Lower incidence of grade II-IV acute Graft-versus-Host-Disease in pediatric patients recovering with high Vδ2+ T cells after allogeneic stem cell transplantation with unmanipulated bone marrow grafts: a prospective single-center cohort study

Gamma delta (γδ) T cells represent a minor fraction of human T cell repertoire but play an important role in mediating anti-infectious and anti-tumorous effects in the context of allogeneic hematopoietic stem cell transplantation (allo-HSCT). We performed a prospective study to analyze the effect of different transplant modalities on immune reconstitution of γδ T cells and subsets. CD3, CD4 and CD8 T cells were analyzed in parallel. Secondly, we examined the impact of γδ T cell reconstitution on clinical outcomes including acute Graft-versus-Host-Disease (aGvHD) and viral infections. Our cohort includes 49 pediatric patients who received unmanipulated bone marrow grafts from matched unrelated (MUD) or matched related (MRD) donors. The cohort includes patients with malignant as well as non-malignant diseases. Cell counts were measured using flow cytometry at 15, 30, 60, 100, 180 and 240 days after transplantation. Cells were stained for CD3, CD4, CD8, CD45, TCRαβ, TCRγδ, TCRVδ1, TCRVδ2, HLA-DR and combinations. Patients with a MRD showed significantly higher Vδ2+ T cells than those with MUD at timepoints +30, +60, +100 (p<0.001, respectively) and +180 (p<0.01) in univariate analysis. These results remained significant in multivariate analysis. Patients recovering with a high relative abundance of total γδ T cells and Vδ2+ T cells had a significantly lower cumulative incidence of grade II-IV aGvHD after transplantation (p=0.03 and p=0.04, respectively). A high relative abundance of Vδ2+ T cells was also associated with a lower incidence of EBV infection (p=0.02). Patients with EBV infection on the other hand showed higher absolute Vδ1+ T cell counts at days +100 and +180 after transplantation (p=0.046 and 0.038, respectively) than those without EBV infection. This result remained significant in a multivariate time-averaged analysis (q<0.1). Our results suggest a protective role of γδ T cells and especially Vδ2+ T cell subset against the development of aGvHD and EBV infection after pediatric HSCT. Vδ1+ T cells might be involved in the immune response after EBV infection. Our results encourage further research on γδ T cells as prognostic markers after HSCT and as possible targets of adoptive T cell transfer strategies.

Gamma delta (gd) T cells represent a minor fraction of human T cell repertoire but play an important role in mediating anti-infectious and anti-tumorous effects in the context of allogeneic hematopoietic stem cell transplantation (allo-HSCT).We performed a prospective study to analyze the effect of different transplant modalities on immune reconstitution of gd T cells and subsets.CD3, CD4 and CD8 T cells were analyzed in parallel.Secondly, we examined the impact of gd T cell reconstitution on clinical outcomes including acute Graft-versus-Host-Disease (aGvHD) and viral infections.Our cohort includes 49 pediatric patients who received unmanipulated bone marrow grafts from matched unrelated (MUD) or matched related (MRD) donors.The cohort includes patients with malignant as well as non-malignant diseases.Cell counts were measured using flow cytometry at 15, 30, 60, 100, 180 and 240 days after transplantation.Cells were stained for CD3, CD4, CD8, CD45, TCRab, TCRgd, TCRVd1, TCRVd2, HLA-DR and combinations.Patients with a MRD showed significantly higher Vd2+ T cells than those with MUD at timepoints +30, +60, +100 (p<0.001,respectively) and +180 (p<0.01) in univariate analysis.These results remained significant in multivariate analysis.Patients recovering with a high relative abundance of total gd T cells and Vd2+ T cells had a significantly lower cumulative incidence of grade II-IV aGvHD after transplantation (p=0.03 and p=0.04, respectively).A high

