system. These processes are carefully controlled by the balance of costimulatory and coinhibitory signals T cells receive. The CD28/CTLA-4 pathway is the prototypic co-signaling pathway in T cells, with CTLA-4 co-inhibition acting as the counter-signal to CD28 costimulation as they compete for the same ligands (CD80 and CD86). Since CD28 costimulation is necessary for optimal T cell activation, immunomodulation via blockade of this pathway has been a promising approach to prevent inappropriate T cell activation in the setting of transplantation.

Belatacept, a recombinant CTLA-4Ig fusion protein, which binds to CD80/86 thus preventing CD28-mediated T cell activation was the first costimulation blocker to be FDA approved for use in clinical transplantation [1 2]. Belatacept-based immunosuppression confers significantly improved long-term graft function and fewer toxicities compared to calcineurin inhibitors (CNIs), with a 43% reduced risk of death or graft loss at 7 years post-transplantation [3]. However, treatment with belatacept is also associated with a significantly higher incidence and severity of acute rejection episodes following transplantation [4]. It is possible that break-through rejection in some individuals could be due to lack of CTLA-4 coinhibition, as evidenced by basic studies demonstrating that CTLA-4 suppresses both CD8⁺ alloreactive memory T cell responses [5 6] as well as Th17 responses in a cell intrinsic manner [7]. Additionally, CTLA‐4‐driven signals are crucial to the suppressive function of Tregs in a cell extrinsic manner [8]. Thus, blocking CD28 while leaving CTLA-4-mediated coinhibition intact could be an effective strategy for modulating immune responses. Indeed, studies have shown that selective targeting of CD28 inhibited autoimmunity [9], graft‐versus‐host disease [10] and prevented acute and chronic rejection in organ transplants in both rodents [11 12] and non-human primates [13]. Additionally, the recent discoveries of the inhibitory interaction between PD-L1 and CD80 [14] and the costimulatory interaction between CD28 and ICOSL [15] reinforce the potential advantage of targeting CD28 over CD80/86.

Recently, an Fc-silent, non-crosslinking anti-CD28 domain antibody (dAb), termed lulizumab, has been developed that monovalently binds to and antagonizes the human CD28 receptor. Lulizumab (BMS-931699) has been shown to be equipotent at inhibiting CD80‐driven T‐cell proliferation and is roughly five times more potent than belatacept against CD86-driven T cell proliferation [16]. Importantly, unlike the early CD28 agonist, TGN1412 [17], no agonist activity, as measured by T‐cell proliferation or cytokine release, was observed with lulizumab in a Phase II clinical trial of patients with systemic lupus erythematosus (SLE; ClinicalTrials.gov: NCT02265744), confirming the preclinical studies [18].

However, given the potent immunosuppressive effects of the anti-CD28 dAb described above, understanding the impact of selective CD28 blockade on protective immunity is an important clinical question, particularly in the context of EBV immunity. Of note, clinical use of belatacept did not increase rates of infection with CMV and BK polyomavirus compared to CNI-treated patients [4], but was associated with increased rates of EBV-associated post-transplant lymphoproliferative disorder (PTLD) in Epstein-Barr Virus (EBV) negative patients [19]. In immunocompetent individuals, EBV-infected lymphocytes are controlled by the immune surveillance activities of virus-specific CD8⁺ T cells [20].

While these data would suggest an important role for CD28 in the priming of naïve EBV-specific CD8⁺ T cell responses, the role of CD28-mediated signals in the generation of EBV-specific CD8⁺ T cell responses and control of viral recrudescence is still controversial. Specifically, two recent studies found that while immune control of MHV-68 was profoundly impaired in CD80/86^−/− animals, immune control of MHV-68 in CD28^−/− animals at greater than day 50 post-infection was similar to that observed in WT animals [21 22]. Thus, these data suggest that blocking CD28 is fundamentally different than blocking CD80/CD86 with regard to EBV-specific adaptive immune responses and viral control. To test this, in this report we directly compared the effects of selective CD28 blockade using murine anti-CD28 dAb with CTLA-4Ig treatment on the T and B cell response to MHV-68, a murine homolog of EBV.

