Acute GVHD in patients receiving IL-15/4-1BBL activated NK cells following T-cell–depleted stem cell transplantation

Natural killer (NK) cells can rapidly kill virally infected cells and tumor cells, drawing interest for a role in cancer immunotherapy.^1-3  The potential for NK cells to mediate antitumor effects has been of particular interest in allogeneic hematopoietic stem cell transplantation (HSCT) (reviewed in Foley et al,⁴  Leung,⁵  and Locatelli et al⁶ ), fueled by animal studies demonstrating that NK cells can facilitate engraftment and augment graft-versus-tumor effects without mediating graft-versus-host disease (GVHD).^7-9  Current models hold that this results from differential expression of ligands for NK-activating receptors on malignant cells and hematopoietic cells vs healthy nonhematopoietic tissues.^10-12  Numerous clinical studies report improved transplant outcomes for HSCT recipients whose donor and/or recipient genotype predict diminished signaling of inhibitory NK receptors or increased NK-activating receptor activity.^11,13-25  Although NK cells recover early following allogeneic HSCT because of high levels of homeostatic cytokines, especially interleukin-15 (IL-15), NK cell–mediated graft-versus-tumor effects may be limited by impaired functionality related to developmental immaturity and inadequate education or licensing of NK cells undergoing post-HSCT reconstitution.^26-33 

One strategy to overcome limitations associated with natural NK immune reconstitution following allogeneic HSCT is to employ adoptive transfer. Several groups have adoptively transferred haploidentical NK cells following lymphodepleting preparative regimens without HSCT and observed transient expansion without evidence for GVHD.^34-39  NK cells used in these studies have comprised resting,^37-39  IL-2–cultured^34,35,39  or IL-15 plus hydrocortisone–cultured cells,³⁶  and, in most series, systemic IL-2 was administered following NK infusion. Limited experience using adoptive transfer of donor-derived cells following major histocompatibility (MHC)-mismatched HSCT have used either resting^40-42  or IL-15/IL-21 cultured NK cells.⁴³  Although acute GVHD (aGVHD) was observed in 2 trials, the contribution of NK cells to GVHD was unclear because T cell–replete grafts were administered.^41,43  Thus, experience with the use of donor-derived allogeneic NK cells infusions following allogeneic HSCT is limited.

Recently, several groups have used artificial antigen presenting cells (aAPCs), engineered to deliver costimulatory and/or cytokine signals, to augment expansion and functionality of NK cells.^44-49  Using a K562-based aAPC with membrane-bound IL-15 (K562-mb15-41BBL), Fujisaki first reported,⁴⁷  and we confirmed using a similar aAPC,⁴⁶  that coculture of NK cells with recombinant human IL-15 (rhIL-15) plus aAPC expressing 4-1BBL and IL15Rα results in NK expansion, upregulation of activating receptors, and enhanced cytotoxicity against a wide range of malignant cells, including pediatric solid tumors.^46,48  IL-15/4-1BBL–activated NK cells display a distinct gene expression profile and more potent killing capacity in vitro compared with resting and IL-2–activated NK cells.⁴⁷  We therefore sought to investigate the effects of donor-derived IL-15/4-1BBL activated NK cell infusion (aNK-DLI) following allogeneic HSCT in subjects with high-risk pediatric solid tumors. Unlike previous studies, we incorporated stringent T-cell depletion of the allograft to augment the potential for NK expansion in vivo by diminishing competition for IL-15 by engrafting T cells and to allow clear assessment of the potential for aNK-DLI to mediate GVHD. Because IL-15/4-1BBL–activated NK cells showed potent antitumor activity in the presence of ligands for inhibitory killer-cell immunoglobulin-like receptors (KIRs), presumably because of high levels of activating receptors on the NK cells and activating receptor ligands on tumors,⁴⁶  enrollment did not require a KIR-KIR ligand mismatch. We enrolled only MHC-matched donors, further improving our ability to isolate the effect of NK cells on GHVD. We note a substantial incidence of grade 2 through 4 aGVHD in this small series, suggesting that IL-15/4-1BBL aNK-DLI can contribute to aGVHD.

