Genotyping Natural Killer Immune Checkpoints to Discover Biomarkers of Response

In this issue of Clinical Cancer Research, Erbe and colleagues found that killer cell immunoglobulin-like receptors (KIR) and KIR-ligand genotypes influenced clinical outcomes for antibody dinutuximab (ch14.18)-based immunotherapy of neuroblastoma (1). Natural killer (NK) cells and myeloid cells are key effectors in antibody-dependent cell-mediated cytotoxicity (ADCC) of neuroblastoma (2). To ensure self-tolerance and recognition of unhealthy cells, NK cells are educated through the interaction of KIRs with HLA class I (MHC class I) ligands (Fig. 1; ref. 3). As KIRs and HLA genes segregate independently, two general types of NK cells may exist in the same person: educated NK cells whose inhibitory KIR receptors have matching cognate self-HLA ligands, and uneducated NK cells that lack receptors specific for self-HLA or whose receptors are missing cognate ligands. Educated NK cells react to activation signals on targets when HLA is absent or downregulated but are inhibited by cells that express cognate HLA; uneducated NK cells require more potent activation signals (e.g., CD16A) but are resistant to HLA-mediated inhibition. In patients undergoing allogeneic hematopoietic stem cell transplantation, those lacking HLA ligands for donor-inhibitory KIR, particularly KIR3DL1, or exhibiting weak binding between HLA and inhibitory KIR, are more efficient in killing leukemia cells, resulting in fewer relapses and better overall survival (4).

High-risk neuroblastoma is an aggressive cancer of the sympathetic nervous system in children. More than 50% of the patients present with metastatic disease to bone marrow, bone, and soft tissues (2). With the addition of anti-GD2 antibody therapy (ch14.18 or 3F8) after dose-intensive induction with or without autologous stem cell transplant (SCT), long-term survival is now possible (5, 6). Previous studies in 245 patients using 3F8 plus granulocyte macrophage-colony-stimulating factor (GM-CSF) identified the positive clinical impact of uneducated NK cells in patients with missing KIR ligands (7). Similar observations have been made in smaller patient cohorts (8, 9). The inverse condition of having educated NK cells due to the presence of all KIR ligands was associated with poor patient survival. These studies suggest that neuroblastomas have enough HLA expression in the postinduction tumor microenvironment to interact with inhibitory KIRs and turn off educated NK cells.

Immunotherapy consisting of ch14.18, IL2, GM-CSF, and isotretinoin demonstrated improved patient survival over isotretinoin alone following autologous SCT (6). IL2 was chosen to accelerate NK-cell recovery and to stimulate cytotoxicity, whereas GM-CSF was included to expand the myeloid compartment, to induce inflammation, and to activate myeloid ADCC. Erbe and colleagues found no benefit from immunotherapy plus isotretinoin over isotretinoin alone among patients with missing KIR ligand (1). Among the ligand-present groups, namely, “all KIR ligands present,” “KIR2DL2 plus ligand (HLA-C1),” or “KIR3DL1 plus ligand (HLA-Bw4),” immunotherapy improved survival.

These observations could imply that neuroblastoma tumor cells following induction therapy have no functional HLA. Without HLA on tumor targets, inhibitory KIRs are no longer relevant checkpoints, and both educated and uneducated NK cells would kill neuroblastoma, equalizing outcomes between the missing-ligand and ligand-present groups. However, in patients treated without immunotherapy, Erbe and colleagues saw the opposite: Survival of the all ligand-present group (educated NK cells) was far worse than the missing-ligand group (uneducated NK cells).

Whether neuroblastomas have HLA expression is critical if classic T cell–based therapies, including vaccines and immune checkpoint inhibitors or activators, are to be pursued. Neuroblastoma cell lines and tumors at diagnosis are known to have low expression or absence of HLA (10). However, treatment-induced upregulation of HLA has been reported with IFN, chemoradiotherapy, isotretinoin (10, 11), and NK-dependent immunotherapy (12) in preclinical models. The use of proteomics or IHC to confirm HLA expression among tumor samples before immunotherapy could provide the definitive answer. If the lack of HLA expression in vivo is confirmed, NK-cell adoptive therapy or NK activation therapy should take priority, and selecting inhibitory KIR–ligand mismatch in adoptive NK therapy would be unnecessary. On the other hand, if HLA expression in vivo is confirmed, one will have to postulate that the particular regimen of ch14.18 + IL2 + GM-CSF + isotretinoin (6) was superior to hu14.18-IL2 (8), ch14.18/CHO (9), or 3F8 + GM-CSF (5) by supercharging NK cells beyond the inhibitory KIR barrier. Alternatively, immunotherapy could have upregulated stress ligands recognized by activating NK receptors, which overrode the inhibitory KIRs even when cognate ligands were present. However, neither of these conjectures could explain the provocative observation by Erbe and colleagues that immunotherapy provided no benefit beyond isotretinoin alone among the missing-ligand patients, who represented two thirds of the entire study population (1). Is it possible that for these patients, neither the anti-GD2 antibody nor IL2 or GM-CSF has any role in the posttransplant setting? If this is the case, oncologists may need to look for alternative therapies that will benefit patients following transplant.