MATERIALS AND METHODS

Mice and Viral Infections

C57BL/6 mice were obtained from the National Cancer Institute repository at Charles River Laboratories (Wilmington, MA) at 6 to 8 weeks of age. Mice were infected i.p. with 2×10³ PFU of a transgenic strain of MHV-68 harboring a fusion protein composed of the enhanced yellow fluorescent protein (EYFP) coding sequence fused to the histone H2b open reading frame, referred to as MHV-68–H2bYFP (kindly provided by Dr. Samuel Speck, Emory University, Atlanta, GA) [23]. All animals were housed in BSL2 animal facilities at Emory University and maintained in accordance with Emory University Institutional Animal Case and Use Committee guidelines (Atlanta, GA).

Antibodies and Flow Cytometry

Splenocytes were processed and T cells were stained for CD8-BV786 (Biolegend, clone 53–6.7), CXCR3-BV421 (Biolegend, clone173), CD27-BV605 (BD Biosciences, clone LG.3A10), PD-1-PE-Dazzle (Biolegend, clone 29F.1A12), TIM-3-PerCP-Cy5.5 (Biolegend, clone RMT3–23), TIGIT-PE (BD Biosciences, clone 1G9), or –BV421 (Biolegend, clone 1G9), 2B4-APC (eBioscience, clone 244F4), as well as APC conjugated H-2K^b/p79 (TSINFVKI) tetramers (NIH Tetramer Core, Atlanta, GA). The FITC channel was left open in order to measure YFP fluorescence of virally infected cells as previously reported [23]. B cells were stained for CD19-PE-Cy7 (Biolegend, clone 6D5), GL7-Pacific Blue (Biolegend, clone GL7), and CD95-APC (Biolegend, clone SA367H8). Absolute numbers were calculated using TruCount bead analysis according to the manufacturer’s instructions (BD Biosciences). Samples were analyzed using a multicolor LSRII (BD). Data were analyzed using FlowJo software (Treestar, San Carlos, CA) and by ViSNE and CITRUS analysis (Cytobank, Inc.) as described in Supplemental Information.

Intracellular Cytokine Staining and CD107a degranulation assay

Splenocytes were restimulated ex vivo in complete media with either 30nM p79 MHV-specific peptide or with 1 μg/ml PMA and 1 μg/ml Ionomycin (Sigma Life Sciences) in the presence of Brefeldin A (BD Biosciences) at 37°C for 4h. Subsequently, cells were surface stained followed by fixation with the Fix/Perm intracellular staining kit (BD Pharmingen) at 4°C for 20 min, and then stained with antibodies against IFN-γ (PE-Dazzle, BioLegend, clone XMG1.2), and TNF (APC, Biolegend, clone MP6-XT22) per manufacturer’s instructions. For assessment of degranulation activity, an anti-CD107a antibody (BV421, Biolegend, clone 1D4B) was added along with the stimuli outlined above, in the presence of Brefeldin A and Monensin (BD Biosciences) and cultured at 37°C for 4h followed by surface staining at 4°C for 20 min.