This trial is registered at Clinicaltrials.gov (NCT01287104), and the National Cancer Institute Institutional Review Board approved the protocol. All subjects or guardians provided written informed consent in accordance with the Declaration of Helsinki, and minor assent was obtained when appropriate. Primary objectives were to assess the feasibility and toxicity of infusing escalating doses of IL-15/4-1BBL donor-derived aNK-DLI following allogeneic HLA-matched T cell–depleted peripheral blood stem cell transplantation (PBSCT) and to monitor for acceptable rates of aGVHD (<25% incidence of grade 3/4). Eligibility criteria for enrollment included adequate performance status, organ function, and availability of a 5 to 6/6 HLA-matched sibling donor (MSD) or a 10/10 HLA-matched unrelated donor (MUD). Disease-specific eligibility criteria were similar to those used in a previously reported trial of allogeneic HSCT for ultra-high-risk pediatric sarcomas and included metastatic disease at diagnosis involving bone/bone marrow, early disease recurrence, or persistent/progressive disease after completion of standard frontline therapy.⁵⁰  A second eligibility screen was conducted following immune-depleting chemotherapy, which required stable disease and adequate clinical status to proceed to transplant.

The trial employed a conventional 3+3 phase 1 dose escalation design using aNK-DLI (1 × 10⁵/kg, 1 × 10⁶/kg). Any of the following were considered a dose-limiting toxicity if they were possibly related to NK-DLI: grade 3 through 5 allergic reactions; grade 4 neutropenia lasting >28 days; grade ≥4 organ toxicity; or grade ≥3 steroid-refractory acute GVHD. Stopping rules were based upon unacceptable rates of nonengraftment, steroid-refractory grade 3 or 4 aGVHD, or 100-day treatment-related mortality. aGVHD observed in 2 of the first 3 subjects was steroid-responsive and therefore did not constitute a dose-limiting toxicity. The protocol, however, was amended at that time to hold the NK dose at 1 × 10⁵ cells/kg to gain more insight into the potential for aGVHD with this regimen. Based upon the absence of GVHD in 4 sequential MSD transplants, an amendment upon enrollment of subject 9 allowed dose escalation for MSD recipients. Following development of GVHD in this patient, accrual was halted and the trial was modified to include GVHD prophylaxis. This report summarizes experience before the incorporation of GVHD prophylaxis. Data were primarily analyzed by the first, second, and senior authors on this publication, with all authors having full access to available data.

Subjects received 1 to 3 cycles of pretransplant immune-depleting chemotherapy (etoposide, vincristine and Adriamycin, with prednisone, cyclophosphamide, and fludarabine [EPOCH-F]), as previously described,^50,51  to facilitate donor engraftment and provide tumor control. The decision to administer subsequent cycles of EPOCH-F following the first was dictated by donor availability, disease response, and disease status, but was not dictated by levels of lymphodepletion. All transplant recipients received a nonmyeloablative preparative regimen consisting of cyclophosphamide 1200 mg/m² per day (days −6 to −3); fludarabine 30 mg/m² per day (days −6 to −3); and melphalan 100 mg/m² per day (day −2) (Figure 1A), and filgrastim (5 mcg/kg per day) starting on day 0, which continued until recovery of the absolute neutrophil count to 5000/mcL. No patient received GVHD prophylaxis.

Grafts were filgrastim-mobilized peripheral blood stem cells (PBSCs) from 6/6 HLA-matched siblings or 10/10 MUDs. PBSC products underwent CD34⁺ selection (CliniMACS, Miltenyi Biotec), and an aliquot of the CD34^–/CD3⁺ fraction was added back to the CD34⁺ graft to result in a T-cell dose of 1 to 2 × 10⁴ CD3⁺ cells/kg, then cryopreserved until infused for transplant (Figure 1B). NK cells were immunomagnetically selected from the CD34^– apheresis product by CD3^– selection followed by CD56⁺ selection (CliniMACS CD3 Microbeads and CD56 Microbeads, Miltenyi Biotech), and then cryopreserved. To generate aNK-DLI, NK cells were thawed and placed in culture with irradiated KT32.A2.41BBL.64⁴⁶  (aAPC:NK ratio of 10:1) plus rhIL-15 (10 ng/mL, CellGenix GmbH, Antioch, USA). Coculture was maintained for 9 to 11 days until achievement of the target dose meeting the prescribed release criteria (CD56⁺ ≥90%, CD3⁺ ≤0.2%, CD14⁺ ≤5%, viability ≥70%, 4-1BBL⁺ ≤1%, negative for endotoxin, mycoplasma, or bacterial contamination). Per Food and Drug Administration guidance, the first 2 patients received aNK-DLI after engraftment (day 24 and day 15 post-HSCT, respectively). For subsequent patients, aNK-DLI was planned for day 7 ± 2 and day 35 ± 2 posttransplant, with allowance for delays to allow for resolution of clinical complications related to transplant that might incur increased risk or confound our ability to accurately attribute toxicities related to the aNK-DLI.

