Alloanergization of Human T Cells Results in Expansion of Alloantigen-Specific CD8+CD28 Suppressor Cells

Authors


Abstract

Allostimulation with concurrent costimulatory blockade induces alloantigen-specific hyporesponsiveness in responder T cells (“alloanergization”). Alloanergized responder cells also acquire alloantigen-specific suppressive activity, suggesting this strategy induces active immune tolerance. While this acquired suppressive activity is mediated primarily by CD4+FOXP3+ cells, other cells, most notably CD8+ suppressor cells, have also been shown to ameliorate human alloresponses. To determine whether alloanergization expands CD8+ cells with allosuppressive phenotype and function, we used mixed lymphocyte cultures in which costimulatory blockade was provided by belatacept, an FDA-approved, second-generation CTLA-4-immunoglobulin fusion protein that blocks CD28-mediated costimulation, as an in vitro model of HLA-mismatched transplantation. This strategy resulted in an eightfold expansion of CD8+CD28 T cells which potently and specifically suppressed alloresponses of both CD4+ and CD8+ T cells without reducing the frequency of a range of functional pathogen-specific T cells. This CD8-mediated allosuppression primarily required cell–cell contact. In addition, we observed expansion of CD8+CD28 T cells in vivo in patients undergoing alloanergized HLA-mismatched bone marrow transplantation. Use of costimulatory blockade-mediated alloanergization to expand allospecific CD8+CD28 suppressor cells merits exploration as an approach to inducing or supporting immune tolerance to alloantigens after allogeneic transplantation.

Abbreviations
CFSE

carboxyfluorescein diacetate succinimidyl ester

CMV

cytomegalovirus

CSB

costimulatory molecule blockade

EBV

Epstein–Barr virus

FOXP3

forkhead box P3

GvHD

graft-versus-host disease

HCT

hematopoietic cell transplantation

IFN-γ

interferon gamma

PBMC

peripheral blood mononuclear cell

SEB

Staphylococcal Enterotoxin B

Treg

regulatory T cell

VZV

varicella zoster virus

Introduction

Strategies to induce immune tolerance have widespread therapeutic applications, including prevention of rejection after solid organ transplantation, reduction of graft-versus-host disease after allogeneic hematopoietic cell transplantation (HCT) and control of autoimmunity. As pharmacologic immunosuppression in these clinical settings is limited in both efficacy and specificity and associated with significant toxicity, alternative approaches to achieve immune tolerance are of considerable interest [1-3].

Alloanergization by ex vivo allostimulation of responder T cells with concurrent blockade of CD28-mediated costimulation can induce immune tolerance [4-7]. This approach limits expansion of alloreactive effector cells by induction of alloantigen-specific hyporesponsiveness via cell-intrinsic mechanisms, and has been used to reduce alloreactivity in the setting of HLA-mismatched HCT, both in a murine model [8] and Phase I clinical trials [9, 10].

In a mismatched murine HCT model, costimulatory blockade-induced alloanergization resulted in the acquisition of alloantigen-specific suppressive capacity in responder cells [11]. This phenomenon also occurred in vitro after alloanergization of human T cells [12-14], and alloantigen-specific CD4+ regulatory donor T cell (Treg) expansion has been observed in vivo after infusion of alloanergized donor mononuclear cells during human HCT [15]. Murine studies have shown that high dose, polyclonal CD4+ Tregs can suppress alloresponses [16], and pilot clinical trials of polyclonal donor CD4+ Treg infusion to control alloresponses after HCT have shown promise [17-19]. Most alloantigen-specific suppression after alloanergization of donor T cells also appears to be mediated by CD4+FOXP3+ (forkhead box P3) Tregs and the removal of CD4+ cells prior to alloanergization prevents induction of most, but not all, acquired suppressive activity [15, 20].

The role of non-CD4+ suppressor T cells present after alloanergization has not been characterized. CD8+ T cells with suppressive activity have been identified in other settings. CD8+CD28 cells are the best-characterized CD8+ suppressor cell subset. Such cells have been associated with preserved graft function and decreased pharmacological immunosuppressant usage following solid organ transplantation [21-23]. Furthermore CD8+CD28 cells can limit autoimmune responses in both mice [24, 25] and humans [26] suggesting a clinically significant role in immune tolerance. Studies have also implicated CD8+FOXP3+ cells in control of both autologous and allogeneic immune responses, and there appears to be a close relationship between these cells and CD8+CD28 cells [27-31].

Using an in vitro model of alloanergization of HLA-mismatched human peripheral blood mononuclear cells (PBMCs) with concurrent blockade of CD28/B7 family-mediated costimulation, we found that this strategy supported expansion of human CD8+CD28 cells with potent and specific allosuppressive capacity. A subpopulation of cells expressed FOXP3. Moreover, the frequency and absolute number of CD8+CD28 cells also increased markedly in vivo in patients after alloanergized HCT. This study provides additional insight into the allosuppressive capacity of alloanergized human T cells, potentially providing opportunities to generate immune tolerance in various clinical situations.

