Monocytes were isolated and differentiated into dendritic cells (DCs) as previously described (24) (ethical approval EK 1880/2012 in accordance with the Declaration of Helsinki). On day 6, DCs were stimulated with 50 ng/mL lipopolysaccharide (LPS, Escherichia coli O111:B4 LPS, Merck, Darmstadt, Germany) and 10³ U/mL human recombinant IFN-γ (Peprotech, Rocky Hill, NJ, USA) for 24 h.

Peripheral blood mononuclear cells (PBMCs) were isolated from buffy coats (Rotes Kreuz, Vienna, Austria) and CD3⁺ T-cells were negatively selected by MACS sorting (Miltenyi, Bergisch Gladbach, Germany). For proliferation studies, T-cells were stained with carboxyfluorescein succinimidyl ester (CFSE; Sigma-Aldrich, St. Louis, MO, USA).

CD3⁺ T-cells were sorted (BD FACSAria™ Fusion; BD Biosciences, San Jose, CA, USA) for naive (CD45RA⁺CD45RO⁻) and memory (CD45RA⁻CD45RO⁺) T-cells, excluding dead cells and duplets (Figure S1A in Supplementary Material). The antibodies CD45RA-PE (clone Hl100), CD45RO-BV605 (clone UCHL1; BD Biosciences) were used.

As depicted in Figure 1, activated allogeneic DCs and CFSE-stained T-cells were co-cultured for 7 days at a ratio of 1:5 (2 × 10⁴ DCs:1 × 10⁵ Tc) with or without 10 µg/mL α-huCD28 (Figure S1B in Supplementary Material) (15–17, 21, 22, 25) (FR104; OSE Immunotherapeutics, Nantes, France) in RPMI 1640 GlutaMAX™ (Thermo Fisher Scientific) supplemented with 2% Octaplas^® (OP, Octapharma, Zurich, Switzerland). T-cells were recovered, rested for 2 days, re-stained with cell proliferation dye 670 (CPD; eBiosciences, San Diego, CA, USA), and counted and re-stimulated again at a ratio of 1:5 (2 × 10⁴ DCs:1 × 10⁵ Tc) with fresh allogeneic DCs (Figure 1A), autologous DCs loaded with UV-inactivated Candida albicans (kindly provided by K. Kuchler, MFPL, Vienna, Austria) (Figure 1B), or third-party allogeneic DCs (Figure 1C). A total of 10 or 100 U/mL human recombinant IL-2 (Peprotech) was added to secondary mixed leukocyte reactions (MLRs) to test for the reversibility of tolerance. Different recipient–donor pairs were used as biological replicates for all experiments performed.

T-cells were harvested on day 7 (primary MLR) or on days 1, 3, and 7 (secondary MLR). They were stained for cell surface or intracellular markers with the Foxp3/Transcription Factor Staining Buffer Set (eBiosciences) according to the manufacturer’s instructions with the following monoclonal antibodies: CD3-PerCPefl710 (clone SK7), CD25-PeCy7 (clone M-A251), CD28-PeCy7 (clone CD28.2), IFNγ-PE (clone 45.B3), and CTLA-4-PerCPefl710 (clone 14D3) (eBiosciences); CD4-PerCP (clone SK3), CD8 APC-Cy7 (clone SK1) PD-1-BV650 (clone EH12), and CD3-PerCP (clone SK7) (BD Biosciences); IL-2-BV421 (clone MQ1-17H12) (Biolegend, San Diego, CA, USA); and anti-human-F(ab′)₂-Alexa Fluor 488 (Jackson Immuno Research, West Grove, PA, USA). Cellular proliferation was determined by CFSE/CPD dilution. The percent of proliferated cells was determined by gating on the CFSE^neg/CPD^neg population. The number of proliferating cells was determined as follows: cells were harvested and 10 µL AccuCheck counting beads (Thermo Fisher Scientific) were added shortly before flow cytometric measurement. After determining the total count of CFSE^neg/CPD^neg T-cells and the total bead count, the number of proliferating T-cells was calculated as follows: (total count CFSE^neg CPD^neg/total bead count) × (beads per μL/analyzed volume) = number of proliferated cells/μL. The same calculation was used throughout the manuscript. For calculating the cell number of CTLA-4⁺⁺ or PD-1⁺ CTLA-4⁺⁺ the same calculation approach was used: (total count CTLA-4⁺⁺ or PD-1⁺ CTLA-4⁺⁺/total bead count) × (beads per μL/analyzed volume) = number of CTLA-4⁺⁺ or PD-1⁺ CTLA-4⁺⁺ cells/μL. T-cell phenotyping was performed on a BD LSRFortessa™ (BD Biosciences) flow cytometer and was analyzed with the FlowJo Software (FlowJo LCC).

