A retrospective analysis to evaluate if KIR B haplotype donors associate with a reduced risk of relapse in patients with haematological malignancies following haploidentical transplantation at the Blood and Bone Marrow Transplant Unit at Hammersmith Hospital ICHNHST

Relapse is a major cause of treatment failure in haploidentical haematopoietic progenitor cell transplant (HPCT) with PTCy. Natural killer cells suppress graft versus host disease and mediate the graft versus leukaemia effect, driven by killer cell immunoglobulin‐like receptors (KIRs). Emerging research suggests that donor KIR genotype may influence graft outcome in haploidentical transplants with varying impacts between patient cohorts. This study investigates whether donors with greater KIR B motifs associate with outcomes such as greater relapse‐free survival (RFS), overall survival (OS), nonrelapse mortality (NRM), acute graft versus host disease (GvHD) and infection. The study cohort included 98 haploidentical donor–recipient (D/R) pairs (myeloablative n = 37, RIC n = 61) with various haematological malignancies, receiving primary T‐cell replete haploidentical HSCT with PTCγ. Following KIR SSO genotyping, donors are categorised into neutral (n = 63) or better and best (n = 35), based on KIR B motif content. Kaplan–Meier and Cox regression survival functions are performed to investigate associations with outcomes. Our results show that the better and best category has significantly poorer RFS (p = 0.013; hazard ratio [HR] 3.16, 95% CI 1.21–8.24: p = 0.018). The greater risk of relapse associated with poorer OS (p = 0.011; HR 2.24, 95% CI 1.18–4.24: p = 0.01) in the better and best category. The competing KIR receptor‐ligand and missing licensing proof models failed to predict transplant outcomes. Here, we show neutral donors associate with favourable outcomes in T‐cell replete haplo‐HPCT with PTCγ after categorisation using the KIR B content model, due to the increased risk of relapse associated with the use of better and best donors.

are performed to investigate associations with outcomes.Our results show that the better and best category has significantly poorer RFS (p = 0.013; hazard ratio [HR] 3.16, 95% CI 1.21-8.24:p = 0.018).The greater risk of relapse associated with poorer OS (p = 0.011; HR 2.24, 95% CI 1.18-4.24:p = 0.01) in the better and best category.The competing KIR receptor-ligand and missing licensing proof models failed to predict transplant outcomes.Here, we show neutral donors associate with favourable outcomes in T-cell replete haplo-HPCT with PTCγ after categorisation using the KIR B content model, due to the increased risk of relapse associated with the use of better and best donors.

