High proportions of regulatory T cells in PBSC grafts predict improved survival after allogeneic haematopoietic SCT

Regulatory T cells (Tregs) modulate immune responses and improve survival in murine transplant models. However, whether the Treg content of allogeneic cell grafts influences the outcome in human haematopoietic stem cell (HSC) transplantation is not well established. In a prospective study of 94 adult allogeneic PBSC transplants (60% unrelated; 85% reduced intensity conditioning), the median Treg (CD3+CD4+CD25+FOXP3+CD127dim/−) dose transplanted was 4.7 × 106/kg, with Tregs accounting for a median of 2.96% of CD4+ T cells. Patients transplanted with grafts containing a Treg/CD4+ T-cell ratio above the median had a 3-year overall survival of 75%, compared with 49% in those receiving grafts with a Treg/CD4+ T-cell ratio below the median (P=0.02), with a 3-year non-relapse mortality of 13% and 35%, respectively (P=0.02). In multivariate analysis, a high graft Treg/CD4+ T-cell ratio was an independent predictor of lower non-relapse mortality (hazard ratio (HR), 0.30; P=0.02), improved overall survival (HR, 0.45; P=0.03) and improved sustained neutrophil (HR, 0.52; P=0.002), platelet (HR, 0.51; P<0.001) and lymphocyte (HR, 0.54; P=0.009) recovery. These data support the hypothesis that the proportion of Tregs in allogeneic HSC grafts influences clinical outcome and suggest that Treg therapies could improve allogeneic HSC transplantation.


INTRODUCTION
Allogeneic haematopoietic stem cell transplantation (HSCT) is a curative therapy for many haematological disorders. However, despite continued improvements, HSCT is associated with significant morbidity and mortality. 1,2 In particular, the immunological disparity between donor and recipient can cause graft rejection or GvHD. 3 T-cell depletion and systemic immunosuppressive therapy reduce the incidence and severity of these complications, but delay immune reconstitution and increase the risk of infection. 4,5 Furthermore, these strategies impair the GvL response, increasing the risk of relapse. 6 Reducing the harmful immunological effects while sparing the GvL response could, therefore, improve HSCT outcomes.
Regulatory T cells (Tregs) have an important role in allogeneic HSCT. In animal models, cotransfer of CD4 + CD25 + Tregs and CD4 + CD25 − effector T cells into MHC-mismatched mice with leukaemia prevented lethal GvHD but allowed an effective GvL response. 7,8 In humans, reduced CD4 + CD25 high cells, CD4 + CD25 high FOXP3 + cells and FOXP3 mRNA in blood and tissues have been observed in patients with GvHD, [9][10][11][12][13] whereas several strategies used to treat GvHD have been associated with in vivo Treg expansion. [14][15][16][17] Early phase clinical trials of adoptive transfer of ex vivo isolated and expanded Tregs have also been commenced. [18][19][20] However, few studies have examined the influence of Tregs in HSC grafts with clinical outcomes. In this study, we hypothesised that higher proportions of Tregs (Tregs/ CD4 + T cells) in PBSC grafts are associated with improved haematological, immunological and clinical outcomes following allogeneic HSCT.
(NRM) was defined as death without relapse. Overall survival was from the day of transplantation to death or last follow-up.

Statistics
Categorical variables were tested using the χ 2 test and continuous data using independent t-test or Mann-Whitney test. Correlations were analysed using Spearman's correlation. Backward linear regression was performed keeping variables with Po0.05. Overall survival was analysed using Kaplan-Meier. Cell recovery, GvHD, CMV, relapse and NRM were assessed using cumulative incidence with competing risk (relapse for NRM; death for other outcomes). Continuous variables were categorized using median or quartiles. Multivariate analysis was performed using the Cox proportional hazards (overall survival) or the Fine and Gray method (cumulative incidence). Variables with Po0.10 in univariate analysis were included and the model developed using a backward approach, keeping variables with Po 0.05. As donor (sibling/unrelated) and alemtuzumab were strongly confounded, alemtuzumab was not included in analyses containing donor type. Statistical analysis was performed using SPSS 21 (IBM Corporation, New York, NY, USA) and R 3.0.2 (R Foundation for Statistical Computing, Vienna, Austria).
To test the hypothesis that higher proportions of Tregs in the PBSC grafts are associated with improved outcomes following allogeneic HSCT, the study cohort was divided into two: above and below the median Treg/CD4 + T-cell ratio ( Table 2). All multivariate analysis between Treg/CD4 + T-cell ratio and outcomes was adjusted for significant differences between these groups (donor age and donor type).

