This study was approved by the Ethics committee of the University Regensburg, Germany (approval no. 02/220) and carried out in accordance to the Declaration of Helsinki. All subjects gave written informed consent in accordance with the Declaration of Helsinki. In this study, data were collected from 2009 to 2012. In sum, 107 patients were included into this study. Patient characteristics are described in Table 1. HLA typing was always performed with the same strategy using high resolution HLA-A, B, DR, and DQ for sibling and additional HLA-C typing for unrelated donors according to the standard of the European Federation for Immunogenetics.

Patient blood samples were collected prior to and following allo-SCT at different time points during regular outpatients visits. The frequency of sample collection was part of the individual follow-up plan for each patient at fixed time points and additionally when specific events occurred during the first 200 days after their treatment by allo-SCT. Exclusion criteria were the development of chronic GVHD or an overlap syndrome with signs of both acute and chronic GVHD.

Peripheral blood mononuclear cells of each sample were freshly isolated by density gradient centrifugation (Pancoll human, Pan-Biotech) and within 24 h analyzed by flow cytometry (FACS). Samples of healthy controls were processed in exactly the same way as patient samples. The following antibodies were used: CD3-FITC (SK7), CD16-PE (B73.1), CD14-PerCP (MΦP9), and CD56-APC (N-CAM 16-2), all from BD Biosciences. The focus of this analysis was on immature CD56^high and mature cytotoxic CD56^dim NK cells. At each time point, patients were considered to be “with aGVHD” or “without aGVHD” based on clinical findings. The evaluation of aGVHD was performed on a weekly basis applying the modified Keystone and NIH criteria (19, 20). All patients suffering from aGVHD, at least at one sample drawing, were included in the aGVHD group, those without any signs of aGVHD at any sampling time point were included in the group without aGVHD.

First, patient characteristics were summarized (Table 1) and the Fisher-Exact test or Mann–Whitney test was used as appropriate for comparisons. Classification of GVHD was performed according to the NIH consensus criteria (20). Additionally, the cumulative incidence (CI) of aGVHD has been assessed considering relapse and death without relapse as competing event. In the same way, we analyzed the CI considering the grade of aGVHD.

With the aim to perform a longitudinal analysis of the NK cell numbers associated with aGVHD development and severity, a mixed effect regression model with a linear spline model was used. The reconstitution profile analyzed in this study comprised the dynamics of absolute values of NK cells, CD56^high cells, CD56^dim cells, and the ratio CD56^dim cells:CD56^high cells within the first 200 days after allo-SCT. Patients were previously classified into two groups according to their clinical and laboratory findings of aGVHD as described before. All measurements from patients who did not develop any signs of aGVHD at any collection time point measured during 200 days following allo-SCT were included in the longitudinal analysis. For patients who developed aGVHD during the 200 days period after allo-SCT, data measured from allo-SCT until achievement of aGVHD remission were included in the analysis. For each NK subset, the incidence of aGVHD and the severity of aGVHD were considered separately to be dependent variables. Furthermore, to evaluate the effect of steroid administration on NK cell reconstitution, the aGVHD patients were classified regarding GVHD stage and steroid treatment (ST) at each observation. Afterward, linear spline regression models were fitted separately for each NK cell subpopulation.

To optimize the residual distribution of the regression analysis, cell counts of NK subsets and the analyzed ratio were previously log₁₀-transformed. Note that the statistical regression methods used here account for repeated measurements at different time points.

All statistical tests were two-sided with a significance level of 5% representing the 95% confidence interval. Data analysis was performed using the R software for statistical computing, version 3.1.1 (R Foundation for Statistical Computing, Vienna, Austria, http://www.R-project.org/). Figure 3 was generated using GraphPad Prism 6 software, version 6.0.4.

A total of 342 samples have been analyzed from 107 patients (78 prior to SCT, 264 after SCT), 37 (35%) females, and 70 (65%) males, were included in this study. Fifty-five (51%) patients suffered from acute leukemia, 19 (18%) from multiple myeloma, 12 (11%) from MDS, 10 (9%) from lymphoma, 6 from chronic leukemia (6%) and 5 from other diseases (5%), as shown in Table 1. The majority of the patients underwent HLA-identical allo-SCT (n = 88, 82%). Thirty patients had related sibling donors (28%), 77 had unrelated donors (72%). The stem cell source was peripheral blood stem cells in 98 patients (92%), bone marrow (BM) in eight patients (7%), and cord blood in one patient (1%). A reduced-intensity conditioning (RIC) regimen was administered in 86 patients (80%), and 21 patients were treated using myeloablative regimens (20%). Fifty-eight patients received anti-thymocyte globulin treatment (54%). GVHD prophylaxis consisted of CSA + MTX in 84 patients (79%). Others received immunosuppressive monotherapy or individual combination treatments (see Table 1). While 45 (42%) patients did not show signs of aGVHD, 62 (58%) patients developed aGVHD (grade 1: n = 22, grade 2: n = 21, grade 3: n = 14, grade 4: n = 5) at a median of day 28 (11–182) after allo-SCT. Symptoms that led to the classification of aGVHD and/or cGVHD were defined according to the NIH criteria as recently published (20).

No major difference in age, underlying disease, donor type, stem cell source, HLA match, conditioning regimen, and GVHD prophylaxis between the GVHD group and the group of patients without GVHD was observed. The only parameter contributing to a significant difference between the two mentioned groups was gender with a higher proportion of male patients in the GVHD group (p = 0.013).

