Neuroblastoma (NBL) is the most common extracranial solid tumor that occurs in children. For children with tumors that either have high risk biologic features or with metastatic disease, overall survival is still poor despite an aggressive treatment regimen that includes chemotherapy, surgery, autologous hematopoietic stem cell transplant (HSCT), radiation, and maintenance therapy with cis-retinoic acid (1). The addition of the monoclonal antibody (moAb) dinutuximab (which targets the NBL-associated antigen GD2), interleukin-2 (IL-2) and granulocyte-monocyte colony stimulating factor (GM-CSF) improved event-free and overall survival (2), but is not curative for the majority of patients who will ultimately relapse and die. Another treatment approach is needed that can potentially improve survival further and lead to a long-term cure.

Allogeneic HSCT was initially used in children with NBL about 30 years ago with some reports of clinical responses (3), but was never shown superior to autologous HSCT (4–6). Because there has not been convincing evidence of a graft-versus-tumor (GVT) effect against NBL (6–8), and because allogeneic HSCT introduces the life-threatening potential for graft-versus-host-disease (GVHD), autologous HSCT remains the standard of care. In fact, current protocols are incorporating tandem autologous HSCTs as consolidative therapy to improve event-free survival (9). Because of both preclinical evidence (10, 11) and case reports suggesting some clinical benefit of allogeneic HSCT in NBL, particularly in the haploidentical setting (12, 13), the objective of this preclinical study was to incorporate haploidentical HSCT as a platform for a combined immunotherapy regimen to enhance the GVT effect against NBL.

Until 2019, dinutuximab was given with GM-CSF and IL-2 in the Children’s Oncology Group (COG) as separate treatments as part of upfront maintenance therapy for NBL. Due to excessive toxicity associated with systemic IL-2 administration without clear added benefit, COG eliminated usage of IL-2 and now dinutuximab and GM-CSF are used. One means by which to maintain the beneficial activation of IL-2 for antibody-dependent cellular cytotoxicity (ADCC) without systemic toxicity is to restrict its use to the immune synapse. Hu14.18-IL2 is a fusion protein consisting of human IL-2 genetically linked to the carboxyl-termini of each human IgG1 heavy chain of the GD2-specific hu14.18 moAb. This “immunocytokine” (IC) provides a local source of IL-2 at the immunological synapse between the effector cell and the NBL, activating immunity against GD2⁺ tumors. Hu14.18-IL2 has been used in both phase I and phase II trials in children with refractory NBL and melanoma, with reversible toxicities and complete responses observed in both phase II NBL trials (14, 15). However, hu14.18-IL2 therapy is not curative when used as a single agent to treat macroscopic refractory or recurrent NBL, and has never been tested after allogeneic HSCT. The mechanism of action for hu14.18-IL2 is thought to be, at least in part, from ADCC from natural killer (NK) cells (16).

Because of the availability of clinical grade cytokines and artificial antigen presenting cells (aAPCs), infusion of high numbers of purified, ex vivo activated NK cells are emerging from preclinical models into clinical trials. NK cells have already been shown to have cytotoxicity in vitro against a variety of NBL cell lines (17) and primary patient tumors (18, 19) as well as in vivo with xenograft NBL models (20). In addition, the lymphopenic environment induced from the conditioning regimen for allogeneic HSCT is conducive for NK cell expansion given the presence of high levels of IL-15 (21). Lastly, NK cells produce growth factors like IL-1β, IL-6, G-CSF and GM-CSF that can support engraftment (22).

NK cells possess inhibitory receptors on their cell surface that can “turn off” the cells when they engage major histocompatibility complex (MHC) antigens (23, 24). Our current standard of administering an anti-GD2 moAb (dinutuximab) after autologous HSCT is limited in that the patient’s own NK cells must engage the antibody to eliminate the tumor, and risk engaging self-MHC on the tumor that could “turn off” the NK cell. In fact, two studies in children with NBL who were treated with anti-GD2 based therapies (one with hu14.18-IL2 and one with the moAb 3F8) reported a better response to therapy in those patients that were self-killer immunoglobulin-like receptor (KIR)/KIR ligand mismatched (25, 26), something that can be easily achieved if NK cells came from an appropriately selected allogeneic donor. In this study, we explore haploidentical allogeneic HSCT in NBL-bearing mice as a means of insuring that some of the inhibitory Ly49 receptors on donor murine NK cells do not engage their cognate MHC ligand, potentially “turning on” the NK cells and maximizing anti-tumor activity after hu14.18-IL2 IC administration.

