Infusion of Alloanergized Donor Lymphocytes after CD34-selected Haploidentical Myeloablative Hematopoietic Stem Cell Transplantation

Allogeneic hematopoietic stem-cell transplantation (HSCT) remains the only curative therapy for many patients with high-risk hematologic malignancies, but not all will have fully human leucocyte antigen (HLA)-matched donors. In contrast, most patients will have haploidentical family donors, the use of which would significantly increase availability of HSCT as a treatment modality (1). Historically, haploidentical HSCT has resulted in increased frequency and severity of acute graft-versus-host disease (aGvHD; refs. 2, 3), mediated by alloreactive donor T cells (4). Various methodologies to deplete donor T cells have significantly reduced incidences of severe aGvHD after haploidentical HSCT (5, 6). However, these nonselective T-cell depletion (TCD) strategies also remove pathogen- and tumor-antigen-reactive T cells, which contributes to delayed immune reconstitution, and increased infection and disease relapse rates (7–9).

Several approaches have therefore been developed to selectively deplete T cells most likely to mediate GvHD from haploidentical donors, including depletion of α/β T cells (10), or selective removal of alloreactive T cells (11, 12). Although some of these strategies have been shown to be feasible in early clinical studies (13, 14), they are technically complex and may also deplete desirable effector and CD4⁺ regulatory T cells (Treg) from the donor pool. More recently, posttransplant cyclophosphamide (PTC) after haploidentical HSCT has been successfully adopted to inhibit proliferation of alloreactive donor T cells in vivo (15). Although early reports suggest this approach results in less viral reactivation and infection than nonselective TCD strategies (16), infections common to immunocompromised populations continue to be reported and the degree and temporal sequence of broad immune reconstitution after PTC-haploidentical HSCT have yet to be determined.

An unmet need therefore remains for strategies that facilitate haploidentical transplantation without negatively affecting beneficial immune reconstitution. One such approach is alloanergization, the induction of alloantigen-specific hyporesponsiveness in donor T cells by recipient alloantigen presentation with concurrent costimulatory blockade (17). This strategy was successfully used in two prior clinical studies in which very large doses of haploidentical alloanergized donor T cells were infused en mass with donor bone marrow (BM), resulting in less severe aGvHD than historical control recipients of non-TCD haploidentical BM transplantation (BMT; refs. 18, 19). Furthermore, the approach does not require the generation of specialized recipient antigen-presenting cells or sorting technologies, retains the majority of pathogen- and tumor-antigen–specific T cells in vitro (19), and results in expansion of allospecific donor Tregs in vivo, which may in turn contribute to the control of harmful alloresponses (20).

However, the previous studies of alloanergized BMT were not designed to determine the impact of alloanergized donor T cells on immune reconstitution or to define the minimal or optimal dose of alloanergized donor T cells able to confer clinically beneficial virus-specific immunity whist maintaining an acceptable toxicity profile. To address these issues, we therefore conducted a novel multicenter phase I clinical study examining toxicity and immune reconstitution after administration of escalating doses of alloanergized donor lymphocyte infusion (aDLI) in patients with high-risk acute leukemia or myelodysplasia after CD34⁺-selected (i.e., nonselectively T-cell–depleted) myeloablative HSCT from haploidentical related donors.

This was an open-label phase I trial to establish the feasibility and safety of delayed administration of aDLI after CD34⁺-selected myeloablative haploidentical HSCT in patients with hematopoietic malignancies. The primary aim of the study was to assess the feasibility and safety of this strategy during two consecutive study stages defined for each patient. The first comprised the period prior to aDLI infusion during which the toxicity of the conditioning regimen, the likelihood of engraftment, and the subsequent percentage of individuals who would be eligible to receive aDLI were determined. The second comprised the period from aDLI infusion through D+100 (as the at-risk period for aGVHD). Preliminary efficacy was addressed in the secondary aim with determination of immune reconstitution and viral infection rates in patients receiving aDLI. Adult and pediatric patients with high-risk hematologic malignancies (acute leukemia in first remission with high-risk cytogenetics or second or subsequent remission, high-risk myelodysplasia or acute myelogenous leukemia with induction failure after at most two regimens) without either available genotypically HLA-matched (6/6) unrelated donors (after a search of longer than 2 months or with high-risk of progression during a search) or fully matched/single HLA-antigen-mismatched related donors, but with available haploidentical related donors were eligible for inclusion at five US centers (Supplementary Methods). The study was conducted in accordance with the Belmont Report, and all patients and/or legal guardians gave consent and/or assent as age appropriate. The study was approved by local institutional review boards and by the FDA and registered at ClinicalTrials.gov (identifier NCT00475384).

