Natural killer cell licensing after double cord blood transplantation is driven by the self-HLA class I molecules from the dominant cord blood

Natural killer (NK) cells gain functionality through an educational process (“licensing”) that depends on killer immunoglobulin-like receptor (KIR) recognition of self-HLA class I molecules. In allogenic double umbilical cord blood stem cell transplantation (2CBT), the respective roles of the two cord blood (CB) units and the recipient’s HLA class I molecules in NK cell licensing of the dominant CB (DCB) unit have yet to be characterized. In this study, following 2CBT, NK cells from the DCB unit were found to be licensed by their own HLA class I molecules (HLA-C1/C2 allotypes). HLA molecules from the recipient and the non-dominant CB (ndCB) did not show a significant role.

In allogenic hematopoietic stem cell transplantation (allo-HSCT), NK cells are able to recognize and eliminate leukemic cells.1 NK cells constitute the first lymphocyte subset to be reconstituted after allo-HSCT, with the rapid expansion of the CD56 NK cell subpopulation and the subsequent expansion of the dominant CD56 subset with a high cytotoxic function. The activity of NK cells is regulated by the balance between activating and inhibitory signals. The latter are mediated by the binding of major histocompatibility complex class I (MHC-I) molecules to the target cell on NK cell inhibitory receptors, including, in humans, the lectin-like heterodimer CD94/NKG2A (which interacts with HLA-E) and some of the polymorphic KIRs. The KIRs specifically bind to HLA-C allotypes with asparagine 80 (group C1 alleles), HLA-C allotypes with lysine 80 (group C2 alleles), and Bw4-expressing HLA-A and HLA-B allotypes.2 Observations in patients having undergone HLA-haploidentical HSCT highlighted a correlation between donor KIR and host KIR ligand expressions in tumor relapse occurrence.31

Recognition of self-MHC-I molecules by inhibitory KIRs is involved in the calibration of the NK cell effector capacities during differentiation and maturation. This enables cells lacking MHC-I to subsequently be recognized as a “missing-self” target.4 We have shown that after KIR ligand-mismatched conventional allo-HSCT in humans, the NK cell licensing reiterates the responsiveness pattern determined by the donor’s HLA ligands.5 Two retrospective studies of the effect of KIR ligand incompatibility in single76 and double7 unrelated CB transplantation for hematological malignancies showed conflicting results in terms of leukemia-free survival and incidence of relapse. Usually, one CB unit dominates the other in the 2CBT. It is presumed that alloreactive graft-versus-graft rejection occurs, depending on the status of CD4 T cells8 or CD8 T cells.9 The dose of CD34 cells10 and the cell banking procedures11 may also influence CB unit dominance which appears to be independent of the KIR-HLA mismatch.12 The co-existence of 3, and then 2, allogeneic partners after 2CBT raises the issue of the NK cell education. The objective of this study was to assess ex vivo NK cell licensing after 2CBT in humans, without studying the in vivo graft-versus-leukemia effect. This study focused on KIR2DL1/S1 and 2DL2/L3/S2 receptor function that recognize C2/C1 HLA-C allotypes.

This study included 20 patients having received a non-manipulated 2CBT to treat a hematological disorder (patients and transplant characteristics are summarized in Online Supplementary Table S1; methods in Online Supplementary Methods). We first analyzed the distribution of NK cell subpopulations after transplantation in ex vivo samples. CD3CD56 NKG2A expressing NK cells were much more abundant than CD3CD56 NKG2A NK cells (mean: 103 +/− 65 /μl vs. 36 +/− 33/μl). Distribution of NK cell subsets was estimated in thawed samples as below: in CD3CD56 NKG2A population, monoKIR2DL1/S1 and monoKIR2DL2/L3/S2 cells represent 9.5±5.8% and 13.6±9.1%, respectively. In CD3CD56 NKG2A population, monoKIR2DL1/S1 and monoKIR2DL2/L3/S2 cells represent 14.1±12.4% and 18.3±17.2%, respectively (Online Supplementary Table S2).