Because IC have never been used after allogeneic HSCT, we established a MHC-mismatched haploidentical allogeneic HSCT model (H-2^b ➔ H-2^b x H-2^a) whereby lethally irradiated B6AJF1 recipients were transplanted with T cell depleted B6 BM and 0 – 2.5 x 10⁶ B6 T cells on day +0 ( Figure 1A ). Because the donor and host cells are MHC-mismatched (in the GVH direction), the presence of T cells in the BM graft leads to weight loss ( Figure 1B ) and lethal GVHD in less than 30 days ( Figure 1C ). The addition of IC following such a transplant is safe in the absence of T cells ( Figure 1C ), but in the presence of 2.5 x 10⁶ T cells there is still GVHD lethality ( Figures 1C, D ). Decreasing the amount of T cells in the donor graft reduces GVHD lethality ( Figure 1D ), suggesting there is a T cell threshold where one could safely administer IC. In fact, 2.5 x10² and 2.5 x 10³ T cells are well tolerated with IC and do not induce lethal GVHD ( Figure 1D ). Analysis of immune reconstitution shows a marked decrease in B220⁺ B cells in allogeneic HSCT recipients of 2.5 x 10⁶ T cells ( Figure 1E ), a surrogate of GVHD in other murine allogeneic HSCT models (32, 33), as well as marked decreases in NK cells ( Figure 1F ) and CD8⁺ T cells ( Figure 1G ), the cells that would typically respond to IC bound to tumor (16, 25). Both T cell depleted grafts and T cell replete grafts generate a low percentage of regulatory T cells ( Figure 1H ).

While human neuroblastomas ubiquitously express GD2, murine neuroblastomas show variable expression of GD2 ( Figure S1A ). NXS2 was selected so that allogeneic donor cells could be used from a C57BL/6 background, allowing for potential usage of knockout mice as NK donors in future experiments. Because children with solid tumors have the best outcomes when transplanted in remission, GD2⁺ NXS2 inoculation was performed on Day +10 after HSCT to mimic tumor relapse post-HSCT. Initially we compared syngeneic and allogeneic HSCT and found that both groups developed tumors, but NXS2 tumors in allogeneic HSCT recipients were smaller ( Figure S1B ), supporting rationale for a GVT effect in this model. We next examined allogeneic HSCT recipients (with add back of a nonlethal dose of T cells in the graft to avoid lethal GVHD but maintain some residual dose to mimic clinical T cell depletion). We performed NXS2 inoculation on Day +10, followed by 3 doses of IC on Days +14-16 ( Figure 2A ) to provide anti-GD2 tumor targeting for donor cells from the graft. Without IC, NXS2 tumors became large ( Figure 2B ). Administering IC significantly enhances the GVT effect by reducing tumor growth after T cell replete allogeneic HSCT, but small tumors still develop ( Figure 2B ). No differences were seen after T cell depleted allogeneic HSCT ( Figure 2B ), suggesting both donor T and NK cells are needed for optimal GVT effects of the IC. Importantly the IC mediates GVT without exacerbating GVHD ( Figure 2C ) after T cell replete allogeneic HSCT. Ex vivo activation of additional effector cells (e.g. donor-derived NK cells) that can recognize the tumor as allogeneic and/or respond to the IC via ADCC may enhance the GVT effect and potentially prevent tumor growth entirely.

Human aAPCs that express the co-stimulatory molecule 4-1BB ligand (CD137L) have been shown to potently expand and activate human NK cells (34–37), however has not been explored on murine NK cells. Using a murine aAPC transfected with CD137L in the presence of IL-15/IL-15Rα expands purified murine NK cells ex vivo ( Figure 3A ), with the highest yields after 1 week at a 1:1 ratio of NK:aAPC ( Figure 3B ). Activating NK cells with the CD137L⁺ aAPC without IL-15/IL-15Rα is insufficient to sustain NK cell growth (data not shown). The purity of NK cells after 7 days of ex vivo expansion is 90% ( Figure 3C ). While the percentage and absolute numbers of NK cells increase after ex vivo expansion, the percentage of NK cell subsets within that expanded population also changes. There is a mild but statistically significant increase in the percentage of Ly49C+I⁺ NK cells after ex vivo activation with IL-15/IL-15Rα alone or with CD137L/IL-15/IL-15Rα compared to unexpanded NK cells ( Figure 3D ). In contrast, we did not see any differences in the percentage of Ly49H⁺ NK cells after ex vivo activation ( Figure 3E ). Ex vivo activation also occurs, as evidenced by enhanced NK cytotoxicity as measured by potency assays in vitro ( Figure 3F ) and in vivo ( Figure S2 ), and augmented TNF-α production ( Figure 3G ). There are no significant changes in the percentage of cytotoxic (TRAIL⁺, FasL⁺ or CD107a⁺) or TNF-α producing Ly49 NK subsets ( Figure S3 ). Interestingly, ex vivo activated (H-2^b) NK cells demonstrate lysis of various syngeneic (H-2^b: 9464D) and allogeneic murine NBL cell lines (H-2^a: Neuro-2A, N18TG2, NXS2), however no significant improvement is seen with the addition of IC in vitro ( Figure 3H ). Because GVT/GVHD is a complex phenomenon that cannot be recapitulated in vitro, this observation led us to test if there were characteristics of the allogeneic HSCT milieu that could enhance a NK-mediated GVT effect with the addition of IC against GD2⁺ NBL in vivo.