For clonotyping, DNA was isolated from T-cells harvested on day 7 (primary and secondary MLR from a donor–recipient pair) with the AllPrep DNA/RNA Mini Kit (Qiagen, Hilden, Germany) and analyzed by Adaptive Biotechnologies by T-cell receptor (TCR) β deep sequencing (Illumina HiSeq, Adaptive Biotechnologies, Seattle, WA, USA). Data were normalized by Adaptive Biotechnologies (26) and analyzed using the immunoSEQ analyzer software.

T-cells from co-culture experiments (1 × 10⁵ T-cells seeded) were harvested with one part of 1× PBS and one part of Novex^® buffer (Thermo Fisher Scientific) containing 3% dl-dithiothreitol (Sigma-Aldrich). Samples were run on 10% SDS PAGE gels and then blotted on an Amersham protan nitrocellulose transfer membrane (Sigma-Aldrich). Phospho-protein phosphatase 2A (pPP2A) (1:7,000, clone E155; Abcam, Cambridge, UK), protein phosphatase 2A (PP2A) (1:1,000), pAkt (Ser 473, 1:2,000, clone D9E), pAkt (Thr 308, 1:1,000, clone C31E5E), Akt (1:1,000, clone C67E7, all Cell signaling technology, Danvers, MA, USA), and vinculin (clone hVIN-1, 1:40,000 Sigma-Aldrich) protein expressions were determined by immunoblotting of CD3⁺ T-cells at intervals up to 96 h after allogeneic stimulation with DCs. Western blots for pPP2A and pAKt (Ser473 and Thr308) were developed by chemiluminescence imaging using the Super signal west femto maximum sensitivity substrate (Thermo Fisher Scientific) and chemiluminescent detection films (Sigma-Aldrich). Western blots for PP2A, Akt, and vinculin were developed by fluorescence imaging using Goat-anti-mouse IgG Dylight 800 and Goat-anti-rabbit IgG Dylight 800 (both Thermo Scientific) on an Odyssey Fluorescence imager (Licor, Licoln, NE, USA). Western blots were analyzed by densitometry using the ImageJ software.

The model consists of bone marrow-derived DC (BMDC) from female BALB/c (H2^d) mice and CD3⁺ T-cells from female C57BL/6 (H2^b) mice as described (27). The mice (Charles River Laboratories, Sulzfeld, Germany), aged 6–10 weeks, were maintained under specific pathogen-free conditions at the Biomedical Research Institute, Medical University of Vienna, Austria. All animal experiments were approved by the Austrian Ministry of Science (BMWFW-66.009/0174-WF/V/3b/2015).

Allogeneic BALB/c (H2^d) bone marrow (BM) was harvested and differentiated to BMDCs as previously described (28) and co-cultured with C57BL/6 (H2^b) CD3⁺ T-cells for 5 days in the presence or absence of 20 µg/mL α-muCD28 (29) (α28scFv, kindly provided by Pfizer, New York, NY, USA). T-cell proliferation was assessed by CPD dilution and analyzed by flow cytometry (BD LSR Fortessa; BD Biosciences). For immune phenotyping of murine T-cells, the following antibodies were used: CD4-PerCP (clone RM4-5) and CD8-APC-Cy7 (clone 53-6.7, both BD Biosciences). Cell number of proliferated cells was determined as indicated above. Cytotoxicity assays, using murine MC38 tumor cells as targets, were performed as previously described (30). Analysis was done on a BD LSRFortessa™ (BD Biosciences) flow cytometer and was analyzed with the FlowJo Software (FlowJo LCC).

On day −1, BALB/c recipient mice were lethally irradiated by two 4 gray fractions (200 kV X-ray radiation) spaced 4 h apart. On day 0, ex vivo treated T-cells were harvested from MLR cultures, washed and sorted for viable cells by FACS. BM from C57BL/6 mice was isolated. A total of 10 × 10⁶ BM cells (i) alone, (ii) together with 0.5 × 10⁶ control (without α-muCD28) T-cells, (iii) T-cells allo-tolerized with α-muCD28, or (iv) non-tolerized (freshly isolated) T-cells were injected intravenously into irradiated BALB/c recipient mice which were monitored daily. A GvHD score was calculated based on four criteria—posture, activity, fur texture, and skin integrity—each scored 0, 1, or 2, as described (31). Upon sacrifice, livers were fixed in 4% paraformaldehyde (Sigma-Aldrich), sectioned, stained with hematoxylin (Dako Agilent Technologies, Glostrup, Denmark) and eosin (Sigma-Aldrich) and histologically analyzed for tissue necrosis, vasodilation, and immune cell infiltration (32). Tissue necrosis was quantified by measuring necrotic areas using Image J software and vasodilation by calculating the mean of the maximum diameters of all vessels in a high power field. Outliers were removed by applying Grubbs’ test (alpha = 0.1). For image acquisition, a Nikon Eclipse 80i microscope with a Nikon DS-Fi1 camera (object lenses 500 μm = 4 × 100 μm = 10×) and NIS-Elements F4.30.00 software were used.