Subgroup analysis
Owing to the potential confounding influence of donor type and alemtuzumab, subgroup analysis was performed for the T-replete (no alemtuzumab) and T-deplete (alemtuzumab) transplants ( Table 3). The association between Treg/CD4 + T-cell ratios in the graft above the median with improved neutrophil (P = 0.02), platelet (P = 0.02) and lymphocyte (P = 0.004) recovery remained significant in the T-deplete transplants. Conversely, an NS trend (P = 0.13) towards a lower incidence of acute GvHD (II-IV) with high Treg/CD4 + T-cell ratios was observed in T-replete transplants only. CMV reactivation and disease relapse were not significantly associated with the Treg/CD4 + T-cell ratio in the graft in either subgroup. The association between high Treg/CD4 + T-cell ratios in the graft with improved NRM and overall survival remained in the same direction for both the T-replete and T-deplete transplants, although not reaching statistical significance in these smaller subgroups.

DISCUSSION
Our study of Tregs in PBSC grafts in allogeneic HSCT demonstrates that patients receiving higher proportions of Tregs have substantially better outcomes. Specifically, patients receiving grafts with Treg/CD4 + T-cell ratios above the median had a 3-year overall survival of 75% compared with only 49% in the remainder (P = 0.02). This profound difference in survival was because of a reduction in NRM, with a 3-year NRM of 13% and 35%, respectively (P = 0.02). Our data showing that successive quartiles of increasing Treg/CD4 + T-cell ratios in the graft are associated with progressively better overall survival is consistent with a causal relationship. Furthermore, although not powered for formal subgroup analysis, our data suggest that the association between Treg/CD4 + T-cell ratios in the graft with NRM or survival may be observed both T-replete and T-deplete transplants. What accounts for the improved NRM in patients receiving higher proportions of Tregs? Although less infective-related deaths were observed in patients receiving higher Treg/CD4 + T-cell ratios, larger studies will be required to determine whether this is related to improved neutrophil and/or lymphocyte recovery and/or accounts for lower NRM. Nevertheless, the potential mechanisms accounting for the association between graft Treg/ CD4 + T-cell ratios and haematological recovery are intriguing. In mice, recipient Tregs residing in the bone marrow form a protective niche for transplanted allogeneic HSC, preventing HvG responses. 24 Moreover, donor Tregs may also facilitate donor haematopoiesis and promote faster lymphocyte recovery by increasing HSC proliferation and cell cycling, possibly through the inhibition of conventional CD4 + T cells. 25 Although, to our knowledge, no direct evidence for such an effect has been published in humans, our observational data suggest that Tregs may influence stem cell growth and development after HSCT. High proportions of Tregs in the graft may also aid immune tolerance, improving immune reconstitution during thymicindependent homeostatic expansion of mature donor T cells and/or thymic-dependent T-cell production. [26][27][28] In our study, there was a trend towards less aGvHD and fewer deaths because of GvHD (1 vs 9) in patients receiving grafts with higher Treg/CD4 + T-cell ratios, suggesting that that lower incidence of NRM in this group is due, at least in part, to reduced GvHD-related mortality.
Few clinical studies have examined the effect Tregs in allogeneic HSC grafts on transplant outcomes, although, the published evidence is consistent with our findings. Pastore et al. 29 demonstrated, in 65 allogeneic HSC transplants, that a CD3/Treg ratio in the graft 436 was associated with an increased aGvHD. In an extended study (n = 74), the cohort with graft CD3/Treg ratios o36 had improved 3-year overall survival (65% vs 31%; P = 0.001) and lower NRM (5% vs 75%; P o 0.0001). 30 There were 10 deaths from GvHD in the high CD3/Treg group, compared with just one in the low cohort. Other studies suggest a link between absolute Treg counts in allogeneic PBSC grafts and outcome. Wolf et al. 31 showed an association between low Treg doses in the graft and aGvHD in 58 patients. Similarly, Pabst et al. 32 showed that grafts with less than median Tregs had more aGvHD. 