Materials and Methods

Alloanergization of PBMCs

Alloanergization cultures (with costimulatory molecule blockade [CSB]) were set up as described previously [14]. PBMCs were isolated by Ficoll-Paque density-gradient centrifugation (GE Healthcare, Piscataway, NJ) from plateletpheresis filter collars obtained under an IRB-approved protocol (#05-321) from healthy volunteer blood donors. The 107 responder PBMCs were alloanergized by coculture at 106/mL in complete media (RPMI-1640, HEPES, L-glutamine, penicillin/streptomycin, gentamicin; Life Technologies, Grand Island, NY), 10% human AB serum (Gemini Bio, Woodland, CA) in upright culture 25 cm2 flasks (Corning, NY) for 72 h with equal numbers of 35 Gy 137Cs gamma-irradiated (GammaCell-1000; Best Theratronics, Ottawa, Canada) unrelated PBMCs (first party stimulators) in the presence of 40 µg/mL Nulojix® (belatacept; Amerisource Bergen, Valley Forge, PA), or 10 µg/mL each humanized anti-B7.1 and anti-B7.2 monoclonal antibodies (clones h1F1 and h3D1; Wyeth, Madison, NJ) [32]. Control cocultures were responders and irradiated stimulators without CSB handled identically. After the 72 h incubation, CSB cocultures were washed twice with complete medium to remove CSB reagents prior to further rounds of allorestimulation. For allorestimulation, washed alloanergized PBMCs were adjusted to 106/mL with equal numbers of new, irradiated stimulator PBMCs for 3–4 days (without the addition of CSB reagent).

Flow cytometry, immunomagnetic cell isolation and CFSE labeling

Details are available in Supplemental Methods section.

Suppression assays

Purified CD8+CD28 (or CD8+CD28+) cells from allorestimulated alloanergized cultures or untreated PBMCs were added at ratios from 1:1 to 1:32 to 1 × 105 untreated autologous PBMCs (responders) in triplicate wells of 96-well plates to which 1 × 105 35-Gy-irradiated first party or third party stimulators were also added. Proliferation was measured after 5 days by 3H-thymidine incorporation, with 1 µCi/well added during the last 18 h. The percentage of allosuppression (% sup) was calculated as

display math

In mitogen-stimulated assays, 1 µg/well of anti-CD3 antibody was added to wells containing 1 × 105 carboxyfluorescein diacetate succinimidyl ester (CFSE)-labeled autologous MACS-isolated CD4+ lymphocyte responders and CD8+CD28 or CD8+CD28+ cells at a ratio of one per four CD4+ cells. Proliferation was analyzed by FACS using CFSE (Invitrogen, Carlsbad, CA) dye dilution. In assays in which cell contact dependency of CD8+CD28 cell effects on alloproliferation was determined, where applicable, responders and stimulators together were incubated separated from CD8+ cells at ratios of 1:4, 1:16 and 1:64 CD8+CD28 cells: PBMCs in 96-well tissue culture Millicell plates with 0.4 µM cell screens (Millipore, Bedford, MA), and then CD4+ or CD8+ responder proliferation was measured by flow cytometry.

Assessment of CD8+ anti-viral immune responses

For intracellular cytokine flow cytometry, 106 PBMCs from cytomegalovirus (CMV), adenovirus, varicella zoster virus (VZV) or Epstein–Barr virus (EBV)-reactive healthy donors were incubated with cell lysates (Microbix, Ontario, Canada), followed by FACS staining for combined intracellular interferon gamma (IFN-γ) and CD107a-PECy5 surface staining (BD Biosciences, San Jose, CA) as previously described [32]. Positive controls were stimulated with Staphylococcal Enterotoxin B (SEB; Sigma-Aldrich, St. Louis, MO), and negative controls contained no added viral lysate.

Analysis of patient samples from human alloanergized HCT

Cryopreserved PBMCs from patients who had undergone haploidentical HCT after ex vivo alloanergization via allostimulation and CSB with humanized anti-B7.1 and B7.2 antibodies on a previously reported IRB-approved Phase 1 clinical study (#05-030) [10] were thawed, washed and stained with surface antibodies to CD3, CD4, CD8, CD28 and where appropriate, fixed, permeabilized and stained with FITC-conjugated rat anti-FOXP3 (clone PCH101; Ebioscience, San Diego, CA). Single-color stained cells were used as compensation controls. Events were acquired on an FC500 (Beckman-Coulter, Indianapolis, IN) flow cytometer and analyzed using FlowJo 7.2.4 software (Tree Star, Inc., Ashland, OR).