| INTRODUCTION
Haematopoietic progenitor cell transplant (HPCT) offers a potentially curative therapy to patients with haematological malignancies.HPCT aims to replace recipient bone marrow with healthy HPCs and to eliminate residual malignant cells.The current gold standard method of HPCT is transplantation involving a fully HLA genotypically identical sibling.However, only 25%-30% of patients have a HLA-matched sibling donor. 1 Furthermore, identification of matched unrelated donors (MUDs) differs significantly between differing races and ethnicities 2 ; therefore, alternative donor options are frequently explored.Haploidentical donors, those who share half of the recipient HLA alleles, represent a source of rapidly identifiable donors for 95% of patients. 3Haplo-HPCT began with poorer outcomes due to elevated rates of graft versus host disease (GvHD) and graft failure. 4The advent of post-transplant cyclophosphamide (PTCγ) has since mitigated such outcomes through the elimination of alloreactive CD4 + T cells. 5dministration of PTCγ on days +3 and +4 results in the elimination of donor-derived T lymphocytes, which direct the graft versus leukaemia (GvL) effect. 6Threeyear overall survival (OS) rates in allogeneic HPCT are between 43% and 51%, with relapse representing a major cause of treatment failure. 7Therefore, to diminish occurrence of relapse, exploitation of alternate immune cell types with alloreactive potential is imperative.
Historically, natural killer (NK) cells are the first PBMC to reconstitute postallogeneic HPCT, irrespective of stem cell source. 8NK cells exist in a heterogeneous population, with the capacity for alloreactivity without inducing GvHD, likely due to the elimination of recipient dendritic cells. 9Elimination of virally infected or malignant cells by donor-derived NK cells influences HPCT outcome.The loss of proliferating T cells in vivo and expansion of regulatory T cells, as a result of PTCγ, creates a therapeutic window in haploidentical transplantation for NK cells to effectively target residual malignant cells. 10K cells do not require previous sensitisation for effector function, regulated by activating and inhibitory receptors.One such family of receptors are killer cell immunoglobulin-like receptors (KIR), which are either activating (aKIR) or inhibitory (iKIR) in nature.The polymorphism of the KIR gene complex rivals that of HLA, and it segregates independently from the HLA genes. 11KIR are expressed in a stochastic and clonal manner, ensuring that one each NK cell clone will express an iKIR specific for and capable of education by self HLA, potentially alongside iKIR specific for nonself HLA.The number of known KIR alleles is constantly increasing, driven by high-resolution sequencing methods.This allelic diversity is critical to KIR function as base alterations at specific sites may alter the target specificity and binding affinity of a KIR molecule to its ligand.An example being KIR2DL1*003 encoding inhibitory receptors specific for C2 epitopes, whereas KIR2DL1*004 is weakly specific for C1 epitopes due to amino acid substitutions within the D2 domain at positions 114, 154, 163 and 182. 12KIR may be divided into haplotypes, the 'A' haplotype considered inhibitory and 'B' activating, based on the presence of inhibitory or activating motifs within the centromeric and telomeric regions.Only four framework genes are common to both: KIR2DL4, 3DL2, 3DL3 and the pseudogene KIR3DP1.NK cell licensing is a process by which tolerance is induced by KIR engagement with 'self' HLA class I molecules, namely HLA-A3/A11, Bw4, C1 or C2 groups as cognate ligands. 12Virally infected or malignant cells typically downregulate the expression of HLA class I molecules, which results in a lack of or impaired binding to iKIR and shifts the NK cell activation threshold towards alloreactivity. 13Additionally, aKIR, which are dominantly inhibited by iKIR and engaged by putative targets on heterologous target cells, further tilt the NK cell reactivity threshold towards an activated state.
Although HLA matching remains most pivotal to donor selection, a number of models have since been developed in order to assess NK cell alloreactivity, all developed with a central tenet that greater cytotoxic potential is beneficial to outcomes of allogeneic HPCT.The KIR ligand-ligand model was designed first, where a greater number of mismatches is favoured and defined as the donor possessing an inhibitory KIR ligand which the recipient does not. 14The receptor-ligand model (RLM) was a step further and employs the definition of donor KIR and 'missing-self' recognition, with the aim to mismatch donor iKIR and recipient KIR ligands to enhance cytotoxic activity. 15The missing licensing proof model is another method currently gaining traction, which antithetically to the receptor-ligand model, proposes that only recipients possessing cognate ligands for which donor NK cells are licenced can achieve a fully licenced GvL effect. 16he most widely utilised KIR model implemented for donor selection is the KIR B content scoring model.HPCT donors possessing KIR B haplotypes are favoured, particularly in myeloid conditions. 17This algorithm employs a scoring system developed by Cooley et al. to categorise patients as neutral, better or best and is provided open access by the European Bioinformatics Institute.The results demonstrated by the use of this algorithm are conflicting, with some indicating greater overall survival (OS), relapse-free survival (RFS) and disease-free survival (DFS) 18,19 whereas others suggest greater KIR B content confers poorer OS, 20 significant nonrelapse mortality 21 and no impact on RFS. 22The impact of donor KIR haplotypes therefore remains controversial, with a paucity of literature investigating this model in haplo-HPCT.Our study retrospectively analyses the effect of the KIR B content model on transplant outcomes in T-cell replete haploidentical transplants using PTCγ at Hammersmith Hospital Blood and Bone Marrow Transplant Unit.

