This study included a total of 30 consecutive patients ( Table 1 ) affected by hematologic malignancies who underwent a T-cell replete h-HSCT with post-transplant Cyclophosphamide (16) and their relative donors who were enrolled at the Bone Marrow Transplant Unit, IRCCS Humanitas Research Hospital, Rozzano, Milan, Italy between 2011 and 2016.

Written informed consent was obtained from all patients and donors prior to sample collection, in accordance with the Declaration of Helsinki. Protocol approval was obtained from the Institutional Review Boards of the Clinical and Research Institute Humanitas (Approval 24/18).

The HCMV-I/R was evaluated in the peripheral blood of both donors and recipients before the transplant and monitored weekly in the recipients after h-HSCT by assessing the HCMV viral load through real-time PCR (CMV R-GENE, Argene, Biomérieux). When a viral load greater than 2,000 IU/mL was detected, recipients were considered as HCMV-reactivated. Preemptive therapy (Foscarnet 90 mg/kg or Ganciclovir 5 mg/kg and/or Valganciclovir, twice a day for 2 weeks) was given to reactivated patients with a viral load exceeding 4,000 IU/mL. All patients included in this study were enrolled before the introduction of Letermovir as HCMV prophylaxis.

Peripheral blood samples were obtained from both donors and recipients before h-HSCT, and from the recipients monthly until 1-year post-h-HSCT. Peripheral blood mononuclear cells (PBMCs) were isolated using standardized density gradient techniques (Lympholyte-H, Cedarlane) and frozen in liquid nitrogen according to standard procedures.

To minimize variability, the flow cytometry experiments were conducted in batches on frozen cells. Frozen cells were thawed and stained for 15 minutes at room temperature (RT) with Zombie Aqua™ Fixable Viability Kit (Biolegend) and 20 minutes at RT with fluorescent-conjugated monoclonal antibody mix (16).

Samples were acquired at BD FACSymphony A5 flow cytometer (BD Bioscience).

Flow Cytometry Standard (FCS) 3.0 files were imported into FlowJo software (version 9.9.6., TreeStar) and analyzed by standard manual gating strategy. NK cells were identified among viable single cells as Lineage^neg (CD3^neg/CD14^neg/CD4^neg/CD15^neg/CD20^neg/CD19^neg/CD33^neg/CD34^neg/CD203c^neg/FCϵR^neg) lymphocytes. Two thousand NK cells/sample were imported into FlowJo (v 10.2.0) to perform a biexponential transformation. Data were then exported in Python (v 3.7.3) and subjected to the Cytophenograph pipeline (http://github.com/luglilab/Cytophenograph) (16). The K-value was set at 45. New CSV files were then saved, and data were further analyzed in FlowJo (v 10.2.0). Clusters with a frequency lower than 0.5% were excluded from the analysis. Metaclustering was then performed using the gplots R package and PhenoGraph clusters were visualized using the Uniform Manifold Approximation and Projection (UMAP) dimension reduction technique.

The EndoGRO HUVEC (MilliporeSigma) were cultured in cell flasks precoated with Collagen I (Corning) at 10⁵-10⁶ cells/mL by using the endothelial cell growth medium-2 (EGM-2) (Lonza) containing 5% FBS (Lonza), 1% Penicillin-Streptomycin (Invitrogen), and 1% UltraGlutamine (Lonza). The bacterial artificial chromosome (BAC) clone of the HCMV strain TB40/E expressing enhanced green fluorescent protein (EGFP) (17) was used to infect HUVEC at a multiplicity of infection (MOI) of 1 for 12 hours.

CD158b1b2j^neg NK cells were purified from PBMCs of NR and R recipients at 8-12 months after h-HSCT ( Supplementary Table 2 ) by using a BD FACSAria III cell sorter (BD Bioscience). FACS-sorted CD158b1b2j^neg NK cells were then incubated overnight at 37°C in 5% CO₂ in RPMI medium containing 10% FBS, 1% Penicillin-Streptomycin, and 1% UltraGlutamine (RPMI complete) and supplemented with 100 IU/mL rhIL-2 and 100 IU/mL rhIL-12. CD158b1b2j^neg NK cells were then cocultured with infected HUVECs for 72 hours in RPMI complete medium containing rhIL-2 (200 IU/mL). The images were acquired with the DMi8 Wide-Field Microscope (Leica). For each well, four representative fields were analyzed by counting the number of fluorescent-infected HUVECs by using ImageJ software (NIH).