Statistical analysis

Statistical analysis was performed with GraphPad Prism Version 5 (GraphPad, La Jolla, CA). Unpaired Mann–Whitney and Wilcoxon matched paired tests were used. A p-value of <0.05 was used to reject the null hypothesis. Two-tailed tests were performed throughout.

Results

Allorestimulation after alloanergization increases the frequency of CD8+CD28 T cells

In vitro alloanergization of HLA-mismatched human PBMCs using blockade of CD28-mediated costimulation with belatacept resulted in a 10-fold reduction in residual alloproliferation (Figure S1). As 20–30% of human CD8+ T cells lack CD28 expression we hypothesized that CD8+CD28 T cells might retain capacity to proliferate after alloanergization [32]. We therefore measured the proportion of CD28 cells within the CD8+ population in untreated, allostimulated and alloanergized PBMCs. Neither allostimulation nor alloanergization altered the proportion of CD28CD8+ cells observed in untreated cells. However, a single round of allorestimulation after alloanergization resulted in a significant increase in the proportion of CD28CD8+ cells compared to that observed in untreated (median 24% vs. 16%, p = 0.03) or allorestimulated cells (median 24% vs. 17%, p = 0.03). In contrast, there was no significant increase in CD28CD8+ cells in allostimulated PBMCs after one round of allorestimulation (Figure 1A).

Figure 1.

Allorestimulation of alloanergized human PBMCs increases the proportion of CD28 CD8+ cells. (A) Frequencies of CD28 cells, expressed as a proportion of CD8+ cells, are shown as box plots depicting median, 25th and 75th centiles and maximum and minimum values in baseline (untreated) PBMCs and in HLA-mismatched allostimulated and alloanergized cells and after one round of allorestimulation using first party allostimulators. Aggregate data from six experiments with unique stimulator–responder pairs are shown. (B) Upper panels—Left-hand panel shows dot plot of allorestimulated PBMCs labeled with CFSE to identify proliferating and nonproliferating cells. Middle- and right-hand panels show CD28 expression on nonproliferating and proliferating CD8+ cells, respectively. Figures on these plots denote the percentage of CD8+ cells with negative or dim expression of CD28. Lower panels—Corresponding dot plots are shown for allorestimulated alloanergized PBMCs. Events were gated on viable CD3+ cells throughout. One representative experiment (of six) is shown. (C) Frequencies of CD28 cells, expressed as a proportion of CD8+ cells, are shown as box plots depicting median, 25th and 75th centiles and maximum and minimum values in dividing (CFSEneg) and nondividing (CFSEpos) CD8+ cells. A significantly higher percentage of dividing CD8+ cells were negative or dim for CD28 in allorestimulated alloanergized PBMCs when compared with either the nondividing allorestimulated alloanergized PBMCs or with dividing allorestimulated cells. Results are shown for six independent experiments with unique stimulator–responder pairs. CFSE, carboxyfluorescein diacetate succinimidyl ester; PBMC, peripheral blood mononuclear cell.

Using CFSE labeling, we determined that the proportion of CD28 cells in residual proliferative alloanergized CD8+ cells after restimulation was significantly greater than in either nonproliferating CD8+ cells (median 30% vs. 18%, p < 0.01) or proliferating allorestimulated CD8+ cells (median 30% vs. 20%, p = 0.03, Figure 1B,C).

Expansion of CD8+CD28 cells following allorestimulation after alloanergization was also observed when individual anti-B7 antibodies were used for costimulatory blockade during the alloanergization process, indicating that this is a class effect not limited to belatacept (Figure S2A). In addition, we observed similar expansions of CD8+CD28 cells when haploidentical stimulator–responder pairs were alloanergized and allorestimulated, extending upon our findings from the fully HLA-mismatched model to a more clinically relevant haploidentical setting (Figure S2B).

Repeat exposure of alloanergized PBMCs to alloantigen results in further expansion of CD8+CD28 cells

We next characterized the effects of repeated rounds of allorestimulation of alloanergized PBMCs on the frequency of CD8+CD28 cells. The median frequency of CD28 cells within in the CD8+ compartment in untreated PBMCs was 22.5% (range: 5.8–37.9%) and expanded modestly to a median of 31.5% (range: 15.3–40%) of CD8+ T cells following anergization. Subsequent rounds of allostimulation resulted in further increases, with a median frequency of 52% (range: 25–82%) following a second restimulation, thereby generating a median ninefold increase in frequency (Figure 2A,B). This resulted in a median eightfold increase in absolute numbers of CD8+CD28 cells in cultures following alloanergization and two rounds of allorestimulation (Figure 2C).

Figure 2.

