Selective CD28 Antagonist Blunts Memory Immune Responses and Promotes Long-Term Control of Skin Inflammation in Nonhuman Primates Free

Current systemic treatments for autoimmune diseases include those inhibiting the whole immune system with immunosuppressive drugs and those that dampen inflammation as a whole. Novel anti-inflammatory therapeutic agents (such as anti-TNF or anti–IL-17) added to the therapeutic arsenal are very efficient at treating symptoms but fail to prevent relapses because they do not prevent memory cells from rapid reactivation or survival (1). Immunological memory generated after T cell activation is induced by specific Ag recognition and engagement of costimulatory molecules. Costimulation by molecules of the CD28 family dominantly controls the priming of naive T cells and their differentiation into pathogenic effector cells after initial Ag exposure. Compared to their naive counterparts, memory T cells are long-lived and have rapid recall effector function with reduced activation requirements. There was a consensus that memory T cells do not require CD28 costimulation for expansion during secondary responses (2), based on a few studies in vitro with strong and sustained nonphysiological TCR stimulation (3, 4) and in vivo models in CD28-deficient mice (5, 6) that present abnormal immunity. Some preclinical studies in macaques suggested also that memory T cells promote allograft rejection, particularly in costimulation blockade-based immunosuppressive regimens (7, 8). However, in contrast to humans or baboons, CD4⁺ effector memory T lymphocytes (Tem) of macaques do not express CD28 (9, 10) and might show different costimulation requirements for their activation than do human and baboon counterparts. Furthermore, CD28 blockade was initially based on the use of CD80/86 antagonist (CTLA4-Ig, abatacept or belatacept), which on the one hand blocks CD28 access to its ligands CD80/86, and on the other hand also inhibits CTLA-4 and PD-L1 signals crucial to the function of regulatory T cells (11, 12) and to the self-inhibition of autoreactive and memory T cells such as Th17 (13). It has been deduced from these data that CD28 blockade failed to block memory T cell responses. However, another possible explanation for past observations is that macaque CD4⁺ Tem might be different from humans (and baboons) and that CTLA-4 and PD-L1 signals might significantly participate in the control of Tem responses.

Selective inhibition of CD28 refers to a blockade of CD28-CD80/86 by targeting CD28 itself, sparing both CTLA-4 and PD-L1 interactions with CD80/86. Selective CD28 blockade was described to induce immune tolerance and regulatory cells in transplant experimental models (14–18). We previously described that selective monovalent CD28 versus CD80/86 antagonists differentially control human effector and regulatory memory T cell activation in vitro at the immune synapse level (19). In the present study, we evaluated a novel selective antagonist of CD28 (FR104, a humanized pegylated anti-CD28 Fab′ Ab fragment), devoid of agonist activity (20, 21), in Ag-specific memory responses in vitro and in nonhuman primate models.

FR104 is a monovalent humanized Fab′ Ab fragment antagonist of CD28, pegylated to prolong its half-life as described previously (20). Briefly, nucleotidic sequence corresponding to the V_H and V_L domains of the CD28.3 mAb were humanized using the des-immunization technology (Algonomics, Gent, Belgium) and fused with human CH1 and Ck Ig domains, respectively. A C-terminal cysteine was added to the H chain to serve as specific branching site for pegylation. The resulting Fab′ Ab fragment was then conjugated with a 40-kDa polyethylene glycol moiety.

Baboons (Papio anubis; 7–14 kg) were obtained from the Centre National de la Recherche Scientifique Centre de Primatologie (Rousset, France). The animals were housed at the large animal facility of the INSERM unit 1064. Animal studies were approved by the French National Ethics Committee (no. CEEA-2011-51).