MHV-68 Specific Serum IgG ELISA

Serum was collected at various time points post-MHV-68 infection for the assessment viral-specific serum antibody as has been previously described [24]. Briefly, NIH/3T3 cells (ATCC CRL-1658) were infected with 0.03–0.1 MOI of MHV-68 and incubated at 37°C for 48 hours. Subsequently, monolayers were washed with PBS and subjected to mechanical disruption to allow for the collection of MHV-68 infected cell suspensions. Cell suspensions were then sonicated twice for 30 seconds to obtain a homogenous cell lysate. Cellular lysate was diluted 1:10 in sterile carbonate-bicarbonate buffer (Invitrogen) and 100ul per well was used to coat 96-well flat bottom microtiter plates (Immulon 4HBX, VWR). Plates were incubated for 48 hours at room temperature. A standard ELISA was then performed utilizing serum dilutions of 1:50 in blocking buffer and goat anti-mouse IgG-HRP (Poly4053, Biolegend) was used for the detection of MHV-68-specific immunoglobulins. Colorimetric development was performed using the TMB substrate system (Thermo Scientific) and acquired at 450nm on a Spectra MAX 340PC Microplate reader (Molecular Devices).

Statistical Analyses

For single comparisons, students unpaired t test was performed. For multiple comparisons, one-way ANOVA tests were performed, followed by Tukey post-test. All analyses were done using GraphPad Prism software (GraphPad Inc, San Diego, CA).

RESULTS

Selective CD28 blockade further inhibits MHV-specific CD8⁺ T cell accumulation but not cytolytic function as compared to CTLA-4Ig

C57BL/6 mice were infected with MHV-68, a murine EBV homolog, and starting at day 0 post infection (dpi), animals were treated with either PBS, 200ug CTLA-4Ig, or 100ug anti-CD28 dAb intraperitoneally every other day until sacrifice. Animals were sacrificed at the peak T cell response (14 dpi) or at a memory time point (28 dpi) and MHV-specific CD8⁺ T cell responses were assessed (Figure 1A). Using a tetramer specific for the immunodominant MHV epitope ORF61/p79/K^b [25], we observed that at 14 dpi, MHV-specific CD8⁺ T cells accumulated less in animals treated with CTLA-4Ig or anti-CD28 dAb when compared to the no treatment control (Figure 1B, top and D). In addition, the anti-CD28 dAb treated group contained a significantly lower frequency and number of MHV-specific T cells compared to the CTLA-4Ig treated group (Figure 1D), suggesting that preservation of CTLA-4 co-inhibition further limited MHV-specific T cell expansion. By 28 dpi, frequencies of tetramer⁺ cells in the CTLA-4Ig treated group were similar to those observed in untreated animals, but they remained significantly lower in the anti-CD28 dAb treated group (Figure 1B and E). Additionally, we also utilized viSNE analysis, from the Cytobank platform (Cytobank.org) to independently analyze the MHV-specific CD8⁺ T cell response (described in Supplemental Methods). viSNE is an algorithm that reduces high-dimensional cytometry data down to two dimensions for ease of visualization and interpretation of exploratory analyses. The resulting viSNE map provides a visual representation of the single-cell data that is similar to a biaxial plot, but the positions of cells reflect their proximity in high-dimensional rather than two-dimensional space. Color is then used as a third dimension to interactively visualize features of these cells [26]. When applied to our data set, viSNE generated a map that clearly separated the MHV-specific tetramer⁺ population in space (p79⁺ cells are shown in red color in Figure 1C) and demonstrated reduced accumulation of this population in the setting of treatment with CTLA-4Ig or selective CD28 blockade (Figure 1C).

Figure 1. Selective CD28 blockade further inhibits MHV-specific CD8⁺ T cell accumulation but not effector function as compared to CTLA-4Ig.