The median age of the GVHD group was 52.7 years (17.5–70.1) and of the group of patients without signs of aGVHD 50.3 years (24.3–70.7). In both groups, the majority of patients suffered from AML (49%). Additional underlying diseases are indicated in Table 1.

The median survival of the group that did not develop aGVHD was 34.1 months (ranging from 0.9 to 93.4 months), whereas the median survival of the GVHD group was 21.6 months (ranging from 1.2 to 65.7 months). The estimated CI of overall aGVHD was 57.9% (95% CI: 48.5–67.30%); the estimated CI for aGVHD grade I was 18.88% (95% CI: 11.42–26.34%), 19.73% (95% CI: 12.2–27.38%) for aGVHD grade II, 13.19% (95% CI: 6.75–19.64%) for aGVHD grade III, and 3.76% (95% CI: 0.68–8.78%) for severe aGVHD grade IV (Figure S1 in Supplementary Material). A critical phase with very high numbers of incidences occurred between days +25 and +50 after allo-SCT.

A density plot gated on all CD56⁺CD3⁻CD14⁻ NK cells further distinguished two major subsets including immature CD56^highCD16^dim (hereinafter referred to as CD56^high) and mature cytotoxic CD56^dimCD16^high NK cells (hereinafter referred to as CD56^dim). In healthy adult controls, only 12.5 ± 9.6% of all NK cells belong to the immature CD56^high NK population whereas the majority of 83.1 ± 12.2% represents mature CD56^dim NK cells that coexpress CD16 (data not shown). Importantly, in our study cohort, there was no significant preexisting difference in NK cell subset distribution prior to allo-SCT among patients that later developed GVHD or not (Figure S2 in Supplementary Material). The appearance of immature CD56^high NK cells that do not yet express the Fcγ-receptor CD16 on their surface is indicative of an efficient immune reconstitution post-allo-SCT. In this study, following allo-SCT, the immature subset of CD56^high NK cells was highly increased and reached up to 45% of all NK cells in patients without aGVHD, but was significantly reduced in patients who developed aGVHD (Figure 1).

Next, we addressed a possible impact of aGVHD on the NK cell numbers and early NK cell reconstitution throughout an observation period of 200 days after allo-SCT. Patients without any signs of aGVHD during the whole observation period had considerably higher NK cell numbers than patients who developed aGVHD, with significant differences in the CD56^high NK cell subset (Figure S2 in Supplementary Material). In contrast, there was only a trend of reduced numbers of CD4⁺ T cells around week 12 post SCT, but overall numbers of CD3⁺ T cells did not significantly differ in GVHD patients compared to patients without GVHD following allo-SCT (Figure S3 in Supplementary Material).

Moreover, as shown in Table 2, the longitudinal analysis of the logarithmically transformed numbers of the different NK cell subsets revealed a negative association between the occurrence of aGVHD and the early expansion of the total NK cell population (p = 0.06), and in particular the CD56^high NK cell subset (p = 0.009). In contrast, there was no significant correlation between the CD56^dim NK cell subsets or the calculated ratio of CD56^dim:CD56^high NK cells and aGVHD (Table 2). Furthermore, a significant inverse correlation between the severity of aGVHD and the frequency of CD56^high NK cells during aGVHD could be demonstrated by the linear spline mixed effect model (Table 3).

Figure 2 shows the recovery of the CD56^high NK subset in patients without signs of aGVHD (solid black line) compared to patients with aGVHD (dashed black line) and their corresponding 95% confidence intervals. We observed significantly lower numbers of CD56^high NK cells in the aGVHD group, suggesting that aGVHD might cause impaired CD56^high NK cell regeneration in the early phase after allo-SCT.

Furthermore, the reconstitution of the CD56^high subpopulation was analyzed with regard to the severity of aGVHD. We observed a significant correlation between the grade of aGVHD and the degree of disturbance of CD56^high NK cell reconstitution (Figures 2 and 3). There was almost no difference in NK cell immune reconstitution between patients with grade I GVHD and those without GVHD. In contrast, patients with severe grade of aGVHD (III–IV) showed lower numbers of CD56^high NK cells in comparison to patients with milder forms of aGVHD. In summary, not only the occurrence but also the severity of aGVHD correlated with a delayed reconstitution of the CD56^high NK cell subset.

To elucidate the impact of steroid treatment (ST) on NK cell recovery, the frequencies of the different NK cell subpopulations have been calculated for all patients who developed aGVHD at least at one sampling time during the entire observation period of 200 days (Figure 3). All data on the NK cell numbers measured in these GVHD patients have been classified and analyzed according to the stage of disease at the given time point, as being obtained either before or during development of GVHD and with or without ST. Reasons for ST without occurrence of GVHD were in most cases inappetence, nausea, gastroenteritis, and weakness treated with moderate dosing of prednisolone. Only few patients received high dose prednisolone due to, e.g., a severe capillary leak, vasculitis, or ARDS. GVHD patients were also treated with prednisolone (1 mg/kg BW).

Figure 3 and Table S1 in Supplementary Material show the results of the linear regression model with smoothing spline performed with the above mentioned disease and treatment classifications of all GVHD patients. All results are shown in relation to measurements at time points without signs of aGVHD and without ST that served as base line (=100%) of each individual patient. In sum, we observed an additive impact of steroids and aGVHD on reconstitution of NK cell subsets. Of note, based on these analysis, we can exclude that the before mentioned negative impact of aGVHD on the CD56^high subpopulation recovery could be completely explained by ST.