Introduction
Allogeneic hematopoietic stem cell transplantation (HSCT) is a curative treatment for various malignant and nonmalignant diseases.Delayed immune reconstitution (IR) is a main risk factor for morbidity and mortality in patients undergoing allogeneic HSCT as it is associated with higher rates of relapse, infectious complications and Graft-versus-Host-Disease (GvHD) (1).To further improve transplant-related outcomes in the future, it is critical to identify the factors that influence speed and quality of immune recovery after HSCT.
In the past two decades, more focus has been placed on the reconstitution of gamma delta (gd) T cells as growing evidence suggests that these cells have beneficial effects in the context of HSCT, by mediating innate and adaptive immune responses independent of HLA-antigen presentation and by exerting potent antitumor activity via various receptors e.g.NKG2D or DNAM-1 (2).
While the majority of circulating CD3 lymphocytes carries an ab T cell receptor, only 1-10% in the peripheral blood are gd T cells.The gd T cell receptor consists of g and d chains that are encoded by 6 Vg genes on chromosome 6 respectively 8 Vd genes on chromosome 14 (3).gd T cells are subclassified based on their Vd chain; Vd2+ T cells are the predominant fraction found in the peripheral blood of healthy adults, whereas non-Vd2+ T cells (mainly Vd1+) are primarily found in epithelial tissue like skin or intestines (4,5).
Nowadays, peripheral blood stem cells represent the main stem cell source for HSCT in adults.In contrast, unmanipulated bone marrow grafts among peripheral blood stem cells and cord blood are still frequently used in children.In addition to graft source and graft manipulation, donor selection has an important impact on IR.
In a recent meta-analysis of 11 studies (919 patients) on gd T cell reconstitution after allogeneic HSCT, Arruda et al. reported that high gd T cell counts were associated with less disease relapse, fewer viral infections and higher overall and disease-free survival (6).Most of these studies included only adult patients with partially mismatched related donors (PMRD).There is only few data on IR of gd T cells and especially their subsets in children undergoing allogeneic HSCT from matched unrelated (MUD) or matched related donors (MRD).
In our prospective single-center cohort study, we report on the reconstitution of gd T cells and subsets until day +240 after transplantation in a cohort of pediatric patients receiving unmanipulated bone marrow grafts.We studied the impact of transplant modalities on the IR and the effect of high versus low gd T cells on HSCT outcomes.

Patients
This study includes 49 patients who underwent their first allogeneic HSCT between August 2016 and January 2019 at the Department of Pediatric Oncology and Hematology, Charité-Universitätsmedizin Berlin.The median patient age at HSCT was 7 years (0-19 years).All patients received unmanipulated bone marrow as graft source.The cohort consisted of 29 patients with malignant hematological disorders and 20 patients with various non-malignant HSCT-indications, mostly hemoglobinopathies.Detailed transplant characteristics of our cohort are presented in Table 1.Patients received different conditioning regimens dependent on the transplant indication (Supplementary Figure 1).
For GvHD-prophylaxis all patients received Ciclosporin A (CSA) intravenously twice daily starting from one day prior to transplantation.Patients were switched to oral CSA formulations before discharge from the hospital.In combination with CSA patients received either Mycophenolate Mofetil (MMF) 2x600 mg/ m 2 /d starting on day +1 after HSCT for 30 days or Methotrexate (MTX) 10 mg/m 2 once a day on days +1, +3 and +6.
Serotherapy was administered in 36 cases, 33 patients were treated with anti-thymocyte globulin (ATG) and 3 patients received Alemtuzumab (for detailed information on dosing and administration of serotherapy see Supplementary Figure 1).
In nine patients, immunosuppression with CSA was changed to Everolimus and MMF after transplantation.For one patient, this became necessary because he developed posterior reversible encephalopathy syndrome, while the other patients had acute renal failure.

Ethics
Written informed consent was obtained from all patients or their parents/guardians before HSCT.The study was approved by the local ethics committee (EA2/144/15).

Evaluation
The day of engraftment was defined as the first of three consecutive days with an absolute neutrophil count of at least 500 cells/µl.In patients with leukemia, relapse was defined either morphologically as more than 5% blast cells in the bone marrow or a minimal residual disease of ≥ 1x10 -4 measured by flow cytometry or polymerase chain reaction.Relapse of Myelodysplastic syndrome (MDS) was defined by morphology, cytogenetics, or both.
Acute GvHD was defined and diagnosed according to the modified Glucksberg criteria based on clinical, laboratory and histological findings (7,8).
All patients were screened twice weekly for Cytomegalovirus (CMV), Epstein-Barr virus (EBV) and Adenovirus (ADV) DNA in peripheral blood and ADV DNA in stool with polymerase chain reaction until discharge.After discharge, analysis was performed once weekly.PCR cut-off levels for detection of EBV, CMV and ADV in blood were 550, 300 and 2000 Copies/ml, respectively.A linear range of the copy numbers was provided by the local laboratory between 1000 -2,2x10 8 Copies/ml (EBV), 2000 -3x10 8 (CMV) and 2000 -1x10 8 (ADV).The viral load in positive stool samples was measured semi-quantitatively.

Sample collection
Whole blood samples (3 ml) from all patients undergoing HSCT were collected on ethylenediamine tetra acetic tubes and analyzed the same day at seven different timepoints: once prior to the start of the conditioning regimen and at six timepoints after HSCT on days +15, +30, +60, +100, +180, +240.