A, Schematic of experimental design. B, Representative flow plots of p79/K^b+ tetramer staining at 14 (top) and 28 (bottom) days post infection (dpi). Frequencies shown represent p79/K^b+ populations within the CD8⁺ T cell compartment. C, Representative viSNE plots of p79/K^b+ tetramer staining. Red indicates high intensity of p79/K^b+ tetramer staining and blue represents low intensity, according to scale shown. Summary data of the frequency (left) and absolute number (right) of p79/K^b+ cells at 14 (D) and 28 (E) dpi. F, Representative flow plots of IFNγ and TNF production within CD8⁺ population following MHV-specific p79-peptide stimulation at 14 dpi. G, Summary data of number of CD8⁺ T cells that express both IFNγ and TNF at 14 dpi. H, Summary data of number of p79/K^b+ CD8⁺ T cells that express both IFNγ and TNF at 14 dpi following PMA/ionomycin stimulation. I, Representative flow plots of CD107a expression within CD8⁺ T cell populations of treated groups following MHV-specific p79-peptide stimulation at 14 dpi. J, Summary data of the number of CD8⁺ T cells that express CD107a at 14 dpi. K, Summary data of the proportion of MHV-specific p79/K^b+ cells that express CD107a following PMA/Ionomycin stimulation. Each point represents data from an individual mouse. n = 5–9 mice per experiment. Data are representative of one-three independent experiments. Student’s t-test = *p<0.05, **p<0.01.

Given the significant reduction in MHV-specific CD8⁺ T cell numbers observed in the setting of CTLA-4Ig and anti-CD28 dAb treatment, we next queried the functionality of the CD8⁺ T cell populations in the different treatment groups. To address this question, animals were sacrificed at 14 dpi and splenocytes were restimulated ex vivo with MHV-specific p79 viral peptide and the frequency of interferon gamma (IFNγ) and tumor necrosis factor (TNF) double producing CD8⁺ T cells was measured. Results indicated that numbers of IFNγ⁺ TNF⁺ cells were similar in CTLA-4Ig treated animals but reduced in anti-CD28dAb treated animals as compared to untreated controls (Figure 1 F-G). Further, there was a statistically significant decrease in the number of IFNγ⁺ TNF⁺ cells when comparing the CTLA-4Ig vs. anti-CD28 dAb treated groups (Figure 1G), which is a reflection of the fact that there are fewer overall antigen-specific p79/K^b+ cells (Figure 1E). However, when the number of p79/K^b+ cells is normalized via gating on p79/K^b+ cells and stimulating with PMA/Ionomycin, no significant differences in the frequency of IFNγ⁺ TNF⁺ cells within splenocytes isolated from CTLA-4Ig- vs. anti-CD28 dAb treated animals were observed (Figure 1H).

Cytolytic function as measured by CD107a degranulation assay was also assessed. Numbers of CD107a⁺ degranulating cells were not different between untreated animals and either CTLA-4Ig or anti-CD28 dAb treated animals (Figure 1I-K). In addition, there was no difference in the number of CD107a⁺ cells when comparing the CTLA-4Ig vs. anti-CD28 dAb treated groups (Figure 1I-K). Likewise, when the number of p79/K^b+ cells was normalized via gating on p79/K^b+ cells and stimulating with PMA/Ionomycin, no differences in the frequency of CD107a⁺ degranulating cells between CTLA-4Ig- vs. anti-CD28 dAb treated animals were observed (Figure 1K).

Impact of CTLA-4Ig and selective CD28 blockade on the quality of effector and memory CD8⁺ T cells

We next endeavored to explore expression of cell surface molecules associated with high-quality effectors. Recently, Woodland and colleagues showed that high co-expression of the chemokine receptor CXCR3 and the costimulatory receptor CD27 marked cells that were better able to respond upon re-challenge in a model of viral infection [27], indicating a higher quality response when compared to CXCR3^loCD27^lo cells. Interestingly, at 14 dpi, CTLA-4Ig and anti-CD28 dAb therapy both resulted in equally reduced expression of CXCR3 and CD27 (Figure 2A) on p79/K^b+ CD8⁺ T cells compared to the no treatment group. Additionally, using the unsupervised viSNE algorithm to extract MFI data, we also observed reduced expression of these molecules in both the CTLA-4Ig and anti-CD28 dAb treated animals relative to untreated controls (Figure 2B). Importantly, however, neither frequency nor MFI was significantly different in the CTLA-4Ig- vs. anti-CD28 dAb-treated groups (Figure 2A, B). At 28 dpi, both treatment groups again showed decreased frequencies of CXCR3⁺ and CD27⁺ cells relative to no treatment (Figure 2C, D), and CXCR3 but not CD27 MFI was higher in the CTLA-4Ig treated group as compared to the anti-CD28 dAb treated group (Figure 2C, D).

