In this cross-sectional monocentric study (approval ethic committee Ref. No. 139/09), 174 donors (64 females and 110 males) were included. Residual peripheral blood samples of hematologically healthy children aged 1 month to 18 years were analyzed (patients were hospitalized, e.g., for cleft lip and palate correction). Inclusion criteria involved no incidence of immunodeficiency or infection (defined as >2 severe infections/year, >8 infections/year, persistent fungal infections, post-vaccinal complications, no evidence of acute bleeding, negative for CRP and normal leukocytes, lymphocytes, and neutrophil granulocytes).

The reconstitution of NK cells and their subpopulation CD56^bright, CD56^int, and CD56^dim was analyzed in n = 74 patients (n = 25 females and n = 49 males) transplanted from 2010 to 2016 (see Table 1). Indications for HSCT were high-risk acute lymphoblastic leukemia (n = 51), acute myeloid leukemia (n = 12), and myelodysplastic syndrome (n = 11). Median age at HSCT was 10.4 years (range: 1.3–24.4 years). Grafts were received from matched family donors (MSD; n = 20), matched unrelated donors (MUD) (n = 39), and haploidentical mismatched family donors with <8/10 HLA matches (MMFD; n = 15). No significant differences in the occurrence of GvHD were available in the different donor groups (MSD: 6/20, MUD: 10/39, and MMFD: 3/15). A second or third transplant was administered to n = 9 and n = 2 patients, respectively, because of relapse after first or second HSCT. Stem cell sources consisted of (bone marrow; n = 48), unmanipulated PBSC (n = 11), and T cell-depleted PBSC (n = 15). Post-transplant aGvHD occurred in n = 31 patients with grades I (n = 9), II (n = 3), III (n = 16), and IV (n = 3), cGvHD in 22 patients. Severe viral infections occurred in 18 patients including primary infection or reactivation with cytomegalovirus (CMV) (n = 8), adenovirus (ADV) (n = 5), and Epstein–Barr virus (EBV) (n = 5).

This study was carried out in accordance with the Declaration of Helsinki and was approved by the Medical Ethics Committee of the Frankfurt University Hospital (Ref. No. #198/16). Peripheral blood was collected within the framework of a post-HSCT routine sampling for clinical follow-up from 2010 to 2016. In engrafted patients, samples were collected starting at day 15, within the first year monthly, within the second year three monthly, and afterward every 6 months until 36 months post-HSCT, respectively. In total, n = 925 measurements of n = 74 patients were included in this analysis.

Depending on the occurrence of unexpected events such as aGvHD (only grades III and IV, grades I and II excluded), cGvHD, and severe viral infections, the study group was partitioned. The patient group without aGvHD/cGvHD or any viral infections was chosen for comparison as control group.

Patients showing a distinct CD56^int population in a sufficient absolute amount post-HSCT were elected for more detailed analysis applying a 10-color flow cytometry.

Natural killer cell subpopulations were analyzed on a FC500 flow-cytometer (Beckman Coulter, Krefeld, Germany) applying a five-color panel to estimate CD3⁺ T cells, CD19⁺ B cell, and CD3⁻CD56⁺ NK cells, including the differentiation into CD56⁺⁺CD16⁻, CD56⁺⁺CD16⁺, CD56⁺CD16⁺⁺ NK cells, abbreviated as CD56^bright, CD56^int, and CD56^dim NK cells, respectively. Absolute cell numbers were estimated from peripheral blood samples in a dual platform lyse-no-wash procedure as described previously (13). In brief, a tube of 100 µl EDTA-peripheral blood was labeled with CD45/CD56/CD19/CD3 tetraCHROME (clones B3821F4A/N901/J3-119/UCHT1) multi-color mAb conjugated with FITC, phycoerythrin (PE), phycoerythrin texas red (ECD), and phycoerythrin-cyanine 5 (PC5). CD16 phycoerythrin-cyanine 7 (PC7, clone: 6607118) was additionally added. For the measurement of T cells including T helper and cytotoxic T cells, we applied the tetraCHROME multi-color reagent CD45/CD4/CD8/CD3 (clones B3821F4A/SFCI12T4D11/SFCI21Thy2D3/UCHT1) conjugated with FITC, PE, ECD, and PC5. All reagents were acquired from Beckman Coulter Immutech (Marseille, France).