Data were analyzed by paired, one tailed Student’s t-test. p ≤ 0.05 was considered statistically significant: ***p < 0.001, **p < 0.005, and *p < 0.05.

To test the effect of CD28 co-stimulatory blockade on human T-cell responses, we cultured CD3⁺, naive or memory T-cells with or without α-huCD28 (FR104) in a two-stage culture using allogeneic, LPS/IFN-γ-activated DCs (Figure 1, top). This provides a strong CD80/86 co-stimulatory signal, upon binding to CD28 to T-cells (33). In this primary MLR of allogeneic DCs and bulk CD3⁺ T-cells, down modulation of allogeneic T-cell responses was evidenced by greatly reduced proliferation (Figure 2A, upper panel; Figure 2B, left panel), confirming a similar effect previously observed in MLRs using unseparated PBMCs (25).

Upon stimulation with fresh allogeneic DCs from the same donor (secondary MLR, Figure 1A), T-cell proliferation remained downregulated even though cultures were no longer supplemented with α-huCD28. Allo-inhibition was observed three days after secondary stimulation, the mean fluorescence intensity was higher (Figure 2A, lower panel), and the number of proliferated T-cells was still significantly lower (Figure 2B, right panel) in α-huCD28 allo-tolerized, compared to untreated control cultures. Upon incubation with α-huCD28 in primary MLRs, we found that CD28 receptors were occupied on essentially all T-cells (Figure S2A in Supplementary Material). However, staining for the α-huCD28 fragment with a secondary antibody could not detect its presence on cells before secondary MLRs, indicating that CD28 was unoccupied (Figure S2A in Supplementary Material).

Because memory T-cells have been described to be less susceptible to co-stimulation blockade than naive T-cells (34–37), we investigated the effect of CD28 blockade on previously sorted naive (CD45RA⁺, CD45RO⁻) and memory T-cells (CD45RA⁻, CD45RO⁺). Unexpectedly, CD28 blockade reduced alloresponses of both naive and memory T-cells, shown by a significant and sustained decrease in proliferation in primary (Figure 2B, left panel) and secondary (Figure 2B, right panel) MLRs with fresh allogeneic DCs, using multiple different MHC-mismatched donor–recipient pairs. Furthermore, this alloantigen-tolerizing effect in CD3⁺, naive and memory T-cells persisted, up to 7 days (Figure S2B in Supplementary Material). In detail, both CD4⁺ and CD8⁺ T-cell populations showed significantly decreased proliferation upon α-huCD28 treatment in primary as well as in 3- and 7-day secondary MLRs (Figure S3A in Supplementary Material).

IL-2 has been shown to profoundly influence memory responses and to reverse anergy (38, 39). We therefore asked whether IL-2 might also affect α-huCD28 pre-treated T-cells. Although high IL-2 doses overall elevated alloreactivity of naive and memory T-cells, α-huCD28 allo-tolerized T-cells showed up to a fourfold decrease in proliferation after 3 days (Figure S2C in Supplementary Material) and even after extended exposure to IL-2 for up to 7 days (Figure S2C in Supplementary Material). These data suggest that alloreactivity can be persistently reduced by CD28 blockade, even resisting long-term high-dose IL-2 exposure.

Naive as well as memory T-cells can raise strong alloantigen-specific, GvHD-causing immunity associated with Th1 cytokine expression (40, 41). We determined whether CD28 blockade would alter the cytokine expression profile of those T-cell populations. Following 7 days of stimulation of previously unstimulated naive T-cells, we observed only a minor decrease in the number of IL-2 and of IFN-γ/IL-2 expressing T-cells in primary MLRs under CD28 blockade (Figure 3A, upper panel). Although memory T-cells have been described to be less dependent on co-stimulation (34, 37), hence resistant to co-stimulation blockade, the expression of IL-2 and IFN-γ was reduced in primary MLRs in the presence of α-huCD28 (Figure 3C, upper panel). The effect of prior CD28 blockade was highly pronounced upon stimulation with fresh allogeneic DCs in secondary MLRs, seen as a significant decrease in the number of IL-2 and IFN-γ double-positive naive T-cells (Figure 3A, lower panel, Figure 3B) as well as of memory T-cells (Figure 3C, lower panel, Figure 3D). Quadrants were defined according to unstimulated control naive or memory T-cells (Figure S1C in Supplementary Material). These data demonstrate that CD28 blockade leads to stably impaired proliferation and Th1 cytokine production in T-cells.