Figure 2. Impact of CTLA-4Ig and selective CD28 blockade on the quality of effector and memory CD8⁺ T cells.

Representative flow plots and summary data of the frequency of p79/K^b+ cells expressing CXCR3 (left, top) and CD27 (right, bottom) at 14 (A) and 28 (C) dpi. Representative viSNE plots and summary data of the median fluorescence intensity (MFI) of CXCR3 (left, top) and CD27 (right, bottom) of p79/K^b+ cells at 14 (B) and 28 (D) dpi. Red indicates high expression of a given marker. Each point represents data from an individual mouse. n = 5 mice per experiment. Data are representative of three independent experiments. Student’s t-test = *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

Anti-CD28 dAb does not differentially impact the expression of PD-1, TIM-3, and TIGIT on MHV-specific CD8⁺ T cells vs. CTLA-4Ig

We next examined the expression of key costimulatory and coinhibitory molecules on the surface of MHV-specific CD8⁺ T cells in the different groups using the CITRUS (cluster identification, characterization, and regression) algorithm (Cytobank.org). Traditionally, identifying cellular subpopulations has relied on prior knowledge-driven manual gating of cell subsets within each sample. CITRUS provides a fully automated tool for validation and discovery of statistically significant biological signatures within single cell datasets. CITRUS analysis generates a radial tree that portrays events in a hierarchical manner, with parent clusters giving rise to two or more daughters [28]. We conducted CITRUS analysis on a panel of immune checkpoint markers (PD-1, TIM-3, TIGIT, 2B4) to identify clusters that were significantly more or less abundant between the groups. At 14 dpi, there was a significant decrease in the abundance of cells found within parent cluster 17654 in both the CTLA-4Ig and anti-CD28 dAb treated samples as compared to no treatment (Figure 3A, B), but the abundance of cells in cluster 17654 was not different between the two treatment groups. Intriguingly, this cluster (17654) represented a cell population that co-expressed PD-1, TIM-3 and TIGIT, but not 2B4 (Figure 3A). This expression pattern is consistent with a “cassette” of coordinately regulated coinhibitory receptors recently described in the literature [29]. Of note, CITRUS also identified a separate cluster that expressed high 2B4 but low PD-1, TIM-3, and TIGIT (17651) (Figure 3A). Interestingly, CD8⁺ T cells isolated from either CTLA-4Ig-treated or anti-CD28 dAb-treated animals showed increased abundance of cells within cluster 17651 compared to no treatment. However, the anti-CD28 dAb treated cells had increased abundance of cluster 17651 cells relative to the CTLA-4Ig treated cells (Figure 3B). These results are consistent with our previous findings demonstrating that 2B4 is upregulated in the setting of selective CD28 blockade [12]. At 28 dpi, although the co-expression of the exhaustion cassette of PD-1⁺, TIM-3⁺ and TIGIT⁺ was still present (25466), there was no difference in the abundance of cells within this node between experimental groups (Figure 3C, D). There was also no significant difference between the abundance of cells expressing 2B4 (25457) at this time point (Figure 3D).

Figure 3. Anti-CD28 dAb does not differentially impact the expression of PD-1, TIM-3, and TIGIT on MHV-specific CD8⁺ T cells vs. CTLA-4Ig.

CITRUS plots depicting expression of coinhibitory markers within the p79/K^b+ population at 14 (A) and 28 (C) dpi. Clusters 17654 (D14, green circle) and 25466 (D28, green circle) express high levels of PD-1, TIM-3, and TIGIT. Clusters 17651 (D14, purple circle) and 25457 (D28, purple circle) express high 2B4. CITRUS generated graphs of summary data at 14 (B) and 28 (D) dpi of abundance (top) and percent of cells (bottom) in clusters indicated. Data is representative of three independent experiments. Each point represents one mouse. n = 7–8 mice per experiment. Student’s t-test = *p<0.05, ***p<0.001, ****p<0.0001.