| Study cohort
Ninety-eight retrospective haploidentical transplants as part of the treatment of a haematological malignancy were identified for this study, each separated into donor-recipient (D/R) pairs.Patients with previous autografts or allografts were excluded after an initial D/R pair search.All identified pairs included for this study were transplanted consecutively.DNA from donor-recipient pairs was collected from April 2014 to February 2020 and stored at À20 C. Transplants selected for this study occurred between March 2015 and November 2022 and involved administration of 50 mg/kg PTCγ on days +3 and +4.HLA typing for donor and recipient pairs had been performed at Hammersmith Hospital Histocompatibility and Immunogenetics laboratory by sequence-specific oligonucleotide (SSO), sequencebased typing (SBT) or next-generation sequencing (NGS) methods.Ethics approval was obtained from the NHS Research Ethics Committee (REC reference number: 23/PR/1027).This research was submitted to and approved by the medical director of the Blood and Bone Marrow Transplantation Unit at Hammersmith Hospital.All blood samples and biological data were obtained with prior informed consent.Subjects selected were consecutive without any selection biases.No D/R pairs were excluded from this study due to a lack of genetic material for KIR genotyping.
Patients received either a myeloablative (MAC) or reduced intensity conditioning (RIC) preparatory regimen consisting of combinations of busulfan, fludarabine, thiotepa and/or total body irradiation.Grafts were T-cell replete, without ex vivo manipulation.GvHD prophylaxis for MAC and RIC regimens consisted of PTCγ, ciclosporin (3 mg/kg IV once per day) and short-term mycophenolate motefil (15 mg/kg three times a day) until day +35.Ciclosporin was tapered after 3 months if clinical signs of GvHD were absent.Diagnosis of aGvHD, including late-onset GvHD, was made frequently by evaluation of clinical symptoms relating to pathologies developing in the skin, upper and lower gastrointestinal tract and liver alongside confirmation by histological investigations dependent on the organs in which aGvHD had manifested.Infections from bacteria, fungi and viral pathogens were considered for analysis.Infections were detected by a range of differing methods viral serology, molecular virology, general bacteriology including microscopy, FBC, urea and electrolytes, LDH, calcium, phosphate, blood and urine culture, as well as mouth, nasal and equipment swabs as part of routine screening posttransplant or upon suspicion of infection.CMV PCR was performed twice weekly and EBV and adenovirus PCR once weekly post-transplant for viral surveillance whilst patients received immunosuppression or CD4 + counts reached 300 cells/μL.Diagnosis was accompanied by imaging techniques such as X-ray and magnetic resonance imaging for a limited number of cases.CMV PCR of >2000 copies/mL whole blood was utilised as criteria for CMV infection, confirmed by repeat PCR, tissue isolation or histology.CMV PCR of low positive 1000-2000 copies/mL whole blood was also considered positive for infection and commencement of antiviral therapy if the patient was receiving steroids or symptoms of acute GvHD.A viral load of >400 IU/mL on two consecutive occasions was considered positive for EBV infection and commencing antiviral therapies.Other viral real-time PCR screens for HPCT-associated infections included adenovirus, BK virus, HHV6 virus, Influenza A/B and RSV.Donor and recipient CMV serostatus was determined by the detection of anti-CMV IgG prior to transplant.CMV infection and CMV disease were both included as incidence of infection within this study.EBV serostatus was determined by the detection of anti-EBV IgG and IgM.Antiviral prophylaxis consisted of Letermovir 480 mg PO without concomitant ciclosporin or 240 mg PO with concomitant ciclosporin until day +100 or day +180 in patients considered at high risk for late CMV reactivation, or Valganciclovir 900 mg BD then tapered to 450 mg BD given on day +0.Aciclovir was also used at 400 mg IV injections twice daily.

| DNA extraction
Genomic DNA prior to 2016 was extracted and purified manually from whole blood, using a Nucleon™ Extraction kit (Gen-Probe) according to the manufacturer's instructions or post-2016 using the automated EZ1 Advanced technology (Qiagen) with EZ1 DNA blood 350 μL kits.DNA was quantified by measurement at 260 nm and purity assessed by 260 nm/280 nm ratio with a ND-1000 spectrophotometer (ThermoFisher).

| KIR genotyping and content scoring
KIR genotyping was performed on donor and recipient DNA samples using the KIR sequence-specific oligonucleotide (KIR SSO) genotyping test kit (One Lambda) with half volumes.Samples were stored at À20 C. In a 96-well plate, 2 μL template DNA was mixed with 6.9 μL D-mix (dNTPs, MgCl 2 , buffer), and 0.1 μL Taq polymerase per well and sealed prior to centrifugation and amplification.The PCR protocol consisted of heating to 96 C for 3 min, 5 cycles of 96 C for 20 s, 60 C for 20 s and 72 C for 20 s.This was followed by 30 cycles of 96 C for 10 s, 60 C for 15 s and 72 C for 20 s and then holding at 72 C for 10 min.The product was cooled at 4 C for 10 min in a final step.
In a clean 96-well plate, 2.5 μL of amplification product was incubated with 1.25 μL denaturation buffer for 10 min prior to the addition of 2.5 μL neutralisation buffer to create single-stranded products.Biotinylated probe hybridisation followed by the addition of 2 μL locus-specific bead mixture and 17 μL hybridisation buffer per well prior to heating at 60 C. 1Â streptavidinphycoerythrin prepared from stock solution was added to the product after washing.Repeat wash steps were performed prior to loading onto a Luminex 200 instrument for data analysis using Luminex xMAP technology.Each run was accompanied by a no-template control, with deionised H 2 O replacing template DNA to check for potential DNA contamination.
Proceeding data analysis using HLA Fusion Research version 6.4.(One Lambda) and assignment of KIR haplotypes, donors were categorised into 'neutral', 'better' or 'best' as according to Table 1 using the open access EMBL-EBI KIR B Content Calculator (https://www.ebi.ac.uk/ipd/kir/matching/b_content/).KIR genotypes are denoted as A/A when containing only A haplotypes or B/ when containing at least 1 B haplotype.