Flow cytometry
Flow cytometry was performed using Duraclone technology (Beckman Coulter) and each timepoint was analyzed using the DuraClone IM Phenotyping T cell subtypes panel containing nine conjugated antibodies.Samples were processed according to validated standard operation procedures.All analyses were performed using a NAVIOS flow cytometer (Beckman Coulter).Measurements were performed according to the validated method described previously (9) and analyzed using the FlowJo 10.4.2 software.Cell counts were conducted for cells expressing CD3, CD4, CD8, CD45, TCRab, TCRgd, TCRVd1, TCRVd2, HLA-DR and combinations.

Statistical analysis
Wilcoxon rank test was performed for non-parametric variables, while comparing two outcomes or timepoints.For multiple strata or groups, the Kruskal test was performed.As a measure of effect size Cliffs Delta was computed for variables having two levels and Spearman correlation for continuous variables.When Spearman correlation was computed, the associated Spearman test of correlation was performed (10).Testing for confounded variables was done using a linear model with multiple variables and interaction terms followed by a loglikelihood ratio test with the null hypothesis being a single variable model.A significant log-likelihood ratio test indicates that the variable in question is confounded.
The analysis of outcomes in relation to the qualitative cell counts of interest i.e. high versus low was performed using the Cox hazard regression model via the R survival package (11) in accordance with the EBMT recommendations (12).To test for the statistical significance of the Cox fits, Rank-sum tests were performed.
The stratification of the cohort into patients with high and low cell relative abundance was done by ordering the patients by their time-averaged relative abundance of the cell type of interest, then finding the best cut-off value of this relative abundance that separates the cohort into two groups.This is defined by the largest effect size w^2 comparing the Jaccard distances in the space of the time-averaged relative abundance and the rank of the patient based on it, Supplementary Figure 2.
In all tests, a p-value of 0.05 is considered significant for singlevariable analysis.For multivariate analysis, the p-values were corrected for multiple testing using the Benjamini-Hochberg method ( 13) to get q-values.A corrected q-value of <0.1 is considered significant.
The t-distributed stochastic neighbor embedding (t-SNE) analysis was performed on a maximum of 200K cells randomly sampled from the samples corresponding to each timepoint and each group (grade II-IV GvHD or not).The analysis is performed using FlowJo software, with the coloring of cells corresponding to the gating performed.

Patient outcomes
The median (Interquartile range) follow-up time in our cohort was 737 (474 -873) days.Successful engraftment was achieved in 100% of the patients and none of them experienced graft rejection afterwards.During the follow-up period five patients (10%) died; three of the deaths were treatment-related mortalities (TRM); two patients died from relapse.Patients in the TRM group had severe infectious complications with consecutive multiple organ failure.
Relapse occurred in 13 of 29 cases (44.8%) in the subgroup of patients with malignant disease at a median of 171 days after transplantation.We registered 23 cases of aGvHD (46%).In eight of these patients (16%) higher grade aGvHD (grade II-IV) was diagnosed.The median onset of aGvHD was 17 (15 -31) days after transplantation.Viral infections occurred in 34 cases (69%) and 19 patients (55% of patients with viral infection) experienced infection with more than one virus (CMV, EBV or ADV) during the course of transplantation.A total of 14 patients developed CMV infection (28.5%) at a median of 28 (9 -40) days after transplantation.EBV infection was diagnosed in 25 patients at a median of 50 (42 -108) days after transplant.In 13 patients ADV infection was diagnosed (26%) by positive ADV PCR from a stool sample.In eight of these patients, we also detected ADV replication in the blood.The median onset of systemic adenovirus infection was 24 (16 -43) days after transplantation.

Descriptive analysis of gd T cell immune reconstitution after HSCT
CD3 cells quadrupled between day +30 and day +240 after transplantation.CD8 cell counts exceeded CD4 counts during the first 240 days after transplantation with the lowest CD4/CD8 ratio of 0.3 at day +100 after transplantation.Absolute and relative cell counts of T cells and their subsets are summarized in Table 2.
Counts of gd T cells increased gradually after HSCT until day +240 after transplantation.Between day +30 and day +240 the proportion of gd T cells of total CD3 cells was relatively stable between 4 and 6%.At day +240 the median count of gd T cells (68/ µl) was still lower than published reference values for healthy children ( 14).The Vd2/Vd1-ratio decreased continuously from day +30 to day +240 after transplantation and was < 1.0 after day +100, Figure 1.