Tumor response was measured using Response Evaluation Criteria in Solid Tumors on days +28, +60, +100, +180, and then every 3 months using computed tomography scans and/or magnetic resonance imaging, and positron emission tomography/computed tomography as clinically indicated. aGVHD was graded on a modified Glucksburg Scale.⁵²  Neutrophil and platelet engraftment were defined as the first of 3 consecutive days with absolute neutrophil count >500/mcL and platelets >20 000/mcL, respectively. Chimerism studies using a standard short-tandem-repeat microsatellite DNA-based approach was assessed on peripheral blood at day 14 and repeated every 2 weeks until attainment of complete donor chimerism or a pattern of stable mixed chimerism. Chimerism was also measured in electronically sorted peripheral blood NK cells (>95% CD3^–CD16⁺ or CD56⁺ cells) following infusion of aNK-DLI and ±14 days from the onset of aGVHD. In subjects with a suspicion for GVHD, biopsies were performed as clinically indicated and evaluated using standard hematoxylin and eosin as well as immunohistochemistry for CD3, CD4, CD8, CD56, and γδ T-cell receptor expression.

High-resolution HLA genotyping was performed on all donors and recipients using standard techniques. HLA-C allotypes were classified as C1 for Asn at position 80 or C2 for Lys at position 80. HLA-A and HLA-B allotypes were screened for the Bw4 epitope based on the curated database (hla.alleles.org). Donor/recipient KIR genotypes were analyzed for inhibitory and activating KIRs (KIR2DL1-5, KIR3DL1-3, KIR2DS1-5, and KIR3DS1) by both sequence-specific priming with locus-specific primers and by oligo-typing with the One Λ Luminex kit. KIR2DS4 positivity required the presence of a full-length gene without the common deletion in the second immunoglobulin domain, as determined by sequence-specific priming. The 5′ region of KIR2DL5 was sequenced to distinguish 2DL5A from 2DL5B. Donor haplotypes were defined as A/A in individuals possessing KIR3DL1-2, KIR2DL1, KIR2DL3, ±KIR2DS4 and no other stimulatory KIRs. Donors with stimulatory KIR genes other than KIR2DS4 were classified as carrying B/x haplotypes. KIR-HLA mismatches were defined based on the presence of a long-tailed/inhibitory KIR in the donor and the absence of a known cognate HLA epitope (C1, C2, or Bw4) in the recipient. Known KIR–ligand pairs analyzed included KIR2DL1-C2, KIR2DL2/3-C1, KIR3DL1-Bw4, and KIR3DL2-HLA-A03/A11.

Activating receptor expression of the NK-DLI product was determined using FACS Aria II using FACS Diva software (Becton Dickinson) and analyzed by FlowJo 9.7 (Treestar). The following fluorochrome or biotin-conjugated monoclonal antibodies were used from Biolegend specific for NKG2D (1D11), NKp44 (P44-8), NKp46 (9E2), NKp30 (P30-15), 158a (HP-3E4), 158b (CH-L), and TRAIL (RIK2). Standard chromium release assays were used to measure cytotoxicity of ⁵¹Cr-labeled TC71 cells in the presence and absence of fusion proteins blocking NKG2D, NKp44, and NKp46 (R&D) as previously described.⁴⁶ 

χ-square test was used to analyze results of donor type vs GVHD occurrence and KIR typing in association with development of acute GVHD. Mann-Whitney tests were used to compare phenotype and NK cell numbers in patients based on presence of GVHD. All P values were 2-sided, except where noted, and the level for statistical significance was set at ≤.05.