Repeat rounds of allorestimulation results in relative and absolute expansions of CD8+CD28 cells. (A) Dot plots from a representative experiment showing CD28 expression on CD8+ cells in untreated, alloanergized and alloanergized peripheral blood mononuclear cells after rounds of first party allorestimulation. Values shown in boxes represent the percentages of CD8+ cells that are CD28 positive (upper) and negative (lower). (B) Aggregate data from four experiments. CD28 cells, expressed as a proportion of CD8+ cells, are shown as box plots depicting median, 25th and 75th centiles and maximum and minimum values. (C) Absolute numbers of CD8+CD28 cells in untreated and alloanergized cells following two rounds of first party allorestimulation. Aggregate data are from four experiments. Box plots depict median, 25th and 75th centiles and maximum and minimum values. (D) The effect of depleting CD8+CD28 cells from starting populations on subsequent expansion of CD8+CD28 cells after alloanergization and subsequent allorestimulation. The frequency of CD28 cells, expressed as a proportion of CD8+ cells, are shown as histograms (mean ± SD). Aggregate data from four experiments are shown. Frequencies of CD8+CD28 cells after allostimulation alone (without anergization) are also shown for comparison.

To determine if the expanded populations of CD8+CD28 cells observed after allorestimulation of alloanergized PBMCs originated from CD8+CD28+ or CD28 cells, we depleted starting PBMC populations of CD8+CD28 cells prior to alloanergization. This resulted in an 80% reduction in the frequency of CD8+CD28 cells after alloanergization and subsequent allorestimulation, indicating that the great majority of these cells expanded from the starting pool of CD8+CD28 cells (Figure 2C).

In view of recent reports of phenotypic and functional overlap between CD8+C28 cells and CD8+FOXP3+ cells, we also measured the frequency of CD8+FOXP3+ cells after alloanergization. In agreement with published data, CD8+FOXP3+ cells were detectable at low frequencies (median: 0.11%, range: 0.05–0.24%) in untreated PBMCs [27]. Alloanergization resulted in a threefold increase in the frequency of CD8+FOXP3+ cells to median: 0.29%, and two further rounds of allorestimulation caused a median 24-fold increase in frequency (to 2.7%) and a 17-fold increase in absolute numbers (Figure S3). However, despite this significant expansion, frequencies of CD8+FOXP3+ cells were much lower than that of CD8+CD28 cells after alloanergization and allorestimulation (median 2.6% vs. 52%).

Phenotype of expanded CD8+CD28 cells after alloanergization

Given our prior findings that allorestimulation of alloanergized cultures expanded both CD8+CD28 cells and CD8+FOXP3+ cells, we further examined the relationship between these two phenotypes. Following allorestimulation of alloanergized cultures, the vast majority of CD8+CD28 cells (>90%) remained FOXP3. Only 7.5% expressed FOXP3. However, 90% of these FOXP3+ cells were found within the CD28CD8+ cells, consistent with a small CD8+CD28FOXP3+subpopulation within the larger population of expanded CD8+CD28 cells (Figure 3A).

Figure 3.

Phenotypic characteristics of expanded populations of CD8+CD28 cells. (A) Coexpression of CD28 and FOXP3 on CD8+ cells after alloanergization and subsequent allorestimulation of human PBMCs (right-hand panels). Expression is shown on untreated (baseline) PBMCs (left-hand panels) for comparison. Representative data are shown for one (of four) experiments using unique stimulator–responder pairs. (B) Expression of molecules associated with suppressive cell function on CD8+CD28FOXP3 and CD8+CD28FOXP3+ cells expanded after alloanergization and subsequent allorestimulation. Aggregate data for six experiments are shown as box and whisker plots showing median, 25th and 75th centile and minimum and maximum values. Data are expressed as the percentage of cells positive for each molecule. Isotype controls were used to determine positive levels of expression. (C) Expression of perforin and granzyme on CD8+CD28FOXP3 and CD8+CD28FOXP3+ cells expanded after alloanergization and subsequent allorestimulation (right-hand panels). Expression is shown on untreated (baseline) PBMCs (left-hand panels) for comparison. Representative data are shown for one (of four) experiments using unique stimulator–responder pairs. PBMC, peripheral blood mononuclear cell.

We next examined expanded populations of CD8+CD28 cells after alloanergization and allorestimulation for expression of a number of other molecules associated with suppressive function. The majority of CD8+CD28FOXP3 cells had low expression of CD39, CD45RO, CCR4, TNFR2 and CTLA4 and intermediate expression of CD103 and CD25. In contrast, the minority population of CD8+CD28FOXP3+ cells had a CD103+, CD45RO+, CD39+, TNFR2+, CTLA4+ phenotype (Figure 3B).

We also evaluated the intracellular expression of perforin and granzyme B, two molecules responsible for target cell lysis. CD8+CD28FOXP3+ and FOXP3 cells from untreated and allorestimulated alloanergized PBMCs exhibited similar levels of granzyme B expression. In contrast, perforin expression was reduced in CD8+CD28FOXP3+ cells from allorestimulated alloanergized PBMCs (Figure 3C).