Impact of donor source on immune reconstitution
We analyzed the impact of pre-transplant modalities on gd T cell reconstitution.In univariate analysis patients with a related donor (MSD or MFD) showed significantly higher absolute Vd2+ T cell counts compared to patients with a matched unrelated donor (MUD) at timepoints +30, +60, +100 (p<0.0001,respectively) and +180 (p=0.004), Figure 2. In multivariate analyses this result remained significant for timepoints days +30, +60, +100 (q<0.001,respectively) and +180 (q<0.05), Figure 3.The correlation was also highly significant (q<0.001) in a multivariate time-averaged analysis, Figure 4.
The amount of total gd T cells in patients with a related donor was higher at days +30 and +60 (q<0.001,respectively) after transplantation but not in the time-averaged analysis.The only timepoint with an association between elevated Vd1+ T cells and MSD/MFD as donor source was day +60 (q<0.001) after transplantation, Figure 3.In summary, the donor type seems to especially influence Vd2+ subtype reconstitution.
The donor type was also correlated with a faster recovery of the ab T cell compartment in (uni-) and multivariate analysis.Early after transplantation (day +30) CD4 cells, HLA-DR positive CD4 cells, CD8 cells and HLA-DR positive CD8 cells showed faster recovery in patients with MSD or MFD (q<0.001,respectively).At day +60 after HSCT having a related donor was independently associated with higher CD4 (q<0.001) and CD4 HLA-DR positive cell counts (q<0.1), Figure 3.In the multivariate time-averaged analysis only CD4 HLA-DR positive cells showed a positive correlation with MSD/MFD donor status, Figure 4.

Impact of other transplant modalities on immune reconstitution
In a multivariate analysis serotherapy with ATG or Alemtuzumab was independently associated with lower CD4 T cells at days +30, +60 (q<0.001,respectively) and +100 (q<0.05).This association was also highly significant in the multivariate timeaveraged analysis (q<0,001).The same accounted for CD8 counts at day +30 (q<0.001).We saw no correlation between the use of serotherapy and the counts of total TCR gd cells or subtypes.Supplementary Figure 3 shows a subgroup analysis of Vd2+ T cell reconstitution in patients with malignant disease that received BM from a MRD without previous serotherapy compared to patients with sickle cell disease that received ATG before transplant from a MRD.Both groups presented high Vd2+ T cell counts early after transplantation in contrast to patients with a MUD, highlighting serotherapy did not affect gd T cell reconstitution in our cohort.Higher donor age (age as a continuous variable) was associated with lower counts of total gd, Vd1+, Vd2+, non-Vd1/non-Vd2 T cells (q<0,001, respectively), CD8 HLA-DR+ and CD8 cells (q<0,001, respectively) at day +30 and Vd2+ T cells at day +180 (p<0.001).Donor age was also negatively associated with CD4 counts at day +240 (q<0.1) and in the time-averaged analysis (q<0,001), Figures 3, 4.

Death and relapse
There was no difference in cumulative incidence of death between patients with high versus low TCR gd cells and Vd2+ or Vd1+ subsets respectively.In the subgroup of patients with malignant hematological disorders we observed no statistical difference of relapse incidence between the two groups.

Acute Graft-versus-Host-Disease
In univariate analysis patients without aGvHD (grade I-IV) showed higher absolute median Vd2+ T cell counts at day +60 (p=0.048),day +100 (p=0.023) and day +240 (p=0.027)compared to those with aGvHD (data not shown).Patients with higher grade aGvHD (grade II-IV) had higher Vd2+ T cell counts at day +180 (p = 0.046), Figure 5.This result remained significant in multivariate time-averaged analysis (q<0.1), but not at single timepoints, Figures 3, 4. The grade of aGvHD was independently negatively correlated with the Vd2+ T cell counts at day +100 (q<0.1) and in multivariate time-averaged analysis (q<0.05),Figures 3, 4. We observed no difference for total gd T cells and Vd1+ subset.
In a multivariate time-averaged analysis the grade of aGvHD was independently negatively correlated with the Vd2+ T cell count (q<0.05), Figure 4. Patients with a high relative abundance of total gd T cells (p=0.03) and Vd2+ T cells (p=0.04) had a significantly lower cumulative incidence of grade II-IV aGvHD, Figures 6, 7. Patients with and without grade II-IV aGvHD showed a population inversion concerning the Vd2/Vd1-ratio after HSCT.While patients without grade II-IV aGvHD showed a trend towards recovering pre-HSCT Vd2/Vd1-ratio, those with grade II-IV aGvHD had an ongoing decline until day +240 after HSCT, Figure 8.The correlation (in terms of spearman r or Cliffs d) between the T cell populations, transplant modalities and transplant outcomes in a time-averaged approach, with confounder analysis, confounded correlations are indicated with circles, while significant unconfounded are indicated with asterisks.A matched related donor (MSD or MFD) is the strongest transplantation modality correlating positively with gd T cells and its subpopulations.HSCT, Hematopoietic stem cell transplantation; aGvHD, acute Graft-versus-Host-Disease; GI, gastro-intestinal.Cumulative incidence of grade II-IV acute GvHD in patients with a high vs. low relative abundance of gd T cells.aGvHD, acute Graft-versus-Host-Disease.
We observed a lower cumulative incidence of aGvHD of the skin in patients with high total TCR gd cells compared to those with low numbers (p=0.02),Supplementary Figure 4.No association was found for aGvHD of the liver or the gut.Cumulative incidence of grade II-IV acute GvHD in patients with a high vs. low relative abundance of Vd2+ T cells.aGvHD, acute Graft-versus-Host-Disease. aGvHD (no or grade I aGvHD vs. grade II-IV aGvHD).The CD4/ CD8 ratio is smaller in patients with grade II-IV aGvHD and they show a higher proportion of activated HLA-DR positive lymphocytes on days +180 and +240.The fraction of Vd2+ T cells is diminished in patients with higher grade aGvHD while the Vd1+ and non-Vd1/non-Vd2 subset is present in both strata.

