Phenotypic and Functional Characterization of NK Cells in αβT-Cell and B-Cell Depleted Haplo-HSCT to Cure Pediatric Patients with Acute Leukemia

NK cells can exert remarkable graft-versus-leukemia (GvL) effect in HLA-haploidentical hematopoietic stem cell transplantation (haplo-HSCT). Here, we dissected the NK-cell repertoire of 80 pediatric acute leukemia patients previously reported to have an excellent clinical outcome after αβT/B-depleted haplo-HSCT. This graft manipulation strategy allows the co-infusion of mature immune cells, mainly NK and γδT cells, and hematopoietic stem cells (HSCs). To promote NK-cell based antileukemia activity, 36/80 patients were transplanted with an NK alloreactive donor, defined according to the KIR/KIR-Ligand mismatch in the graft-versus-host direction. The analysis of the reconstituted NK-cell repertoire in these patients showed relatively high proportions of mature and functional KIR+NKG2A−CD57+ NK cells, including the alloreactive NK cell subset, one month after HSCT. Thus, the NK cells adoptively transfused with the graft persist as a mature source of effector cells while new NK cells differentiate from the donor HSCs. Notably, the alloreactive NK cell subset was endowed with the highest anti-leukemia activity and its size in the reconstituted repertoire could be influenced by human cytomegalovirus (HCMV) reactivation. While the phenotypic pattern of donor NK cells did not impact on post-transplant HCMV reactivation, in the recipients, HCMV infection/reactivation fostered a more differentiated NK-cell phenotype. In this cohort, no significant correlation between differentiated NK cells and relapse-free survival was observed.

infections, HCMV greatly influences the NK-cell phenotype, promoting the expansion of highly differentiated NK cells with "adaptive" features, characterized by the expression of NKG2C, self-iKIR (primarily KIR2DL), the marker of terminally differentiated stage CD57, and the lack of NKG2A [27][28][29]. The role of HCMV in accelerating NK-cell maturation has been described in patients after allogeneic HSCT, in different transplantation settings [30][31][32].
As documented in CD34 + haplo-HSCT recipients, the NK cells derived early from HSC display an immature phenotype, characterized by CD56 bright KIR − NKG2A + expression; the emergence of fully functional KIR + mature NK cells, including the alloreactive NK cells, may require at least eight weeks, this resulting into a delay of the NK cell-associated GvL effect [13,33]. Notably, the use of a novel graft manipulation method, based on the selective depletion of αβT cells and B cells, allows the infusion, in addition to HSC, of immunocompetent cells such as mature, donor-derived NK, γδT, and myeloid cells. In addition, in this setting of selective T-cell depletion, no post-transplant pharmacological immune suppression is given, and NK cells can promptly exert an immediate anti-leukemia effect after transplantation, before the wave of NK cells differentiating from donor hematopoietic precursors emerges. Here, we analyzed the cohort of patients transplanted from αβT/B-cell depleted haplo-HSCT (NCT01810120), whose clinical outcome has been already described [34], providing new insights on NK-cell receptor repertoire of donors and transplanted patients.

Criteria for Donor Selection
To select the most suitable donor, when alternative donors (e.g., both parents) were available (58 out of 80 cases), we considered several features found to be correlated, by either in vitro assays and/or clinical studies, with a better anti-leukemia potential. These criteria, defined by genetic and phenotypic analyses, included: (a) presence of NK alloreactivity (i.e., KIR/KIR-L mismatch in GvH direction) and larger size (i.e., ≥5%) of the alloreactive subset [3,33], (b) presence of a KIR B/x genotype especially with B content value ≥2 [35], (c) presence of KIR2DS1 [36,37], (d) higher absolute number of NK and γδT cells [38], and (e) higher expression of NKp46 [21] and presence of NKG2C [32]. Tables 1 and S1 report patient and donor characteristics including: type of disease (acute lymphoblastic leukemia, ALL, or acute myeloid leukemia, AML), presence, and type of donor NK alloreactivity, KIR genotype, B content value, and patient clinical outcome.
All donors were analyzed combining genetic and phenotypic approaches. An example, comparing the parents of a patient, is shown in Figure S1. Based on KIR-ligand and KIR genotype analyses, only the mother could be considered a NK alloreactive donor. Flow-cytometry analysis using different anti-KIR mAb combinations allowed the detection of surface expressed iKIR and aKIR, and the evaluation of the alloreactive subset. In addition, the expression of NKp46, as well as of NKG2C in relation to NKG2A, KIRs, and CD57, was also tested. In this case, the mother was preferred as donor for the transplant, mainly because of the presence of NK alloreactivity and higher B content value.
In the next two sub-sections, the relevance of NK alloreactivity and activating NK cell receptors will be analyzed in detail.