32 These studies are, therefore, broadly consistent with our findings, although we suggest that the ratio of Tregs to other cells (CD3 + or CD4 + T cells) in the graft is more likely to show an association with clinical outcomes than Tregs alone. Our data showing the association of Tregs/CD4 + T cells, but not Treg numbers, with outcomes supports this hypothesis. Furthermore, our findings are consistent with seminal work of Edinger et al. 8 using adoptive transfer of T-cell subsets in mouse models of HSCT to show high Treg/CD4 + T-cell ratios permit a potent GvL response while reducing the clinical impact of aGvHD. 8 The main limitation of our study is the size and heterogeneity of the study cohort and the confounding influence of donor type and alemtuzumab. This detailed phenotypic characterisation of donor grafts was only feasible at one centre, thus restricting the number of patients and dictating the population. However, a larger study would allow clarification of which donor factors influence Tregs in PBSC grafts. Interestingly, in this study, unrelated donors were associated with lower Treg/CD4 + T-cell ratios in the graft. Although the explanation remains unclear, it is speculated that the increased time between harvesting and flow cytometry analysis for the unrelated donor grafts may be associated with preferential loss of Tregs and/or downregulation of Treg phenotypic markers, for example, CD25 and/or FOXP3. A larger study would also address the question of whether the proportion of Tregs in the graft has an association with transplant outcomes across different clinical settings. In particular, it would allow predetermined subgroup analysis to confirm whether the influence of graft Tregs is significant in both T-deplete and T-replete transplants. Given the heterogeneity of our study cohort for diagnosis and conditioning regimen, the influence of donor Tregs on disease relapse also remains uncertain. A larger study may clarify whether higher proportions of donor Tregs in the graft reduce NRM, without adversely impacting on relapse, across different disease settings. Use of methylation-specific quantitative PCR to identify demethylated CpG loci within FOXP3, specific to Tregs, may simplify quantification of putative Tregs in these larger studies, while excluding transiently activated FOXP3 + T cells. 33 If confirmed by future studies, our observations that higher proportions of Tregs in donor PBSC grafts is associated with improved outcomes could have profound implications for HSCT. It may be speculated that selection of donors with higher peripheral blood Treg levels may provide PBSC grafts with higher proportions of Tregs. Alternatively, additional ex vivo isolation of Tregs could be used to increase the proportion of Tregs in the final HSC product. In keeping with this, Martelli et al. 20 reported a study of 43 patients receiving a haploidentical HSCT for acute leukaemia with coinfusion of ex vivo isolated Tregs. In the absence of posttransplant immunosuppression, adoptive immunotherapy with Tregs, produced high levels of engraftment and low incidence of acute GvHD and/or relapse compared with historical controls. Adoptive Tregs were also associated with improved lymphoid reconstitution and immunity to opportunistic infections. Whether these approaches improve survival compared with conventional methods and/or in the setting of HLA-matched transplants remains to be established. In summary, our work shows a robust association between the proportion of Tregs in PBSC grafts and overall survival in allogeneic HSCT. These findings are supported by several other smaller studies. Large-scale, prospective studies of graft composition and outcome are imperative to determine the relationship between Tregs/CD4 + T cells in the grafts with clinical outcomes in different allogeneic HSCT settings. If high proportions of Tregs really are causally associated with lower NRM, it may be possible to manipulate Treg/CD4 + T-cell ratios in the graft or recipient to substantially improve survival.  Abbreviations: Low Tregs/CD4 + = below the median (o 0.0296); High Tregs/CD4 + T = above the median (40.0296). Numbers within parentheses are the percentage of non-relapse mortality only.