EBV infection
We examined the association between virus infection and gd T cell reconstitution.In a first step we compared transplant modalities, HSCT outcomes and IR data of CD3, CD4, CD8, TCR ab, TCR gd, Vd1, Vd2 and non-Vd1/non-Vd2 T cells between patients with and without viral infections.
Patients with EBV infection had a higher median donor age (27.5 vs. 11 years, p=0.024) and more frequently received ATG (88% vs. 45.8%,p=0.004) or had an EBV-positive donor (100% vs. 68.2%,p=0.011) than patients without EBV infection.There was no significant difference concerning the distribution of underlying disease or conditioning regimens between the two groups.Rates of aGvHD and grade II-IV aGvHD were comparable between EBVpositive and EBV-negative patients (p=0.893 and 0.739, respectively).No significant difference was seen in CMV and ADV infection rates (p>0.2,respectively), Table 3.
There was no difference for total gd T cells and Vd2+ T cells in the TCR gd cell compartment at any timepoint, but Vd1+ T cell counts were significantly higher at day +100 and day +180 in patients with EBV infection (p=0.046 and 0.038, respectively), Supplementary Figure 5.This association was even more distinct in the analysis of relative Vd1+ T cell counts (in percent of total TCR gd cells).At timepoints +60 (p=0.036),+100 (p=0.006),+180 (p=0.002) and +240 (p=0.009)EBV-positive patients had a significantly higher percentage of Vd1+ T cells than EBV-negative patients, Supplementary Figure 6.
In multivariate analysis the association between elevated Vd1+ T cell counts and EBV infection remained significant in the timeaveraged model (q<0.1)but not at single timepoints, Figures 3, 4.
As mentioned before, Vd2+ T cell reconstitution was delayed in the whole cohort and especially in patients with grade II-IV aGvHD.We saw the same effect for patients with EBV infection.Vd2/Vd1-ratio decreased during the first 240 days after transplantation.In patients without EBV infection it stabilized and stayed >1, while in patients with EBV infection Vd2/Vd1ratio continuously decreased from day +30 (1.5) until day +240 (0.2), Figure 10.
After stratification of the cohort into patients with high and low relative abundances of total TCR gd, Vd1+ and Vd2+ T cells we registered a lower cumulative incidence of EBV-infection in patients with a high relative abundance of Vd2+ T cells after HSCT compared to those with low Vd2+ T cells (p=0.02), Figure 11.