Advantage of NK Alloreactivity in the Anti-Leukemia Effect
The presence of NK alloreactivity represented the first criterion considered for donor selection, since alloreactive NK cells, expressing as inhibitory receptor(s) only the educated iKIR(s) which do not recognize any ligand on recipient cells, can display a higher anti-leukemia activity and thus potentially exert GvL effect [21,33]. Three types of NK alloreactivity were identified based on donor self-iKIR and mismatched KIR-L. They included Allo C1 (mediated by KIR2DL2/L3 and mismatched HLA-C1), Allo C2 (KIR2DL1 and mismatched HLA-C2), and Allo Bw4 (KIR3DL1 and mismatched HLA-Bw4). In our cohort, 45% of the patients were transplanted from an NK alloreactive donor (Table 1 and Table S1). To functionally assess the advantage of NK alloreactivity, we performed rapid degranulation assays. We focused on Allo C2, since KIR2DL1 displays a more stringent recognition of HLA-C allotypes, and a greater number of HLA-C1/C1 than HLA-C2/C2 leukemia blasts, in addition to the pediatric leukemia cell line NALM-16 (HLA-C1 hemizygous) were available (Table S2). NK cells derived from HLA-C1/C2 donors were significantly more efficient than NK cells from HLA-C1/C1 donors against HLA-C1/C1 pediatric ALL ( Figure 1A, left). Importantly, considering the activity of different subsets in the case of HLA-C1/C2 effector cells, the alloreactive NK cell subset (i.e., cells expressing KIR2DL1 as the only inhibitory receptor, single positive KIR2DL1, sp2DL1) displayed a significantly higher degranulation capability than other subpopulations ( Figure 1A, right). As shown in a representative experiment ( Figure 1B), sp2DL1 was the most efficient subset against HLA-C1/C1 pediatric ALL-1 and AML-2, both characterized by high surface expression of HLA-C as detected by DT9 mAb (Table S2). NALM-16, characterized by low expression levels of HLA class I molecules, appeared to trigger NK-cell degranulation, but still the advantage of the alloreactive subset could be detected ( Figure 1B and Figure S2). The presence of NK alloreactivity represented the first criterion considered for donor selection, since alloreactive NK cells, expressing as inhibitory receptor(s) only the educated iKIR(s) which do not recognize any ligand on recipient cells, can display a higher anti-leukemia activity and thus potentially exert GvL effect [21,33]. Three types of NK alloreactivity were identified based on donor self-iKIR and mismatched KIR-L. They included Allo C1 (mediated by KIR2DL2/L3 and mismatched HLA-C1), Allo C2 (KIR2DL1 and mismatched HLA-C2), and Allo Bw4 (KIR3DL1 and mismatched HLA-Bw4). In our cohort, 45% of the patients were transplanted from an NK alloreactive donor (Tables 1 and S1). To functionally assess the advantage of NK alloreactivity, we performed rapid degranulation assays. We focused on Allo C2, since KIR2DL1 displays a more stringent recognition of HLA-C allotypes, and a greater number of HLA-C1/C1 than HLA-C2/C2 leukemia blasts, in addition to the pediatric leukemia cell line NALM-16 (HLA-C1 hemizygous) were available (Table S2). NK cells derived from HLA-C1/C2 donors were significantly more efficient than NK cells from HLA-C1/C1 donors against HLA-C1/C1 pediatric ALL ( Figure 1A, left). Importantly, considering the activity of different subsets in the case of HLA-C1/C2 effector cells, the alloreactive NK cell subset (i.e., cells expressing KIR2DL1 as the only inhibitory receptor, single positive KIR2DL1, sp2DL1) displayed a significantly higher degranulation capability than other subpopulations ( Figure 1A, right). As shown in a representative experiment ( Figure 1B), sp2DL1 was the most efficient subset against HLA-C1/C1 pediatric ALL-1 and AML-2, both characterized by high surface expression of HLA-C as detected by DT9 mAb (Table S2). NALM-16, characterized by low expression levels of HLA class I molecules, appeared to trigger NK-cell degranulation, but still the advantage of the alloreactive subset could be detected ( Figures 1B and S2).    Figure S3). E:T ratio 10:1. NK cells were pre-incubated with saturating amounts of mAb to the indicated receptors. Data are shown as percentage of lysis, mean + SEM. Each target has been tested with effector cells derived from at least n = 7 donors. Three independent experiments were performed. Significant differences between lysis in the presence of blocking mAb in comparison to "No mAb" are shown: * p < 0.05; ** p < 0.01; **** p < 0.0001 (Friedman test with Dunn's post-test).

Relevance of Activating NK Cell Receptors
In the absence of NK alloreactivity, donor selection criteria were primarily based on the higher B content value, described to correlate with a better LFS [35,39]. Among aKIRs, the presence of KIR2DS1 was positively considered, particularly in HLA-C1 + donor and HLA-C2 + patient pairs, implying its education and its capability to recognize the ligand on leukemia blasts (termed "educated and useful", E/U) [19,33]. Thus, 45% or 35% of transplants were performed from donors characterized by KIR B content value ≥ 2 or by the presence of KIR2DS1 E/U, respectively (Table S1). In addition, assuming a relevant role of NKG2C receptor during HCMV reactivation [30,32], whenever possible, we avoided choosing donors characterized by NKG2C del/del genotype. In this cohort, only 3 out of 80 donors were NKG2C del/del .
Since within the activating receptors involved in natural cytotoxicity NCRs play a major role in leukemia recognition [21], we examined the NK-cell expression of NKp46, which may vary among different donors, privileging those with higher surface density. We analyzed various pediatric primary leukemia cells as targets to define the involvement of different activating pathways in NK-mediated killing. Effector cells were represented by polyclonal activated NK cells, selected according to their killing capacity. As shown in Figure 1C, mAb-mediated blocking of NCR, as compared to NKG2D and DNAM-1, induced a higher inhibition of lysis of most primary leukemia blasts, both BCP-ALL, T-ALL, and AML. This confirmed the primary role of NCR/NCR-L interaction in NK-cell-mediated anti-leukemia activity. The ULBP1 + ALL-2, in which NKG2D-mediated killing was primarily effective, represented an exception. The NALM-16 leukemia cell line was highly susceptible to NK-cell lysis by the engagement of NKG2D, NCR, and DNAM-1 triggering receptors, in order of relevance, consistent with their expression of all NKG2D-L and DNAM-1-L (Table S2 and Figure S3).