| Missing licensing proof model
Donor NK cell licensing status was defined by coexpression of cognate HLA ligands (C1, C2, Bw4 and  1A).Furthermore, choice of conditioning regimen did not impact RFS ( p = 0.34; data not shown).The effect on RFS of the better and best group compared with the neutral group withstood multivariable analysis (HR 3.19, 95% CI 1.18-8.62:p = 0.022).
Overall survival was 61% (60/98) for all patients 3 years post-transplant.Due to poorer RFS, the better and best group associated with significantly reduced overall survival ( p = 0.011; HR 2.24, 95% CI 1.18-4.24:p = 0.01; Figure 1B).The better and best group displayed far poorer 3-year survival rates, with only 16 (36%) (16/35) patients in surviving compared to 70% (44/63) surviving in the neutral donor group.Patients in the better and best group were 2.2 times more likely to be deceased after 3 years following haploidentical HPCT.The effect of the better and best group on OS was conserved after multivariable analysis when EBMT risk score and disease risk status at the time of HPCT were included in a forward stepwise regression (HR 2.19, 95% CI 1.16-4.12:p = 0.016).

| Transplant from donors with greater KIR B content does not impact infection, GvHD and NRM in T-cell replete haplo-HPCT with PTCγ
Following the identification of better and best donors associating with increased likelihood of poorer OS and RFS, assessment for impact upon rates of infection, GvHD and NRM followed using identical statistical methods.Cumulative incidence curves demonstrated no significant difference in rates of infection between groups (p = 0.27; Figure 2A), with only a small increase in hazard ratio observed (HR 1.43, 95% CI, 0.76-2.72:p = 0.271).Overall, 39 patients developed infection in this study (40%) (40/98).Infection or infection-related pathologies were the cause of 23 deaths in this study Overall, 19% (n = 19) of patients developed GvHD within this study, 22% (14/63) patients developed GvHD from neutral donors and five (14%) (5/35) within the better and best donor group with no statistically significant difference in rates of GvHD between donor categories ( p = 0.449; Figure 2B), without observable increase in hazard ratio (HR 0.68, 95% CI 0.24-1.88:p = 0.452).Additionally, 28 (29%) (28/98) incidences of NRM occurred within this patient cohort, 24% (15/63) in the neutral donor group and 79% (13/35) patients within the better and best donor group, with no observable impact by better and best donors on rates of NRM (p = 0.244; Figure 2C).The hazard ratio was not statistically significant (HR 1.55, 95% CI 0.74-3.26:p = 0.248).

| Receptor-ligand and missing licensing proof fail to predict outcomes of T-cell replete haplo-HPCT
Next, we tested models that incorporate HLA genotyping data as well as donor KIR.The RLM and missing licensing proof models were both applied to determine influence on transplant outcomes.Previous studies from Leung et al. 16 and Nowak et al. 17 have shown RLM and missing licensing proof model to effectively reduce relapse risk and increase overall survival, respectively.Our observations did not show improved RFS using RLM mismatch (p = 0.126; HR 0.47, 95% CI 0.17-1.27:p = 0.14; Figure 3A) or increased OS upon licensing proof ( p = 0.518; HR 0.80, 95% CI 0.41-1.57:p = 0.52; Figure 4B).Neither model displayed a significant advantage in any transplant-related outcome, summarised in Table 4.