Peripheral blood from healthy volunteers was obtained by the Etablissement Français du Sang (Nantes, France). All donors were informed of the final use of their blood and signed an informed consent. PBMC were collected from CMV⁺ healthy volunteers and extracted by Ficoll gradient. T lymphocytes were enriched by negative selection with a pan T cell isolation kit (Miltenyi Biotec, Paris, France) and autoMACS separator (Miltenyi Biotec). Enriched human T lymphocytes were then stained with fluorescent mAbs against human CD4 (L200), CD8 (RP4-T8), CD45RA (T6D11), and CCR7 (REA108). Naive (CD4⁺ or CD8⁺ CD45RA⁺CCR7⁺) and memory T cells (CD4⁺ or CD8⁺ without CD45RA⁺CCR7⁺naive cells) were then sorted with a high-speed cell sorter (FACSAria; BD Biosciences, San Jose, CA) and FACSDiva software (BD Biosciences) with a purity >95%. Cells were then stained with the cell proliferation dye eFluor 450 (eBioscience, Paris, France), washed, and counted in TexMACS medium (Miltenyi Biotec). T cells (10⁵) were cocultured with 2 × 10⁵ allogeneic PBMC irradiated at 35 Gy or 10⁵ T cell–depleted autologous PBMC with 2 μg/ml CEF peptide pool “plus” or CEFT peptide pool “MHC II plus” peptides (CTL Europe, Bonn, Germany). Selective CD28 antagonist (FR104; Effimune, Nantes, France) or CD80/86 antagonist (CTLA4-Ig, belatacept; Bristol-Myers Squibb, New York, NY) was added at 10 μg/ml. T cell proliferation was acquired on flow cytometer on day 5 for allogeneic stimulation and on day 9 for peptide stimulation.

Baboons were immunized intradermally twice with a bacillus Calmette–Guérin vaccine (0.1 ml; 2–8 × 10⁵ CFU; Sanofi Pasteur MSD, Lyon, France) in the upper region of the leg, 4 and 2 wk before the delayed-type hypersensitivity (DTH) skin test as previously described (22). Ag-specific T cell frequency was followed with an IFN-γ ELISPOT assay (nonhuman primate IFN-γ ELISPOT kit; R&D Systems, Minneapolis, MN) on freshly isolated PBMC, according to the manufacturer’s instructions. Intradermal reactions (IDR) were performed with duplicate 0.1 ml intradermal injections of two doses (2000 or 1000 IU) of tuberculin purified protein derivative (Synbiotics, San Diego, CA) in the skin on the right back of the animals. Saline (0.1 ml) was used as a negative control. Dermal responses at the injection sites were measured using a caliper square. The diameter of each indurated erythema was measured by at least two observers from days 3 to 12 and were considered positive when >4 mm in diameter. The mean of the reading was recorded and plotted for each time point. To compare multiple experimental conditions, erythema responses were quantified as area under the curve (AUC) using GraphPad Prism software for calculation. Skin biopsies were performed on day 4 on one duplicate. A second IDR was performed after 3 wk and animals received one i.v. injection of either 0.1, 1, or 10 mg/kg FR104 (Effimune) or 10 mg/kg CTLA4-Ig (belatacept) or an equivalent volume of excipient 1 d before the second challenge with tuberculin. Other IDR were performed every 3–4 wk without any further injection of the drug.

Two months before drug administration, baboons were immunized with i.m. injection of 0.6 ml keyhole limpet hemocyanin (KLH) at 9 mg/ml (Merck Milipore, Darmstadt, Germany) and 0.6 ml CFA (Sigma-Aldrich, Saint-Louis, MO). Baboons were treated once i.v. with either 0.1, 1, or 10 mg/kg FR104 or an equivalent volume of excipient. Just after drug administration, baboons received an i.v. administration of 1.5 ml/kg SRBC at 10% (Eurobio, Courtaboeuf, France). Animals were then challenged a second time with SRBC and KLH 5 mo after drug administration using the same protocol, excepted that the second KLH challenge was performed with IFA. Sera from animals were collected over time and IgG titers were determined by serial dilution on SRBC by flow cytometry using a fluorescent anti-human IgG (Dako, Glostrup, Denmark) or on recombinant protein by a KLH ELISA test kit (Stellar Biotechnologies, Port Hueneme, CA).