Infused NK Cells Persist and Are Functional up to One-Month Post-Transplant
In order to assess the possible contribution of the NK cells adoptively infused with the graft, we studied the pattern of receptor expression on NK cells at an early phase (approximately 1 month) after αβT/B-depleted haplo-HSCT. In most instances, CD94:NKG2A, the inhibitory receptor characterizing NK cells at the immature stages, was expressed on a fraction of NK cells. We stratified the patients in two groups according to the number of infused NK cells, namely higher or lower than the median value (i.e., 34.6 × 10 6 /kg body weight). Patients infused with high NK cell numbers displayed a significantly lower % NKG2A + NK cells, as compared to those infused with low NK cell numbers (p < 0.05) (Figure 2A). For comparison, we also analyzed the reconstitution in a cohort of patients after CD34 + haplo-HSCT, which is typically characterized by NK cells with an immature phenotype one-month post-transplant [33]. Notably, only the αβT/B-depleted haplo-HSCT patients receiving a high NK cell dose greatly differed from CD34 + haplo-HSCT in terms of NKG2A expression (p < 0.001).
Thus, the infusion of high NK cell doses resulted in higher frequency of mature NK cells defined as KIR + NKG2A − ( Figure 2B). If the graft contained high NK cell numbers with adaptive phenotype (i.e., high percentages of NKG2C and CD57), the early-reconstituted NK-cell repertoire of the patient showed a pattern of receptor expression very similar to the respective donor. A representative case is shown in Figure 2C. Remarkably, this reflected also in the identification of the alloreactive NK cell subset in the early phase after transplantation (Table 1 and Figure 2D). Importantly, the anti-leukemia activity of NK cells was also preserved. Indeed, NK cells from αβT/B-cell depleted haplo-HSCT patients at one month after the transplant displayed similar degranulation capacity against K562 as compared to NK cells from healthy donors ( Figure 2E). In addition, the sp2DL1 NK cell subset of patients transplanted from Allo C2 donors was the most efficient against NALM-16, similarly to NK cells from healthy donors ( Figure 2F and Figure S2). reconstituted NK-cell repertoire of the patient showed a pattern of receptor expression very similar to the respective donor. A representative case is shown in Figure 2C. Remarkably, this reflected also in the identification of the alloreactive NK cell subset in the early phase after transplantation (Table  1 and Figure 2D). Importantly, the anti-leukemia activity of NK cells was also preserved. Indeed, NK cells from αβT/B-cell depleted haplo-HSCT patients at one month after the transplant displayed similar degranulation capacity against K562 as compared to NK cells from healthy donors ( Figure  2E). In addition, the sp2DL1 NK cell subset of patients transplanted from Allo C2 donors was the most efficient against NALM-16, similarly to NK cells from healthy donors ( Figures 2F and S2). Altogether, these data highlight that engrafted mature and functional NK cells persist in the peripheral blood of the recipient for at least one month.  Altogether, these data highlight that engrafted mature and functional NK cells persist in the peripheral blood of the recipient for at least one month.

Analysis of NK Alloreactivity in the Reconstituted Repertoire after Transplantation
If the detection of mature NK cells at the 1st month post-transplant mainly relies on the phenotype and dose of engrafted NK cells, at the following time points, the NK-cell repertoire can be complemented and thus influenced by newly differentiated NK cells from immature precursors. We compared the presence of alloreactive NK cell subset in the reconstituted NK cell repertoire of the recipients with the respective donors, and analyzed the functional activity of NK cells after transplantation. Table 1 summarizes the 36 cases transplanted from NK alloreactive donors, grouped by the type of alloreactivities: 15 cases of Allo C1 (including two cases with additional Bw4 mismatch), 14 cases of Allo C2, and seven cases of Allo Bw4. Among the activating KIRs, the presence of KIR2DS1 and KIR2DS2 has been reported. In particular, KIR2DS1 E/U was favored in Allo C1 and Allo Bw4 groups (present in 10 out of 22 cases). The presence of KIR2DS2 leads to an imprecise definition of the size of alloreactive subset for Allo C2 and, possibly, Allo Bw4, since it can be co-expressed with KIR2DL1 and KIR3DL1, respectively, but these cells cannot be quoted as alloreactive (due to the lack of an antibody recognizing KIR2DL2/L3 and not KIR2DS2). We analyzed the frequency of alloreactive NK cells present in donors and patients at post-transplant time ranging from 3 to 6 months (3-6 M), in comparison to one month (1 M) ( Table 1). Figure 3A shows the comparison of the percentage of the alloreactive NK cell subsets detected in the donors and in the 3-6 M recipients.

Influence of HCMV Reactivation on Alloreactive NK-Cell Phenotype
The post-transplant HCMV reactivation, which is known to promote the differentiation of mature KIR + NKG2A − [31,40], possibly containing alloreactive cells, was also evaluated ( Figure 3A). In the Allo C1 group, the alloreactive subset (i.e., sp2DL2/3) was substantially preserved post-transplantation and it appeared to be sustained by HCMV reactivation. A strong reduction of alloreactive subset was observed only in two patients (UPN35 and UPN40), who received a graft containing low numbers of NK cells and did not experience HCMV reactivation. In the Allo C2 group, the size of the alloreactive subset almost always appeared reduced in the reconstituted patient repertoire in comparison to the respective donors, and only one exception (UPN24) was observed. Nevertheless, in 6 out of 13 cases analyzed, we observed percentages ≥ 5 (value arbitrarily chosen as threshold for a substantial size of the alloreactive subset). In addition, the possible under-estimation due to the presence of KIR2DS2 should be considered. Notably, HCMV reactivation, observed in three patients, was associated with a reduction of the sp2DL1 subpopulation, possibly due to the lack of NKG2C + cell subset increase (i.e., UPN23 and UPN43) or the expansion of NKG2C + KIR2DL2/L3 + NK cells, limiting alloreactivity (i.e., UPN71). In the Allo Bw4 group, the presence of a substantial post-transplant alloreactive subset was only observed in UPN51, who did not experience HCMV reactivation. In UPN12 and UPN26 (HLA-C1 + donors and recipients), an expansion of NKG2C + KIR2DL2/L3 + NK cells, possibly induced by HCMV reactivation, resulted in sp3DL1 cell subset reduction.