Impact of selective CD28 blockade on MHV viral load

We next queried the impact of anti-CD28 dAb on viral load. We utilized a transgenic MHV-68 virus harboring an EYFP-H2b fusion protein (Figure 4A) which allows for the detection of cells that are latently infected with MHV-68 via flow cytometry, as shown in Figure 4B, where a 10-fold increase in YFP-expressing cells is observed at 14 dpi compared to uninfected controls [23]. This tool has allowed for the ability to correlate surface phenotype with infection and has been shown to accurately and reproducibly reflect viral burden when compared to standard virological techniques. Using this method, at both 14 and 28 dpi, the frequencies of YFP⁺ virally-infected splenocytes were not different between the anti-CD28 dAb- vs. CTLA-4Ig-treated animals (Figure 4B-D). These findings were confirmed by qPCR for expression of the MHV-68 latent viral gene ORF50 as previously described (data not shown) [30].

Figure 4. Impact of selective CD28 blockade on MHV viral load.

A, Generation of the transgenic virus, MHV68-H2bYFP. Modified from Collins and Speck, 2014 [23]. B, Representative flow plots of YFP expression in uninfected (left) vs 14 dpi (right), indicating latently infected cells. C, Representative flow plots of YFP expression amongst total live splenocytes at 28 dpi. D, Summary data of the frequency of latently infected cells expressing YFP at 14 (left) and 28 (right) dpi amongst total splenocytes. CITRUS plots depicting GL7⁺ CD95⁺ germinal center B cell clusters (blue circle) and YFP⁺ (green circle) at 14 (E) and 28 (G) dpi amongst total CD19⁺ cells. Red indicates high expression according to scale shown; blue indicates low expression. CITRUS generated graph of summary abundance of cells within Clusters 74988 (F, D14) and 74970 (H, D28). I, ELISA data of MHV-specific IgG production at the indicated time points post-infection. J, Schematic of experimental design. K, Summary data of the proportion of cells expressing YFP following intranasal viral re-challenge with 1×10⁴ PFU of MHV68-H2bYFP. Data are representative of one-three independent experiments. Each point represents one mouse. n = 5–10 mice per group. Panel D represents pooled data from two independent experiments. Student’s t-test = **p<0.01, ***p<0.001, ****p<0.0001.

It is well established that human EBV infection is found primarily first in the epithelial cell compartment of the nasopharynx, followed by the B lymphocyte compartment where the virus takes up permanent residence [31 32]. The same is true in MHV-68, where it is well recognized that the virus preferentially attacks the germinal center (GC) B cell compartment in the spleen in order to establish latency and gain access to the long-lived memory B cell pool [23]. We thus sought to determine whether the control of latent infection in our treated mice could be due, in part, to a loss of the GC B cell reservoir in the setting of costimulation blockade. To this end we evaluated the CD19⁺ GL7⁺ CD95⁺ compartment, well described markers of GC B cells, at 14 and 28 dpi using CITRUS. Gating solely on CD19+ B cells, results indicated that the abundance of cells in cluster 74988 (14 dpi) and cluster 74970 (28 dpi) that co-express GL7 and CD95 were significantly lower in both treatment groups compared to the no treatment control, indicating fewer cells of the GC B cell phenotype (Figure 4E-H). This analysis tells us that in the setting of gammaherpesvirus infection, there is a significant inhibition of the GC reaction following costimulation blockade with either CTLA-4Ig or anti-CD28 dAb. Interestingly, CITRUS was also able to detect that among all CD19+ B cells, the GC B cell clusters 74988 (D14) and 74970 (D28) co-expressed YFP, indicating that the majority of latently-infected cells were contained within the GC B cell population, confirming previously published reports [23] (Figure 4E, G).