Alloanergization increases the ability of CD8+CD28 cells to suppress mitogen-specific responses

We next sought to determine if CD8+CD28 cells expanded after alloanergization could suppress polyclonal responses of autologous responder cells. CD8+ cells isolated from untreated or allorestimulated alloanergized PBMCs were fractionated into CD28+ and CD28 populations and incubated with CFSE-labeled autologous CD4+ cells stimulated with anti-CD3 antibody. CD8+CD28 cells from untreated PBMCs caused modest suppression of CD4+ proliferation, and CD8+CD28+ cells had no suppressive effect (Figure 4A). In contrast, both CD8+CD28 cells and CD8+CD28+ cells from allorestimulated alloanergized PBMCs suppressed proliferation of autologous CD4+ cells, but with more potent suppression mediated by CD8+CD28 cells (Figure 4B).

Figure 4.

Alloanergization increases the ability of CD8+CD28 cells to suppress mitogen-specific responses. (A) CD8+CD28 and CD8+CD28+ cells were isolated from untreated and 2× allorestimulated alloanergized cultures and tested for the ability to suppress proliferation of autologous CFSE-labeled CD4+ cells stimulated for 5 days with anti-CD3 antibody. Histograms of CFSE are shown and values shown in each panel denote the percentage of proliferating CD4+ cells. The upper left and right panel show negative and positive controls with minimal and maximal cell division with no stimulus and anti-CD3 antibody stimulation in the absence of any CD8+ cells. Middle panels show the percentages of dividing CD4+ cells with addition of CD8+CD28+ (left) and CD8+CD28 (right) cells from untreated PBMCs at a 1:4 ratio of CD8+:CD4+. Bottom panels show the percentages of dividing CD4+ cells with addition of CD8+CD28+ (left) and CD8+CD28 (right) cells from 2× allorestimulated alloanergized PBMCs at a 1:4 ratio of CD8+:CD4+. Data shown are representative of one of four independent experiments. (B) Mean (± standard deviation) of suppression of anti-CD3-stimulated proliferation of CD4+ cells mediated by CD8+CD28 and CD8+CD28+ cells isolated from untreated PBMCs (white bars) and 2× alloanergized PBMCs (gray bars). CFSE, carboxyfluorescein diacetate succinimidyl ester; PBMC, peripheral blood mononuclear cell.

Expanded populations of CD8+CD28 cells after alloanergization potently suppress alloresponses

Having shown that expanded populations of CD8+CD28 cells after alloanergization can suppress mitogen-driven proliferation, we next determined whether these cells could suppress allospecific responses. Proliferation of untreated PBMC responders stimulated with irradiated HLA-mismatched stimulators was measured by 3H-thymidine incorporation in the absence or presence of autologous CD8+CD28 cells from either allorestimulated alloanergized or untreated PBMCs. CD8+CD28 T cells purified from untreated responders modestly inhibited alloresponses (at 1:16 ratio, median: 32.6%, range: 7.6–52%). Potency was increased only at high CD8+CD28 T cell: PBMC ratios, consistent with a polyclonal suppressive response. In contrast, CD8+CD28 T cells from allorestimulated alloanergized PBMCs suppressed alloresponses more effectively (at 1:16 ratio, median: 74.3%, range: 57–95.4%) and at lower ratios (Figure 5A), consistent with activation or enrichment of the allospecificity of these CD8+CD28 T cells. To confirm the specificity of this effect, in a separate set of experiments we also demonstrated that CD8+CD28 T cells from allorestimulated alloanergized PBMC cultures did not significantly reduce proliferative responses to third party allostimulators (Figure 5B).

Figure 5.