Nineteen patients received either total body irradiation (TBI)-based (n = 8) or chemotherapy-based (n = 11) myeloablative conditioning at the site investigator's discretion. After conditioning, patients received CliniMACS-selected CD34⁺ G-CSF–mobilized peripheral blood cells from haploidentical related donors (median age, 39 years; range, 19–50) at D0. Patient and donor characteristics and conditioning regimens are shown in Supplementary Table S1. Of the 10 male and nine female patients, five were children (median age, 7 years; range, 2–17), of whom three were white/non-Latino and two white/Latino; 14 were adults (median age, 39 years; range, 22–50), of whom six were white/non-Latino, three black non-Latino, and five white/Latino. No posttransplant immune suppression was used. Neutrophil engraftment was defined as the first of 3 consecutive days on which the patient's absolute neutrophil count was >500/μL. Chimerism was assessed by D+28 on unselected and CD3⁺ fractions of peripheral blood using DNA polymerase chain reaction (PCR) of short tandem repeats. Supportive care included antibacterial, antifungal, and PCP prophylaxis per institutional practice. For recipients at risk of CMV reactivation (donor and/or recipient CMV IgG-seropositive) systemic anti-CMV prophylaxis was recommended from the initiation of conditioning through D+75 unless contraindicated and thereafter at the discretion of the treating physician. Physicians were encouraged to use a strategy of CMV-DNA monitoring and preemptive therapy upon reactivation for all at-risk patients.

ADLI were generated ex vivo under Good Manufacturing Practice conditions by coculture of fresh nonmobilized peripheral blood mononuclear cells (PBMCs) from the stem cell donor with gamma-irradiated allostimulator PBMC. In order to both ensure adequate numbers of allostimulator cells were available and reduce theoretical concerns regarding possible anergization to tumor-associated antigens present in the patient, where the patient had a suitable second haploidentical family member sharing the haplotype possessed by the patient but not by the stem cell/aDLI donor, fresh, nonmobilized PBMC obtained from the second haploidentical family member were used as allostimulators. Where the patient lacked such a suitable second haploidentical family member, nonmobilized patient PBMCs were obtained prior to transplant and cryopreserved for later thaw and use as allostimulators. Nonmobilized PBMCs from stem cell/aDLI donors were obtained on D+32/39 and alloanergized by coculture with equal numbers of irradiated (33 Gy) allostimulators in the presence of humanized anti-B7.1/2 antibodies as described (clones h1F1 and h3D1, Wyeth; ref. 19). Control flasks with donor PBMCs and irradiated allostimulators without anti-B7 antibodies were set up in parallel. After 72 hours, alloanergized donor lymphocytes were harvested, washed twice, and sampled for viability, contamination (Gram stain, Mycoplasma testing, bacterial and fungal cultures), endotoxin levels, and flow cytometric phenotype (see Supplementary Methods). Residual alloreactivity was determined by measuring residual proliferative stimulator-specific alloresponses in secondary mixed lymphocyte reactions, as previously described and calculated as

ADLIs were infused fresh, prior to phenotypic assessment and determination of residual alloreactivity, if they met release criteria (sufficient T-cell numbers with ≥70% viability, no evidence of bacterial contamination, and endotoxin levels ≤5 EU/kg/). A schema for the study is shown in Supplementary Fig. S1.

Patients who were stably engrafted with no active aGVHD or uncontrolled infection received aDLI on D+35. Patients who did not fulfill these criteria could be delayed for 1 week and reassessed. Twelve patients received aDLI on D+35 as scheduled, while four received their aDLI after a 1 week delay (three due to viral reactivation and one due to active aGVHD). A Bayesian, adaptive study design based on efficacy and toxicity seen in prior patients was used to assign the aDLI dose for each patient (21). For this purpose, efficacy was defined as stable engraftment (absolute neutrophil count > 500/μL by D+100) and toxicity was defined as severe refractory aGVHD (see below) or a grade 4 adverse event (graded per the Common Terminology Criteria for Adverse Events (CTCAE v.3) between the time of aDLI and D+100 that was possibly, probably, or definitely related to the aDLI. The aDLI consisted of PBMCs with the total cell number adjusted to contain the indicated dose of CD3⁺ T cells/kg. The initial starting aDLI dose for the first patient was 10³ CD3⁺ T cells/kg (dose level 1) and subsequent available dose levels were 10⁴ CD3⁺ T cells/kg (dose level 2) and 10⁵ CD3⁺ T cells/kg (dose level 3). In order to implement this adaptive trial design, prior patients had to reach D+100 before the subsequent aDLI dose assignment. The study design mandated stopping if no acceptable aDLI dose could be identified for a patient based on prior patient outcomes.