Differences in the expression profiles of CD56^bright, CD56^int, and CD56^dim cells were analyzed applying two panels on a Navios™ 10-color flow cytometer (Beckman Coulter, Krefeld, Germany). Panel 1: CD226 = DNAM-1 (FITC; clone: KRA236), NKG2A (PE; clone: Z199), DUMP = CD3&CD14&CD19 (ECD; clones: UCHT1, RMO52, J3-119), CD117 (PC5.5; clone: 104D2D1), CD27 (PC7; clone: 1A4CD27), CD56 [allophycocyanin (APC); clone: N901], CD127 (APC-A700; clone: R34.34), CD16 (APC-A750; clone: 3G8), CD57 [Pacific Blue (PB); clone: NC1], CD45 [Krome Orange (KrO); clone: J.33]. Panel 2: CCR5 (FITC; clone: 2D7), killer cell immunoglobulin-like receptor (KIR) mix (CD158 + CD158b + CD158e1; clones: EB6B, GL183, Z27.3.7), DUMP = CD3 + CD14 + CD19 (ECD; clones: UCHT1, RMO52, J3-119), CD117 (PC5.5; clone: 104D2D1), CX3CR1 (PC7; clone: 2A9-1), CD56 (APC; clone: N901), CD62L (APC-A700; clone: DREG56), CD16 (APC-A750; clone: 3G8), CCR7 (PB; clone: G043H7), and CD45 (KrO; clone: J.33). All antibodies were purchased from Beckman Coulter Immutech except CCR5 (BD Biosciences, Heidelberg, Germany) and CX3CR1 (Biolegend, San Diego, CA, USA). Staining was performed, using 100 µl of peripheral blood for each tube followed by 15 min of incubation at room temperature and erythrocyte lysis applying NH₄Cl reagent (Beckman Coulter, Marseille, France).

Stained Cyto-Comp™ cells were applied to compensate the fluorescence overlap. The flow cytometer fluidic stability and the optical alignment were daily tested using Flow-Check™ Fluorospheres (Beckman Coulter, Krefeld, Germany). For verification, Immunotrol cells (Beckman Coulter) were applied three times a day. Furthermore, we participate in an external quality assessment for the detection of T-, B-, and NK cells (INSTANT e. V.—provider for German round robin test, No 213).

Data evaluation was performed using Kaluza and CXP-software (Beckman Coulter).

Statistical analysis was performed using GraphPad Prism 6 (GraphPad Software, San Diego, CA, USA). The NK subpopulation reference values of healthy children and adolescents were calculated with non-linear exponential regression analysis (equation: one-phase decay, least square fit; function: Y = (Y0 – Plateau) × exp(– K × X) + Plateau). Significant differences between groups were assessed by a non-paired two-tailed Mann–Whitney U test. p-Values <0.05 were regarded as significant and are indicated as *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

The distribution of NK cell subpopulations post-HSCT is divergent from healthy individuals showing a high proportion of CD56^bright cells shifting toward CD56^dim NK cells. To estimate the reconstitution time to a normal CD56^bright, CD56^int, and CD56^dim NK cell distribution and absolute cell counts post-HSCT, we first generated age-matched reference values of NK cell subpopulation frequencies and absolute NK cell counts for healthy children and adolescents (Figure S1 in Supplementary Material). Interestingly, the proportion of CD56^bright, CD56^int, and CD56^dim in early childhood changed from 15, 8, and 78 to 6, 4, and 90% in adults, respectively. Subsequently, we matched the NK cell reconstitution of all patients of the control group without severe events (e.g., GvHD, infections) with the newly generated reference values (Figure 1). The frequency of CD56^bright NK cells is elevated directly after HSCT but reaches the upper reference range following 2 months. However, the levels of CD56^bright appear slightly increased remaining between the 50th and 90th percentile until 12 months post-HSCT (Figure 1A). Surprisingly, the percentage of CD56^int cells matches the reference range within the first 2 months post-HSCT but increases starting from 3 months exceeding the 90th percentile of the reference range following 5 months post-HSCT. After this period, the CD56^int fraction declines and converges to the 50th percentile of the reference values 12 months after HSCT (Figure 1B). By contrast, the CD56^dim fraction remains below the 10th percentile until 8 months and reaches the 50th percentile of normal reference values after 12 months post-HSCT (Figure 1C). Comparable reconstitution profiles are also apparent for absolute values of CD56^bright, CD56^int, and CD56^dim NK cells. However, absolute values after HSCT seem to be lower than NK cells of healthy children (Figures 1D–F).