Leukocytes were from baboon blood and prepared by RBC lysis. Fluorescent mAbs against human CD3 (SP34-2), CD4 (L200), CD25 (MA251), CD28 (28.6), and CD127 (hIL-7R-M21) were from BD Biosciences. An anti-human Foxp3 staining kit (236A/E7) and anti-human CD95 (DX2) were purchased from eBioscience. FR104 staining was performed with a polyethylene glycol rabbit mAb (Epitomics) followed by fluorescent goat anti-rabbit IgG (Invitrogen, Grand Island, NY). CD28 receptor occupancy by FR104 on T lymphocytes was determined by performing the ratio of mean fluorescence intensity of FR104 staining between an unmodified blood sample and a blood sample incubated for 30 min at room temperature with a saturating concentration of FR104 (5 μg/ml) as previously described (23). This represents the percentage of maximal binding. Samples were acquired on a BD FACSCanto flow cytometer (BD Biosciences) and analyzed with FlowJo software.

Frozen sections (10 μm) were prepared from surgical skin biopsies. Slides were air-dried at room temperature overnight before acetone fixation for 10 min at room temperature. Sections were saturated with PBS containing 10% baboon serum, 2% normal goat and donkey serum, and 4% BSA. Sections were incubated overnight with primary Abs at 4°C, followed by fluorescent secondary Abs as previously described (15, 22). T cell infiltration analysis was performed with rabbit anti-human CD3 (Dako) and macrophages infiltration with mouse anti-human CD68 (clone KP1; Dako). Slides were analyzed with the AxioVision imaging software (Carl Zeiss, Le Pecq, France).

DNA was isolated from PBMC extracted by Ficoll gradient, saliva, spleen, and kidney using the DNeasy blood and tissue kit (Qiagen, Manchester, U.K.) and viral RNA from sera using the QIAmp viral RNA mini kit (Qiagen) and from urine using the QIAmp MinElute virus spin kit as per the manufacturer’s instructions. Samples were taken, at set intervals, before and after treatment with FR104, and quantitative PCR was used to test the levels of virus. All assays used the TaqMan gene expression system (Applied Biosystems) in a final volume of 25 μl using the ViiA 7 real-time PCR system (Applied Biosystems). Analysis of CMV, lymphocryptovirus (LCV), SA12, and hepatitis E virus (HEV) were as described previously by Haanstra et al. (24). Analysis of herpesvirus papio 2 (HVP-2) in saliva used the method as previously described by Rogers et al. (25). The pUC19–HVP-2 plasmid used as the standard was provided by Prof. R. Eberle (Oklahoma State University, Stillwater, OK). Nucleic acid integrity was assessed using the 18S rRNA assay (Eurogentec, Southampton, U.K.).

Continuous variables were expressed as the mean ± SEM, unless otherwise indicated, and compared with the Mann–Whitney nonparametric test. A p value <0.05 was considered statistically significant. All statistical analyses were performed on GraphPad software.

We have recently reported that selective blockade of CD28 with FR104 is effective in naive nonhuman primates to prevent allograft rejection (23) and protects from acute fatal experimental autoimmune encephalomyelitis (24), the elected experimental model of multiple sclerosis. To evaluate the role of CD28 on memory T lymphocyte reactivation, we purified naive (CD45RA⁺CCR7⁺) and memory (CD45RA⁺CCR7⁺-depleted cells) human T lymphocytes from healthy volunteers and stimulated these cells with either irradiated allogeneic PBMC or autologous T cell–depleted PBMC pulsed with pools of 32 HLA class I–restricted peptides and 23 HLA class II–restricted peptides from CMV, EBV, influenza virus, and tetanus toxin. As expected, allogeneic stimulation induced both naive and memory T cell proliferation, and autologous presentation of viral peptides induced proliferation of memory T lymphocytes (Fig. 1). Selective blockade of CD28 during allogeneic or viral peptide stimulus significantly prevented proliferation of both naive and memory T lymphocytes. CTLA4-Ig prevented similarly alloantigen-induced proliferation of naive and memory T cells but failed to control proliferation of memory T lymphocytes stimulated with viral peptides. These findings indicate that human memory T cell reactivation in vitro is CD28-dependent and might be controlled by immune checkpoints (e.g., CTLA-4 and/or PD-L1, which interact with CD80/86 molecules) (11).

