|Home | About | Journals | Submit | Contact Us | Français|
T-cell responses to allogeneic targets arise predominantly from the naïve pool. However, in man the risk of graft-versus-host disease (GvHD) is increased if the donor has circulating T cells recognizing multiple persistent DNA viruses, suggesting that memory T cells also contribute to the alloresponse. To examine HLA alloreactivity we used flow cytometry-based proliferation and cytokine production assays. Clonal identity of virus specific T cells cross-reacting with HLA-disparate targets was substantiated by sequencing the T-cell receptor-β (TCR) chains in virus-specific T cell lines restimulated with cognate and with HLA-disparate targets, and sorted according to cytokine response. We confirmed that naïve T cells from cord blood and adult individuals responded to HLA-mismatched target cells. In addition, in adults both in direct assays and after eight days culture with allogeneic stimulator cells, we identified memory T cells responding by cytokine release to human leukocyte antigen (HLA)-mismatched targets. EBV- and CMV-specific T cells, tested against a panel of 30 T-cell antigen-presenting cells with a broad coverage of the most prominent HLA types, displayed specificity for certain mismatched HLA alleles. Sequencing of the TCRβ chain demonstrated clonotypic identity of cells that responded to both viral and allogeneic stimulation. These findings conclusively show that alloresponses in man are not confined to the naïve T cell subset and that memory viral antigen-specific T cells can cross-react with specific mismatched HLA-peptide complexes not presenting CMV or EBV peptides.
Transplantation of donor hematopoietic cells or solid organs into a partially matched recipient activates CD4+ and CD8+ T cells recognizing allogeneic tissues. The high frequency of such alloresponses, in the order of 0.1–10% of all T cells (1) has puzzled investigators. This T cell alloresponse has been proposed to represent either MHC- (2) or peptide-focused (3) recognition by the T cell receptor. The consensus is that such alloreactivity is both MHC-restricted and peptide-specific, with T cells recognizing either a peptide in the non-self MHC (4–9), or alternatively a non-self MHC-derived peptide presented and recognized in the context of self-MHC (10–13).
Allloreactivity can be identified in murine and human T cells directly ex vivo, and in murine models naïve but not memory T cells display alloreactivity in vivo and in vitro(14–17), although recent data in animal models of GvHD suggest that the memory pool can exert non-self MHC reactivity as well (18, 19). Based on the findings in murine T cell allo-stimulations, where naïve T cells produce tumor necrosis factor-α (TNFα) but not interferon-γ (IFNγ), it was assumed that any TNFα produced by human T cells stimulated ex vivo with HLA-mismatched targets originated from naïve T cells (20)However, evidence using cloned T cells suggests that virus-specific T cells can recognize non-self peptide-MHC (21–28). Since the human T cell memory pool is largely dominated by reactivities against common DNA viruses such as EBV, CMV, HSV and VZV (29–32), the possibility of frequent cross-reactivity of antigen-experienced T cells with foreign pMHC is high, despite the relative rarity of individual cross-reactivities.
The distinction between naïve and memory T cell alloreactivity is important in allogeneic stem cell transplantation (SCT). Although umbilical cord blood (UCB) SCT contain over 99% naïve T cells, which should be capable of strong alloreactivity, they confer less graft-versus-host disease (GvHD) than transplants from similarly mismatched adult sources of bone marrow or peripheral blood, conversely suggesting a role for memory T cells in alloresponses causing GvHD. Indeed, clinical observations in HSCT indicate an association between DNA virus reactivity and GvHD (33, 34).
Here we evaluated the ability of both naïve and antigen-experienced CD4 and CD8 T cell subsets to recognize and respond to MHC-mismatched APC. Our findings indicate that both memory and naïve T cells recognize allogeneic targets.