We examined CD56 CD94/NKG2A KIR NK cell subsets by degranulation assays at several time points after 2CBT (Online Supplementary Table S1). CD56 NK cells were divided into four subsets by flow-cytometry, according to their CD94/NKG2A and KIR expressions: NKG2AKIR cells; NKG2A panKIR cells (defined as KIR2DL1/S1 KIR2DL2/L3/S2 KIR3D); KIR2DL1/S1 monoKIR NK cells (defined as KIR2DL1/S1 KIR2DL2/L3/S2) and KIR2DL2/L3/S2 monoKIR NK cells (defined as KIR2DL2/L3/S2 KIR2DL1) (Online Supplementary Figure S1). The NK cell populations were analyzed for degranulation activity (CD107a is mobilized to cell surface following a co-culture with the K562 cell line) at 3, 6 and 12 months post-2CBT. As a baseline measurement, we evaluated the degranulation of CD56 NKG2AmonoKIR cells lacking their cognate HLA-C ligand in the two CB units and the recipient. Accordingly, CD56 NKG2AKIR2DL1/S1 and CD56 NKG2AKIR2DL2/L3/S2 NK cells were unlicensed in the absence of HLA-C2 and HLA-C1 ligands, respectively (degranulation mean ± SD after K562 challenge: 3.1±3.3%). As expected, CD56 NKG2A KIR NK cells also responded poorly during the first year post-2CBT (Figure 1A). In contrast, the CD56 NKG2A panKIR NK subset was the most responsive (degranulation mean: 28.1±8.8%) independently of the KIR/KIR ligand status of the recipient and CB units. We examined NK cell subsets at several time points after 2CBT (Online Supplementary Table S1). The licensing pattern was stable at two of the time points tested (Figure 1B).

Figure 1.Responsiveness of NK cell subsets after 2CBT. (A) The degranulation capacity of NK cell subsets was evaluated in all samples after incubation with K562 targets. The NKG2A^neg KIR^neg and the NKG2A^neg monoKIR NK cell subsets, in which the 2 cord blood units and the recipient lack a cognate KIR ligand (KIRL^neg), are hyporesponsive in comparison to the NKG2A^pos panKIR^pos subsets. **P<0.001 (Kruskal-Wallis). (B) Degranulation at 3 months (M3), 6 months (M6) and 12 months (M12) of the NKG2A-KIR⁻ NKG2A-panKIR⁺ and NKG2A⁺panKIR⁺ NK cell subsets was respectively evaluated in 12, 6 and 9 samples. ns.: not significant (Kruskal-Wallis).

We subsequently focused on the NKG2AKIR population to determine the influence of the KIR ligand source on NK cell education. To that end, we evaluated the degranulation activity of the monoKIR-expressing NK cells according to the presence, or absence, of the corresponding KIR ligand in the DCB, the ndCB and/or the recipient.

Firstly, to evaluate the recipient’s involvement in NK cell licensing, we studied monoKIR NK cell subsets in the context of HLA-C group-matched or -mismatched transplantation between the recipient and the DCB. In the HLA-C group-matched transplantation, the presence of KIR ligand in the recipient and in the DCB (R+DCB+) was associated with a significantly higher percentage of degranulation for NKG2A monoKIR NK cells (19±7.5%; P<0.001) compared to NKG2AKIR cells (Figure 2A). We then studied cases with one HLA-C group mismatch between the recipient and the DCB. The presence of KIR ligand in the DCB alone (R-DCB+; n = 6 cases) conferred responsiveness on the corresponding NKG2A monoKIR NK cells (15.8±5.6%) similar to that of the matched situation (Figure 2B). In contrast, in the presence of KIR ligand only in the recipient (R+DCB−; n = 4 cases), the NKG2AmonoKIR subset showed a lower degranulation capacity than NKG2AmonoKIR cells in an HLA-C group-matched situation (8.8±3.2% and 18.3±8.4%, respectively; P<0.001) (Figure 2C). This led us to the conclusion that recipient KIR ligand played a minor role in the NK education process after 2CBT.