Patients suffering from aGvHD with grade higher than III show significant differences in the reconstitution of NK cell subpopulations compared to patients without any severe events post-HSCT. The median time of first symptoms of aGvHD was 22 days post-HSCT (range: 13–84). Within the first, second, and third month 68, 89, and 100% of the affected patients showed first signs of aGvHD, respectively. Patients without events show conspicuously higher frequency of CD56^bright NK cells within the first 3 months following HSCT compared to patients suffering from aGvHD with the most significant difference already 15 days following engraftment (p < 0.0001; Figure 2A). This could also be shown for the absolute CD56^bright NK cell amount 1 month after HSCT (p < 0.01; Figure 2D). This tendency was less pronounced when analyzing NK cells in patients with lower grades aGvHD (data not shown). Furthermore, the reconstitution of NK cell subpopulations seems to be delayed in patients suffering from aGvHD. For the CD56^int NK cell population, a displacement in time could be shown, which leads to a longer increase in CD56^int frequency and absolute amount (Figures 2B,E). To reach the 50th percentile of normal reference values, patients with aGvHD need a prolonged reconstitution time taking at least two times longer compared to patients without events. Patients with lower GvHD grades were lying in between (data not shown). Even after 3 years of monitoring, a trend toward higher frequency of CD56^bright and CD56^int NK cells concomitant with lower CD56^dim was seen in patients suffering from severe aGvHD (not significant, Figures 2A–C). These differences in the NK cell development post-HSCT were also found evaluating the absolute amounts of NK cell subpopulations (Figures 2D–F). Noteworthy, patients affected with aGvHD following HSCT also show a reduced absolute amount of CD56^dim NK cells after 3 years post-HSCT (Figure 2F). Analyzing absolute NK cell count (including all three subgroups), we did not see a correlation between patients with and without aGvHD (Figure S2A in Supplementary Material). Analyzing cytotoxic T cells, we detected that patients reaching levels of cytotoxic T cells above 1,500/μl within the first year post-HSCT developed in almost all cases an aGvHD (Figure S2B in Supplementary Material). But this fact was only true for less than 25% of the total aGvHD patient cohort. Furthermore, we evaluated the NK cell regeneration of patients with primary aGvHD grade III or IV that became chronic. Thereby, we detected important differences in the development of NK cell subpopulations from CD56^bright above CD56^int to CD56^dim NK cells. The development from CD56^int to CD56^dim NK cells is delayed for at least 2 years. Furthermore, patients with chronification of aGvHD with grade >III have a markedly elevated CD56^int frequency (Figure 3A), which is also clearly visible in absolute cell count of CD56^int NK cells (Figure 3B).

As part of the patients had low absolute cell counts post-HSCT, a detailed phenotyping of the CD56^bright, CD56^int, and CD56^dim NK cell populations was only possible for an elected cohort of patients. To get an understanding of the function of CD56^int NK cells, we analyzed surface molecules linked to NK cell cytotoxicity, adhesion, and immune regulatory functions (e.g., chemokine and cytokine receptors) and compared the expression of KIRs, CD62L, NKG2A, CD127, CD117, CX3CR1, CD226, and CD57 on all three NK subpopulations. The CD56^dim population showed a higher expression of KIRs, whereas the CD56^int and CD56^bright population did not. However, not all KIRs applied within the mix were expressed with equal density (Figure 4). Regarding the homing receptor CD62L, the CD56^int, and the CD56^bright fraction showed an increased expression compared to CD56^dim population. The expression of NKG2A was highest on CD56^int and CD56^bright cells but bipartite in CD56^dim NK cells. CX3CR1 is involved in adhesion and migration of NK cells and to a small extent higher presented on CD56^dim NK cells. As already described, we could also show that CD57 was only detectable on the CD56^dim subpopulation. Low expression of CD127 (IL7α chain), CD117 (c-Kit), and CD226 (DNAM-1) could be seen on all NK cell subpopulations; however, CD56^bright intend to have higher expression than CD56^dim, whereas CD56^int was always lying in between. In summary, the expression profiles of CD56^int and CD56^bright NK cells were nearly congruent, but differed to CD56^dim cells in KIR, CD62L, NKG2A, CX3CR1, and CD57 expression (Figure 4).

The immune reconstitution of NK cell subpopulations post-HSCT was analyzed in patients without events and patients suffering from ADV (n = 5), EBV (n = 5), and CMV infection (n = 8). Infection was detected by the routine analysis of DNA copies in peripheral blood. Patients with elevated viral load at the day of transplantation were excluded from the study, resulting in a cohort of patients with occurrence of a positive viral load between 30 and 90 days post-HSCT. Interestingly, we observed a slight reduction in CD56 and CD16 expression in patients suffering from viral infection in between day 30 and day 60 post-HSCT measured by mean fluorescence intensity (Figure 5). After viral clearance in most patients, a considerable loss in absolute CD56^dim NK cell count occurred followed by continued regeneration of CD56^int NK cells, which was lower in patients without events post-HSCT on day 150 post-HSCT (Figure 5).