Interferon-γ (IFN-γ) production was quantified on the supernatant of purified CD158b1b2j^neg NK cells co-cultured for 72 hours in the presence of HCMV-infected HUVEC using the Human IFN-gamma DuoSet ELISA (R&D Systems), according to manufacturer’s instructions.

FACS-sorted CD158b1b2j^neg NK cells were lysed in 49 μL of RLT buffer (Qiagen) containing 1 μL of RNAse inhibitor (Thermo Fisher Scientific) and stored at -80°C.

The MicroRNAeasy KitTM with DNAse (Qiagen) was used to purify total RNA, which was then quantified by a Nanodrop 2000 (Thermo Fisher Scientific). Starting from 0.5 ng of high-quality total RNA with an RNA Integrity Number (RIN) greater than 6, assessed using a 4200 Tape Station (Agilent), libraries were prepared with the SMART-Seq Stranded Kit (Clontech-Takara). Libraries were then multiplexed in equimolar pools and sequenced by using a NextSeq-550 Illumina Platform. At least 60 million 75bp-paired-end reads per sample were generated. Read alignment, differential gene expression, and functional enrichment analyses were performed as previously described (16).

GraphPad PRISM (version 9.0) and R statistical software were used to perform statistical analysis. Different variables among samples were compared by unpaired, two-tailed Student’s t-test. Different variables within the same sample were compared by using a paired, two-tailed Student’s t-test. Results are presented as the mean ± standard deviation (SD) or as the mean ± standard error of mean (SEM) when repeated sampling was considered. Two-sided or one-sided P values were considered significant when P ≤ 0.05.

A total of 30 patients affected by hematologic diseases and receiving a T-cell replete h-HSCT with PT-Cy were enrolled in this study ( Table 1 ; Supplementary Table 1 ). Peripheral blood mononuclear cells (PBMCs) were collected from recipients at different time points after transplant till one year after and from their relative HSC donors (16). HCMV-I/R was observed in 10/30 (33.3%) haplo-HSCT recipients with median onset at +40.5 days after h-HSCT (range 21-116).

Acute GvHD requiring systemic therapy was observed in 6/10 HCMV not-reactivated and 12/20 HCMV reactivated patients with a median day of onset at day 41 (range 19-104). Chronic GvHD was observed in 4/30 patients, three of whom belong to the HCMV reactivated group.

We first performed a longitudinal analysis of high-dimensional flow cytometry data on circulating NK cells. To investigate the differences in NK cell immune reconstitution according to HCMV-I/R, h-HSCT recipients were stratified into two groups, namely either experiencing (R, n=20) or not (NR, n=10) the HCMV-I/R that generally occurs within the first 2-3 months after h-HSCT. NK cell data were next concatenated, analyzed with Cytophenograph pipeline, and visualized using Uniform Manifold Approximation and Projection (UMAP) ( Figure 1A ). These analyses identified 28 clusters of phenotypically distinct NK cells (16). We next focused our attention on four clusters (2, 3, 7, and 8) detectable at high frequencies during the first weeks after h-HSCT prior to HCMV-I/R in NR and R ( Figure 1B ). We observed that the percentages of these NK cell clusters started to decline in R at 3-4 months after h-HSCT, while their frequencies remained high in NR even after 12 months from the transplant ( Figure 1C ). A common phenotype of clusters 2, 3, 7, and 8 was the absence of Killer Immunoglobulin-like Receptors (KIR) CD158b1b2j, low expression of NKG2C, and high levels of NKG2A, NKp30, and NKp46.

The manual gating strategy analysis confirmed that CD158b1b2j^neg NK cells were significantly enriched in NR patients, who also expressed high levels of NKG2A, NKp30, and NKp46 and lower levels of NKG2C ( Figure 2A ).