AGVHD was graded according to the Consensus Grading Scale (22). Hyperacute GvHD was defined as occurring within 14 days of aDLI. Refractory aGVHD was defined as aGVHD that did not resolve within 3 weeks of treatment with corticosteroids. AGVHD treatment was at the discretion of site investigators. Patients who were placed onto other randomized studies or given unconventional therapy for refractory aGVHD were removed from study at the time of these interventions and followed only for survival. Chronic GvHD was diagnosed and graded as per contemporaneous NIH criteria (23).

Immune reconstitution assays were performed on patient PBMCs at 1, 2, 3, 4, 6, 9, and 12 months after transplant. Patients were eligible for immune reconstitution analysis if they had engrafted and had samples available for a minimum of two time points. Patients were excluded from subsequent immune reconstitution studies from time of relapse or administration of experimental or unconventional therapy for refractory aGvHD. T-cell subsets, naïve and memory, regulatory and effector cell phenotypes, and antigen-specific T-cell responses were determined by surface and intracellular cytokine flow cytometry as previously described (19, 20). The frequencies of CD4⁺ Tregs and T effectors (Teff) were measured by flow cytometry and defined as the percentage of CD4⁺ T cells with a CD25^hiCD127^lo and CD25^hiCD127⁺ phenotype, respectively. Further details are contained in Supplementary Methods.

For correlations of phenotypic characteristics of aDLI and residual alloresponses, the nonparametric test of significance was used. For assessing the significance of differences between groups, two-tailed t tests were used throughout. For unpaired t tests, equal variance was not assumed and Welch correction was used where appropriate. Area under the curve (AUC) was calculated using the trapezoid rule. Statistical analysis was performed on GraphPad Prism V4. Overall survival was calculated using the method of Kaplan and Meier

Stem cell infusions contained a median CD34⁺ cell dose of 9.9 × 10⁶/kg (range, 5.2–17.5), with a median residual CD3⁺ T-cell dose of 2.0 × 10⁴/kg (range, 0–3). Eighteen of 19 patients (95%) engrafted (median time to neutrophil engraftment D+12; range, 8–19). Only one of 18 engrafted patients received G-CSF. All patients who engrafted achieved full donor chimerism at first testing (median, D+28; range, D+17–D+40). The patient who failed to engraft had autologous reconstitution (Supplementary Table S2).

We analyzed the phenotypic and functional characteristics of aDLI prior to infusion. Half of the viable infused cells were CD4⁺ T cells, with smaller proportions of CD8⁺ T cells, CD56⁺ NK cells, CD19⁺ B cells, and CD14⁺ monocytes. Infused T cells were predominantly central and effector memory cells (Fig. 1A and B). A more detailed phenotypic characterization was performed on 7 aDLI products and on corresponding unmanipulated donor PBMCs. Consistent phenotypic changes seen in alloanergized CD4⁺ T cells included increased expression of activation markers CD25 and CD69, positive costimulatory receptors CD40L and CD30, and negative costimulatory receptors BTLA and CTLA4. Additionally, there was a consistent increase in the frequency of CD25⁺CD127^lo CD4⁺ Tregs in accordance with our previous observations (20). In contrast, less consistent changes occurred in CD8⁺ T cells, although five of seven products had increased expression of PD1 on CD8⁺ T cells (Fig. 1C). We determined residual alloreactivity in aDLI to quantify the efficiency of alloanergization. The median residual alloreactivity was 25% (range, 1%–65%) for all assessable products. There was no significant difference in residual alloreactivity in products generated at different study sites or for different dose levels (Supplementary Fig. S2) or using patient or nonpatient allostimulators (Fig. 1D). Finally, as aDLI were infused fresh before residual alloreactivity results were available, we correlated phenotypic characteristics of aDLI with residual alloreactivity to identify real-time correlates of residual alloreactivity. Levels or changes in expression of individual molecules did not correlate closely with residual alloresponses. However, there was a significant negative correlation with the ratio of CD4⁺ Tregs to CD4⁺ Teffs in aDLI products and residual alloresponses (Fig. 1E).