To investigate the role of CD28 in vivo on memory T cell reactivation, we first used a DTH model in baboons, having demonstrated that expression of CD28 on baboon memory T cells has a similar pattern to humans, which is not the case for macaques (9, 10). Baboons previously sensitized with bacillus Calmette–Guérin vaccine were challenged by intradermal injections of tuberculin to induce memory Th1-mediated skin inflammation. One month after a first positive IDR, animals were treated with a single dose of selective CD28 antagonist (FR104) or control excipient, and were challenged again by IDR with tuberculin 24 h later. Tuberculin IDR were repeated every month for 6 mo, including after drug elimination according to pharmacokinetic and receptor occupancy monitoring (Supplemental Fig. 1). Whereas control animals developed consistent erythema after IDR challenges during six months, a single administration of FR104 at 10 mg/kg i.v. prevented skin inflammation for at least 2 mo (Fig. 2A), and this was accompanied by a 100% receptor occupancy on blood T cells (Fig. 2C). Surprisingly, long-term responses to IDR challenges remained significantly weak (both in terms of intensity and duration of erythema) even several weeks after complete drug elimination (>17-fold terminal half-life) (Fig. 2). At month 5, at a time when IDR was still inhibited, these animals were nevertheless immunocompetent because they were able to mount an Ab response after injection of sheep RBC (Supplemental Fig. 2).

Skin histological analyses on biopsies performed on day 4 after tuberculin challenge revealed high dermal perivascular infiltration by T lymphocytes and macrophages in control animals as opposed to naive skin from the same animals (Fig. 3). In contrast, in animals treated with FR104 at 10 mg/kg, we observed no T lymphocyte infiltration and normal subepidermal macrophage infiltrates, similar to naive skin (Fig. 3).

Long-term control of IDR by FR104 was dose-dependent, because the 1 mg/kg dose was effective during the period of high receptor occupancy (immediately after drug administration) but was ineffective at inducing maintenance of hyporesponsiveness when receptor occupancy was partial and after drug elimination (Fig. 2C, green panel). Animals treated with 0.1 mg/kg did not display reduced IDR intensity even though a 90% receptor occupancy in blood had been reached shortly after drug administration. Receptor occupancy in tissue (not determined in this study) might be weaker, which could explain why a low dose was ineffective although achieving nearly complete receptor occupancy in blood. To confirm that the long-term induction of hyporesponsiveness seen after injection of 10 mg/kg FR104 was triggered by tuberculin purified protein derivative Ag, as opposed to a possible nonspecific hyporesponsiveness, animals received a dose of 10 mg/kg FR104 in the absence of tuberculin challenge. After drug elimination, on month 5, animals were rechallenged with tuberculin IDR and showed normal responses (Fig. 2C, gray panel). Finally, confirming in vitro observation, animals treated with 10 mg/kg i.v. of the CD80/86 antagonist CTLA4-Ig (belatacept) did not display reduced IDR intensity (Fig. 2C, orange panel) whereas this drug prevented in vitro baboon T lymphocyte proliferation with similar efficacy to FR104, and that dose allowed high concentration of CTLA4-Ig in sera and significant binding on blood circulating cells during the IDR experiment (Supplemental Fig. 3). Altogether, selective blockade of CD28 with FR104, but not CTLA4-Ig, dose-dependently prevented memory Th1-mediated skin inflammatory responses in vivo and induced Ag-specific hyporesponsiveness persisting after drug elimination.

FR104 did not induce general lymphocyte depletion, and no significant modification of regulatory T cells number or frequency was observed in the blood (Supplemental Fig. 4). No significant modulation of total blood lymphocytes numbers was observed, including in the CD4⁺ or CD8⁺ memory T cell subsets (Fig. 4A–C). Additionally, the frequency of tuberculin-specific memory T lymphocytes in the blood, assessed by ELISPOT, remained unmodified (Fig. 4D). Addition of IL-2 or depletion of CD25⁺ T cells in this assay failed to demonstrate implication of these factors in a potential suppressive event. However, addition of FR104 in vitro to the ELISPOT assay did reduce the frequency of IFN-γ–secreting cells, suggesting that CD28 is required for reactivation of these memory cells ex vivo (Fig. 4D). Finally, mRNA transcripts analyses of skin biopsies after tuberculin challenge showed an increase in expression of TGF-β and a decrease in expression of CD28 and TCR β constant chain after FR104 administration (Fig. 5A), confirming histological observation that T lymphocytes poorly infiltrate the skin when CD28 is blocked. After normalization to T cell infiltrates with TCR C-β as a housekeeping gene, to compare the quality of T cell infiltrates, we found an accumulation of Helios and PD-1 transcripts (Fig. 5B), suggesting that T cells, which still infiltrated the skin after CD28 blockade, harbored a more regulatory phenotype.