Anti-Leukemia Activity
We tested the functional activity of NK cells derived from representative transplanted patients against primary leukemia blasts in degranulation assays. In UPN7 and UPN29, representative of Allo C1 group, NK cells degranulated when co-cultured with ALL-2 (characterized by HLA-C2/C2), and the alloreactive subset was the most efficient ( Figure 3B). In the Allo C2 group, a good degranulation capacity, especially from the alloreactive subset, was observed in UPN58 (and to a less extent in UPN43) NK cells against the HLA-C1/C1 leukemias, AML-2 and ALL-1 ( Figure 3C). (A) Percentages of alloreactive NK cell subsets (gating on CD3 − CD56 dim ) in patients 3-6 months (3-6 M) after haplo-HSCT were compared with those observed in the donors, considering the different type of alloreactivity. Filled black circles refer to donor/recipient pairs in which patients experienced HCMV reactivation. The dotted lines are set at 5%, considered a cut-off for a good size of alloreactive subset; (B,C) degranulation activity of NK cells from patients at six months after haplo-HSCT against pediatric leukemia blasts. E:T ratio 2:1. CD107a expression was evaluated on CD3 − CD56 + NK cells (All NK) and on different NK cell subsets, as indicated. Allo Subset represents the alloreactive NK cell subset, No Allo Subset consists of NK cells expressing iKIRs specific for ligands on target cells. Numbers represent the ΔCD107a. In (B), NK cells from two representative patients, transplanted with Allo C1 donors, were co-cultured with HLA-C2/C2 ALL-2. In (C), NK cells from two representative patients, transplanted with Allo C2 donors, were co-cultured with HLA-C1/C1 AML-2 or ALL-1.

Evaluation of Differentiated vs. Naïve NK-Cell Repertoires and Possible Correlation with the Clinical Outcome
Similar to the approach described by Bjorklund et al. [41], we performed an unsupervised hierarchical clustering of relevant NK-cell immunophenotyping data of both donors ( Figure S4) and recipients at 3-6 months after transplant (Figure 4). The frequencies of CD56 bright , and, among the CD56 dim , of NKG2A + , CD57 + , NKG2C + , and KIR + NKG2A − cells have been evaluated. In these analyses, cases with NKG2C del/del donors (n = 3) have been excluded. Two main NK cell clusters, associated with a naïve (high NKG2A and/or CD56 bright ) or more differentiated (high NKG2C, KIR + NKG2A − and/or CD57 + ) pattern, were identified in both donors and recipients. (A) Percentages of alloreactive NK cell subsets (gating on CD3 − CD56 dim ) in patients 3-6 months (3-6 M) after haplo-HSCT were compared with those observed in the donors, considering the different type of alloreactivity. Filled black circles refer to donor/recipient pairs in which patients experienced HCMV reactivation. The dotted lines are set at 5%, considered a cut-off for a good size of alloreactive subset; (B,C) degranulation activity of NK cells from patients at six months after haplo-HSCT against pediatric leukemia blasts. E:T ratio 2:1. CD107a expression was evaluated on CD3 − CD56 + NK cells (All NK) and on different NK cell subsets, as indicated. Allo Subset represents the alloreactive NK cell subset, No Allo Subset consists of NK cells expressing iKIRs specific for ligands on target cells. Numbers represent the ∆CD107a. In (B), NK cells from two representative patients, transplanted with Allo C1 donors, were co-cultured with HLA-C2/C2 ALL-2. In (C), NK cells from two representative patients, transplanted with Allo C2 donors, were co-cultured with HLA-C1/C1 AML-2 or ALL-1.

Evaluation of Differentiated vs. Naïve NK-Cell Repertoires and Possible Correlation with the Clinical Outcome
Similar to the approach described by Bjorklund et al. [41], we performed an unsupervised hierarchical clustering of relevant NK-cell immunophenotyping data of both donors ( Figure S4) and recipients at 3-6 months after transplant (Figure 4). The frequencies of CD56 bright , and, among the CD56 dim , of NKG2A + , CD57 + , NKG2C + , and KIR + NKG2A − cells have been evaluated. In these analyses, cases with NKG2C del/del donors (n = 3) have been excluded. Two main NK cell clusters, associated with a naïve (high NKG2A and/or CD56 bright ) or more differentiated (high NKG2C, KIR + NKG2A − and/or CD57 + ) pattern, were identified in both donors and recipients. Unsupervised hierarchical clustering of NK-cell repertoires in 60 patients at 3-6 months after HSCT allowed the identification of two main clusters, namely cluster 1 with "naïve" and cluster 2 with a more differentiated repertoire. Data were based on multi-color flow-cytometry, analyzing the frequencies of CD56 bright , or, among the CD56 dim , the frequencies of NKG2A + , CD57 + , NKG2C + , and KIR + NKG2A − NK cells. Information on clinical and biological observations is annotated in the top bars above and to the right of the heat map. Z-scores were calculated and used to scale for visualization across rows in the heat map. High (red) and low (blue) frequencies of each subset are represented in the color scale. Underneath the heatmap, red and black squares indicate the cases whose respective donors showed NK cells with "differentiated" (identified by cluster 1 + 3) and "naïve" (identified by cluster 2) phenotype, respectively, as shown in Figure S4, while white squares indicate when there is no available donor data (NA).
We could not prove any association between the different donor clusters and HCMV reactivation after haplo-HSCT, suggesting that the differentiated phenotype of donor NK cells represents neither a predisposing nor a protecting feature (p = 0.79) ( Figure 5A). In contrast, the evaluation of the recipients' repertoire led to the evidence that the cluster with differentiated phenotype was associated with HCMV reactivation (p = 0.002) ( Figure 5C). These data are consistent with the notion that HCMV reactivation accelerates NK-cell differentiation, thus proving the accuracy of this clustering approach. Therefore, we assessed whether a certain cluster identified in both donors and transplanted patients might be associated with the clinical outcome. However, we could not find any significant correlation with the overall survival (OS), LFS, and cumulative incidence of relapse ( Figure 5B,D). Stratifying by disease type of the transplanted patients (ALL and AML), we could observe an opposite trend toward a better clinical outcome, considering the donor cluster: differentiated for ALL, and naïve for AML patients ( Figure S5). Unsupervised hierarchical clustering of NK-cell repertoires in 60 patients at 3-6 months after HSCT allowed the identification of two main clusters, namely cluster 1 with "naïve" and cluster 2 with a more differentiated repertoire. Data were based on multi-color flow-cytometry, analyzing the frequencies of CD56 bright , or, among the CD56 dim , the frequencies of NKG2A + , CD57 + , NKG2C + , and KIR + NKG2A − NK cells. Information on clinical and biological observations is annotated in the top bars above and to the right of the heat map. Z-scores were calculated and used to scale for visualization across rows in the heat map. High (red) and low (blue) frequencies of each subset are represented in the color scale. Underneath the heatmap, red and black squares indicate the cases whose respective donors showed NK cells with "differentiated" (identified by cluster 1 + 3) and "naïve" (identified by cluster 2) phenotype, respectively, as shown in Figure S4, while white squares indicate when there is no available donor data (NA).
We could not prove any association between the different donor clusters and HCMV reactivation after haplo-HSCT, suggesting that the differentiated phenotype of donor NK cells represents neither a predisposing nor a protecting feature (p = 0.79) ( Figure 5A). In contrast, the evaluation of the recipients' repertoire led to the evidence that the cluster with differentiated phenotype was associated with HCMV reactivation (p = 0.002) ( Figure 5C). These data are consistent with the notion that HCMV reactivation accelerates NK-cell differentiation, thus proving the accuracy of this clustering approach. Therefore, we assessed whether a certain cluster identified in both donors and transplanted patients might be associated with the clinical outcome. However, we could not find any significant correlation with the overall survival (OS), LFS, and cumulative incidence of relapse ( Figure 5B,D). Stratifying by disease type of the transplanted patients (ALL and AML), we could observe an opposite trend toward a better clinical outcome, considering the donor cluster: differentiated for ALL, and naïve for AML patients ( Figure S5).