Immune checkpoint expression on T-cells is associated with a tolerogenic T-cell phenotype (42, 43). Therefore, we assessed CTLA-4 and PD-1 levels of CD28-blocked allogeneic T-cells after secondary alloantigen stimulation. α-huCD28 pre-treatment of MLR cultures resulted in a significant increase in CTLA-4 expression on alloantigen-primed naive T-cells, with a tendency toward increased PD-1/CTLA-4 co-expression (Figure 4A). Furthermore, CTLA-4-expressing cells essentially did not proliferate, whereas PD-1/CTLA-4 co-expressing cells did (Figure 4B). Therefore, we deduced that CD28 blockade caused either cellular arrest or substantial exhaustion of previously proliferating allo-stimulated naive T-cells (44, 45).

Blockade of CD28 by α-huCD28 is proposed to spare the CD80/86-CTLA-4 signaling axis (Figure 1, top) (16). We therefore investigated both the signaling molecule PP2A that releases CTLA-4 upon phosphorylation and the key downstream molecule Akt (protein kinase B) (46), which is then bound by pPP2A, preventing its (Akt) own phosphorylation. In our experiments, PP2A activity, induced by its phosphorylation, was enhanced while phosphorylation of Akt, both on Thr and Ser residues, was decreased upon stimulation with alloantigen under CD28 blockade, compared to untreated cultures (Figure 4C). Decreased phosphorylation on Thr residues was more pronounced than on Ser residues, which is in line with previous observations that PP2A preferentially targets Thr phosphorylation sites (47, 48). These results combined are presumably the cause of inhibited growth and proliferation, as we observed for CTLA-4 positive allo-tolerized T-cells (Figure 4B). In summary, α-huCD28 allo-tolerized T-cells showed increased expression of CTLA-4 and PD-1 consistent with enhanced phosphorylation of the CTLA-4 inhibitor PP2A and its downstream molecule Akt, suggesting development of a tolerogenic T-cell phenotype.

Infusion of donor lymphocytes with retained pathogen reactivity after alloantigen-tolerization can be pivotal for HSCT patients upon pathogen encounter (49, 50). To test the reactivity of α-huCD28 allo-tolerized T-cells against C. albicans, washed CD3⁺ T-cells recovered from primary MLRs with α-huCD28 were re-stimulated with autologous DCs loaded with UV-inactivated C. albicans (Figure 1B). This resulted in significant T-cell expansion after allo-tolerization upon re-stimulation with C. albicans (Figure 5A). In unstimulated cultures (autologous DCs without C. albicans), T-cells did not show significant proliferation (Figure 5B), suggesting that the proliferation was Candida antigen-specific and excluding an antigen-independent mechanism for the observed proliferative response to Candida. Furthermore, T-cell re-stimulation with foreign third-party allogeneic DCs (Figures 1C and 5C) also resulted in significant expansion after allo-tolerization, confirming specificity of allo-tolerization against primary alloantigen. The same effect was observed when analyzing proliferative responses of CD4⁺ and CD8⁺ T-cells to foreign antigen (Figure S3B in Supplementary Material). Thus, despite effective and specific tolerization to the primary alloantigen, CD28 blockade did not impede C. albicans or third-party-specific reactions in re-stimulation cultures.

To investigate, whether clonal expansion of allo-tolerized T-cells to unrelated antigen derived from unique C. albicans and third-party reactive clones, or from cross-reactive clones, we sequenced TCRβ chains of allo- and Candida-antigen-reactive T-cells and searched for cross-reactive and unique clones in the T-cell repertoire, the latter representing highly specific foreign antigen reactive clones.

First, to confirm inhibition of alloreactive clones upon CD28 blockade, we compared the clonal frequency of expanding T-cell clones in primary and secondary MLRs. Vigorous clonal expansion in the primary MLR without co-stimulation blockade was observed, as expected (51). However, 9 of the top 10 most expanded alloreactive T-cell clones were suppressed by up to 90% upon blockade of CD28 (Figure 6A, left graph; Figure S4A in Supplementary Material left table). Eight of these same clones remained up to 80% inhibited in a secondary MLR in the absence of CD28 blockade (Figure 6A, right graph; Figure S4A in Supplementary Material right table). In total, we observed persistent inhibition of 192 of the 200 top-ranked alloreactive T-cell clones.