Between July 2011 and March 2013, 14 subjects enrolled, and 9 proceeded to HSCT. Four did not proceed to HSCT because of progressive disease and a fifth became ineligible because of a donor anaphylactoid reaction following the first dose of filgrastim.⁵³  All subjects with pretransplant disease progression during EPOCH-F subsequently died from progressive neoplastic disease.

Demographics for HSCT subjects are shown in Table 1; median age at study enrollment was 21 years (range, 14.4-34.5 years); median number of EPOCH-F cycles was 2 (range 1-3). Five subjects received MSD grafts; 4 received MUD grafts. Median CD34⁺ dose was 6.8 × 10⁶ CD34⁺ cells/kg (recipient weight); median T-cell dose was 1.4 × 10⁴ CD3 cells/kg (range, 1.1-2.0 × 10⁴ CD3 cells/kg). Production of an aNK-DLI that met preestablished release criteria was feasible for all subjects and all HSCT recipients received at least 1 aNK-DLI cell infusion; 5 received a second aNK-DLI. Subjects 1 through 8 received an NK cell dose of 1 × 10⁵ cells/kg at dose level 1. Based upon the observation that 0/4 MSD recipients developed aGVHD, the dose of NK cells was escalated for subject 9 to 1 × 10⁶/kg, who had an MSD donor. All subjects engrafted briskly with median posttransplant neutrophil and platelet engraftment of 9 days (range, 8-10 days) and 10 days (range, 6-12 days), respectively. All subjects achieved >95% donor myeloid chimerism by day 14.

Five subjects developed aGVHD; grade 4 in 3 cases (Table 1, Figure 2). Because the total dose of CD3⁺ T cells administered was low (≤2 × 10⁴/kg), this rate of high-grade aGVHD in completely HLA-matched recipients was unexpected. Four of the 5 subjects with MUD donors developed aGVHD, whereas only 1 of 4 subjects with MSD developed aGVHD (P = .02), raising the prospect that aNK-DLI exacerbated subclinical T cell–mediated aGVHD. Consistent with this hypothesis, donor T-cell engraftment was significantly higher in subjects who experienced aGVHD. Median donor CD3 chimerism at day 14 in recipients who developed aGVHD was 96% (range, 73% to 100%) vs 7% (range, 2-36%) in those without GVHD (P = .03). Similarly, at day 28, donor CD3 chimerism was 100% (range, 92%-100%), and 58% (range, 8%-92%) in recipients who did and did not experience GVHD, respectively (P = .04) (Table 1). In 3 subjects, aNK-DLI preceded the day 14 analysis and therefore could have contributed directly to the higher CD3 engraftment rate, whereas in 2 subjects, high donor lymphoid engraftment preceded the aNK-DLI.

Interestingly, all 3 patients in this series who developed grade 4 GVHD had symptom onset within 24 to 48 hours of the aNK-DLI. Among the 5 patients with aGVHD, 4 had fever on the day of aNK-DLI infusion or the day before infusion and patient 9, who developed biopsy-proven hepatic GVHD without concomitant rise in bilirubin, had mild transaminitis at the time of aNK-DLI. In contrast, none of the subjects without GVHD had fever in the 48 hours preceding NK cell infusion or other signs of an inflammatory process.

Among the 5 subjects with aGVHD, 5 skin biopsies, 6 biopsies from the upper gastrointestinal tract (2 esophageal, 3 gastric, and 1 duodenal), 6 biopsies of the colon or rectum, and 1 liver biopsy were obtained. All but 1 skin biopsy and 1 stomach biopsy confirmed aGVHD, which varied from mild to very severe, with ulceration of affected sites. Histologically, observed aGVHD could not be distinguished from that observed historically in patients who had not received aNK-DLI. Lymphocytic infiltrates were mainly composed of CD3⁺CD8⁺ T cells, with a minority of CD3⁺CD4⁺ T cells. In 2 cases (1 skin biopsy and 1 gastrointestinal biopsy), numerous CD56⁺ cells were seen (Figure 2), which were negative for the γδ T-cell receptor and thus judged likely to represent NK cells. In GVHD biopsies from the 3 additional subjects, CD56⁺ cells were present and typically represented less than 10% of the lymphocytes, consistent with historic reports of NK cells within aGVHD lesions occurring in patients who did not receive adoptive NK cell therapy.^54,55  NK chimerism studies, performed on electronically sorted NK cells from 5 samples from 4 subjects who developed aGVHD following aNK-DLI, uniformly demonstrated 100% donor origin of the NK cells.