MHV-specific antibody responses are required for long-term control of viral recrudescence [33 34]. Thus, in light of previously published reports suggesting that CD28 and CD80/86 deficient animals exhibit an impaired antibody response [33 35], we next measured MHV-specific antibody titers in plasma isolated from CTLA-4Ig or anti-CD28 dAb treated recipients at serial time points post-infection by ELISA. Interestingly, MHV-specific IgG antibody titers in both the CTLA-4Ig and anti-CD28 dAb treated groups were not lower than in untreated animals. Further, no difference was observed in the production of viral-specific IgG antibody at any time point when comparing the CTLA-4Ig and anti-CD28 dAb treated groups (Figure 4I).

Finally, to determine whether the observed T and B cell responses following MHV-68 infection were protective in the setting of anti-CD28 dAb treatment (relative to CTLA-4Ig treated animals) we subjected mice that were previously infected with MHV-68 to a intranasal rechallenge and measured their viral load via expression of YFP (Figure 4J). While viral load on day 3 following in vivo rechallenge with MHV-68 was increased in both the CTLA-4Ig and anti-CD28 dAb treatment groups relative to untreated controls, we observed equivalent expression of YFP between the CTLA-4Ig and anti-CD28 dAb treated animals, indicating similar levels of latent viral burden (Figure 4K). We also found there to be no significant differences in the MHV-specific T cell response in regards to p79/K^b+-tetramer staining, cytokine production or degranulation between the CTLA-4Ig vs. anti-CD28 dAb treatment groups (Figure S1A-F).

Discussion

The data presented here directly compare the effects of selective CD28 blockade with CTLA-4Ig treatment on the primary CD8⁺ T cell response to gammaherpesvirus infection, a transplant-relevant pathogen. We show that despite increased inhibition of MHV-specific CD8⁺ T cell accumulation in the setting of anti-CD28 dAb treatment, selective CD28 blockade did not inhibit MHV-specific CD8⁺ T cell cytolytic function, reduce the quality of the memory T cell response, differentially impact the expression of the PD-1, TIM-3, or TIGIT coinhibitory receptors, the MHV-specific antibody response or viral load when compared directly with CTLA-4Ig treatment. These results indicate that selective CD28 blockade is not inferior to CTLA-4Ig in its ability to preserve protective immunity to MHV.

Due to the potent immunosuppressive effect of selective CD28 blockade on alloreactive CD8⁺ T cells compared to CTLA-4Ig, observed in several transplant models [12 13], it was not surprising that we observed greater inhibition of the antigen-specific CD8⁺ T cell response following treatment with anti-CD28 dAb in the setting of infection. We recently showed that the increased efficacy of selective CD28 blockade is due to preservation of CTLA-4-mediated inhibition [12], leading to cell intrinsic inhibition of the antigen-specific CD8⁺ T cell response [36 37]. It is also now well established that CTLA-4 signaling plays both an intrinsic inhibitory role on effector T cells [36 37] and an extrinsic role [8] in the suppressive function of T regulatory cells (Tregs). Indeed, the use of a CD28 monovalent fusion protein, termed sc28AT, in a model of non-human primate transplantation, showed that treatment with selective CD28 blockade was able to directly impair effector T cell function while also increasing the suppressive activity of Tregs and favoring their infiltration into the graft [13].