In this randomized trial, only 77% of the patients had tissue samples available for genotyping (88 samples in the immunotherapy arm and 86 in the no-immunotherapy arm). In addition, 21% of patients on study never completed the intended immunotherapy (6). Thus, the actual number of patients available for subset analysis was relatively small. If accepted prognostic risk factors such as clinical stage (stage III vs. IV), patient age at diagnosis (12–18 months vs. ≥18 months), tumor MYCN amplifications, presence/absence of marrow disease, and tumor biology happened to be distributed unevenly among the different KIR/HLA subsets, the conclusions could be flawed. A multivariable analysis was not feasible due to the small patient numbers and low event frequencies. Given these limitations, it could be risky to run multiple comparisons, even with Bonferroni correction. As intriguing as the findings are, both KIR2DL2 and KIR2DL3 are known to bind HLA-C1, making it somewhat arbitrary to focus on the former and disregard the latter. Moreover, it is not clear whether the authors differentiated between homozygous KIR2DL2 + HLA-C1 versus heterozygotes (KIR2DL2/2DL3), which comprise the majority of all KIR2DL2(⁺) individuals. Interestingly, many KIR2DL2 alleles also bind HLA-C2 alleles, which are universally recognized by KIR2DL1. Why KIR2DL3 and KIR2DL1, which exist with comparable if not higher frequencies as KIR2DL2, appear to lack survival impact is unclear. Furthermore, the definition of certain contributing Bw4 alleles in NK education is somewhat controversial. HLA-A alleles exhibiting the Bw4 epitope are not known to contribute in a significant way to NK education through KIR3DL1, although in overexpressed systems, they may inhibit. In addition, the strength of interaction of the KIR3DL1 alleles (7) with Bw4 could be informative. And finally, activating receptor–ligand interactions could further complicate the definition of risk groups. It is notable that the activating KIR2DS2 shares strong linkage disequilibrium with KIR2DL2, although the ligand for KIR2DS2 remains unclear.

Although KIR typing can provide useful biomarkers of response, the underlying biology should drive treatment strategies. The absence of benefit from immunotherapy among the missing-ligand groups might suggest sparing these children of futile immunotherapy. Given its importance for both NK- and T-cell biology, HLA expression in postinduction neuroblastoma deserves attention in future analysis. Although the findings of Erbe and colleagues (1) seem to contradict the standard model of inhibitory KIRs in neuroblastoma, they have added another dimension to the complexity of inhibitory receptors for NK cells. More testing will be needed to define the utility of genotyping of KIR/HLA as response predictors and to shed light on the underlying immunobiology of NK cells. The day will come when precision biomarkers are available to guide practicing oncologists to choose curative treatments while minimizing side effects in metastatic neuroblastoma.

N.-K. Cheung reports receiving commercial research grants from Y-mAbs Therapeutics; holds ownership interest in Abpro-Labs and Y-mAbs Therapeutics, both of which have licensed patents from Memorial Sloan Kettering Cancer Center; and is an advisory board member for Abpro-Labs and Eureka Therapeutics. Except for patents filed by Memorial Sloan Kettering Cancer Center where both N.-K. Cheung and/or K.C. Hsu are named as inventors, there are no other potential conflicts of interest.

Conception and design: N.-K. Cheung, K.C. Hsu

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N.-K. Cheung, K.C. Hsu

Writing, review, and/or revision of the manuscript: N.-K. Cheung, K.C. Hsu

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): N.-K. Cheung, K.C. Hsu

This work was supported by NIH CA182526 (to N.-K. Cheung), NIH CA164365 (to K.C. Hsu), FDA FD-R-005415 (to K.C. Hsu), Alex's Lemonade Stand Foundation Innovation (to K.C. Hsu), and NIH CA008748.