Acute GVHD was steroid-responsive in all patients. In subject 4, grade 4 GVHD occurred within 48 hours of the aNK-DLI infusion, but rapidly abated after initiation of steroids, allowing rapid discontinuation of immunosuppression and administration of a second NK infusion. Following the second NK infusion, symptoms of grade 3 GVHD recurred within 48 hours (Table 1). This subject ultimately weaned off all immunosuppression and remains disease-free more than 2 years following PBSCT. In all 3 subjects with grade 4 GVHD, multiagent immunosuppression was initiated following steroids and maintained for a minimum of 5 months. No patient developed chronic GVHD.

To investigate a potential role for KIR-KIR ligand mismatch in the development of GVHD, donor KIR and recipient HLA genotype were analyzed for the presence of an inhibitory KIR in the donor genome with an absence of the cognate ligand in the recipient. Based on the hypothesis that GVHD would be associated with KIR mismatch, mismatches were identified in 4 of 5 subjects who developed GVHD and in 1 of 4 subjects who did not develop GVHD (Table 1, P = .1 by χ-square 2-tailed; P = .05 by χ-square 1-tailed). Donor KIR haplotypes were also evaluated and were classified as an inhibitory haplotype, A/A (having inhibitory KIRs with either KIR2DS4 or no activating KIRs), or as an activating haplotype, B/x (having any activating KIR other than KIR2DS4). Four of the 5 subjects that developed GVHD had donors with the inhibitory A/A haplotype, whereas 3 of the 4 subjects without GVHD had donors of the activating B/x haplotype (Table 1, P = .1 by χ-square, 2-tailed).

Peripheral blood lymphocytes and NK cell subsets were analyzed serially following HSCT (Figure 3A). We observed no significant difference in circulating CD3⁺ T cell or CD3^–CD16⁺ and/or CD56⁺ NK cell numbers between subjects who did or did not experience aGVHD (P = NS at all time points). Similarly, analysis of the aNK-DLI product and circulating NK subsets for activating receptor expression and 158a and 158b KIR expression showed no differences in subjects who did or did not experience GVHD (P = NS for all products). Interestingly, aNK-DLI phenotypes differed significantly from NK cells circulating following HSCT with respect to NKp44 expression (high on aNK-DLI but lower in vivo) and with respect to NKp30, TRAIL, and KIR 158a/158b expression (low on aNK-DLI but higher in vivo)(Figure 3A). As expected, aNK-DLI products showed significant killing capacity when tested against the Ewing sarcoma line TC-71 (Figure 3B), which was used as a reference line for these experiments, although we could identify no difference in potency between aNK-DLI administered to subjects who did or did not develop GVHD (mean ± standard error of the mean percent lysis at 20:1 of 46 ± 9.9% in GVHD⁺ subjects [n = 5] vs 70.5 ± 9.0 in GVHD^– subjects [n = 4], P = .11). aNK-DLI showed diminished killing upon blockade of NKG2D, NKp44, and NKp46 activating receptors as previously described (Figure 3B).

Antitumor activity and survival of HSCT recipients is shown in Table 2. Two HSCT recipients (3 and 5) with measurable disease at the time of transplant attained complete remission following HSCT and aNK-DLI. In patient 3, this was sustained for 8.7 months, but he ultimately died of sepsis related to malnutrition and complications from previous surgeries without definitive evidence for neoplastic disease or GVHD at the time of death. With a median follow-up of 23.1 months (range 12.5-27.4), 4 HSCT recipients remain alive, 2 of which are in continuous complete remission at 12.5 and 27.4 months following transplant.