UCB cells for research were provided by the New York Blood Center. Peripheral blood cells were collected from hematopoietic stem cell transplant donors and from healthy paid volunteers under National Heart, Lung, and Blood Institute (NHLBI) institutional review board-approved protocols. Informed consent was obtained in accordance with the Declaration of Helsinki. UCB and adult PBMC were isolated using Ficoll Hypaque density gradient centrifugation and cryopreserved in liquid nitrogen using standard procedures. PBMC were thawed and rested overnight at 37°C/5%CO2 in complete medium (IMDM [Cambrex, Walkersville, MD] supplemented with 10% heat-inactivated human AB serum [Gemini Bio-Product, Woodland, CA], 2 mM L-glutamine, 100 U/ml penicillin and 100 μg/ml streptomycin [Invitrogen, Carlsbad, CA]) with 50 units DNase I (Roche)/ml before functional assays.
The following fluorochrome-conjugated monoclonal antibodies (mAbs) were purchased from commercial vendors: (i) αCD3 Cyanin-7-allophycocyanin (Cy7APC), αCCR7 biotin, αCD45RA allophycocyanin (APC), αCD45RO APC, αCCR5 phycoerytherin (PE), αCD3 Cy7PE, αCD4 peridinin chlorophyll protein (PerCP), αCD4 APC, αCD137 PE, αCD137 biotin, αCD38 Cy5PE, αCD69 fluorescein isothiocyanate (FITC), αinterleukin-2 (IL-2) FITC, and APC- or Alexa 647-conjugated IFNγ (BD Biosciences, San Diego, CA); (ii) αCD4 Cy5.5PE, αCD8 Alexa 750-APC, αCD14 pacific blue, αCD19 pacific blue, αCD57 FITC, αTNFα PE, and αIL-2 APC (Invitrogen, Burlingame, CA); (iii) αCD27 Cy5PE and αCD45RO PE (Beckman Coulter, Miami, FL); and (iv) αCD4 Cy5.5PerCP (Biolegend, San Diego, CA). For some experiments, mAbs were conjugated in-house: αCD8 and αCD45RA (BD Biosciences, San Diego, CA) were conjugated to Quantum Dot (QD) 585 and QD705 (Invitrogen), respectively, and αCD107a to Alexa 594. Streptavidin PE (SA-PE; Becton Dickinson) was used to identify biotinylated mAb-labeled cells. The fixable violet amine reactive dye (ViViD; Invitrogen/Molecular Probes, Eugene, OR) or Via-probe (7AAD, BD Biosciences) were used to eliminate dead cells from the analysis (35). For ICD-mixed lymphocyte reaction (MLR) experiments, cells were labeled with the green fluorescent dye carboxyfluorescein diacetate, succinimidyl ester (CFSE; Invitrogen)(36). Magnetic beads coated with mAb towards CD27, CD45RO, CD45RA, CD57, and CD62L were obtained from Miltenyi (Bergisch Gladbach, Germany).
PBMC were stimulated overnight in complete medium with irradiated (75 Gy) autologous Epstein Barr virus-transformed lymphoblastoid cell lines (EBV-LCL) or custom synthesized CMV pp65 peptide library as described (37). Controls were unstimulated PBMC. The next day the cells were stained with αCD14 and αCD19 pacific blue, αCD69 FITC and either αCD137 PE or αCD137 biotin followed by SA-PE, washed, and resuspended in sterile FACS buffer (PBS supplemented with 2% fetal calf serum), to which 7AAD was added. Live (7AAD-) CD14-CD19- lymphocytes brightly expressing CD69 and CD137 were sorted and further expanded by restimulation with the same antigen in complete medium supplemented with 100 international units of IL-2 (rhIL-2; Tecin; Roche Pharmaceutical, Indianapolis, IN). After four weeks, the T cells were tested for reactivity with a panel of HLA-disparate PBMC or activated T cells (T-APC) as APC (37) using intracellular cytokine detection (Table 1).
PBMC were labeled with magnetic bead-coupled mAb (Miltenyi) specific for CD27, CD45RO, CD45RA, CD57, and CD62L in combinations of CD57 plus CD45RO, CD57 plus CD45RA, CD62L and CD57, and CD27 plus CD45RO to obtain naïve, central memory, effector memory, and effector T cells, respectively. These fractions were applied to an LS column (Miltenyi) and the unbound, i.e. unlabeled, cells were collected, washed, and cultured for eight days with irradiated (50 Gy) HLA mismatched targets. The purity of the fractions exceeded 90%.