To investigate the role of CD28 in vivo on memory T-dependent humoral responses, we first used the T-dependent SRBC immunization model. Indeed, Old World primates (such as baboons and macaques), apes, and humans differ from other mammals because they contain significant levels of serum IgM and IgG Abs directed against anti-Gal carbohydrate residues expressed by other species (i.e., sheep) and microorganisms (26). We performed the first i.v. SRBC injection 24 h after excipient or FR104 administration at different doses. All animals significantly increased their serum anti-SRBC IgG titers within 2 wk (Fig. 6), but not baboons treated with 10 mg/kg FR104. To assess possible induction of Ag-specific humoral hyporesponsiveness, we challenged these animals a second time after complete drug elimination. Upon second SRBC challenge, all baboons demonstrated increased serum anti-SRBC IgG titers, including FR104-recipent baboons, which did not respond after the first challenge.

To investigate whether FR104 disrupted unrelated humoral immunity, we used a second T-dependent humoral response model. Two months before drug treatment, baboons were immunized against KLH. All animals developed specific anti-KLH IgG within 2 wk and titers were monitored over time (Fig. 6). Animals were then treated with FR104 but not challenged with KLH during exposure to the drug. Existing anti-KLH IgG titers were not modified by FR104 treatment, even at higher doses, and all animals responded similarly to a second KLH challenge performed after drug elimination.

Finally, to evaluate whether FR104 impaired memory immunity against chronic latent virus infection, which could translate into reactivation of quiescent viruses, we examined, by quantitative PCR, a number of viruses relevant to human immune-related complications: HEV, polyomavirus (SA12, closely related to BK virus), HVP-2 (equivalent to human HSV), and CMV and LCV (equivalent to human EBV). For HVP-2, saliva extracts were used to determine viral copy number. HVP-2 was detected in all animals during the experimental time period, with little nonsignificant changes observed in treated animals versus controls (Fig. 7A). LCV was also detected in the PBMC of all groups, but viral copy levels did not differ in the treated groups when compared with the controls (Fig. 7B). Similarly, DNAemia in serum showed variation in levels during the time period with a slight increase demonstrated in the control animals and those treated with 0.1 mg/kg, but values were not indicative of significant reactivation (Fig. 7C). Assessment of CMV copy number in PBMC demonstrated no detectable virus in either the control group or animals treated with 0.1 mg/kg FR104. In the 1 mg/kg group, although viral copies were not detectable at the initiation of the study, we observed a modest increase at the end of the experiment after drug elimination. In contrast, in the 10 mg/kg group, significant viral copy numbers of CMV were measured before treatment, but they did not change significantly during the 20-wk period of follow-up (Fig. 7D). For SA12 polyomavirus, we did not observe significant viral copy number modification in the treated groups when compared with control animals in either PBMC or urine during the 20 wk of follow-up. Virus was either undetectable or below the limit of quantification, that is, <10–100 copies/μg DNA (data not shown). Similarly for HEV, no virus was evident in the serum for any of the treated groups. Only in the control group was virus observed at week 15 (range, 591 ± 48 to 3412 ± 184 IU/ml) and no viral shedding was observed in the feces at any time point and in any animal tested (data not shown).

Altogether, these results suggested that CD28 is required in the baboon for reactivation of T-dependent humoral memory responses in vivo. However, it has not been possible to induce long-lasting hyporesponsiveness of T-dependent humoral memory with FR104, and the treatment did not disrupt preformed pathogen immunity.