Discussion
Several platforms of allogeneic HSCT can be employed to treat leukemia patients, including the use of different type of donor with diverse HLA matching (HLA-identical sibling, SIB; HLAmatched unrelated volunteer, URD; umbilical cord blood, UCB; or haploidentical), different stem cell source (umbilical cord blood unit, UCB; bone marrow, BM; or mobilized peripheral blood, PBSC), various preparative conditioning regimen (reduced intensity, RIC; or myeloablative, MAC), graft manipulation strategies and pharmacological GvHD prophylaxis [1,42]. All these variables, along with the patient age (pediatric or adult), may influence the post-HSCT NK-cell reconstitution affecting both phenotypic and functional features. For children with high-risk leukemia in urgent need of a transplant, both parents can represent readily available haploidentical donors. However, to circumvent the risk of graft rejection and GvHD related to the high degree of HLA mismatch intrinsic to the haploidentical setting, different strategies have been exploited, including the use of post-transplant immune suppression or ex vivo graft manipulation to obtain an extensive T-cell depletion [13,[43][44][45]. The haplo-HSCT platform offered to the cohort of pediatric patients analyzed in this study took advantage of an innovative strategy based on the selective depletion of αβT lymphocytes (major responsible of GvHD) and CD19 + B cells (responsible of post-transplant EBVrelated lymphoproliferative disorders) [46]. Notably, this procedure allows for overcoming the delayed immune reconstitution, which is typical of CD34 + haplo-HSCT [47]. In fact, the transfer to the recipient of high numbers of HSCs, CD34 − intermediate precursors and of myeloid cells, including monocyte/dendritic cells, can contribute to the low risk of non-relapse mortality (NRM) [34]. Additionally, the presence of mature NK and γδT cells can protect against life-threatening infections and leukemia relapse [13]. Lastly, the lack of any post-transplant GvHD prophylaxis allows the persistence in the circulation of the infused cells. The study of kinetics and function of reconstituting γδT cells in pediatric patients enrolled in the same clinical trial has been already described [48]. Briefly, γδT cells represent the predominant T-cell population during the first weeks after transplantation, derived from cells infused with the graft and expanded in vivo, particularly the Vδ1 cells in patients reactivating HCMV. The γδT cells were shown to exert an anti-leukemia activity, which could be potentiated especially in Vδ2 cells upon ex-vivo exposure to zoledronic