Allorestimulation of alloanergized PBMCs results in the expansion of CD8+ CD28 cells that potently suppress alloresponses. (A) The percentage suppression of proliferation (measured by 3H-thymidine incorporation) of first party allostimulated responder PBMCs by CD8+CD28 cells isolated from untreated (open symbols) and 2× allorestimulated alloanergized PBMCs (gray filled circles) is shown. Symbols denote mean percent suppression of proliferation and bars denote standard deviation. Aggregated data from six experiments using different stimulator–responder pairs are shown. (B) Mean percentage (± standard deviation) of suppression of first party and third party alloresponses after 5 days of culture in the presence of a 1:16 ratio of CD8+CD28 cells. Aggregate data from four independent experiments are shown. (C) The percentage suppression of proliferation of first party allostimulated responder PBMCs by addition of CD8+CD28+ cells (black circles) or CD8+CD28 cells (gray circles) isolated from 2× allorestimulated alloanergized PBMCs is shown. Symbols denote the mean percent suppression of proliferation and bars denote standard deviation. Aggregated data from six experiments using different stimulator–responder pairs are shown. (D) Mean percentage (± standard deviation) of suppression of first party and third party alloresponses after 5 days of culture in the presence of a 1:16 ratio of CD8+CD28+ cells or CD8+CD28 cells generated in 2× allorestimulated alloanergized cultures (with first party allostimulators). Aggregate data from four independent experiments are shown. (E) The percentage suppression of proliferation of first party allostimulated responder PBMCs by addition of CD8+CD28 cells isolated from either allorestimulated alloanergized cells (gray circles) or allorestimulated cells (black triangles). Symbols denote the mean percent suppression of proliferation and bars denote standard deviation. Aggregated data from four experiments using different stimulator–responder pairs are shown. (F) Effect of CD8+CD28 cells on alloproliferation of CD4+ and CD8+ T cell subsets. CFSE labeled responder PBMCs were cocultured with irradiated allostimulators for 5 days. Allostimulation resulted in 34% of CD4+ cells and 35% of CD8+ cells proliferating, left-hand panels. The addition of autologous CD8+CD28 cells isolated from allorestimulated alloanergized cultures reduced alloproliferation in both CD4+ cells and CD8+ cells when cell:cell contact could occur (middle panels). When 0.4 µM cell screens were used to separate the CD8+CD28 cells from responder PBMCs, alloproliferation was only partially suppressed (right-hand panels). Data from one representative experiment of two experiments done with similar results are shown with added cell ratio of 1:16. CFSE, carboxyfluorescein diacetate succinimidyl ester; PBMC, peripheral blood mononuclear cell.

In view of our finding that CD8+CD28+ cells from allorestimulated alloanergized cultures could suppress mitogen-stimulated proliferation, we also examined the allosuppressive capacity of this cell subset. Although CD8+CD28+ cells could suppress alloproliferation at high cell ratios, they exhibited significantly less potency than CD8+CD28 cells. Moreover, there was no alloantigen-specific suppression, consistent with a polyclonal, nonspecific effect (Figure 5C,D).

In order to determine if the acquisition of increased allosuppressive capacity of CD8+CD28 cells was specific to alloanergization and not just a product of allostimulation, we assessed allosuppression mediated by purified CD8+CD28 cells from either alloanergized or allostimulated cultures after each had been allorestimulated twice. CD8+CD28 cells from allorestimulated cultures could suppress alloresponses at high ratios yet were significantly less allosuppressive at lower cell ratios than were CD8+CD28 cells from restimulated alloanergized cultures (Figure 5E).

We next used CFSE labeling to determine if CD8+CD28 cells suppressed alloproliferation of both CD4+ and CD8+ populations. Allostimulation resulted in proliferation in more than a third of both CD4+ and CD8+ lymphocytes, and the addition of purified CD8+CD28 T cells from allorestimulated alloanergized PBMCs effectively reduced alloproliferation of both CD4+ and CD8+ lymphocyte subsets. The suppressive effect of CD8+CD28 cells appeared to rely in part on cell:cell contact. When cell screens were used to exclude cell contact, suppression of alloproliferation of both CD4+ and CD8+ responders was reduced but not completely abrogated (Figure 5E).

CD8+CD28 cells from allorestimulated alloanergized cultures do not reduce functional anti-viral responses

In view of the highly potent allosuppressive function of CD8+CD28 cells expanded after alloanergization, we sought further evidence of specificity by determining the capacity of these cells to suppress pathogen-specific responses. Donor PBMCs were screened for detectable CD8+ T cell responses to multiple viral lysates, including CMV, adenovirus, VZV and EBV. PBMCs from donors with detectable virus-specific baseline responses were alloanergized using belatacept. CD8+CD28 T cells were isolated from these cultures and added to untreated, autologous PBMCs stimulated with viral lysates. The addition of CD8+CD28 T cells from alloanergized PBMCs at low ratios where potent allosuppression had been observed did not affect the frequency of virus-specific IFN-γ+ or CD107a+ CD8+ T cells (Figure 6A–D). Additionally, a subset of studies showed that CD8+CD28 and CD8+CD28+ cells from the same anergized cultures did not impinge on the CD8+ T cell functional immunity (Figure S4).

Figure 6.

CD8+CD28 cells from allorestimulated alloanergized cultures do not reduce functional anti-viral responses when added at a ratio that potently suppresses alloresponses. (A) Plots show IFN-γ accumulation in CD3+CD8+ cells after stimulation with viral lysates (or a positive control consisting of SEB) in the absence (top panels) or presence (bottom panels) of CD8+CD28 cells (at 1:16 ratio) isolated from allorestimulated alloanergized cultures. Dot plots from one representative donor are shown. Boxed regions represent the IFN-γ+ frequencies as a percentage of CD8+ cells. (B) Plots show CD107a accumulation in CD3+CD8+ cells after stimulation with viral lysates (or a positive control consisting of SEB) in the absence (top panels) or presence (bottom panels) of CD8+CD28 cells (at 1:16) ratio isolated from alloanergized and allorestimulated cultures. Dot plots from one representative donor are shown. Boxed regions represent the CD107a+ frequencies as a percentage of CD8+ cells. (C,D) Collated data from five healthy donors are shown. The frequency of IFN-γ-secreting and CD107a+ CD3+CD8+ T cells are shown in the absence (white boxes) or presence (gray boxes) of CD8+CD28 cells. Horizontal lines are medians, boxes are 25th and 75th percentiles and whiskers are maximum and minimum values. Stimulus-specific responses shown were calculated by subtracting values for mock-stimulated peripheral blood mononuclear cells. IFN-γ, interferon gamma; SEB, Staphylococcal Enterotoxin B.