Donor PBMC and T cell subsets were first primed by stimulation for eight days with irradiated HLA mismatched PBMC in complete medium at 37°C/5% CO2. Cells were then labeled with CFSE and restimulated with the same target cells or the autologous control for 6 hours in the presence of brefeldin A, monensin, and αCD28 and αCD49d (BD Biosciences). The cells were washed, stained with ViViD and cell surface mAb, and fixed and permeabilized using BD Cytofix/Cytoperm fixation and permeabilization kit following the manufacturer’s instructions. The cells were stained with anti-cytokine mAb, washed twice with Cytoperm buffer, fixed for 10 minutes at room temperature with 4% formaldehyde (Tousimis, Rockville, MD), washed once with FACS buffer (PBS supplemented with 2% FCS and 0.05% sodium azide (Sigma, St. Louis, MO) and resuspended in FACS buffer for acquisition. Virus antigen-specific T cells were similarly examined for alloreactivity by ICD using a panel of 30 T-APC (Table 1). For the direct ex vivo assessment of allo-HLA reactivity of CD4+ and CD8+ T cells, the responder cells were labeled with CFSE and stimulated for 6 hrs at a 1:1 ratio with (unlabeled) stimulator cells as described above.
Stained cells were acquired on either a (modified) LSR II or a Canto II flow cytometer (BD Biosciences). A minimum of 2 × 105 events was collected for each condition. Compensation and data analysis was performed as described (35). Forward scatter-area versus forward scatter-height properties were used to exclude cell aggregates; live T cells were separated from dead cells, monocytes and B cells using a ViViD/CD14/CD19 (dump channel) versus CD3 bivariate plot. Lymphocytes were identified in a forward scatter-area versus side scatter-area plot, and responder cells were identified in the CFSEhi fraction following restimulation of CFSE-labeled primed allo- or virus antigen-specific T cells in an ICD experiment. Gating was standardized within individual samples to generate a fully comparative dataset.
Virus antigen-specific T cells were restimulated with the cognate antigen and with T-APC to which they displayed reactivity and electronically sorted on a modified BD Biosciences Aria sorter based on ICD. Sorted cells were collected in a dry collection tube and genomic DNA was extracted as described (38). T cell receptor Vβ (TCRVβ) sequences were amplified, cloned, and sequenced as described (36, 38), and the composition of the third complementarity determining region analyzed using the international Immunogenetics Information System (http://imgt.cines.fr/).
GraphPad Prism v4 (La Jolla, CA) was used to determine the difference in the magnitude of the response of allo-antigen-primed donor T cells toward PBMC or T-APC in the second stimulation using Wilcoxon two-tailed signed rank test.
We first tested the alloreactive potential of naïve T cells by stimulating UCB T cells, which contains predominantly phenotypically naïve T cells (39), with allogeneic stimulator cells and determined the fraction that had proliferated and acquired expression of the activation marker CD38 by day 8. In five UCB we found that a median of 79% and 86% of CD4+ and CD8+ T cells, respectively, were CFSEdimCD38+ by day 8 (Figure 1), confirming that naïve T cells respond to HLA-disparate target cells.
Next, adult PBMC were separated into various functional subsets by immunomagnetic depletion and stimulated with a pool of irradiated allogeneic peripheral blood mononuclear cells (PBMC). We used an approach where T cell subsets were purified free of bound antibody by negative immunomagnetic selection (40). T cell subsets were identified using well-established marker combinations (40). Thus, naïve T cells were defined as CD45RA+CD27+ and lacking CD57 and CD45RO; central memory T cells were CD45RO+CD27+, lacking CD45RA and CD57; effector memory T cells were CD45RO+CD27+/−, lacking CD62L and CD57; and effector T cells were CD27−CD45RO−CD45RA+CD57+. Naïve T cells were thus obtained by depleting cells expressing CD57 and CD45RO; central memory T cells by depleting cells expressing CD45RA and CD57; effector-memory T cells by depleting cells expressing CD62L and CD57; and effector T cells by depleting cells expressing CD27 and CD45RO.