Discussion
Several platforms of allogeneic HSCT can be employed to treat leukemia patients, including the use of different type of donor with diverse HLA matching (HLA-identical sibling, SIB; HLA-matched unrelated volunteer, URD; umbilical cord blood, UCB; or haploidentical), different stem cell source (umbilical cord blood unit, UCB; bone marrow, BM; or mobilized peripheral blood, PBSC), various preparative conditioning regimen (reduced intensity, RIC; or myeloablative, MAC), graft manipulation strategies and pharmacological GvHD prophylaxis [1,42]. All these variables, along with the patient age (pediatric or adult), may influence the post-HSCT NK-cell reconstitution affecting both phenotypic and functional features. For children with high-risk leukemia in urgent need of a transplant, both parents can represent readily available haploidentical donors. However, to circumvent the risk of graft rejection and GvHD related to the high degree of HLA mismatch intrinsic to the haploidentical setting, different strategies have been exploited, including the use of post-transplant immune suppression or ex vivo graft manipulation to obtain an extensive T-cell depletion [13,[43][44][45]. The haplo-HSCT platform offered to the cohort of pediatric patients analyzed in this study took advantage of an innovative strategy based on the selective depletion of αβT lymphocytes (major responsible of GvHD) and CD19 + B cells (responsible of post-transplant EBV-related lymphoproliferative disorders) [46]. Notably, this procedure allows for overcoming the delayed immune reconstitution, which is typical of CD34 + haplo-HSCT [47]. In fact, the transfer to the recipient of high numbers of HSCs, CD34 − intermediate precursors and of myeloid cells, including monocyte/dendritic cells, can contribute to the low risk of non-relapse mortality (NRM) [34]. Additionally, the presence of mature NK and γδT cells can protect against life-threatening infections and leukemia relapse [13]. Lastly, the lack of any post-transplant GvHD prophylaxis allows the persistence in the circulation of the infused cells. The study of kinetics and function of reconstituting γδT cells in pediatric patients enrolled in the same clinical trial has been already described [48]. Briefly, γδT cells represent the predominant T-cell population during the first weeks after transplantation, derived from cells infused with the graft and expanded in vivo, particularly the Vδ1 cells in patients reactivating HCMV. The γδT cells were shown to exert an anti-leukemia activity, which could be potentiated especially in Vδ2 cells upon ex-vivo exposure to zoledronic acid.
Our present study represents the first in-depth characterization of the post-transplant NK-cell dynamics, through the analysis of phenotypic receptor repertoire, alloreactive NK cell subset, and anti-leukemia activity. One month after αβT/B-cell depleted HSCT, especially when high numbers of NK cells had been infused, we could detect a relatively high proportion of CD56 dim KIR + NKG2A -NK cells, and sizeable alloreactive and/or adaptive subsets, resembling the donor NK-cell phenotypic repertoire ( Figure 2). Remarkably, we also demonstrated their anti-leukemia activity. These findings highly suggest that donor derived NK cells, adoptively infused with the graft, persist in recipient peripheral blood. This represents a remarkable peculiarity of the αβT/B-cell depleted haplo-HSCT. In fact, early after either CD34 + haplo-HSCT [33] and T cell-replete haplo-HSCT [49] (which is based on the infusion of an unmanipulated graft and post-transplant high-dose cyclophosphamide (PT-Cy) followed by other immunosuppressive drugs) NK cells exclusively display an immature phenotype. In the PT-Cy platform, it has been recently documented that infused NK cells first proliferate in response to the high systemic levels of IL-15 and then become sensitive to Cy-mediated killing, resulting in early elimination of all mature NK cells, including the alloreactive subset [49]. Similarly, we observed a dominant CD56 bright KIR − NKG2A + NK cell phenotype in both adult [50] and pediatric (unpublished) patients given PT-Cy haplo-HSCT.
In αβT/B-depleted haplo-HSCT, we speculated that the choice of a donor whose NK-cell repertoire contains alloreactive NK cells represent a relevant advantage to boost anti-leukemia activity. Indeed, in the absence of an "off" signal, these NK cells can lyse target cells expressing ligands of triggering receptors, mainly NCR-L, NKG2D-L, and DNAM-1-L [21,23,33]. Here, we compared the functional capability of different NK cell subsets on the basis of rapid degranulation assays against leukemia blasts as target cells. The use of primary leukemia blasts represents a major technical challenge. Nevertheless, especially in the combination of NK cells from HLA-C1/C2 donors and leukemia from HLA-C1/C1 patients, we could clearly show that sp2DL1 NK cells (i.e., Allo C2 subset) are endowed with the highest effector function. This was observed by analyzing NK cells from both donors and transplanted patients. These findings are consistent with previous functional data obtained with NK cell clones in chromium-release assays [33]. Leukemia cell lines such as K562 and NALM-16, characterized by absent/low HLA class I expression and a wide panel of ligands for triggering receptors, can stimulate NK cell cytotoxicity and degranulation with higher efficiency than that induced by primary leukemia blasts. By comparing the HLA-C1 hemizygous NALM-16 with HLA-C1/C1 primary leukemia blasts as target cells, a marked difference could be detected. Remarkably, sp2DL1 cells exceeded the other NK cells in degranulation capacity even with NALM-16 target cells. Thus, the use of NALM-16 allowed for documenting the superior functional activity of sp2DL1 NK cells isolated from patients at early time points after HSCT from Allo C2 donor. It is conceivable that a similar advantage may be detected also with primary leukemia blasts. While the surface detection of the NKG2D-L MICA and ULBPs, of the DNAM-1-L PVR and Nectin-2, and of the NKp30-L B7-H6 is possible thanks to the use of specific mAb, reagents are missing to detect other NCR-L. In this context, receptor blocking experiments suggested a primary role of NCR (in particular NKp46) in NK-cell recognition and killing of different pediatric primary leukemia. On the basis of these in vitro data, NK cell populations characterized