Expansion of human CD8+CD28 cells occurs in vivo after alloanergized haploidentical HCT

Finally, we sought to determine if the expansion of CD8+CD28 cells that occurred after alloanergization and subsequent allorestimulation in vitro also occurred in vivo. We have used ex vivo alloanergization with either first generation CTLA4-Ig or humanized anti-B7.1 and B7.2 monoclonal antibodies to reduce donor T cell alloreactivity prior to haploidentical HCT in two clinical studies [9, 10]. We chose to phenotype cryopreserved unmanipulated donor PBMCs and post-HCT PBMCs recovered from engrafted patients who had received anti-B7.1/2-alloanergized haploidentical HCT in one of our clinical studies.

The expression of CD28 on CD8+ cells from peripheral blood was tracked in three evaluable patients in whom there were sufficient cells for analysis (Figure 7A). Early posttransplant (15–30 days after infusion of haploidentical bone marrow containing alloanergized donor T cells) the majority of peripheral blood CD8+ cells expressed CD28. However in all three patients a marked increase in the proportion of CD8+ T cells which were CD28 occurred 20–40 days posttransplant. This was accompanied by rapid expansion of absolute numbers of CD8+CD28 cells in patient peripheral blood (Figure 7B).

Figure 7.

Expansion of human CD8+CD28 cells and CD8+FOXP3+ cells occurs in vivo after alloanergized haploidentical HCT. (A) CD28 cells, expressed as the proportion of CD8+ T cells in peripheral blood of patients after infusion of alloanergized haploidentical HCT. (B) Absolute numbers of CD8+CD28 cells in peripheral blood of the same patients. (C) FOXP3+ cells, expressed as the proportion of CD8+ T cells in peripheral blood of the same patients (D) Absolute numbers of FOXP3+ cells in peripheral blood of the same patients. HCT, hematopoietic cell transplantation.

In view of our findings in vitro that a smaller expansion of CD8+FOXP3+ cells also occurred after alloanergization, we also determined the frequency of these cells in the same patients. Although 10- to 20-fold relative expansions of CD8+FOXP3+ cells were also observed in all three patients, absolute numbers of these cells remained a log lower in peripheral blood than the absolute numbers of CD8+CD28 cells (Figure 7C,D).

Discussion

Alloanergization (allostimulation of human PBMCs with concurrent blockade of CD28-mediated costimulation) causes induction of cell-autonomous, alloantigen-specific T cell hyporesponsiveness as a consequence of activation of specific intracellular signaling pathways [33, 34] and emergence of one or more suppressor cell population(s). CD4+ Treg frequencies increase in human PBMCs after alloanergization via blockade of CD28-mediated costimulation, with more marked increases after subsequent rounds of allorestimulation [15]. Furthermore, the high frequencies of CD4+ Treg in patients undergoing HLA-mismatched HCT after ex vivo alloanergization suggested that this phenomenon is recapitulated in vivo. We now show that alloanergization of human PBMCs also induces an additional tolerogenic mechanism by supporting the expansion of CD8+CD28 T cells with potent allospecific suppressor function.

CD8+ suppressor cells were first characterized by Gershon and Kondo four decades ago [35]. The phenotype and frequency of CD8+ suppressor cells differs by anatomical location [27, 36] and age [37]. Advances in the understanding of their role in alloresponses have lagged behind that of CD4+ Tregs, which have already entered clinical allogeneic HCT trials. Multiple populations of induced/adaptive CD8+ suppressor cells have been described with evidence that CD8+CD28 and CD8+FOXP3+ T cells may play a role in regulating alloreactivity after human allogeneic transplantation. Higher frequencies of CD8+ suppressor cells have been associated with less acute graft rejection and better graft function in liver transplantation [22, 38]. Of particular note, homeostatic repopulation by CD8+CD28 cells prior to CD4+ effector cell recovery following alemtuzumab depletion in kidney transplant recipients has been associated with reduced maintenance levels of immunosuppression and lower rejection rates [39].