Incidences of aGvHD pre-aDLI (from the start of conditioning to immediately prior to aDLI infusion) and post-aDLI (from aDLI infusion to D+100) are shown in Table 1. One of 18 evaluable patients developed aGVHD pre-aDLI (grade 2, D+22). Three patients did not receive aDLI; one died from bacterial sepsis at D+32, one donor was unable to donate, and one patient failed to engraft. Sixteen patients received aDLI at dose levels 1 (n = 4), 2 (n = 8), and 3 (n = 4; Supplementary Table S3). No infusional reactions or hyperacute GvHD occurred after aDLI. Six patients developed aGVHD of any grade post-aDLI at a median D+63 (range, D+52–D+98). One patient (P005) had developed and resolved grade 2 aGvHD pre-aDLI, and then presented again after aDLI with grade 2 on D+98. Five additional patients developed aGVHD post-aDLI: two patients at dose level 2 (one grade 1 and one grade 4); and three patients at dose level 3 (all grade 3). There was a trend to greater residual alloreactivity of aDLI in patients who went on to develop aGvHD than those who did not but this did not reach statistical significance (median 35% vs. 20%, P = 0.15; Supplementary Fig. S3). None of the four patients receiving aDLI at dose level 1 and only two of the eight patients who received aDLI at dose level 2 developed grade 2 or higher aGVHD, compared with three of four patients who received aDLI at dose level 3. Two patients had refractory aGvHD, while the others responded to steroids/other systemic immunosuppressive therapies. Four patients went on to develop chronic GVHD.

A high proportion of patients (12 of 19; 63%) had documented bacteremia, and CMV reactivation occurred in nine of 17 (53%) at-risk patients during the pre-aDLI period. None of these nine had received CMV prophylaxis, and all developed peripheral blood CMV reactivation prior to aDLI at median D+24 (range, 16–33 days). Eight at-risk patients did not develop CMV reactivation pre-aDLI. Five of these received CMV prophylaxis as recommended, two received prophylaxis after stem cell infusion but not during conditioning, and one received HSV prophylaxis with famcyclovir and valacyclovir. Five patients had BK viruria, and five patients had symptomatic viral infections in the pre-aDLI period; three had oral/nasal HSV infection, one had nasopharyngeal RSV infection, and one had systemic adenoviral infection.

Only four of 14 at-risk patients (29%) reactivated CMV post-aDLI, with an overall cumulative incidence of CMV reactivation below 60% at D+180. Of these, three were receiving immunosuppression for GvHD treatment, and three successfully cleared CMV with antiviral therapy. CMV disease (retinitis) was confirmed in one patient. Four patients had EBV reactivation in peripheral blood with no confirmed cases of posttransplant lymphoproliferative disorder. Two patients were electively treated with rituximab. One had become EBV DNA-negative prior to treatment, while the other did so on treatment. Two patients became EBV DNA negative without any treatment. Other viral reactivations/infections occurred in four of four assessable patients who received aDLI at dose level 1/no aDLI, compared with four of eight patients receiving dose level 2, and three of four patients who received aDLI at dose level 3. Two patients developed aspergillus pneumonia, and one patient was positive for aspergillus by PCR in bronchoalveolar lavage.

Sixteen patients were eligible for immune reconstitution studies. Median values and ranges for absolute lymphocyte count, CD3⁺, CD4⁺, and CD8⁺ T-cell recovery are shown in Fig. 2A–D. CD4⁺ T cells recovered more rapidly than CD8⁺ T cells. Reconstituting T cells early posttransplant were predominantly central and effector memory cells reflecting the phenotype of infused aDLI, with no emergence of naïve T cells until 9 to 12 months (Fig. 2E–F).