Figure 2.Role of KIR ligand of the dominant CB unit and the recipient in NK cell licensing after 2CBT. Recipient KIR ligand involvement in NK cell licensing was evaluated in a total of 33 assays performed on 27 samples. (A) The responsiveness of CD56^dim NKG2A^pos monoKIR^pos and CD56^dim NKG2A^neg monoKIR^pos NK cells was evaluated when appropriate KIR ligands (HLA-C allotypes) were present in the recipient and the two cord bloods when compared to CD56^dim NKG2A^neg KIR^neg NK cells; (n = 31 tests; **P<0.001, Kruskal-Wallis). (B) The responsiveness of CD56^dim NKG2A^pos/neg monoKIR^pos NK cells was evaluated when appropriate KIR ligands (HLA-C allotypes) were absent in the recipient but present in the DCB (R- DCB⁺, n=6 tests; cases 15, 16, 17, 18, 19, 20). MonoKIR mismatched cells represent NK KIR^pos cells which recognize HLA-C present only in the DCB. MonoKIR matched cells represent NK KIR^pos cells which recognize HLA-C present in the recipient and the DCB. NS: Not Significant (P>0.05). (C) The responsiveness of CD56^dim NKG2A^pos or neg monoKIR^pos NK cells was evaluated when appropriate KIR ligands (HLA-C allotypes) were present in the recipient but absent in the DCB (R+ DCB-, n=4 tests; cases 11, 12, 13, 14). MonoKIR mismatched cells represent NK KIR^pos cells which recognize HLA-C present only in the recipient. MonoKIR matched cells represent NK KIR^pos cells which recognize HLA-C present in the recipient and the DCB. CD56^dim NKG2A^neg monoKIR mismatched compared to CD56^dim NKG2A^neg monoKIR matched NK cells were hyporesponsive (**P<0.001; Kruskal-Wallis). NS: Not Significant (P>0.05).

In 2CBT, the HLA class I environment transiently includes the ndCB HLA class I molecules. Indeed, in our cohort, whilst all patients displayed full DCB chimerism at three months after 2CBT, 8 patients underwent mixed ndCB/DCB chimerism for at least one month post-graft (See Online Supplementary Table S1), thus potentially impacting licensing of the DCB. We therefore evaluated the education pattern for monoKIR NK cells according to HLA-C group-matched or -mismatched CB units (i.e. the presence of HLA-C1 or HLA-C2 in the DCB or ndCB unit alone) at our earliest available time point, 3 months after 2CBT. The presence of KIR ligands only in the ndCB was unable to confer responsiveness to NKG2A monoKIR NK cell subsets whilst the presence of KIR ligand in the DCB was associated with a significantly higher percentage of NK cell degranulation. The presence, for example, of C2 ligands in the ndCB did not license monoKIR2DL1/S1 cells (Figure 3).

Figure 3.KIR ligand involvement of the ndCB in NK cell licensing after 2CBT. The responsiveness of CD56^dim NKG2A^pos/neg monoKIR2DL1/S1^pos or monoKIR2DL2/L3/S2^pos NK cells in a degranulation assay was evaluated as a function of the presence of KIR ligands HLA-C2 only in the ndCB. CD56^dim NKG2A^neg monoKIR2DL1/S1^pos mismatched were hyporesponsive. We present one representative example of two cases (cases 8, 9 see Online Supplementary Table S1).

In the present study, NK cell education was investigated 3, 6 and/or 12 months after KIR ligand matched or mismatched 2CBT. NKG2AKIR NK cells had a very low CD107a expression after stimulation by target cells consistent with the known hyporesponsive state of NK cells lacking inhibitory receptors.4 In contrast, the NKG2A panKIR CD56 NK subset was the most responsive, with a stable licensing pattern at two of the time points tested, highlighting the pivotal role of CD94/NKG2A and KIRs in maintaining a functional NK cell repertoire after transplantation. Regarding the more mature NKG2AKIR subset, the presence of KIR ligand only in the recipient conferred a poor responsiveness on the corresponding monoKIR NK cells in contrast to its presence in DCB units.

We previously reported that in conventional allogeneic HSCT, including 11 single CBT, NK cell education in the recipient is supported by the donor HLA-genotype for at least 3 years post-transplantation.5 To our knowledge, only the study by Foley et al. has addressed the issue of NK cell education after 2CBT showing that NK cells expressing self-KIR were (i) fully educated after 2CBT in a CD107a degranulation assay and (ii) expressed IFNγ.13 However, Foley et al. did not report the source of the ligand (recipient vs. DCB or ndCB) involved in NK cell licensing. In view of the reported plasticity of the education process, the potential impact of the ndCB on this process also needs to be assessed. Indeed, in murine models, unlicensed NK cells became functional after transfer into a MHC-I-sufficient environment, whereas previously educated NK cells became hyporesponsive in an MHC-I-deficient environment.14 However, reports on patients having undergone allo-HSCT showed the existence of a stable donor-type licensing pattern5 as observed here, suggesting that the cell population required for NKG2ANK cell education is transferred with the graft which will successfully reconstitute the recipient.155

In conclusion, although based on a small cohort of patients, these data provide proof of principle rules governing NK licensing after 2CBT with a potential impact on CB choice.