During allogeneic HSCT, the GVT effect is mediated by T cells and NK cells while GVHD is mainly mediated by α/β⁺ T cells. To determine the contribution of NK alloreactivity to a GVT effect without contribution of donor T cells, we designed a F1 into parent allogeneic HSCT model so that (1) any residual donor T cells in the BM graft would be tolerized to host MHC and minor histocompatibility antigens in the thymus and thus not mediate GVHD (38), and (2) donor NK cells could still mediate alloreactivity since the host would lack cognate MHC ligands needed to engage donor Ly49 inhibitory receptors (39) ( Figure 4A ). When we infused F1 NK cells into one parent strain (H-2^bxd ➔ H-2^b), we observed that allogeneic ex vivo activated NK cells could mediate a mild weight loss ( Figure 4B ), but no lethality was observed (data not shown). Lethality was observed after infusion of F1 NK cells into the other parent strain (H-2^bxd ➔ H-2^d), with significantly more lethality observed with NK cells activated with CD137/IL-15/IL-15Rα after allogeneic HSCT than infusing NK cells activated with IL-15/IL15Rα alone ( Figure 4C ), indicating the contribution of CD137L during NK expansion and host MHC molecules in driving toxicity. More weight loss was seen with CD137/IL-15/IL-15Rα NK cells ( Figure 4D ). The infusion of ex vivo activated NK cells in a fully MHC-mismatched, T cell depleted, allogeneic HSCT model (H-2^b ➔ H-2^a) leads to lethality with or without hu14.18-IL2 ( Figure 4E ), suggesting IC does not contribute to lethality. Interestingly, when the allogeneic HSCT donor and recipients were MHC-matched, minor histocompatibility antigen-mismatched, no differences in weight loss (data not shown) or lethality were observed ( Figure 4F ). Histopathologic examination of classic acute GVHD target tissues (liver, gut, skin) did not show any lymphocytic infiltrate (data not shown), suggesting there was no direct attack of host tissues. Analysis of serum cytokines, however, did show cytokine release syndrome (CRS) with statistically significant increases in IL-6, IL-10 and IL-12p70 and a decrease in TNFα noted 1 and/or 2 weeks after ex vivo activated NK infusion as compared to recipients of allogeneic HSCT alone ( Figure 5 ). No differences in IFNγ, IL-1β, IL-2, IL-4, IL-5 and CXCL1 were observed ( Figure S4 ).

Immune profiling of allogeneic HSCT recipients showed mild increases in B cells and CD8⁺ T cells after tumor inoculation, but no changes in NK cells ( Figure 6A ). While adoptive transfer of wild type NK cells did not increase the total percentage of NK cells in the host, total NK cells did increase after IC administration but without enrichment of inhibitory Ly49C/I⁺ or activating Ly49H⁺ NK subsets ( Figure 6A ). Because ex vivo activated NK cells showed superior cytotoxicity in vitro ( Figure 3F ) and in vivo ( Figure S2 ), as well as high levels of TNFα production ex vivo ( Figure 3G ), we wanted to determine if the GVT effect was mediated by contact-dependent killing (via perforin) or contact-independent cytokine release (via TNF-α release), and whether abrogating these pathways would impact GVT. Infusion of Perforin^-/- ex vivo activated NK cells with IC did lead to a slight delay in tumor growth, but ultimately tumors overtook the mice ( Figure 6B ). But when we infused ex vivo activated TNFα^-/- NK cells after allogeneic HSCT with IC, we observed improved tumor control compared to ex vivo activated TNFα^+/+ NK cells ( Figure 6C ), suggesting TNFα may be contributing to CRS in a manner that attenuates the GVT potential of the infused NK cells. Flow cytometric analysis of splenocytes of mice treated with ex vivo activated TNFα^-/- NK cells showed a decrease in T cells and NK cells, with specifically less CD69⁺, CD107a⁺, and TRAIL⁺ NK cells seen in the periphery ( Figure 6D ). However, there were no differences between these NK subsets within the tumor ( Figure S5 ). Depletion of Ly49H⁺ NK cells, which represent an NK subset bearing an activation receptor that can engage MHC (H-2^b) on B6AJF1 host tissues, after NK infusion also led to improved tumor control early after tumor development ( Figure 6E ), suggesting blockade of TNFα-producing or depletion of activated NK cell subsets may help regulate toxicity while preserving GVT responses against NBL.