We next analyzed immune reconstitution in relation to the dose of aDLI received. Reconstitution of CD4⁺ and CD8⁺ T-cell numbers was accelerated in patients receiving aDLI at dose levels 2 and 3 compared with patients receiving dose-level 1/no aDLI (Fig. 2G and H), although differences in T-cell counts at individual time points did not reach statistical significance. In order to better evaluate the impact of aDLI on reconstitution of T cells over the first 4 months after transplant, we also performed an AUC analysis in patients surviving until this point. The D+120 AUC for both CD4⁺ T-cell and CD8⁺ T-cell recovery was significantly greater in patients who received aDLI at either dose level 2 or 3 compared with those who received dose level 1/no aDLI [P = 0.04 and 0.02 (CD4⁺) and P = 0.04 and 0.04 (CD8⁺) respectively]. Additionally, D+120 AUC for both CD4⁺ and CD8⁺ T-cell recovery in patients receiving aDLI at dose level 3 was significantly greater than patients receiving aDLI at dose level 2 (P = 0.03 and 0.01, respectively). Importantly, there was no significant difference in D+120 AUC for either CD4⁺ or CD8⁺ T-cell reconstitution in patients grouped by age, donor age, TBI, or chemotherapy-only conditioning, or higher or lower contaminating T-cell doses within the stem cell transplant. Beyond D+120, CD4⁺ T-cell counts in recipients of aDLI at dose-level 3 fell transiently, reaching levels similar to those of recipients of aDLI at dose levels 1 and 2, although a similar dip in CD8⁺ T-cell reconstitution was not observed. This dip in CD4⁺ T-cell levels may have reflected the lympholytic effect of systemic steroid therapy for aGvHD initiated in three of four recipients of aDLI at dose level 3.

As virus-specific T cells identified by multimer technology may lack functionality after HSCT (24), we assessed the frequency of functional CMV- and VZV-specific CD4⁺ and CD8⁺ T cells by measuring intracellular IFNγ accumulation after stimulation with virus-infected lysates. Thirteen patients who received transplants from CMV and/or VZV-seropositive donors were studied. In several patients, greatly expanded (>50-fold) frequencies of IFNγ⁺ virus-specific T cells were detectable posttransplant compared with levels determined in donor PBMC. None of three assessable patients receiving dose 1/no aDLI had detectable functional CMV-specific T cells in the first 4 months after transplant despite three of four having detectable peripheral blood CMV DNA. In contrast, three of six patients receiving aDLI at dose level 2, and all assessable patients receiving aDLI at dose level 3 had functional CMV-specific T cells detectable by this time point (Fig. 3A and B).

To assess a broader spectrum of virus-specific T-cell reconstitution, we grouped the earliest time points posttransplant, according to dose of aDLI, that functional CMV or VZV-specific T cells were detectable in patient peripheral blood (Fig. 3C and D). The median time when virus-specific CD4⁺ IFNγ⁺ T cells were first detectable was 3 months and 2.5 months for patients who received aDLI at dose levels 2 and 3, respectively, significantly earlier than patients receiving dose 1/no aDLI (median, 9 months; P = 0.03 for comparison with either dose level 2 or 3). Similarly, functional virus-specific CD8⁺ T cells were first detectable at a median of 4 and 3 months in recipients of aDLI at dose levels 2 and 3, considerably earlier than in patients receiving dose 1/no DaLI (median 9 months), although only dose level 3 versus dose level 1/no aDLI reached statistical significance.

We also determined the frequency of adenovirus-specific T cells in P018, who prior to aDLI had developed symptomatic adenoviral infection with a high and rising copy number of adenovirus by PCR despite systemic antiviral treatment with cidofovir. aDLI was followed by marked expansion of donor-derived adenovirus-specific T cells coincident with a fall and subsequent eradication of adenovirus DNA (Fig. 3E).

Donor-derived T cells specific for the tumor-associated antigen WT1 can provide graft-versus-leukemia effects after HSCT (25, 26). We therefore measured WT1-specific CD4⁺ and CD8⁺ IFNγ⁺ T cells in patient peripheral blood. WT1-specific T cells were largely undetectable in peripheral blood from donors. However, large expansions of WT1-specific CD4⁺ and CD8⁺ T cells were detectable in patients' peripheral blood after HSCT. Furthermore, WT1-specific T cells were detectable earlier and at higher frequencies in patients receiving aDLI at dose levels 2 and 3 (Fig. 3F and G).