Our study questioned the requirement of CD28 costimulation for memory T lymphocyte reactivation at the preclinical level. We assessed the efficiency of selective CD28 blockade in a nonhuman primate skin inflammatory model on the control of cellular and humoral memory responses and the effect on latent viral infections. It had been shown previously that CD28 blockade induces immune regulation and prevents allograft rejection or autoimmune attacks in rodents (14) and nonhuman primate (15, 23, 24) models. However, in these models, animals were immunologically naive toward immunizing Ags. Homeostatic proliferation (27) and heterologous immunity (cross-reactivity between Ags and environmental pathogens) of memory T lymphocytes (28) have been identified as a major barrier of tolerance induction by costimulatory blockade, in particular in monkey models (7, 8, 29). Memory T lymphocyte reactivation was therefore considered as less dependent or indeed independent of costimulation, in particular for the CD28 pathway. However, CD28 blockade has so far been experimentally achieved using antagonists of CD80/86, the molecular ligands of CD28, that in addition to inhibiting CD28-mediated costimulation also prevent interaction of CD80 with crucial checkpoints such as CTLA-4 and PD-L1 (11). We have previously reported that selective blockade of CD28 differentially controls in vitro human effector and regulatory memory CD4⁺ T lymphocyte activation in comparison with CD80/86 antagonist (19). CTLA-4– and PD-L1–dependent mechanisms were shown to play a key role in the stability of the immune synapse and on the velocity and motility of memory T lymphocytes. Similarly, CTLA4-Ig, a CD80/86 antagonist, was described recently to inhibit in vitro stimulation of Th1 responses but to increase Th17 cell responses due to higher expression of CTLA-4 by this memory subset and greater sensitivity to coinhibition by CTLA-4 (13).

To understand the requirement of CD28 for memory T lymphocyte reactivation, we used an anti-CD28 Fab′ antagonist devoid of agonist or superagonist activity as previously reported (20, 21). We first purified human naive and memory T lymphocytes from healthy volunteers and evaluated their proliferation after stimulations with alloantigens or with 55 HLA class I– and class II–restricted peptides from CMV, EBV, influenza virus, and tetanus toxin. We found that selective blockade of CD28 in vitro is as effective on memory as it is on naive T cells to prevent proliferation induced by either alloantigens or viral peptides. However, this was not the case using an antagonist of CD80/86 (CTLA4-Ig), which inhibits similarly naive and memory T cell proliferation stimulated with alloantigens but not when stimulated with viral peptides. This is in accordance with a recent study reporting that CTLA4-Ig efficiency decreased in increasingly matured human T cells when stimulated with a CMV peptide or alloantigens (30). Collectively, our results suggest that human memory T cell reactivation in vitro is still CD28-dependent but tightly controlled by CTLA-4 and/or PD-L1 coinhibitory signals as described in mouse models of candida or Listeria monocytogenes infection (31, 32).

We next evaluated the role of CD28 costimulation on memory cellular immune responses in vivo in the baboon, because these animals express CD28 on effector memory CD4⁺ T lymphocyte as do humans, which is not the case for rhesus or cynomolgus macaques classically used in nonhuman primate experimental models (9, 10). We investigated the CD28 requirement of Ag-specific reactivation of memory immune responses using a previously described DTH model in baboon (22). As observed in vitro on human cells, we found in vivo that memory cellular immune response in nonhuman primates is also CD28-dependent and probably controlled by CTLA-4 and/or PD-L1, because administration of FR104 dose-dependently prevented erythema development and inflammatory skin infiltrates by T lymphocytes and macrophages, whereas CTLA4-Ig had no effect in same model. Unexpectedly, whereas all animals treated with excipient had reproducible DTH responses during several months, and animals treated with a dose of 1 mg/kg FR104 recovered normal inflammatory skin responses after a month, in those animals treated with a dose of 10 mg/kg FR104, no full recovery was observed even several weeks after drug elimination. This is suggestive of an induced immune regulation. However, we did not identify the exact mechanism: 1) general or clonal memory T cell deletion was excluded because no significant modification of memory T lymphocyte frequencies or numbers was observed in blood, nor significant change of the IFN-γ response after ex vivo reactivation of tuberculin-specific cells; 2) anergy was excluded because addition of exogenous IL-2 did not significantly increase the frequency of IFN-γ–producing cells in response to tuberculin; 3) peripheral regulatory T cell induction or accumulation was excluded in blood because we did not observe significant modulation of regulatory T cell numbers or frequencies, and elimination of these cells did not modify the ex vivo IFN-γ response to tuberculin after treatment with FR104 as compared with control; and 4) we and others have previously reported that selective CD28 blockade induced regulatory T cell infiltrates at the site of inflammation (15, 17, 18, 23). In this study, we did not observe important T cell infiltrates in skin biopsies after selective CD28 blockade, suggesting that if regulatory T cells directly control effector T cell activation, they would preferably be located in draining lymph nodes. However, we could not exclude that regulatory T cells are at play in the skin and modified the microenvironment, because mRNA transcripts analyses showed that the remaining T cells that are still present in the skin after FR104 treatment display an immunoregulatory phenotype (overexpression of Helios, PD-1, and TGF-β). Finally, CD28 was also described to regulate memory T lymphocytes trafficking to extralymphoid tissue (33), and recently it was reported that selective CD28 blockade with FR104 prevents skin allograft rejection in humanized mice by suppressing cutaneous lymphocyte–associated Ag (CLA) expression (presentation by Issa et al. at the European Congress of Organ Transplantation, September 13–16, 2015, Brussels, Belgium, no. O337). If so, it is possible that FR104 modified immunologically challenged memory T cells in such a way that they became unable to migrate into the skin. These observations are reinforced by previous studies showing that another anti-CD28 mAb (FK734) reduced T cell–mediated skin allograft rejection in humanized mice and reduced epidermis thinning as well as lymphocyte infiltrates of human psoriasis plaques transplanted to SCID mice (34, 35).