While haploidentical allogeneic HSCT is effective against leukemia (40), despite the publication of preclinical data (10, 11) and clinical data from case series describing the impact of allogeneic HSCT on NBL (12, 13), significant barriers are preventing allogeneic HSCT from more widespread testing as potential therapy for children with high risk or metastatic NBL. Barriers include the absence of conclusive evidence of a GVT effect against NBL and the development of GVHD that contributes to treatment-related mortality (41). We hypothesized that these barriers may be overcome by: 1) using T cell-depleted haploidentical allogeneic HSCT to enhance GVT and minimize GVHD; 2) focusing the localization and activity of the GVT inducing cells in the allogeneic HSCT via co-administration of the anti-GD2 IC hu14.18-IL2; 3) augmenting the capability of the GVT causing cells by selecting donors with the appropriate haploidentical relationship to the patient to enable NK allo-recognition in the GVT direction; and 4) co-infusing ex-vivo activated NK cells.

We show for the first time that usage of an IC, in this case hu14.18-IL2, is feasible and effective after allogeneic HSCT; IC induces GVT without GVHD as long as the T cell dose is minimized in the donor bone marrow graft. Because IL-2 could activate alloreactive T cells and exacerbate GVHD, but also expand regulatory T cells and abrogate GVHD, it was not clear what the effect of infusing IC would be after allogeneic HSCT. With higher T cells doses, the IL-2 present on the IC may have unintentionally stimulated alloreactive T cells from the donor, leading to GVHD. Also, poorer immune reconstitution was observed, which could reflect immunosuppression from GVHD or reduced spleen size as allogeneic HSCT recipients were dying from GVHD. With lower T cell doses, while anti-tumor activity against GD2⁺ NBL was observed as compared to allogeneic HSCT recipients without IC, tumors still developed. Because the mechanism of action of IC involves ADCC by NK cells, allogeneic HSCT recipients are lymphopenic, and the post-allogeneic HSCT milieu has high levels of IL-15 (21, 42), we hypothesized that infusions of ex vivo activated NK cells from the donor could enhance the GVT effect of the IC. Instead, we observed that adoptive transfer of ex vivo activated NK cells led to lethality in the presence or absence of IC.

NK cells have been adoptively transferred to recipients of allogeneic HSCT in preclinical models, with promising anti-tumor activity observed (43–46). Adoptive transfer of NK cells can also inhibit acute GVHD by limiting expansion and infiltration of donor T cells (47–49), producing TGF-β (45), controlling infections (50), depleting recipient dendritic cells (39), and improving lymphopenia (51). One limitation of applying these murine studies to our data is that all but two of these studies infused inactivated NK cells, and none of those studies used a co-stimulatory molecule like CD137L to activate the NK cells. Using F1 into parent HSCT models, where T cells cannot cause GVHD or alloreactivity, we observed weight loss or lethality depending on the recipient strain, suggesting ex vivo activated NK cells can mediate toxicity independent of T cell allorecognition. While this has not been observed in prior preclinical studies of adoptively transferred NK cells, it is possible that the biology of NK cells is different in vivo after activation by ex vivo as compared to inactivated NK cells. In vitro, ex vivo activated human NK cells can overcome KIR-mediated inhibitory signals (52). In clinical studies, infusion of NK cells expanded with either IL-2 (53, 54), or IL-15 and IL-21 after HLA-mismatched allogeneic HSCT (55–57) induced low rates and/or grades of GVHD, whereas infusion of NK cells activated with CD137L/IL-15 after HLA-matched allogeneic HSCT led to higher rates and grades of GVHD (58). The exact mechanism of NK-mediated GVHD is unclear but our data suggests it could have been in part driven by CRS.