We have previously shown that infusion of alloanergized T cells en bloc with haploidentical BM results in an expansion of CD4⁺ Tregs in patients' peripheral blood after transplant, and that expanded populations of Tregs may contribute to the control of alloresponses in this setting (20). We therefore measured the frequencies of CD4⁺ Tregs in aDLI pre- and post-alloanergization and in the peripheral blood of patients after aDLI. CD4⁺ Treg frequencies were significantly increased in aDLI compared with unmanipulated donor PBMC (median, 6% vs. 3.5%, P = 0.004). A further increase in the frequencies of CD4⁺ Tregs occurred in patients' peripheral blood after infusion of aDLI, with a peak frequency occurring at D+120 after HSCT. Infusion of higher dose levels of aDLI was associated with significantly higher frequencies of CD4⁺ Tregs in patient peripheral blood at D+60 and D+120 when compared with frequencies in patients receiving dose level 1/no aDLI (Fig. 4A and B). Expanded populations of CD4⁺CD25^hiCD127^lo Tregs in patients' peripheral blood after transplant expressed high levels of FOXP3 consistent with suppressive function (Supplementary Fig. S5). Several patients who had demonstrated large expansions of virus-specific T cells in peripheral blood after transplant (including P016, CMV-specific T-cell expansion and P018, adenovirus-specific T-cell expansion) also had large expansions of peripheral blood CD4⁺ Tregs.

However, there was no clear association with relative or absolute numbers of CD4⁺ Tregs in patients' peripheral blood after transplant and occurrence of aGvHD. As recent studies have shown that aGvHD is more clearly associated with the ratio of CD4⁺ Treg:Teff than with CD4⁺ Treg numbers alone after T-cell–depleted HSCT (27), we calculated the CD4⁺ Treg:Teff ratio in aDLI preinfusion, and in patients' peripheral blood after haploidentical HSCT and aDLI. Patients who developed grade 2 or higher aGvHD had significantly lower CD4⁺ Treg:Teff ratios in the peripheral blood posttransplant. In addition, the CD4⁺ Treg:Teff ratio was also significantly lower in aDLI products administered to patients who went on to develop aGvHD (Fig. 4C).

All 16 patients who received HSC and aDLI were evaluable for survival (Supplementary Table S4). Two patients relapsed posttransplant (P010 at D+180, P012 at D+150); both subsequently died. Only one patient (P023) died as a result of viral infection. One patient died from aGvHD (P026) after receiving aDLI at dose level 2. There were seven patient deaths from regimen-related toxicity [pulmonary veno-occlusive disease (one patient, D+78), respiratory distress syndrome (five patients, D+59–D313), and sepsis/multiorgan failure (one patient at D+312)]. An additional patient (P002) died from bleeding after a liver biopsy 2 years after transplant. Four of 13 assessable patients developed chronic GVHD—limited in one (P027) and extensive in three patients (P015, P016, and P033). The actuarial overall survival of the 16 patients receiving aDLI was 38% at 1 year. Three patients (P005, P015, and P018) are known to be currently alive and disease-free, with a median follow-up of 9 years (range, 8.9–9.3). Patient 004 was alive and disease-free through 5.8 years after transplant, after which she was lost to follow-up (Table 2).

We have shown that generation and administration of aDLI after CD34-selected myeloablative haploidentical HSCT in patients with acute leukemia and myelodysplasia are feasible across multiple clinical sites. While this study highlights the potential influence of regimen-related toxicities in high-risk patients on the interpretation of early-phase studies, we demonstrated that a modest dose of aDLI is well tolerated and positively affects a range of immune reconstitution parameters.