T cells can proliferate in response to common gamma chain (γc) signaling cytokine stimulation alone (41). Such cytokines, including IL-2, are abundantly produced in a mixed pool of allogeneic PBMC and could contribute to a TCR-independent T cell proliferation in the MLRs. In the next set of experiments we therefore first primed PBMC and T cell subsets from three donors with HLA mismatched targets for eight days, followed by restimulation with the original donor PBMC and measured antigen response based on the flow cytometric detection of cytokine (IL-2 and/or TNFα) production (ICD) in response to secondary stimulation. Unstimulated allo-HLA primed T cells served as a negative control. In all three individuals (Figure 2a) naïve, central and effector memory, and in some cases effector T cells were primed and responsive to restimulation with allogeneic target cells in the secondary 6 hr stimulation. To exclude the possibility that the memory T cells were merely responsive to EBV or CMV viral antigens present in PBMC, the same experiment was repeated using activated T cells as APC (T-APC) which do not present these viral antigens. The pattern was largely the same except that the magnitude of the response to restimulation with T-APC was significantly greater than with PBMC. (Figure 2b; Wilcoxon two-tailed signed rank test, p<0.05), probably because T-APC express high levels of costimulatory molecules plus HLA class II (37, 42). Collectively, these results show that both naïve and memory T cells contribute to the alloresponse, suggesting that while removing undesired self-reactivity from the T cell repertoire, negative thymic selection does not prevent responses to antigens presented by non-self MHC.
Since we found that memory T cells recognized unrelated peptide-MHC complexes, they are likely to exert their effector function with fast kinetics and should be detected in a short-term stimulation-based assay, using a standard overnight ICD procedure. First we stimulated PBMC from six responders individually overnight with five HLA mismatched and one autologous control PBMC in the presence of cytokine secretion inhibitors, and then stained for IL-2, IFNγ, and TNFα. Five of six and two of six CD4+ and CD8+ T cell fractions, respectively, showed reactivity with HLA mismatched targets (Figure 3a). We next assessed the alloreactivity of naïve, memory, and effector CD4+ and CD8+ T cells. T cells readily responded to the HLA-mismatched target cell stimulation by producing cytokines (Figure 3b), amounting to up to 2.35% of CD8+ T cells and 0.23% CD4+ T cells. Direct alloreactivity predominantly by naïve (CD27+CD45RO−CD57−) and end-stage effector (CD27−CD45RO−CD57+) T cells was demonstrated in the CD4+ and CD8+ T cell populations, respectively (Figure 3b). Thus, both naïve and effector T cells can respond with fast kinetics to HLA-disparate target cells.
To extend these observations to T cells with known specificity, we stimulated donor T cells for up to 24 hours with either autologous EBV-LCL or a pp65 peptide library, known to elicit both CD4+ and CD8+ T cell responses (37), and sorted and expanded activated (CD69+CD137+) cells for four weeks. These antigen-specific T cell lines were then tested against a panel of T-APC expressing the most frequent HLA class I and II molecules (Table 1) in an ICD experiment. Either the CD4+ or CD8+ T cell subset displayed reactivity with a panel of 18–29 T-APC (Figure 4A and Figures S1, S2). The virus antigen-specific CD4+ and CD8+ T cells displayed specificity for certain restricted target cells, recognizing between one and four APC. The CMV pp65-specific CD4+ T cells cross-reacted with targets 20 and 33 (HLA-identical siblings). Although stimulators 20 and 33 were genotypically HLA identical it is notable that the alloresponse to APC #33 was 3-fold greater than to the HLA identical sibling, suggesting differences in the self-peptides presented by the two individuals.
To demonstrate that the same T cell clones can respond to both viral and allogeneic peptide-MHC stimulation, we electronically sorted CD4+ and CD8+ T cells from the EBV- and CMV-responsive T cell lines described above, based on the intracellular detection of both TNFα and IFNγ after stimulation with cognate (viral) antigen and with an allogeneic T-APC to which they also responded. Figure 4B shows the response of two such T cell lines to cognate antigen (pp65 or EBV) and to allogeneic targets in the CD4 and/or CD8 T cell subsets. To restrict the study to pure viral-specific T cells we selected only T cells producing both TNFα and IFNγ, since no unstimulated cultured T cells were double positive on restimulation (Figure 4B). Alloresponding CD4+ T cells from a third, anti-EBV-LCL T cell line were selected based on expression of TNFα and/or IFNγ since this gave the greatest difference with unstimulated cells (Figure S3). Sequencing the TCRβ chains in these samples using our well-established methodology (38, 43) identified identical sequences in viral- and allo-antigen-stimulated cells (Figure 4C and S3), providing strong evidence of cross-reactivity of virus-specific CD4+ and CD8+ T cells with unrelated antigens.