CMV infection
Analogously, we studied the association of T cell subset reconstitution and the risk of CMV infection after HSCT.Patients with CMV infection had higher mortality (p=0.03) and higher rates of grade II-IV aGvHD compared to patients without CMV infection (p=0.006).The amount of CD45 positive cells in the bone marrow graft was significantly higher in the group without CMV infection (p=0.045).There was no difference concerning counts of CD3 and CD34 positive cells in the grafts and pretransplant CMV status between the two groups, Table 4.
Patients with CMV infection had significantly higher CD8 counts at days +60, +100 and + 180 after transplantation  (p=0.026,p=0.006, p=0.006, respectively).The same accounted for CD3 positive cells at the same timepoints (p=0.03,p=0.012, p=0.005, respectively) while there was no difference at any timepoint for CD4 positive cells.In the gd T cell compartment we observed significantly higher amounts of total gd T cells at day +100 (p=0.038) and of Vd1+ T cells at days +100 and +180 (p=0.038 and p=0.041, respectively), Supplementary Figures 7-10.No difference was seen for Vd2+ T cells.
Relative counts of gd T cell subsets significantly differed between patients with and without CMV infection at day + 180.Counts of Vd1 cells (in percent of total TCR gd cells) were higher in patients with CMV infection (p=0.008), while relative counts Vd2+ T cells were decreased consecutively (p=0.004),data not shown.
In multivariate analysis including the impact of various pretransplant modalities and transplant outcomes CD8 cells were still significantly higher in patients with CMV reactivation at days +100 and +180 (q<0,05, respectively).Total gd T cells and Vd1+ T cells were higher at day +100 (q<0,1, respectively).CD8 counts were also higher at day +180 (p<0.05).In the multivariate time-averaged model the result remained significant for CD8 cells, Figures 3, 4.
After stratification into patients with high or low relative abundance of total TCR gd, Vd1+ and Vd2+ T cells, we saw a lower cumulative incidence of CMV infection in patients with high total gd T cells (p=0.02),Supplementary Figure 11.No difference was observed for patients with a high relative abundance of Vd1+ or Vd2+ T cells.

ADV infection
The group of patients that experienced systemic ADV infection (blood PCR positive) showed a significantly higher rate of aGvHD grade II-IV (p=0.022) and had been hospitalized significantly longer after HSCT than those patients without or with gastrointestinal ADV infection (66 days (60-88 days) vs. 48 days (44-58 days), p=0.009), data not shown.
We found no significant association between ADV infection (systemic as well as only gastrointestinal infection) and the reconstitution of any of the examined T cell subtypes.