We also explored the requirement of CD28 costimulation for memory humoral T-dependent immune responses using the selective monovalent CD28 antagonist. Indeed, recent studies reported that CTLA-4 controls B cell responses by intrinsic and extrinsic mechanisms (36–38), in particular at the T follicular helper cell level, a specialized subset of memory lymphocytes (39–41). Although SRBC challenge is considered as a T-dependent primary humoral response model in rodents, baboons (but also other Old World primates, apes, and humans) are naturally immunized against anti-Gal carbohydrate residues (26) expressed by microorganisms and also other mammals (e.g., sheep). Indeed, as described previously (42–44), baboon sera were positive for preformed xenoreactive anti-SRBC IgG before any sensitization. Similar to the control of memory cellular immune response, we found that selective blockade of CD28 dose-dependently prevented anti-SRBC IgG induction after i.v. administration of SRBC, as it was previously reported with CTLA4-Ig (45). However, FR104 did not induce immune regulation in this memory humoral response model because, after complete elimination of the drug, all animals responded normally to a second challenge. This suggests that Ag type, its presentation, or its route of administration determines whether immune regulation will take place after Ag reactivation of memory cells.

A potential adverse effect of immunosuppressed memory T cells is the reactivation of latent viral infections. To investigate whether selective CD28 blockade impairs immunity, leading to reactivation of quiescent virus, we assessed CMV, LCV, HVP-2, SA12, and HEV status in these baboons. In this study, we did not observe exacerbation of latent viruses over the time course of the experiment (until 20 wk after treatment), even at higher treatment doses. It is possible that preformed Abs against these viruses have contributed to maintaining low viral titers even though some memory cellular immunity was probably impaired. Indeed, our KLH experiment showed that even if selective CD28 blockade could prevent memory humoral recall response during exposure to the drug, it did not modify preformed Ig titers (46, 47). Furthermore, it is possible also that in primates some memory T cells in charge of viral surveillance are CD28-independent as it was originally described in CD28 knockout mice, which develop normal immune response against lymphocytic choriomeningitis virus infection (48). Nonhuman primate immunology is close but could be different from human immunology. Any clinical translation of this novel therapeutic strategy will have to pay attention to viral immunity in patients, in particular to primary viral infections as described with CD80/86 antagonist for EBV⁻ patients (12, 49).

Our study shows that memory cellular and humoral immune responses are CD28-dependent in vivo in nonhuman primate models and in vitro in humans. Selective blockade of CD28 prevents in vivo reactivation of memory T lymphocytes and efficiently controls inflammatory skin responses in a preclinical model.

We thank the Nantes Structure Federative de Recherche François Bonamy MicroPICell imaging platform (Nantes, France) and Sabrina Pengam (INSERM UMR-S 1064, Nantes, France) for cell sorting with the FACSAria flow cytometer.

N.P., C.M., N.D., L.B. and B.V. are employees of Effimune, a company developing CD28 antagonists. N.P., C.M., and B.V. are shareholders of Effimune. The other authors have no financial conflicts of interest.