by a large size of the alloreactive subset and by a NCR bright phenotype are likely to exert an optimal anti-leukemia activity. In addition, B/x KIR genotype and high B content value, the presence of KIR2DS1 E/U, and the presence of NKG2C, indicating a receptor repertoire well equipped with activating receptors, were positively considered in our donor selection criteria [4,12,35,37,39,51,52]. Unexpectedly, no significant correlation with LFS was observed considering donor NK alloreactivity, KIR genotype, B-content score and KIR2DS1 E/U [34]. As a general comment, since αβT/B-depleted haplo-HSCT is an extremely successful procedure, with LFS probability around 70%, it is difficult to establish correlations with single biological features. It should be also emphasized that the donors have been selected upfront based on the presence of positive immunogenetic variables, thus already influencing the donor cohort. Therefore, this selection bias may be at least in part responsible of the lack of statistical correlation with clinical outcome. Here, we also evaluated if naïve versus differentiated phenotype, considering the NK cell repertoire in both donors and recipients undergoing αβT/B-cell depleted haplo-HSCT, influences the clinical outcome. Since these characteristics were not included in donor selection criteria, cluster distribution in donors did not deviate from a random choice, as it appears similar to that shown in the pivotal study, describing an unrelated donor cohort [41]. Moreover, this clustering analysis could be considered correct, since a more differentiated phenotype in the transplanted patients was associated with HCMV reactivation, consistent with a previous report in the same transplantation platform [32]. Bjorklund et al. reported that donor NK cell repertoires dominated by naïve NK cells were protective against leukemia relapse in AML and myelodysplastic syndrome (MDS) patients after RIC and HLA-matched graft, focusing on a poor anti-leukemia ability of terminally differentiated NK cells. In our cohort, we could not find any significant correlation between differentiated/naïve NK-cell phenotype, considering both donors and transplanted patients, and relapse-free survival. Stratifying patients by type of leukemia, a trend toward a better clinical outcome could be observed in AML patients receiving cluster naïve graft, while in ALL receiving cluster differentiated graft. Larger cohorts will be necessary to eventually confirm the observed trends, possibly reaching a statistical significance. In another study, Cichocki et al. reported, in patients receiving RIC and UCB, a direct association of the expansion of adaptive NK cells after HCMV reactivation associated with better LFS [53,54]. However, even if we confirmed an impact of HCMV on NK-cell maturation, in our study, we could not identify any clinical correlation between this more differentiated NK cell phenotype in transplanted patients and LFS, also differentiating by the type of leukemia. The different results in our cohort might be related to the use of myeloablative conditioning, the choice of HLA-haploidentical donors, and the presence of mature NK cells in the graft. Since adaptive NKG2C + NK cells usually express a specific iKIR (particularly self-KIR2DL) and lack NKG2A, it is crucial to understand whether HCMV reactivation can trigger or not the expansion of the alloreactive subset. We found a certain degree of heterogeneity regarding the size of alloreactive NK cell subsets in the reconstituted NK cell repertoire, possibly depending on the type of alloreactivity (Table 1 and Figure 3A). During NK-cell development, the first expressed iKIR is KIR2DL2/L3 [55,56], which can frequently associate with NKG2C when early HCMV reactivation occurs. In Allo C1 transplants, sp2DL2/L3 donor-derived NK cells can be sustained. Thus, high percentages of sp2DL2/L3 NK cells were detected especially in patients who experienced HCMV reactivation and showed an adaptive phenotype. In Allo C2 transplants, the size of sp2DL1 NK cells detected in patients was almost always lower than in donors, and HCMV reactivation seemed not to be helpful in its increment or even maintenance. In Allo Bw4 transplant, the size of sp3DL1 is quite low in most donors (lower than 5%). It is possible that an HCMV-related expansion of NK cells expressing NKG2C in association with self-KIR2DL can contrast this type of NK alloreactivity. To address this issue in more detail, it will be necessary to analyze a larger cohort of αβT/B-cell depleted haplo-HSCT and a more precise identification of adaptive features, defined by KIR + NKG2C + CD57 + phenotype and/or lacking FcεRγ, EAT-2, and SYK expression.
The adoptive transfer of mature NK and γδT cells with the graft and their persistence in the recipient as functional effector cells not impaired by any pharmacological post-HSCT GvHD prophylaxis is crucial and can explain the clinical success of αβT/B-depleted haplo-HSCT. So far, to further improve the outcome of αβT/B-depleted haplo-HSCT, two new approaches involving either innate and adaptive immune system have been explored. The use of zoledronic acid for promoting Vδ2 cell differentiation and cytotoxicity against leukemia blasts has been recently tested in a cohort of pediatric patients [57]. A clinical trial, based on post αβT/B-depleted haplo-HSCT adoptive transfer of donor-derived T cells genetically modified with a suicide gene [58], has been conducted with the aim of accelerating recovery of adaptive immunity (NCT02065869). However, no significant influence on NK cell reconstitution has been observed [59] in this latter group. A further attempt to optimize graft engineering in the haploidentical setting might include the depletion of CD66b + polymorphonuclear myeloid derived suppressor cells (PMN-MDSC), which were recently demonstrated to exert an inhibitory effect on NK cells [60]. Finally, the donor selection criteria can be continuously refined, following the increased knowledge on NK-cell mediated GvL effect, as well as on specific features correlating with better clinical outcome in transplanted patients. In fact, more recently, in the evaluation of the KIR genotype, we started to consider the proposed centromeric and telomeric KIR score (ct-KIR score) reported to correlate Cen B/x and Tel A/A donor KIR repertoires with low relapse risk in ALL pediatric patients undergoing allo-HSCT [61].