We observed a relatively low incidence of clinically significant aGvHD at dose levels 1 and 2 (2/12 patients), whereas aGvHD occurred in three of four recipients of aDLI at dose level 3 (10⁵ CD3⁺ T cells/kg). This should be set within the context of reported frequencies of aGVHD after CD34-selected haploidentical HSCT without DLI of between 0% and 10% in adults and children (9, 28). The maximum tolerated dose of aDLI given in the context of this study was therefore 10⁴ T-cell/kg (dose-level 2), lower than T-cell doses administered in published reports of CD25-allodepleted DLI and our previous studies using alloanergization to reduce alloreactivity of T cells within BM grafts (18, 19, 29). Several potential explanations are possible. First, in our prior studies, 2–3-log greater doses of alloanergized cells were infused at the time of BMT, which may have optimized homeostatic expansion and the presence and activity of relevant immunomodulatory cell populations such as Tregs. Second, proinflammatory environments have been shown to both reverse anergy in alloreactive T cells in vitro and adversely influence the functional stability of potentially tolerogenic CD4⁺ Tregs (30, 31). In our current study, a notably large proportion of these high-risk patients (63% overall and 56% who went on to get aDLI) had proven bacteremia in the first period of the study (pre-aDLI) and all at-risk patients who did not receive CMV prophylaxis demonstrated reactivation. This high rate of infectious comorbidity could have created a proinflammatory, cytokine-rich microenvironment affecting adversely on both passive and active tolerance mechanisms in the period post-aDLI resulting in an increased frequency of aGvHD. Finally, alloanergization of DLI resulted in less efficient and more variable reduction in residual alloreactivity compared with our preclinical scale-up studies (19) or alloanergization of T cell–replete donor BM in our previous clinical studies (18, 19). This may in turn have contributed to breakthrough aGvHD at lower doses of infused T cells. It is possible that the increased variability in the efficiency of alloanergization resulted from either the use of nonpatient allostimulators or from center effects, although the small numbers in this study are insufficient to firmly support or reject these hypotheses. The most generalizable means to address the variability would be to implement a real-time assay of residual alloreactivity prior to infusion of aDLI. As the ratio of CD4⁺ Treg:Teff in aDLI significantly correlated with residual alloreactivity, use of this phenotypic correlate might improve quality control of aDLI whilst facilitating the use of fresh cells to preserve optimal viability.

Infusion of aDLI at a dose above 10³ CD3⁺ T cells/kg resulted in faster T-cell reconstitution than reported after CD34-selected haploidentical HSCT without T-cell addback (9, 32), and comparable reconstitution to that reported after TCR α/β depletion prior to haploidentical HSCT (33). While CD34-selected grafts infused prior to aDLI contained up to 3 × 10⁴/kg contaminating CD3+ T cells, these unmanipulated donor T cells should not have contributed to immune reconstitution as they would likely have been destroyed by ATG administered in the conditioning regimen.

We observed a high incidence of CMV reactivation (53% of at-risk patients) pre-aDLI, similar to that reported (60% by D+42) in the largest published study of CMV reactivation rates in the early posttransplant period after CD34-selected TCD haploidentical HSCT (34). Only 29% of at-risk patients on our study had further episodes of CMV reactivation after aDLI, with one first reactivation occurring in this period so the cumulative incidence of CMV reactivation remained below 60% at D+180, in contrast to the study reported by Lilleri and colleagues, who found that CMV reactivation continued beyond D+40 to reach a cumulative incidence above 80% by D+100 posttransplant (34).

Encouragingly, functional virus-specific T cells became detectable early after aDLI. In contrast, virus-specific T-cell responses remained undetectable until 12 months after TCD haploidentical HSCT in the absence of additional T-cell infusions in previously published studies (35). The median time to reconstitution of detectable CD4⁺ and CD8⁺ CMV and VZV-specific T cells in recipients of aDLI at the lowest dose level (10³ CD3⁺ T cells/kg) was 9 months and was significantly earlier in recipients of 10⁴ CD3⁺ T cells/kg (median, 3 and 4 months, CD4⁺ and CD8⁺ T cells). This is consistent with the provision of a threshold number of viral-specific T cells contained within aDLI that are able to undergo antigen-specific expansion in vivo. Although variance in the strategies used for CMV prophylaxis at different clinical centers makes it impossible to draw firm conclusions regarding the impact of aDLI on CMV reactivation and CMV-specific T-cell reconstitution, the rates of CMV reactivation after aDLI are comparable with that recently reported with the PTC approach (16), suggesting that the emergence of virus-specific T cells after aDLI might have potential to provide functional virus-specific immunity.

Importantly, we also observed expansion of donor-derived tumor–antigen-specific T cells which could contribute to graft-versus-leukemia (GvL) effects. While our study was not designed to assess the impact of aDLI on disease relapse and there were confounding effects of early mortality, only two of 16 patients who received aDLI subsequently relapsed, despite uniformly high-risk disease characteristics. This contrasts with some reported outcomes after haploidentical HSCT with PTC, where relapse rates of 35% to 50% have been reported, suggesting that this approach may result in impairment of donor-derived GvL effects (15, 36). However, it is important to stress that our study was not designed to determine the clinical efficacy of aDLI in terms of provision of functional immunity posttransplant. Thus, while no conclusions can be drawn on the impact of aDLI on infections or relapse, the observed results support the hypothesis that aDLI may exert beneficial effects on these endpoints and provides a rationale for addressing these endpoints in larger subsequent studies.