| DISCUSSION
Currently, the most important factor in donor selection for HPCT is matching of HLA alleles at class I and II. 23n the incidence of haploidentical transplant as the most viable option for treatment of a haematological malignancy, patients are disadvantaged due to a greater degree of HLA mismatching.Therefore, secondary donor characteristics such as donor KIR repertoire, which may provide benefit in the HLA mismatched setting due to donor KIR and recipient HLA interactions, could be important in donor selection.Current selection criteria relating to non-HLA factors: age, male sex, CMV matching and nonparous donors are implemented for selecting a haploidentical donor, yet in some instances do not differentiate a single best donor.The addition of KIR genotyping may further aid in haploidentical donor selection.
Haploidentical donor transplant achieves similar survival rates to that of matched related donor transplants, 24 yet in haploidentical transplant relapse incidence ranges from 22% to 30% 25 and use of an RIC regimen increases rates further. 5Previous studies in matched and unrelated donor transplants have demonstrated that selection of donors with a greater KIR B content score (≥2) can reduce the risk of relapse by up to 50% in myeloid conditions, 17,26 with positive effects on relapse also observed in paediatric ALL. 27Relapse risk was also reduced when using ex vivo T-cell depletion in a haploidentical cohort. 28he previous studies have demonstrated favourable effects of greater KIR B content under certain criteria.However, those aforementioned had not assessed the effect of KIR B content either on a haploidentical cohort or using patients conditioned by the currently widespread Hopkins protocol, involving PTCγ for in vivo T-cell depletion as GvHD prophylaxis.This study differs in that it consists of only a haploidentical cohort receiving PTCγ and focusses on both myeloid and lymphoid conditions.The results demonstrated by this study contrast with those in favour of selecting better or best donors as defined by the content model as they are associated with greater relapse and infection risk, consequently decreasing OS.To our knowledge, this is the first observation of a significant decrease in RFS and OS as a potential result of an increased presence of activating KIR motifs in a haplo-HPCT with PTCγ cohort.However, the results align with those observed by Kröger et al., who attributed reduced relapse risk and better disease-free survival in myeloid conditions to donors carrying a lower number of activating genes after unrelated in vivo T-cell-depleted HPCT. 29is discrepancy could be attributed to a number of factors.Perhaps most prominent is that of PTCγ targeting alloreactive leukocytes.Not only are alloreactive T cells eliminated in the early post-transplant period, but recent studies display proliferating NK cells, which do not express ALDH, are purged by day +5. 30Although NK cells exist in differing subsets, only subsets of mature NK cells express KIR, it is these KIR expressing cells which proliferate predominantly in response to allogeneic HLA class I targets. 31These reports suggest that mature NK cells present in the early post-transplant period are eliminated and cannot impart alloreactivity.
Concurrently, a total of 56% (5/9) of the patients who relapsed in the better group relapsed prior to 100 days post-transplant, whereas only 17% (1/6) relapsed before 100 days in the neutral donor group.This evidences a greater rate of elimination of alloreactive NK cells in the better and best group early post-HPCT.In contrast, donor KIR haplotype A appears to improve T-cell expansion. 20s a result, greater T-cell expansion in the neutral donor group may have protected from relapse within the first 100-day post-HPCT.
Additionally, IL-15 appears to be an essential component not only in NK cell and KIR reconstitution but also to regulatory T-cell expansion. 32Donor-derived CD4 + Foxp3 + regulatory T cells are resistant to elimination from cyclophosphamide and expand following allogeneic HPCT. 33onsequently, a competitive environment between immature NK cells and expanding Treg cells may ensue, hindering donor-derived NK cell maturation.It must be noted that previous in vivo expansion of adoptively transferred NK cells from haploidentical donors in the presence of a high-dose fludarabine and cyclophosphamide leucodepletion regimen has been observed. 34However, this was outside the allogeneic transplant setting and was supported by infusion of IL-2, a molecule that signals through the IL-2/ IL-15 receptor beta chain. 35This signifies the importance of IL-15 and its requirement for in vivo expansion of donor-derived NK cells and resistance to immunosuppression, which may have been lacking in the better and best group.A lack of immunophenotypic analysis, to assess NK cell subsets using CD16/CD56 markers as well as assessment of KIR repertoire expression, is a limitation of this study as well as measurement of serum IL-15 concentrations, which was unavoidable due to its retrospective nature.