Because we used T cell depleted bone marrow and did not detect NK cells in host tissues, we hypothesized that ex vivo activated NK cells mediated CRS that inhibited GVT. In fact, elevated levels of IL-6, IL-10 and IL-12p70 and decreased levels of TNF-α were observed in allogeneic HSCT recipients who received ex vivo activated NK cells than in uninfused allogeneic HSCT recipients. We did observe a minor population of regulatory T cells after IC administration that was not influenced by the number of T cells in the donor graft. Given the high IL-10 production observed after NK infusion, future studies should examine the contribution of IC in activating regulatory T cells and their role in GVT/GVHD/CRS in this model. In addition, increased TNF-α production by NK cells has been previously observed after haploidentical allogeneic HSCT (59), yet the decreased level noted in our model was still clinically significant. To determine if CRS was attenuating the GVT effect, adoptive transfer of purified, ex vivo activated TNFα ^-/- NK cells was performed and significantly attenuated tumor growth, suggesting that TNF-α production from ex vivo CD137L/IL-15/IL-15Rα activated NK cells may be contributing to a CRS that hinders anti-tumor effects. One potential mechanism could have been disruption of TNF-α mediated priming of regulatory T cells through TNFR2 (60, 61), reducing tolerance by/to the tumor. Future studies should examine if TNF-α may be activating regulatory T cells which in turn suppress elimination of NBL by T and NK cells. In addition, depleting ex vivo activated NK cells that express the activating receptor Ly49H after infusion improves early anti-tumor responses, overall suggesting that hyperactivated NK cell subsets may have to be carefully monitored after allogeneic HSCT as they may contribute to toxicities like CRS than can undo GVT effects. To our knowledge, this is the first example of CRS using adoptive transfer of ex vivo activated NK cells in the allogeneic HSCT setting. Because of the clinical availability of TNF-α inhibitors like infliximab, or soluble TNF-α receptor like etanercept, TNF-α blockade could be explored clinically to improve the GVT potential of ex vivo activated NK cells, but additional agents would likely need to be explored to better control CRS, like tocilizumab.

The preclinical data shown here provide preliminary groundwork for more mechanistic studies to enable clinical translation and evolution of existing pediatric trials using T cell depleted (e.g. α/β T cell depletion) haploidentical allogeneic HSCT for NBL by demonstrating the safety and efficacy of the combination of IC and ex vivo activated NK cell infusions to induce GVT effects against NBL. Clinical trials incorporating α/β T cell depletion haploidentical HSCT are underway at several pediatric centers as a means of depleting GVHD-causing α/β T cells while enriching the donor graft with GVT-promoting γ/δ T cells and NK cells (62), including for children with NBL (63, 64) (NCT02508038). A pilot trial testing the combination of IC and ex vivo activated haploidentical NK cell infusions in non-transplanted NBL patients is also underway (NCT03209869) (65). Further studies are warranted with these clinically available therapy platforms given the poor prognosis for high risk NBL and lack of effective salvage regimens.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The animal study was reviewed and approved by University of Wisconsin-Madison IACUC.

PB, SR, JH, and MC performed research, collected, analyzed, and interpreted data, and revised the manuscript. AR performed research and revised the manuscript. RO developed and provided the aAPC and revised the manuscript. MB, TF, SG, and PS analyzed and interpreted data and revised the manuscript. CC designed and supervised research; analyzed and interpreted data, drafted and revised the manuscript. All authors contributed to the article and approved the submitted version.

This work was supported by grants from the NIGMS/NIH T32 GM008692 and NCI/NIH T32 CA009135 (MC), NCI/NIH T32 CA009614 (SR), St. Baldrick’s – Stand up to Cancer Pediatric Dream Team Translational Research Grant SU2C-AACR-DT-27-17 (AR, TF, PS, CC), the Intramural Research Program at the NIH (TF), NCI/NIH R35 CA197078 (PS), the NCI/NIH R01 CA215461, St. Baldrick’s Foundation, American Cancer Society Research Scholar grant RSG-18-104-01-LIB, Hyundai Hope on Wheels and the MACC Fund (CC). We would like to thank the UWCCC flow cytometry core laboratory, small molecule screening facility and experimental animal pathology core laboratory, who are supported in part through NCI/NIH P30 CA014520. The St. Baldrick’s Foundation collaborates with Stand Up To Cancer.

SG is an employee of Provenance Biopharmaceuticals, and has a patent related to hu14.18-IL2. CC reports honorarium from Nektar Therapeutics and Novartis. These companies had no input in the study design, analysis, manuscript preparation or decision to submit for publication.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.