It is well accepted that naïve T cells mount a T cell response to any new antigen, whether pathogen-derived or an allogeneic pMHC complex; however, the contribution of memory cells to allo-HLA reactivity is less well defined. Several reports demonstrated alloreactivity of virus antigen-specific CD4+ (23, 26–28, 44) and CD8+ (21–25, 44) T cell clones (reviewed in (45), but none of these studies addressed this issue at the population level as we did here. Furthermore, depletion of naïve T cells has been proposed as a method of avoiding GvHD after allogeneic SCT (15, 46). We measured cytokine production at the single cell level by flow cytometry in allo-MHC primed T cells following short-term (6 hrs) restimulation with the same allogeneic PBMC and confirmed their alloreactive potential. However, the functional read-out may not accurately reflect the true number of alloresponding T cells. In these experiments we chose to analyze TNFα and IL-2 production after allogeneic or syngeneic secondary stimulations. It is possible that responder frequencies might be different when analyzing other effector functions. While these results do not therefore reflect overall allo-HLA responder T cell frequencies they do show that responder frequencies between naïve and memory T cells to allo-HLA stimuli are comparable.
Assays to demonstrate ex vivo the alloreactive potential of T cells have made the assumption that such reactivity resided exclusively in the naïve pool (14, 20). We confirmed using ICD that such allo-pMHC reactivity can be identified in healthy donor T cells ex vivo. Importantly, however, by combining such assays with the phenotypic identification of naïve, memory, and effector T cell subsets in a polychromatic flow cytometry approach, we demonstrate conclusively that predominant subsets reacting with pMHC targets are naïve CD4+ T cells and effector CD8+ T cells.
Since the DNA viruses are latently present in circulating B (EBV (47)) and myeloid (CMV (48)) cells, the analysis of allo-pMHC reactivity of memory T cells could be confounded by responses to viral antigens presented by the HLA disparate stimulators. However, activated T cells, which do not carry EBV or CMV, also elicited alloresponses, excluding the possibility that the responses were directed against these common DNA viruses.
The identification of alloreactivity in the memory pool raises the question whether cross-reactivity of viral antigen-specific CD4+ and CD8+ T cells occurred with mismatched virus antigen-free pMHC complexes. Most studies have only examined a few clones against a limited panel of allogeneic EBV-LCL (27, 28, 49) or PBMC (20) as APC. We earlier established that the chance of finding alloreactivity depended on the complexity of the responder cell population (Melenhorst et al., unpublished observations) and the extent of HLA diversity of the T-APC panel. Activation marker-selected and expanded virus-specific T cell lines were analyzed for clonal composition upon restimulation with the cognate antigen (the immunodominant antigen from CMV, pp65, or a more complex source of antigen: EBV-LCL, which are known to express approximately eight EBV-encoded proteins (50). To refine our selection of pure virus-specific T cells we sorted only the T cells that produced both TNFα and IFNγ in response to cognate or allo-antigen stimulation since the proportion of T cells producing both of these inflammatory cytokines in the absence of restimulation was negligible, whereas single cytokine producers were present in the unstimulated populations. TCRβ sequence analysis of CD4+ and CD8+ T cells responding to cognate antigen stimulation showed that the responder populations were highly oligoclonal, confirming previous reports (51, 52). The examination of the TCRβ sequences in the same T cell lines responding to an allogeneic target identified shared clonotypes with the cognate antigen-responsive population. To expand our findings with the clonotype analysis of virus-specific T cell lines reacting with allogeneic APC we subjected a third, CD4+ EBV-LCL-reactive T cell line to the same procedure. Here we also identified a shared clonotype between virus- and alloresponding cells (Fig. S3). However, since the background production of cytokines by this cell line, i.e. in the absence of antigen stimulation, was substantial, we cannot formally prove the dual reactivity of the clone identified in both reactivities. Collectlively, our data do allow the conclusion that DNA virus antigen-specific T cells, which can make up 10% or more of the circulating post-HSCT T-cell population (51) and in elderly donors (32) are commonly reactive with unrelated pMHC complexes.