Patients and Donors
Our cohort refers to 80 pediatric patients with ALL (n = 56) and AML (n = 24) (Table S1), enrolled in the phase I/II study of haplo-HSCT after TCRαβ/CD19 negative selection, from September 2011 to September 2014 with the follow-up until December 2019 [34]. This clinical trial was approved by the Ethical Committee of the Ospedale Pediatrico Bambino Gesù (OPBG, Rome, Italy, TCR αβ haplo-HSCT-OPBG; Prot. n. 424/2011) and registered at ClinicalTrial.gov website (NCT01810120). Written informed consent was obtained from patient's parents in accordance with the Helsinki Declaration. Peripheral blood mononuclear cells (PBMC) from donors and patients at different time points after transplantation, isolated by density-gradient centrifugation, were freshly analyzed and cryopreserved in FBS containing 10% DMSO. Primary leukemia blasts were derived from peripheral blood (PB) or bone marrow of some pediatric patients at diagnosis.

NK Cell Phenotype of Donors and Post-Transplant Patients
Surface phenotype of NK cells from donors and patients after HSCT was analyzed on freshly derived or thawed PBMC by immunofluorescence. Appropriate combinations of specific antibodies were used to identify iKIR and aKIR, and various NK-cell subsets including the alloreactive one, by multi-parametric flow cytometry performed on Gallios flow-cytometer (Beckman Coulter, Brea, CA, USA) or MACSQuant-analyser (Miltenyi-Biotec, Bergisch Gladbach, Germany). All antibodies used are listed in Table S3. Indirect labeling was performed for KIR2DL3 staining using ECM41 (anti-KIR2DL3) mAb followed by FITC-conjugated goat anti-mouse IgM (Southern Biotech, Birmingham, AL, USA). The alloreactive NK cell subsets can be identified using appropriate mAb combinations as described in Table S4. Data were analyzed using FlowJo Version 10 (BD Biosciences, San Jose, CA, USA).
Primary AML, BCP-ALL, and T-ALL blasts collected from PB or BM of pediatric patients at diagnosis after density-gradient centrifugation were analyzed by immunofluorescence and cryopreserved. Five samples that were selected on the basis of KIR-L expression, displayed ≥90% of leukemia blasts, as documented by cytofluorimetric analysis. In addition, leukemias were also characterized for the expression levels of activating receptors ligands (i.e., DNAM-1-L, NKG2D-L, and NKp30-L) and of HLA class-I molecules (Table S2 and Figure S3) by indirect immunofluorescence and cytofluorimetric analysis using specific mAb and appropriate secondary reagents (Southern Biotech). Staining index, used to indicate expression levels, is calculated as the difference between the median fluorescence intensity of cells stained with the relevant mAb and that of the negative control divided by two times the standard deviation of the negative control.
In degranulation assays, 2 × 10 5 PBMC obtained from HSCT donors or transplanted patients, cultured 3-5 days in the presence of IL-2 (600IU/mL), were incubated for 3 h either in the absence or in the presence of 1 × 10 5 target cells with the addition of PE-conjugated anti-CD107a mAb. Thawed leukemia blasts, used as targets, were rested overnight in complete medium at 4 • C before the co-culture with effector cells. Golgi Stop (BD Biosciences Pharmingen) was added after the first hour of incubation. Thereafter, cells were collected, washed in PBS with 2% FBS and 2mM EDTA and stained with anti-CD3, -CD56, -CD107a mAbs, and the appropriate antibody combinations allowing the identification of NK cell subsets (Tables S3 and S5). Samples were analyzed by Gallios flow-cytometer (Beckman Coulter Brea, CA, USA). Data analysis was performed using FlowJo Version 10 (BD Biosciences, San Jose, CA, USA); gating strategy is shown in Figure S6. Data referred as ∆CD107 represent the difference between the % of CD107a + NK cells following target stimulation and the % of CD107a + NK cells after incubation with medium alone.

Unsupervised Hierarchical Cluster Analysis
The NK immunophenotyping data from flow cytometry including the percentages of CD56 bright , NKG2A + , CD57 + , NKG2C + , and KIR + NKG2A − were collected for both donors and recipients at 3-6 months after transplant. The unsupervised hierarchical clustering was performed as previously described by Bjorklund et al. [41] adopting the R package pheatmap (version 1.0.8) setting "Ward" as clustering method and "Manhattan" as distance metric. The data were normalized computing a Z-score normalization and scaled on samples (rows). The results were showed as heatmap with the corresponding annotations on the top.

Statistical Analysis
Three or more groups of paired or not paired samples were analyzed with Friedman or Kruskal-Wallis tests, respectively. Dunn's post-test was used for multiple comparison. Two groups of paired samples were compared with the two-tailed Wilcoxon Mann-Whitney non-parametric test. Graphic representations and statistical analysis were performed using GraphPad Prism 6 (GraphPad Software, Inc. La Jolla, CA, USA). In addition, p-values above 0.05 were considered not significant.
Survival analysis (OS and LFS) was performed using the Kaplan-Meier method (R package survival, functions survfit and survdiff ). The cumulative incidence of relapse was analyzed adopting the Fine and Gray's (FG) implemented in the cmprsk R package.

Conclusions
The present study provides the first in-depth evaluation on the NK-cell receptor repertoire (combining genotypic and phenotypic profiles) in donor/recipient pairs of αβT/B-cell depleted haplo-HSCT whose clinical data have been already reported (clinical trial NCT01810120) [34]. The donor selection process was performed by applying an algorithm based on NK-cell characteristics known to be associated with a greater anti-leukemia activity. While designing this algorithm, we hypothesized that the selection of the more appropriate donor is particularly relevant in the αβT/B-cell depleted haplo-HSCT platform, since NK cells are adoptively infused with the graft. In vitro functional assays against pediatric primary leukemia cells clearly documented that the alloreactive NK cell subset is endowed with the highest anti-leukemia activity, supporting the benefit of selecting haploidentical donors with NK alloreactivity. The evidence of the primary role of NCR in leukemia cell recognition and lysis induction was also shown, with NKG2D and DNAM-1 being minor contributors. In addition, we provided a detailed analysis of NK-cell repertoire on the different stages of reconstitution. We demonstrated that, already one month after HSCT, donor-derived mature and functional NK cells circulate in the recipients' peripheral blood, including the alloreactive and/or adaptive subsets. Conversely, at this early time point, NK cells derived from HSCs display an immature phenotype. At a later stage, through unsupervised hierarchical clustering, we identified a naïve and differentiated pattern of NK-cell phenotype. The occurrence of HCMV reactivation was associated with a more differentiated NK-cell phenotype, variably affecting the size of alloreactive subsets. However, the two different clusters did not identify subgroups with differences in OS, LFS, or relapse incidence, but the investigation in a bigger cohort of αβT/B-depleted haplo-HSCT recipients will be necessary to better elucidate this issue.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6694/12/8/2187/s1, Figure S1: Genetic and phenotypic analyses of alternative donors, Figure S2: Degranulation capacity of NK cells stimulated by NALM-16 cell line, Figure S3: Expression of ligands for activating NK-cell receptors on leukemia cells, Figure S4: Unsupervised hierarchical clustering of NK cell phenotype of donors, Figure S5: Correlation analysis of donor clusters with clinical outcome, stratifying by type of leukemia, Figure S6: Gating strategy to define NK cell subsets in degranulation assays, Table S1: Patient and donor characteristics, Table S2: Characterization of pediatric leukemia blasts and cell line, Table S3: Antibodies used in immunofluorescence and flow cytometry, Table S4: Antibody combinations used to define Alloreactive NK cell subsets, Table S5: Antibody combinations used to define Alloreactive NK cell subsets in degranulation assays. Acknowledgments: This study is dedicated to Alessandro Moretta who sadly passed away on February 2018 at age of 64, while this study was in progress. His seminal discoveries of KIRs and of other important inhibitory and activating human NK receptors, such as NCRs, represent major milestones in Immunology. Important achievements in the cure of high-risk leukemias in the haplo-HSCT setting stem from Alessandro's seminal discoveries. His role in planning and discussing the present study has been invaluable. We sorely miss his continuous scientific insight as well as his humanity, irony, and smile.
Conflicts of Interest: Alessandro Moretta was the founder and shareholder of Innate-Pharma (Marseille, France). The remaining authors declare no conflicting financial interests.