After allo-tolerization in primary MLRs in the presence of α-huCD28 and then re-stimulation with C. albicans, we detected 146 enriched T-cell clones (Figure 6B, left graph; Figure S4B in Supplementary Material left table). To exclude antigen-independent driven expansion, we subtracted the frequency of T-cell clones that non-specifically expanded in control cultures without Candida from the frequency of clones in Candida stimulated cultures. A similar pattern was observed upon re-stimulation with third-party alloantigen: 192 of the top 200 third-party alloantigen-enriched T-cell clones expanded up to 10-fold, compared to primary allo-tolerized cultures (Figure 6B, right graph; Figure S4B in Supplementary Material right table). Since the dominant clone (CASSPDLNSPLHF) was also the one most enriched under first party, Candida, and third-party alloantigen re-stimulation (Figures S4A,B in Supplementary Material), we selected for unique T-cell clones exclusively enriched after Candida or third-party alloantigen and not present upon first-party alloantigen re-stimulations. Indeed, we detected 27 unique and 110 cross-reactive Candida and 26 unique and 174 cross-reactive third-party-specific expanded T-cell clones (Figure 6C). Although the abundance of these unique clones was low (up to 0.45% for Candida and 0.51% for third-party re-stimulation; Figure S4C in Supplementary Material), they expanded more than 200-fold upon foreign antigen re-stimulation compared to their abundance of <0.002% in the original T-cell pool before allo-tolerization.

In summary, we confirmed C. albicans and third-party-specific clonal expansions after CD28 allo-tolerization that arose from neo-primed or cross-reactive T-cells. This suggests that the ex vivo treatment we explored does not eliminate responsiveness to new antigen exposure, even though response to primary alloantigen that could cause GvHD is markedly reduced.

Taken together, our findings using human alloreactive T-cells suggest that it is possible to selectively induce tolerance by CD28 blockade ex vivo and that this can persist in secondary MLRs in vitro. To assess the in vivo significance of these findings, we tested the effect of CD28 co-stimulation blockade, once again ex vivo, on allogeneic donor T-cells in a murine MHC-mismatched GvHD model (Figure 7A). This model has been previously described to be suitable for in vivo evaluation of co-stimulation blockade (15, 52). We co-cultured donor T-cells with allogeneic DCs from recipient mice in the presence of α-muCD28, a single chain fragment blocking murine CD28 (29). As an in vitro control for inhibited proliferation by CD28 blockade, we showed that α-muCD28 allo-tolerized T-cells proliferated significantly less than did T-cells not exposed to α-muCD28 before infusion (Figure 7B, left graph). CD8⁺ alloreactive T-cells, known to have high cytolytic capacity in GvHD (53), and also CD4⁺ alloreactive T-cells, showed significantly reduced cytotoxicity when CD28 had been blocked (Figure 7B, right graph). Additionally, we showed that treatment of murine DC/T-cell cultures with α-muCD28 increased the frequency of regulatory T-cells on day 5 in primary MLRs (Figure S5 in Supplementary Material).

We injected equal numbers of FACS-sorted, living, ex vivo allo-tolerized, washed T-cells intravenously into lethally irradiated, BM-transplanted recipient mice. GvHD was monitored for 44 days. We observed that infused T-cells that were pre-treated with α-muCD28 did not induce GvHD, as shown by a low GvHD score (Figure 7C). While we observed only a minor difference in the GvHD score in mice receiving T-cells that were pre-treated with α-muCD28 (▲) versus not pre-treated (○) (Figure 7C), the hepatic histological findings were striking. The liver, one of the main target organs in GvHD (54, 55), was not affected by the α-muCD28-modulated T-cells (Figures 7D,E); there were significantly less necrotic lesions (Figure 7D) and a marked decrease of vasodilation (Figure 7E) in livers of mice treated with α-muCD28-modulated T-cells (+) compared to control mice injected with non-tolerized (−) or freshly isolated T-cells. Histologically, livers of mice treated with α-muCD28-modulated T-cells showed no signs of organ damage and resembled the livers of mice that received only BM and no added T-cells (in which no damage was expected) (27). We did not detect differences in immune cell infiltration among the treatment groups (data not shown). These findings show that alloreactive T-cells, blocked for CD28 ex vivo, do not induce GvHD in mice.