Our study also demonstrated a dose-dependent expansion of donor-derived CD4⁺ Tregs in the peripheral blood of patients after aDLI, confirming our earlier findings after alloanergized BMT. We and others have previously shown that donor CD4⁺ Tregs generated by alloanergization play an important role in the control of alloresponses mediated by human and murine T cells (20, 37). Observational studies have shown higher numbers of CD4⁺ Tregs after transplant are associated with freedom from aGvHD (38). Adoptive therapy with donor-derived CD4⁺ Tregs has been shown to reduce the incidence of aGvHD after haploidentical HSCT (39). Higher ratios of CD4⁺ Treg:Teff in patients who did not develop aGvHD after aDLI suggest that CD4⁺ Tregs may indeed have a central role in controlling clinically significant alloresponses in this setting. While limiting numbers of cells prevented us from characterizing the suppressive function of expanded populations of CD4⁺ Tregs, large expansions of CD4⁺ Tregs were seen in several patients who also had vigorous expansion of pathogen-specific T cells demonstrating that expanded CD4⁺ Tregs did not result in global immunosuppression. Strategies to potentiate the expansion of CD4⁺ Tregs after aDLI may further reduce toxicity and improve the therapeutic window of this approach. Additional studies of the suppressive capacity and in vivo stability of these Tregs would be of interest.

We are encouraged that only one patient in the study died from aGvHD and only one patient from viral infection. However, despite using a chemotherapy-only conditioning regimen in the majority of patients, the incidence of non-GvHD treatment-related mortality remained disappointingly high. The high rate of infection observed in patients during conditioning and immediately thereafter suggests that the baseline risk of the study population was high and may have affected study endpoints. With the greater acceptance of haploidentical HSCT since the study was conducted, due in large part to the successes of the approaches reviewed above, it should become increasingly easier to study novel approaches at times when patients are earlier in their disease course.

In summary, we have shown that low doses of aDLI can provide beneficial immune reconstitution after CD34-selected myeloablative haploidentical HSCT with acceptable rates of aGvHD. As beneficial effects can be observed after aDLI at low cell numbers, this may be a valuable strategy to bolster antiviral and potentially antitumor effects after HSCT undertaken using one of a variety of available approaches to the problem of aGvHD. The approach has the advantage of rapid in vivo expansion of donor CD4⁺ Tregs, which may provide an additional mechanism to control alloreactivity in this setting.

L.J. Cooper is an employee of and has ownership interests (including patents) in Ziopharm Oncology. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J.K. Davies, L.J. Cooper, P.F. Thall, R.E. Champlin, L.M. Nadler, E.C. Guinan

Development of methodology: J.K. Davies, L.M. Nadler, E.C. Guinan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.K. Davies, L.L. Brennan, J.R. Wingard, C.R. Cogle, N. Kapoor, A.J. Shah, B.R. Dey, T.R. Spitzer, M. de Lima, R.E. Champlin, E.C. Guinan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.K. Davies, J.R. Wingard, M. de Lima, P.F. Thall, R.E. Champlin, E.C. Guinan

Writing, review, and/or revision of the manuscript: J.K. Davies, L.L. Brennan, J.R. Wingard, C.R. Cogle, B.R. Dey, T.R. Spitzer, M. de Lima, L.J. Cooper, P.F. Thall, R.E. Champlin, E.C. Guinan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.K. Davies, L.L. Brennan, E.C. Guinan

Study supervision: J.K. Davies, L.M. Nadler, E.C. Guinan

The authors thank DFCI Cell Manipulation Core Facility (CMCF), John Daley, John McMannis, and cell processing centers at all sites.

This study was funded by NIH R21CA137645 and U19CA100265. J.K. Davies was funded by Leukemia and Lymphoma Society (Special fellow in Clinical Research), MRC (Clinician Scientist Fellowship G0902269), and CRUK center grant (C16420/A18066). C.R. Cogle was awarded Scholar in Clinical Research by the Leukemia and Lymphoma Society (2400-13).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.