Our findings have clinical relevance. Firstly, the adoptive transfer of virus-specific T cell lines and clones given to treat reactivating viruses in (partially) matched recipients (53) may risk sporadic but powerful alloreactions against the recipient. In our study we used T-APC and PBMC as a representative of GvHD targets because they were readily available. However it is possible that these targets are not as representative of GVHD targets as fibroblasts or cells from GVHD target tissues such as the skin, gastrointestinal tract, and the liver. Amir et al. (44) have recently used an EBV-LCL, HLA-transduced erythroleukemic cell line, and PHA blasts. They demonstrated that a large proportion of virus antigen-specific T cell clones reacted with non-self peptide-MHC complexes. However, these findings should be interpreted with caution - in vitro reactivity of virus-specific T cells may not directly translate to the alloreactivity in the form of GvHD since other factors such as target antigen expression, homing, and in vivo expansion can affect the clinical outcome. Indeed the adoptive transfer of virus-specific T cell lines in the partial HLA-disparate setting (53, 54) has not resulted in GvHD (55). Furthermore, a recent study by the Riddell group (56) suggests that even minor histocompatibility antigen-specific T cells selected for reactivity with hematopoietic targets (EBV-transformed B cells) but non-reactivity with patient fibroblasts still exerted anti-lung reactivity in the patient, indicating that even fibroblasts may not express the relevant repertoire of target antigens of GvHD. Secondly, as others have demonstrated, cross-reactivity by viral antigen-specific T cell receptors with various non-self pMHC complexes (24), implies that, when targeting an epitope presented in one particular HLA, (for example WT1RMF presented in the context of HLA-A*0201), TCRs from multiple clones should be available for such therapeutic applications since, depending on the HLA make-up of the TCR donor, these complexes may recognize unrelated pMHC complexes. The same principles may apply to any TCR challenged with pMHC complexes not encountered during thymic selection in the original host (57). Lastly, since alloreacting memory T cells with their rapid proliferation kinetic may kill their target within hours of engagement, they may contribute to the hyperacute GvHD mostly observed in partially matched SCT (58). Though end-stage effector T cells are short-lived in vitro (59), they arise from the memory pool and may continue to replicate (60).
To conclude: T cell reactivity with unrelated pMHC complexes can originate from any post-thymic T cell population; alloreactivity is common, and mediated by the same TCR that recognizes viral and possibly self antigens.
Either CD4+ or CD8+ T cells dominate in the allo-HLA reactivity of virus-specific T cell lines. The CMV pp65-specifc T cell lines from donors 19 and 33 were examined for cross-reactivity with the panel of T-APC described in materials and methods using intracellular cytokine detection by flow cytometry. The responses for the CD4+ T cell subset of donor 19 and the CD8+ T cell subset of donor 33 are shown in figure 4A, but here shown again alongside the other T cell subsets of the respective donors to allow direct comparison.
Additional evidence for allo-HLA reactivity by EBV-LCL-specific CD8+ T cells from donors 2 and 23. The T cell lines were generated and tested against TAPC as described in materials and methods. Both cell lines contain small but detectable reactivities against one or more allogeneic TAPC.
Identification of the same clonotype in the anti-EBV-LCL T cell line from donor 12 in the response to cognate and allo-HLA target antigens. The T cell line from this donor contained EBV-reactivity in both the CD4+ and CD8+ T cell subsets (S3A), but only the CD4+ population contained a small but sizable proportion of allo-HLA reactive cells. These cells were sorted after stimulation with cognate and allo-antigens and subjected to clonotype analysis as outlined in the materials and methods section and in the legend to figure 4C. The data presented here (S3B) suggest that the same clone recognized both antigens.
This work was supported by the Intramural Research Program of the National Institutes of Health, National Heart, Lung, and Blood Institute, and the Vaccine Research Center, National Institute of Allergy and Infectious Diseases.
The authors declare no competing financial interests.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.