Discussion
In the past, various studies reported favorable outcomes in patients recovering with high gd T cells after allogeneic HSCT, highlighting their potent anti-infectious and anti-tumorous abilities.Meanwhile data from pediatric cohorts and especially those with unmanipulated, HLA-matched grafts remain scarce.Despite promising results of transplant settings using haploidentical TCRab/CD19-depleted grafts in children (15), unmanipulated bone marrow grafts remain the standard of care for many transplant indications in the pediatric field.Both strategies differ fundamentally in terms of gd T cell reconstitution due to the high content of these cells in the manipulated grafts.We performed a study on the dynamics and the clinical impact of gd T cell reconstitution in the context of unmanipulated bone marrow transplantation.
One of the challenges in studying IR is defining thresholds for high and low gd T cells.The definitions used in earlier studies differed widely and depended on the timepoint of assessment.Two studies defined high gd T cell reconstitution as reaching a cut-off of 150 respectively 175 cells/µl at two consecutive timepoints in a timeframe of one year post HSCT (16,17).Only about 10% of the included patients reached this cut-off, which bears the risk of an underestimated effect of robust gd T cell reconstitution.A different approach is the use of median gd T cell counts at certain timepoints for stratification.Minculescu et al. recently reported improved overall survival and relapse-free survival with lower incidence of aGvHD in adult patients with above median concentrations of gd T cells at day +56 after HLA-matched, T cell replete stem cell transplantation (18).Other studies used relative cell counts at different single timepoints for stratification (19).Both strategies have their limitations as it is not clear which timepoint is most suitable for stratification because HCST complications occur at different stages after transplantation.Our approach to addressing the problem of stratification was using the time-averaged ratios R = TCR gd/TCR ab, Vd2/TCR ab and Vd1/TCR ab cells.The cutoff for high and low gd T cells and subsets was found using clustering in the space of the ratio and the ranking of the patients based on it.
Few studies investigated the impact of graft sources on the reconstitution of gd T cells.Perko et al. found significantly higher gd T cell counts in patients with matched related donors (MRD) compared to matched unrelated donors (MUD) (17).Eyrich et al. described a subgroup of 12 pediatric patients that received bone marrow grafts from MSD in which some of the patients showed transient gd T cell expansion early after transplantation, that did not occur in the other subgroup with CD34 + selected PBSC grafts from unrelated donors (20).Others reported faster gd T cell reconstitution in MSD/MUD transplantation compared to cord blood (21) or in patients with ab-depleted PBSC grafts compared to CD34 + selection (22).
In our study we investigated not only the correlation between donor source and reconstitution of total gd T cells but also its Cumulative incidence of EBV infection in patients with a high vs. low relative abundance of Vd2 cells.subsets Vd1+, Vd2+ and non-Vd1/non-Vd2 T cells.We could demonstrate that especially Vd2+ T cell reconstitution is severely hampered in patients with non-related donors while patients with MSD or MFD showed fast Vd2+ T cell recovery early after transplantation.This finding was independent of other transplant-related modalities that significantly influenced T cell reconstitution in our cohort like donor age and the use of serotherapy.Although the biological background of this finding needs further research, it bears many implications for transplant design.MSD HSCT remains the preferred donor source for many transplant indications as outcomes including GRFS (composite endpoint of graft-versus-host-disease-free and relapse-free survival), relapse-free survival and non-relapse mortality show better results compared to MUD transplantation (23).As Vd2+ T cells are already known to have important antitumor and antiinfectious capacities, fast reconstitution of these cells might contribute to favorable outcomes in patients with matched related donors.
To what extend immune reconstitution is influenced by the method of stem cell harvest in a T cell replete setting is still unclear.As mentioned above, the study by Minculescu et al. showed superior outcomes in patients with high gd T cells.Their cohort consisted mainly of adult patients receiving PBSC grafts from MRD/MUD donors and in contrast to our cohort, Vd2+ T cells were the predominant subset found in the peripheral blood during the first year after transplantation.Nevertheless, they observed a shift towards Vd1+ subset similar to the distribution in our pediatric cohort during that timeframe (18).If this difference is due to mobilization of peripheral blood stem cells is not clear and further studies are needed to identify the best transplant setting to facilitate fast gd T cell recovery.
We report a lower cumulative incidence of grade II-IV aGvHD in patients recovering with a high relative abundance of Vd2+ T cells.At the same time, we observed an inversed Vd2/Vd1-ratio after HSCT that has been described by other groups for the first few months after transplantation (22).In our cohort the population inversion persisted until day +240 after HSCT.Patients without extensive aGvHD showed a trend towards normalization of the ratio between day +100 and +240, contrary to patients with higher grade aGvHD.This evidence supports a protective effect of gd T cells and especially Vd2+ T cells by sustaining immune homeostasis and thereby avoiding aGvHD in the context of unmanipulated bone marrow transplantation.This hypothesis is supported by findings from pediatric studies with patients receiving TCR ab/B celldepleted grafts with high counts of gd T cells (22).Although patients in this study did not receive GvHD-prophylaxis no severe aGvHD was registered.Still, other reports demonstrated an elevated risk of grade II-IV aGvHD in patients with gd T cellenriched graft composition (24) and several studies in mice reported that donor-as well as host-derived gd T cells can promote aGvHD (25)(26)(27).These conflicting results emphasize the need for further research on the role of the different gd T cell subtypes and the molecular pathways via which gd T cells affect aGvHD pathogenesis.
Viral infections account for another large part of transplantassociated morbidity and mortality.In our cohort we observed a high EBV infection rate of 51%.One explanation of this finding might be the frequent use of serotherapy, mainly with ATG, in 73% of our patients.ATG is known to delay IR after HSCT and increase the risk of EBV infection and EBV-associated post-transplant lymphoproliferative disorder after HSCT (28).
Liu et al. reported a higher incidence of EBV infection in adult patients recovering with low absolute Vd2+ T cell counts at day +30 after haploidentical transplantation (29).In our study we could reproduce the same significant association between EBV infection and reduced Vd2+ counts using the time-averaged ratio R of Vd2/ TCR ab.Our hypothesis, that early robust Vd2+ T cell reconstitution might protect patients from developing EBV infection after HSCT is supported by groups that demonstrated the ability of Vd2+ T cells to recognize and kill EBV-infected cells in vitro (29,30).
Interestingly, we found that absolute Vd1+ T cell counts were significantly increased at day +100 and +180 in patients with EBV infection.This was accompanied by a reversed Vd2/Vd1-ratio in patients with EBV infection between days +60 and +240.Like Vd2+, Vd1+ T cells expand upon activation with EBV-infected cells and skewing towards an oligoclonal Vd1+ T cell repertoire after EBV infection in the context of HSCT has been observed (31).The same study reported that Vd1+ T cells are capable of lysing EBV infected cells in vitro.Taking into consideration that EBV infection in our study occurred at a median of 50 days after transplantation, Vd1+ expansion in patients with EBV infection after day +60 could be a sign of a targeted immune reaction to clear EBV-infected cells.Furthermore, the high rate of EBV infection can serve as an explanation for the inversed Vd2/Vd1-ratio seen in our cohort.Whether Vd1+ T cells detected in the peripheral blood are mobilized from epithelial tissueits natural habitator proliferate upon contact with EBV-infected cells in a different location is not clear.While expansion of Vd1+ T cells after EBV infection is a relatively new finding, several studies have described a possible role of gd T cells and especially Vd1+ subset in CMV infection after HSCT (18,22,32,33).Our findings support the existing evidence, adding data from a pediatric study population.We saw a lower cumulative incidence of CMV infection in patients recovering with a high relative abundance of total gd T cells after HSCT.The Vd1+ T cell subset expanded between day +60 and day +180 in patients with CMV infection, highlighting its capacity to recognize and kill CMV-infected cells (32).Our study cohort was too small to show any association between CMV infection and relapse rates.In 2011 Elmaagacli et al. reported lower relapse rates in adult AML patients who experienced CMV infection and Scheper et al. provided a possible explanation by showing that non-Vd2+ T cells are able to cross-recognize leukemia cells and CMV-infected cells in vitro (34,35).
In contrast to EBV and CMV disease we did not find any significant association between the reconstitution of gd T cell subsets and ADV infection after HSCT.Interestingly, patients with systemic ADV infection had higher rates of grade II-IV aGvHD.This finding is in line with other studies that described aGvHD as a risk factor for ADV infection (36).At the same time, ADV might trigger intestinal aGvHD as hypothesized by a study that found higher rates of intestinal aGvHD in patients that had a positive ADV stool PCR before transplantation (37).Correspondingly, three of four patients in our cohort with intestinal aGvHD also presented with positive ADV PCR in stool and blood.
The results of our study are limited by the single-center design and the heterogeneity of the patient population.Still, it includes a relatively high number of patients for a pediatric HSCT study and all participants uniformly received unmanipulated bone marrow grafts, which is of great importance as graft source has a major impact on immune reconstitution.
In conclusion, we demonstrate that donor choice has a strong influence on gd T cell reconstitution after pediatric HSCT with unmanipulated bone marrow grafts.Especially Vd2+ T cell reconstitution was severely hampered in patients with MUD compared to those with MRD, which can contribute to the understanding of favorable outcomes after MRD transplantation.We report a lower cumulative incidence of grade II-IV aGvHD and EBV infection in patients recovering with a high relative abundance of Vd2+ T cells.This supports a protective role of the Vd2+ subset early after transplantation, a cell population that bears great potential as part of posttransplant adoptive T cell therapy (38,39).First studies are currently performed exploring the potential of allogeneic gd T cell transfer against aGvHD and the relapse of hematological malignancies after HSCT (40).Similar studies using ex vivo expansion and activation of gd T cells are necessary to evaluate the efficacy in pediatric patients.Due to the high variability of gd T cell reconstitution dynamics seen early after transplantation the ex vivo approach seems more promising than in vivo expansion methods.Establishing an effective but safe cell number for reinfusion will be a crucial point of the analysis.
Additionally, we observed expansion of Vd1+ T cell subset in patients with CMV as well as with EBV infection.This new association in the context of EBV infection after HSCT can contribute to the process of understanding the role of gd T cells in anti-virus immunity (41).This opens grounds for future research.