Given the observed attenuation of the antigen-specific CD8⁺ T cell response discussed above, our finding of similar viral burden in anti-CD28 dAb- vs. CTLA-4Ig treated animals was perhaps unexpected. There are several possible explanations for this finding. First, the CD8⁺ T cell function and phenotypic data presented in Figures 1–3 show no significant differences between the anti-CD28 dAb and CTLA-4Ig treated animals, indicating that CTLA-4 is dispensable for cytolytic function of these antigen-specific cells in the setting of selective CD28 blockade. This finding confirms previously published work from our lab in the setting of transplantation, where while numbers of donor-reactive CD8⁺ T cells were significantly reduced in anti-CD28 dAb- vs.CTLA-4Ig-treated animals, effector function was not different between the two groups [38]. Overall, the similar function and phenotype of the CD8⁺ T cell response between the two populations suggests that those MHV-specific CD8⁺ T cells that remain in the anti-CD28 dAb treated animals may be qualitatively similar to those isolated from CTLA-4Ig treated animals.

Secondly, our results show that the anti-CD28 dAb also better controls the GC B cell response over time following MHV infection (Figure 4), which is consistent with previously published results in the setting of transplantation [39]. It is well established that the viral reservoir in MHV-68 infection is the GC B cell pool and that MHV-infected cells drive expansion of this population [23]. Thus, by reducing the reservoir available for infection and further propagation of the virus, anti-CD28 dAb treatment may result in a smaller pool of infected cells that necessitate CD8⁺ T cell-mediated immune surveillance, as we observed in Figure 4D. Indeed, early during the response (day 14) both costimulation blockade groups appeared to have a lower viral burden compared to untreated animals (Figure 4D). We posit that this is due to the fact that blockade of the CD28 signaling pathway leads to a near complete attenuation of the GC B cell response, significantly shrinking the latent viral reservoir for MHV-68 in mice. This interpretation is supported by a previous study in this viral model showing that a lack of T cell help in the GC response resulted in a significant reduction in the number of MHV-68 latently infected B cells [40].

Finally, the seemingly incongruous result of comparable latent viral burden despite marked loss of MHV-specific T cell responses is potentially explained by the preserved MHV-specific antibody response in both the CTLA-4Ig and anti-CD28 dAb treated groups (Figure 4I). Additionally, viral re-challenge suggests that the memory CD8⁺ T cell response generated in the setting of the selective CD28 blocker is not different from that generated in CTLA-4Ig-treated animals (Figure 4K).

It is important to note that one limitation of this mouse model is the inability to study the development of PTLD per se. Previous studies have shown that the rates of development of PTLD in MHV-infected mice is exceedingly low [41], rendering the use of a model in which only a few percent of the mice can be studied unfeasible for both financial as well as animal usage (i.e. IACUC) considerations. Thus, while this model yields important insight into the impact of selective CD28 blockade on virus-specific T and B cell responses and viral load, there may be other parameters that factor in to the risk of development of PTLD in EBV-infected human transplant recipients. Similarly, while we have shown the impact of selective CD28 blockade during priming on the secondary immune response, the impact of CD28 blockade during the secondary response is an important and clinically relevant question that should be addressed in future investigations.

Taken together, the data presented here indicate that use of selective CD28 blockade, in the form of non-crosslinking anti-CD28 dAb is not inferior to CTLA-4Ig in terms of its ability to control viral load during gammaherpesvirus infection. Coupled with our work and that of others showing that use of anti-CD28 dAb confers prolonged allograft survival in both mouse and non-human primates relative to CTLA-4Ig/belatacept [12 13], these data further highlight the clinical promise of this therapy. Indeed, an NIH funded clinical trial is currently underway at Emory University to test the efficacy of selective CD28 blockade-based immunosuppression in renal transplant recipients. Understanding the potential impact of selective CD28 blockade on protective immunity will continue to be an essential aspect of both pre-clinical and clinical investigation as this promising new therapy makes its way through the translational pipeline.

Abbreviations

CNIs

calcineurin inhibitors

dAb

domain antibodies

EYFP

enhanced yellow fluorescent protein

EBV

Epstein-Barr Virus

GC

germinal center

MHV

murine gammaherpesvirus-68

IFNγ

nterferon gamma

PTLD

post-transplant lymphoproliferative disorder

SLE

systemic lupus erythematosus

T cells

T regulatory cells

TNF

tumor necrosis factor