Murine studies dissecting the basis for “hybrid resistance” identified an essential role for “missing self” in NK cell recognition,⁵⁶  and exploitation of this property as a means for augmenting NK function has driven attempts to augment graft-versus-leukemia without inducing GVHD. Proof of principle for the merit of this approach comes from numerous clinical studies reporting decreased leukemia relapse when donor:recipient pairs have a KIR:KIR ligand mismatch.^{13,14,17-20,23}  However, NK cell killing also requires engagement of activating receptors, and several studies have demonstrated an important role for activating receptor genotype on antitumor effects mediated by NK cells.^15,16,22,57  Although the full gamut of activating receptor ligands and their biology remains to be elucidated, it is well-known that cellular stress leads to upregulation of activating receptor ligands, both on malignant and nonmalignant tissues. Recent work also demonstrates an important role for licensing of NK cells, such that NK subsets with varied levels of maturity and differentiation will have differential capacities for cell killing.⁵⁸  With an appreciation of this complex array of factors impacting NK cell killing as a background, we propose that IL15/4-1BBL–activated NK cells are licensed to kill and can contribute to aGVHD if inflammation or other factors upregulate activating ligand expression on nonmalignant tissues, and this may be potentiated by a KIR:KIR ligand mismatch.

In this small clinical series employing the first-in-human application of IL15/4-1BBL aNK-DLI, we observed a higher rate of aGVHD than expected. The dose of T cells administered, ≤2 × 10⁴/kg, was unlikely to be the sole cause for the rate of high-grade aGVHD observed in these HLA-matched recipients.⁸  Furthermore, the T-cell content of the aNK-DLI, following CD3 depletion, was required to be less than <0.2%, corresponding to a maximal additional T-cell dose that could have been administered as part of the aNK-DLI of 2 × 10³/kg, a dose likely insufficient to induce aGVHD. Although the aAPC expressed HLA-A2, 0/5 donor-recipient pairs with GVHD were HLA-A2⁺, ruling out a possibility that small numbers of activated T cells contained in the product were reactive against the HLA-A2 major histocompatibility antigen (Table 1). Interestingly however, aGVHD was more common in MUD allograft recipients, raising the prospect that aNK-DLI enhanced existent T-cell alloreactivity in MUD recipients and contributed to the aGVHD observed. Consistent with this, we observed higher rates of donor CD3⁺ chimerism in patients who developed aGVHD, which is a biomarker of alloreactivity in the GVHD direction. We propose therefore that the IL15/41BBL aNK-DLI exacerbated otherwise subclinical alloreactivity, thus contributing to aGVHD. The clinical observation that subjects who developed GVHD in this series tended to have higher rates of fever or evidence of an inflammatory process immediately before the aNK-DLI further supports the possibility that inflammation and/or cellular stress may have increased activating receptor ligand expression on nonmalignant tissues. This hypothesis is consistent with murine work implicating NK cells as cofactors capable of contributing to and increasing the severity of aGVHD.^59-65  Postulated mechanisms through which NK cells augment T cell–mediated aGVHD include direct killing of tissues expressing activating receptor ligands as well as production of inflammatory cytokines that augment T-cell killing. Indeed, the short latency between aNK-DLI infusion and aGVHD symptoms raises the prospect that NK-mediated cytokine release could potentially contribute to the clinical observations reported here.

Although several previous reports have employed adoptive NK cell transfer in an allogeneic transplant setting, none have implicated NK cells in inducing or exacerbating GVHD. Important distinctions between this trial and others reported thus far could explain these discrepant results. First, this is the first-in-human trial to employ IL15/4-1BBL–activated NK cells. Preclinical studies demonstrate clearly that IL15/4-1BBL–activated NK cells are functionally distinct from resting and IL-2–activated NK cells,⁴⁷  and thus the clinical effects using IL15/4-1BBL aNK-DLI could differ from those employing resting or cytokine–activated NK cells. Second, many previous studies employed adoptive transfer of MHC-mismatched NK cells to hosts that had not undergone allogeneic HSCT.^34-39  In this setting, the NK cells are likely to be rapidly rejected and may not undergo equivalent expansion, as could occur in this trial. Furthermore, underlying T cell–mediated alloreactivity would not be expected in these settings. In contrast, 3 of 4 studies employing donor-derived adoptive NK transfer reported GVHD.^40-43  Because MHC mismatched, T cell–replete allografts were used in 3 studies^40,41,43  with concern for T-cell contamination of the NK product in the fourth,⁴²  no clear conclusions regarding a potential role for NK cells in the GVHD observed can be rendered. Third, several previous trials employing adoptive NK transfer coadministered rhIL-2 following transfer, which may have a protective effect on GVHD because of expansion of regulatory populations.^66,67 