Therefore, a strategy that limits alloreactive effector T cell expansion while favoring expansion of allospecific CD8+ suppressor cells might be particularly advantageous in allogeneic transplantation, and could be used to generate CD8+ suppressor cells for subsequent administration. Our in vitro studies show that one round of allorestimulation of alloanergized PBMCs resulted in a modest expansion of CD8+CD28 cells, whereas much larger expansions occurred with additional cycles. These expanded populations of CD8+CD28 cells originated from starting populations of CD8+CD28 cells, and acquired potent and specific allosuppressive function. Repeated rounds of allostimulation model to some extent the clinical scenario in which alloanergized PBMCs infused into either an allogeneic HCT recipient (donor cells tolerized to recipient) or solid organ recipient (recipient cells tolerized to donor) repetitively encounter the original stimulator alloantigens. Our observations suggest that infused CD8+CD28 cells have the potential for amplification in vivo to numbers sufficient to achieve sustained clinical effect. Moreover, expansion of CD8+CD28 cells might be further potentiated either in vitro or in vivo with the mTOR inhibitor, rapamycin, which has been shown to expand CD8+ suppressor cells in mice [23].

The majority of expanded populations of CD8+CD28 cells observed after alloanergization did not express FOXP3 or other markers common to CD4+and CD8+ FOXP3+ cells that have been associated with suppressive function. Both CD8+CD28 and CD8+FOXP3+ cells have been associated with suppressive/regulatory cell function [21, 25, 27, 28, 37]. Our data demonstrate that a small expansion of CD8+FOXP3+ cells also occurs after alloanergization of human PBMCs, and that these cells express markers of suppressive function in common with classical CD4+FOXP3+ Tregs and consistent with a CD8+ memory subset with immunoregulatory properties [36, 40].

Our functional studies demonstrate that after alloanergization and allorestimulation CD8+CD28 cells acquire potent and specific allosuppressive capacity. As current strategies to determine FOXP3 expression do not preserve cell viability, we were unable to examine suppressive capacity and specificity of the FOXP3+ or FOXP3 CD8+CD28 subpopulations. Identification of a surrogate surface phenotype (akin to the CD25hi CD127lo phenotype used to purify FOXP3+ CD4+ Tregs) to identify CD8+FOXP3+ cells would facilitate dissection of the individual contributions of CD28FOXP3+ and CD28FOXP3 cells to the suppression of alloresponses. However, given that the expanded CD8+CD28 population was much larger than the CD8+FOXP3 population, and most CD8+CD28 cells were FOXP3, it seems unlikely that CD8+CD28 suppressive capacity in this setting is solely associated with the expression of FOXP3. Our observation that removal of cell:cell contact reduces, but does not abolish, allosuppressive capacity of expanded populations of CD8+CD28 cells is consistent with allosuppression produced by both direct contact and soluble factor mechanisms. These may well be mediated by different subpopulations within the CD8+CD28 cell pool. The FOXP3+ subset with reduced levels of perforin identified in this report may be one such example. Of note, CD8+CD28 cells with lower levels of perforin have been previously postulated as contributory to maternal gestational alloimmune tolerance [41].

There is significant interest in exploring ex vivo preparation of regulatory immune cells to limit alloreactivity. To date, human clinical transplant studies have been limited to the use of CD4+ Tregs with polyclonal suppressive capacity expanded from natural Tregs. However, preclinical studies suggest that antigen-specific regulatory cell subsets would be more efficacious [42, 43]. The expanded populations of CD8+CD28 T cells generated via our strategy of alloanergization demonstrate significant in vitro capacity to suppress specific alloreactivity specific and do so without impairing anti-viral responses. These characteristics are highly desirable in clinical transplant settings. Expansion by alloantigen restimulation absent any further manipulation represents an additional useful quality for this potential application, and the observation of markedly increased donor CD8+CD28 and CD8+FOXP3+ cells early posttransplant in recipients of alloanergized, haploidentical HCT supports the potential for in vivo expansion.

In summary, we have demonstrated that expansion of CD8+CD28 cells with potent and specific allosuppressive function is a novel additional mechanism by which alloanergization via costimulatory blockade ameliorates alloreactivity. The identification of an approach for expansion of these cells provides a means to pursue a possible role in improving the outcome of allogeneic transplantation.

Acknowledgments

We thank Larry Turka for thoughtful comments on the manuscript. JKD was funded by an ASBMT New Investigator Award and an MRC Clinician Scientist Award. ECG, JKD, LMN and LLB were funded by NCI U19 CA100265 and NCI 1R21 CA137645-01A1.

Authors' contributions

CMB and JKD designed the study, performed experiments, analyzed data and wrote the manuscript. AV and RK performed experiments, LLB collected samples and data on clinical trial patients, LMN provided vital reagents and critically reviewed the manuscript. ECG designed the study, oversaw the clinical trial and wrote the manuscript.

Disclosure

The authors of this manuscript have no conflicts of interest to disclose as described by the American Journal of Transplantation.

Ancillary