3
FIGURE 3The correlation (in terms of spearman r or Cliffs d) between T cell populations, transplant modalities and transplant outcomes at timepoints (A) day +15, (B) day +30, (C) day +60, (D) day +100, (E) day +180, (F) +240 with confounder analysis; confounded correlations are indicated with circles, while significant unconfounded are indicated with asterisk.A matched related donor (MSD or MFD) is the strongest transplantation modality correlating positively with gd T cells and its subpopulations.HSCT, Hematopoietic stem cell transplantation; aGvHD, acute Graft-versus-Host-Disease; GI, gastro-intestinal.

Figure 9
Figure 9 illustrates the impact of aGvHD on immune reconstitution by using t-distributed stochastic neighbor embedding (t-SNE) dimensionality reduction and clustering of CD3 lymphocytes.Patients are stratified according to the grade of

FIGURE 9 t
FIGURE 9 t-distributed stochastic neighbor embedding (t-SNE) dimensionality reduction and clustering of the CD3 cells, while the color indicates the gating performed on those cells.The cell populations are randomly subsampled from all the stratified samples according to the aGvHD grade status (no or grade I versus grade II-IV aGvHD) to a maximum of 200K events per time point.The counts of the gated T cell subpopulations are shown qualitatively in the sunburst diagrams.

FIGURE 10 Logarithm
FIGURE 10Logarithm of the Vd2/Vd1-ratio stratified over patients with and without EBV infection.HSCT, Hematopoietic stem cell transplantation.

TABLE 1
Patient and transplant characteristics.

TABLE 2
Absolute and relative concentrations of T cells and subsets during the first 240 days after transplantation.

TABLE 3
Distribution of transplant modalities and transplant outcomes between patients with and without EBV infection after transplantation.

TABLE 4
Distribution of transplant modalities and transplant outcomes between patients with and without CMV infection after transplantation.