Given the small numbers of subjects included in this series, we were underpowered to definitively identify factors influence GVHD risk in this setting. For instance, although the incidence of KIR:KIR ligand mismatch did not meet statistical significance, there was a trend in this direction. Similarly, although expansion of adoptively transferred NK cells has been demonstrated in studies using MHC mismatched NK cells,^34,39  we were unable to track the progeny of the aNK-DLI product in vivo, but noted that the phenotype of NK cells expanding in vivo was distinct from that administered as part of the aNK-DLI product (Figure 3). These results could be explained if reconstitution of NK cells was derived from cells other than the aNK-DLI, but it is also possible that the phenotype of aNK-DLI transferred cells evolved in response to signals received in vivo. Despite the limitations, our study represents the first observation that aNK-DLI can be associated aGVHD. Because the rate of aGVHD was higher than we judged to be acceptable for further development of this platform, we modified the study for subsequent patients to incorporate GVHD prophylaxis and to report results of the small series to alert the clinical community of this possibility.

In summary, we report the first-in-human experience using IL15/4-1BBL aNK-DLI. Relatively modest doses of aNK-DLI, administered following nonmyeloablative, MHC-matched, T cell–depleted PBSCT was associated with rapid onset of a substantial rate and severity of acute GVHD. The aNK-DLI expressed high levels of activating receptors, and preclinical models demonstrate that such cells are fully licensed for killing. aGVHD in this series was more common in recipients of MUD grafts than MSD grafts and was associated with increased rates of donor CD3 chimerism. We conclude that IL15/4-1BBL aNK-DLI contributed to the aGVHD observed in this series and propose that the mechanism relates to augmentation of subclinical alloreactivity, which was more prevalent in recipients of MUD grafts. This observation is not inconsistent with the current understanding of essential roles for activating receptor signaling in controlling NK reactivity, the high levels of activating receptor expression on aNK-DLI, and evidence that activating receptor ligand expression is upregulated on nonmalignant tissues in response to stress. Future work is ongoing to include GVHD prophylaxis into a similar platform using aNK-DLI and to further investigate potential antitumor benefits that may be associated with administration of IL15/4-1BBL aNK-DLI in the allograft setting.

The authors acknowledge the exhaustive efforts of Showri Kakumanu, Keith O’Neill, our clinical fellows, attending physicians, and nursing staff and express sincere gratitude to our patients and their families.

This work was supported by the Center for Cancer Research, Intramural Research Program of the National Institutes of Health.

Contribution: N.N.S. wrote the article, analyzed the data, helped with patient care, and is the current principal investigator of the trial; K.B. contributed to study design, was responsible for patient care, and was the principal investigator of the trial; C.P.D. was involved with patient care and implementation and management of the trial; T.A.F., S.R., and K.L. performed the lymphocyte phenotyping; M.E.K. analyzed the KIR data and contributed to the article; H.Z. conducted the natural killer phenotype analysis of the infused product and the preclinical studies; M.S. and D.F.S. were involved with cell processing, activation, and expansion of the cellular products; C.K.H. directed and analyzed the KIR typing; S.P. and D.E.K. evaluated the biopsy specimens and the associated immunohistochemistry; A.S.W., M.S.M., and T.J.F. contributed to patient care, study design, and reviewing the article; and C.L.M. is the lead associate investigator on the trial and the Investigational New Drug Application holder, and contributed to the preclinical data, patient care, and writing of the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Crystal L. Mackall, 10-CRC, 1W-3750, 10 Center Dr, MSC 1104, Bethesda, MD 20892; e-mail: mackallc@mail.nih.gov.