Moreover, the KIR B content model can be considered a crude predictor of NK cell alloreactivity.This model overlooks the fundamental role of licensing in NK cell biology and disregards the role of iKIR.In the haploidentical donor setting, there is an increased probability of a missing HLA ligand involved in the licensing of donor NK cells.In the context of allogeneic HPCT, recipients lacking cognate HLA ligands for donor iKIR result in NK cells unable to fulfil licensing and are hyporesponsive, limiting a GvL effect. 36Nowak et al. argue that increasing presence of aKIR-cognate HLA ligand pairs causes hyporesponsiveness to recipient malignant cells in a quantitative manner, which advocates the selection of donors with fewer aKIR-HLA pairs.Inadvertently, this study by use of the KIR B content model likely segregates recipients receiving haplo-HPCT from donors with greater aKIR-HLA pairs (better and best donors) from those with fewer (neutral donors), with results not dissimilar to that of Nowaks' group albeit in differing donor cohorts.This suggests a model that considers both aKIR and iKIR interactions may better aid haploidentical donor selection.
Due to the nature of haploidentical transplantation, a greater proportion of missing licensing proof recipients (26%) were identified than in previous reports using MUD.However, this group displayed similar outcomes to transplants involving D/R pairs with shared licensing ligands.Additionally, this study did not observe a beneficial effect by RLM.These observations may be due to the lack of consideration of aKIR-HLA ligand pairs, where increasing numbers of aKIR-HLA ligand pair mismatches in a haploidentical recipient reduce alloreactive potential.Another cause is this study is exclusively based on KIR and HLA genotyping, lacking functional assessment of NK cell cytotoxic capabilities dependent on the models implemented, as well as immunophenotypic data to determine the stochastic expression of the KIR repertoire following HPCT.It must be highlighted that the p-values calculated in this study are influenced by a small sample size, consisting only of 98 D/R pairs.Our small sample size results from the studies' retrospective nature and introduction of PTCγ for haplo-HPCT only in 2015, limiting the number of haploidentical transplants available for inclusion in comparison with others.Moreover, caution is further advised due to the heterogeneous mix of pathologies in this cohort, where some previous studies demonstrating positive effects of KIR B haplotypes have focussed on fewer target pathologies and utilised matched related donors or MUDs.
The results of this study are in agreement with Schetelig et al., 37 in which transplants involving donors possessing centromeric or telomeric KIR B/B motifs displayed a greater risk of relapse and inferior OS.Furthermore, this group also identified that missing ligand or additive models could not predict outcomes of patients with MDS or secondary AML.In combination with our results, this proposes a necessity to develop models that assess the impact of KIR/HLA incompatibilities as well as functional or licenced KIR post-transplant, alongside NK cell and T-cell reconstitution based on KIR A/A and KIR B/B motif content.
One such recent study attempted to develop an overall wider picture of iKIR engagement and its effects on haplo-transplantation with PTCγ using a count functional inhibitory score. 38They determined KIR B content score had no effect on RFS or OS, but this does not directly obfuscate our observations.This study involved a larger cohort (n = 354) with the majority of transplants involving a bone marrow graft (84.5%) as opposed to G-CSF mobilised PBSC.We speculate that neutral donors outperform better and best donors, perhaps due to greater longevity of NK cell or CD8 + T-cell subpopulations, which Zou et al. have provided some insight into using Boelen et al.'s study. 39It is possible to speculate that neutral donors with KIR A/A gene motifs have a higher probability of engaging iKIR post-transplant and thus could in theory have a higher count functional inhibitory score which associated with superior progression-free survival and OS in this alternative study.Altogether, this demonstrates the importance of developing models with a wider focus on functional KIR rather than the presence or absence of specific genes or haplotypes.
In summary, the results of this study do not favour the implementation of the KIR B content model for the selection of haploidentical donors for in vivo depleted haplo-HPCT.As of yet, a defining role of donor KIR genotype in HPCT outcomes has not been established.Future research focussed on determining NK alloreactivity using cellular assays based on KIR/HLA incompatibilities combined with the presence or absence of KIR genes by molecular tests will further reveal NK cell immunobiology in HPCT.With relapse remaining a major cause of treatment failure, models designed to improve the selection of donors based on KIR genotype must balance the potential of NK cell-mediated GvL through aKIR and iKIR interactions, whilst maximising NK cell and T-cell reconstitution.
AUTHOR CONTRIBUTIONS CPB designed the project, performed KIR genotyping and data analysis and developed the journal manuscript.AA assisted with project design and reviewed the journal manuscript.AH and IL coordinated the transplants utilised for this study.EO and RP reviewed the journal manuscript and authorised this research project.
Unit at Hammersmith Hospital and was recognised as a service development project.All blood samples and biological data were obtained with prior informed consent.