By combining these markers differentially expressed in R and NR, we identified a novel CD158b1b2j^neg/NKG2A^pos/NKG2C^neg/NKp30^pos/NKp46^pos (KIR^neg) NK cell subset. This NK cell subpopulation was detectable at significantly higher frequencies in NR compared to R at all time points analyzed. Moreover, the percentage of KIR^neg NK cells started to decrease in R after 2 months from the transplant and further declined after the onset of HCMV-I/R. Conversely, circulating KIR^neg NK cells persisted more in NR compared to R even 12 months after the transplant ( Figure 2B ; Supplementary Figure 1 ). Of note, we also found significantly higher amounts of KIR^neg NK cells in the grafts of NR compared to those of R ( Figure 2B ), regardless of their HCMV serostatus and donor age ( Supplementary Figure 2 ).

By assessing the NK cell subset distribution among CD158b1b2j^neg/NKG2A^pos/NKG2C^neg/NKp30^pos/NKp46^pos NK cells, we observed that they mainly belonged to less mature CD56^bright/CD16^neg, CD56^bright/CD16^pos, and CD56^dim/CD16^neg NK cell subsets in both NR and R and at all time points considered (10, 18). Moreover, we demonstrated that CD56^bright/CD16^neg and CD56^bright/CD16^pos NK cells with a CD158b1b2j^neg/NKG2A^pos/NKG2C^neg/NKp30^pos/NKp46^pos phenotype were present at significantly higher frequencies in NR compared to R at 1-2 months and 3-4 months after h-HSCT, respectively. In clear contrast and in line with previous findings (8, 19), our data showed that HCMV-I/R induced a significant expansion of terminally differentiated CD56^neg/CD16^pos NK cells in R, considering both total and KIR^neg NK cells ( Figure 2C ; Supplementary Figure 3A ).

We next focused our attention on CD56^bright/CD16^pos NK cells, a neglected subset of unknown origin, which is present at very low frequency under homeostatic conditions or during post-h-HSCT IR (10, 18, 20). Our data demonstrated that CD56^bright/CD16^pos NK cell frequencies and absolute counts were significantly higher in NR in the first months after the transplant ( Figure 2C ; Supplementary Figures 3A, B ). Phenotypic features of CD56^bright/CD16^pos NK cells place them in an intermediary maturation stage between CD56^bright/CD16^neg and CD56^dim/CD16^pos in term of differentiation markers (CD117, CD127, T-bet, EOMES, and CD57), KIRs (CD158b1b2j and CD158e1e2), activating and inhibitory receptors (NKG2A, NKG2C, NKp30, NKp46, and NKp44), and cytolytic potential (PerforinA and GranzymeB) ( Figure 3 ).

To assess the functional relevance of KIR^neg NK cells in controlling HCMV infection, we investigated their ability to prevent HCMV replication ex vivo (16, 20). To obtain enough purified NK cells that should recapitulate those NK cell clusters highly enriched in NR, CD158b1b2j^neg NK cells were FACS-sorted from NR (n=4) and R (n=4) at 8-12 months ( Supplementary Table 2 ) and co-cultured with HUVEC cells infected with the BAC clone of the HCMV strain TB40/E expressing EGFP for a maximum of 72 hours. Every 24 hours cells were observed with a microscope and the number of green infected HUVEC cells was counted. Our results first showed that CD158b1b2j^neg NK cells diminished the HCMV spread in permissive endothelial cells. Of note, CD158b1b2j^neg NK cells from NR were able to completely suppress HCMV-infection ex vivo ( Figures 4A, B ). Accordingly, our data also demonstrated that CD158b1b2j^neg NK cells purified from NR were able to release higher amounts of INF-γ after co-culture with HCMV-infected HUVEC compared to R ( Figure 4C ).

To better assess the functional properties of our newly disclosed NK cell subset, we investigated the transcriptional profile of purified CD158b1b2j^neg NK cells from NR and R by RNA sequencing. Our analyses identified 1327 differentially expressed genes (DEGs; p-value <0.05), 72 of which had a p-value adjusted<0.05. We next performed a Gene Set Enrichment Analysis (GSEA) and found that CD158b1b2j^neg NK cells from NR were more enriched in pathways associated with oxidative and glucose metabolisms compared to R ( Figure 4D ).