F I G U R E 1
Effects of better and best versus neutral donor groups on transplant outcomes in T-cell replete haplo-HPCT.(A) Univariate probability of relapse-free survival as according to donor categorisation calculated from KIR B motif content with greater KIR B content confer an increased risk of relapse after 3-year post-T-cell-depleted haplo-HPCT ( p = 0.007).(B) Univariate probability of OS is significantly reduced in the better and best group (p = 0.011).F I G U R E 2 Better and best donors do not impact incidence of infection, GvHD or NRM in T-cell replete haplo-HPCT.(A) Cumulative incidence of infection, (B) cumulative incidence of a GvHD between neutral and best donors and (C) cumulative incidence of NRM.Univariate probabilities were not statistically significant.(23%) (23/98) with sepsis occurring in 74% (n = 17) of these deaths.

F
I G U R E 3 Receptor-ligand model-based mismatches do not impact transplant outcomes.(A) RFS according to KIR receptor-ligand compatibility.(B) OS according to KIR receptor-ligand compatibility.(C) Cumulative incidence of infection according to KIR receptor-ligand compatibility.(D) Cumulative incidence of aGvHD according to KIR receptor-ligand compatibility.(E) Cumulative incidence of NRM according to KIR receptor-ligand compatibility.

F
I G U R E 4 Missing licensing proof model, with total NK cell licensing system(s) in donor and recipient (Licensing D(+), HLA ligand(R +)) versus total NK cell licensing system(s) in donor and missing cognate HLA ligand in recipient (Licensing(D+), HLA ligand(RÀ)) fails to predict transplant outcomes.(A) RFS according to missing licensing proof model, (B) OS according to missing licensing proof model, (C) cumulative incidence of infection according to missing licensing proof model, (D) cumulative incidence of aGvHD according to missing licensing proof model and (E) cumulative incidence of NRM according to missing licensing proof model.

Table 2
T A B L E 1 KIR B content definition and scoring.and disease status at the time of HPCT were included in multivariable analysis with KIR B content model score to evaluate OS.Statistical significance was denoted as (p < 0.05) and statistical trend as (p < 0.1).Patients receiving donor lymphocyte infusion (DLI) to treat relapse (n = 2) were censored on the date of infusion due to the presence of infused donor T cells, which may confound NK cell alloreactivity and survival analysis.Additionally, patients having received DLI or mesenchymal stem cells (MSC) were censored in NRM outcome analysis.Donor-recipient pairs characteristics and transplant-associated variables.
Recipients were assessed for the presence or absence of HLA ligand involved in licensing in the donor.Recipients with all HLA ligands present involved in iKIR licensing of the donor were denoted; Licensing D(+), HLA ligand R(+).Recipients lacking any HLA ligands involved in iKIR licensing of the donor were denoted; Licensing D (+), HLA ligand R(À).2.6 | Statistical analyses.Variables established with a pvalue >0.20 independently on transplant outcome were then included in forward stepwise regression to establish hazard ratios.Donor age ≤30 years was included for a multivariable Cox regression analysis alongside donor KIR B content score to determine RFS risk.Disease risk EBMT score neutral donors (52%, n = 33) compared with better and best donors (34%, n = 12).Most patients had myeloid conditions (70%, n = 69), whereas 30% (n = 29) had a lymphoid malignancy.At the time of HPCT, 52% (n = 51) of patients were in CR1, 9% (n = 9) were in CR2, 8% (n = 8) of patients were in PR, 4% (n = 4) wereT A B L E 3Determination of the effect of KIR B content on relapse after haplo-HPCT with PTCγ was the primary goal in this study.Overall, 17 (17%) patients developed relapsed disease.Analysis revealed the better and best group had significantly poorer RFS (p = 0.013; hazard ratio [HR] 3.16, 95% CI 1.21-8.24:p = 0.018).Only eight (13%) Influence of differing KIR models on transplant-related outcomes determined by univariate or multivariable Cox regression.Significantly poorer RFS and OS observed in Better + Best donors using the KIR B content model.Receptor-ligand and missing licensing proof models failed to predict transplant outcomes.
T A B L E 4Note: