Bone marrow and blood demonstrate distinct immune reconstitution patterns and correlations with relapse posttransplant Open Access

T-cell immune reconstitution (IR) correlates with clinical outcomes after allogeneic hematopoietic cell transplantation (allo-HCT).¹ Innate immune populations including natural killer (NK) cells and innate lymphoid cells (ILCs) may also have important roles after transplant,^2-6 and NK cells are typically the first lymphocytes to emerge in the peripheral blood (PB) of transplant recipients.⁷ The bone marrow contains the tumor microenvironment for many hematologic malignancies and is therefore a potentially critical site for alloimmunity. However, much of our understanding of IR relies on analyses of PB,^8-11 and insights into the posttransplant recovery of lymphocytes within the marrow are limited.

In addition to providing a microenvironment for benign and malignant hematopoiesis, the marrow is a site of early lymphoid development, with orchestrated patterns of egress to and migration from the periphery. As such, we hypothesized that the kinetics and phenotypes of lymphocyte IR may be site specific and that characterization of marrow immunity could elucidate factors associated with clinical transplant outcomes. Thus, we evaluated IR patterns of innate and adaptive lymphocytes in longitudinal marrow and PB samples from recipients of allo-HCT for hematologic malignancies to determine whether the compartments were comparable or whether they provided distinct insights to lymphocyte recovery and clinical outcomes after transplant.

We prospectively profiled marrow aspirate and PB samples starting before transplant and continuing up to 1 year after transplant in 33 adults undergoing allo-HCT as treatment for hematologic malignancies with marrow involvement. Transplants included diverse representation of graft sources, conditioning regimens, and graft-versus-host disease (GVHD) prophylaxes to maximize clinical relevance (supplemental Table 1). Clinical chimerism status is included for available samples in supplemental Table 2.

To optimize identification of low-frequency populations such as ILCs in samples with low lymphocyte content early after transplant, we performed flow cytometry on fresh blood/marrow samples using a single condensed panel (supplemental Figure 1). Analyses quantified the distribution of innate and adaptive lymphocytes and selected phenotypic markers.

We selected bone marrow aspirates from patients (9 and 26) with viably frozen cells at day 100 after transplant who received sex-mismatched allografts (unmodified PB stem cell [PBSC] allografts from HLA-matched related donors) for single-cell RNA sequencing (5′ gene expression, 10× Genomics) after fluorescence-activated cell sorting for CD45⁺ lymphocytes. Sequencing data were processed using Cell Ranger v8.0.0. CellBender¹² (v0.3.0) was used to generate denoised expression matrices. CellBender was specifically implemented to avoid ambient RNA leading to false chimerism assignments in downstream analyses. The resulting expression matrices were imported into AnnData objects for further analysis using Scanpy¹³ (v1.10.1) in Python (v3.12.1). Quality control filtering excluded cells with >30% mitochondrial gene expression content. Expression counts were normalized and log scaled. Using the top 4000 highly variable genes, the first 30 principal components were calculated. Phenograph¹⁴ clustering using the Leiden method (k = 30, resolution_parameter = 0.5) was applied to identify cell clusters. Cell-type identification was performed by examining differential gene expression patterns and evaluating the expression of established cell-type marker genes. For chimerism analyses, we examined the expression of genes RPS4Y1 (Y-chromosome) and XIST (X-chromosome inactivation). We then used Souporcell¹⁵ (v.2.5) to attribute cells to genotypes within the sample based on clustering of single-nucleotide variants.

All statistical analyses were performed in R.¹⁶ Wilcoxon signed-rank tests were performed for comparisons of populations measured by flow cytometry. The significance level of all statistical tests was set at 0.05. A description of all statistical tests performed is included in each figure legend. A comprehensive list of all statistical comparisons is included in the supplemental Tables associated with each figure (supplemental Tables 3-9).

For additional methodologic details, see the supplemental Material.

To compare IR in the bone marrow and PB after allo-HCT, we profiled innate and adaptive lymphocyte subsets beginning before transplant and then continuing after transplant (Figure 1A) (n = 110 marrow and n = 115 PB samples, respectively) in 33 adults at our institution (supplemental Table 1). Distinct patterns of IR were observed in the marrow vs PB across major lymphocyte subsets (T cells, NK cells, and ILCs) as a percentage of lymphocytes (Figure 1B). NK cells (including CD56^bright and CD56^dim subsets) emerged early after transplant in the marrow, akin to known patterns in PB.⁷ However, at day 100, when T-cell frequencies matched those of NK cells in PB, the NK cells remained more abundant than T cells in the marrow (Figure 1B). Furthermore, after day 100, T cells became more dominant in PB but not the marrow, where similar T and NK cell frequencies persisted (Figure 1B-D; supplemental Table 3). Although NK cells were particularly abundant in both the blood and marrow of CD34-selected graft recipients early after transplant, the persistent high frequency of marrow NK cells was observed across graft sources (Figure 1C-D).

Within the T-cell compartment, regulatory T cells (Tregs) represented a higher frequency of lymphocytes in PB than the marrow (Figure 2A). This was likely caused by the greater overall T-cell abundance in PB than the marrow, given that Tregs accounted for a similar proportion of CD4⁺ and CD3⁺ T cells in the 2 tissues (supplemental Figure 2C-D). Nonetheless, given that Tregs can affect the function of non-T lymphocytes such as NK cells,¹⁷ this relatively low proportion of Tregs among marrow lymphocytes could be biologically and clinically meaningful.

Tissue-resident ILCs are thought to contribute to tissue recovery in experimental transplant models, and ILC frequency in PB has been shown to correlate with a reduced risk of clinical GVHD in recipients of allo-HCT for leukemia.^3-5 Here, ILCs were detectable in both the marrow and PB. Although rare, ILC frequency was greater in the marrow than in the blood (Figure 2B). Assessing the distribution of major ILC subsets, ILC2s were the least abundant in both the marrow and PB, whereas ILC3s seemed relatively more frequent in the marrow (Figure 2C), aligning with previous studies.¹⁸ 

We hypothesized that, in addition to frequency differences, lymphocyte phenotypes may also differ in the marrow and blood. To investigate this, we focused on CD69 and CD25 because both are upregulated upon T-cell activation and CD69 can also serve as a marker of tissue-resident lymphocytes.^19,20 Accordingly, CD69 expression seemed to correlate with the location cells were isolated from, being more highly expressed by most marrow lymphocytes than by their circulating counterparts, although ILCs expressed CD69 relatively robustly even in PB. In contrast, although CD25 expression did vary, its expression was generally more consistent among corresponding lymphocytes in the marrow and blood (Figure 3A; supplemental Table 8).

Regarding the role of allograft type in patterns of marrow IR after transplant, there were no marked differences in overall T-cell frequencies in the marrow by graft source, even including CD34-selected grafts (supplemental Figure 3A). CD4-to-CD8 ratios in the marrow tended to reflect what was observed in the blood for recipients of mobilized PBSC, CD34-selected, and umbilical cord blood allografts, although on day 100 after PBSCs the marrows skewed toward slightly more CD8s whereas the blood skewed toward slightly more CD4s, suggesting an earlier shift from CD4 to CD8 predominance within the marrow. For cord blood transplant (CBT) recipients, consistent with previous analyses of their PB,²¹ circulating T cells skewed heavily toward CD4s early after transplant, and a similar trend was observed in the marrow. After CD34 selection, the marrow and blood indicated similar trends toward CD8 predominance (Figure 3B).

We next evaluated whether marrow lymphocyte profiles could correlate with relapse of malignancy after transplant. PD1 expression was of particular interest because it can reflect either T-cell activation or T-cell exhaustion in certain contexts, and previous work has correlated PD1 elevation with activated effector function²² and with exhaustion^23-25 after transplant. We quantified T-cell PD1 expression at day 100, before any relapses in this cohort, and found a significantly higher PD1 expression on marrow T cells from patients who subsequently remained in remission. This phenotypic correlation was observed in both CD4 and CD8 T cells in the marrow, including conventional and regulatory CD4 subsets. PB T cells showed a similar but nonsignificant trend (Figure 4A; supplemental Figure 3B-C).

Finally, we investigated subset chimerism within the bone marrow by performing single-cell RNA sequencing analysis on CD45⁺ cells sorted from day 100 posttransplant bone marrow samples (n = 37 074 cells). The samples were prepared from 2 patients who received sex-mismatched allografts (9 and 26). After identification of major cell types (Figure 4B; supplemental Figure 4), we applied 2 complementary approaches to define the donor vs host origin of the different lymphoid and myeloid subsets: assessing the expression of X- or Y-chromosome-associated genes (XIST or RPS4Y1, respectively) and identifying the distinct genotypes of each donor/host pair via Souporcell analysis (see “Methods”).¹⁵ We were unable to resolve 2 genotypes for patient 9 because all cells seemed to be of donor origin. However, for patient 26, who demonstrated evidence of T-cell mixed chimerism in the blood based on clinical chimerism studies (supplemental Table 2), we also found persisting host T cells in the marrow. Notably, this persistent host chimerism was more evident in the T-cell compartment than in other populations. Furthermore, even among T cells, there was distinct chimerism between the CD4s and CD8s (Figure 4C-D).

This study revealed distinct patterns of IR in the marrow and PB, with regard to both cell type and cell phenotype. These findings highlight the relative abundance of innate immune cells, specifically NK cells, within the marrow early after transplant, even 100 days after transplant. This prolonged dominance of NK cells among lymphocytes in the marrow aligns with recent studies underscoring the critical role of NK cells for posttransplant IR and promoting leukemia remission.^2,26 Furthermore, the increased expression of CD69 on both T and NK cells in the marrow compared with the PB points supports that interpretation that at least a proportion of these lymphocytes are indeed tissue resident^19,20 in nature and thus reflect a different compartment of lymphocytes vs those sampled from the circulation, the focus of previous IR studies.^8-11 

T-cell phenotype in the marrow was more closely associated with relapse outcomes than the phenotype in the blood. In particular, we found increased PD1 expression on T cells at day 100 after transplant in patients who remained in remission compared with those who subsequently relapsed, potentially reflecting the importance of T-cell activation status within the leukemia microenvironment for protection from leukemia relapse. Based on our data, we are unable to make definitive conclusions regarding this T-cell population and its relationship to specific memory subsets or other phenotypic markers that have been associated with relapse after transplant.^23,25,27 Given the size of this study and the limited number of relapse events in our cohort, larger studies are needed to further investigate this finding along with a deeper characterization of these PD1⁺ T cells in the marrow to interrogate their repertoire and function. A previous investigation of PD1 expression in T cells in the blood after transplant suggests an overall higher PD1 expression early after transplant in both CD4 and CD8 compartments. However, increased PD1 expression was not associated with reduced T-cell cytotoxicity; this previous study was insufficiently powered to detect associations between PD1 expression and disease relapse.²² Furthermore, follow-up studies will be valuable for understanding potential links between features of the disease itself and patterns of marrow IR, including how the type of malignancy and its immunologic profile^28,29 may shape the T-cell compartment at the time of posttransplant relapse.

In this study, the most striking difference in marrow T-cell reconstitution across allograft sources related to the relative skewing of early T-cell recovery toward CD4s compared with CD8s early in recipients of CBT grafts, as has been noted in the analysis of IR in PB.²¹ T cells were identified in marrows from recipients of CD34-selected grafts, but larger studies are needed to address potentially impactful questions regarding how graft source shapes marrow T-cell frequencies, which may have implications for the efficacy of graft-versus-leukemia responses. Follow-up studies will also be critical to further delineate the relationship between other clinical variables, such as GVHD and infection, and patterns of bone marrow reconstitution and T-cell phenotypes.³⁰ 

To begin to generate insights into the origin of the lymphocytes analyzed here in the marrow, we performed chimerism analysis at the single-cell level via RNA sequencing of marrow aspirates isolated day 100 after transplant from 2 recipients of sex-mismatched allografts. We found mixed chimerism in the bone marrow, which was specifically enriched in the T-cell compartment compared with other myeloid and lymphocyte subsets. Furthermore, there was greater persistence of host T cells among CD4s than among the CD8s. These results suggest that distinct T-cell subpopulations may show different proportions of donor and host origins. Further analysis of the degree of chimerism shaping the reconstitution of naive and memory T-cell subsets will be valuable for improving our understanding of marrow immunity and its impacts on malignant relapse after transplant.

This study presents the results of >30 transplant patients and 200 samples, including transplants performed using unmanipulated mobilized PB, CD34-selected grafts, or CBTs. The cohort also included varying degrees of HLA mismatch. This diversity of the patient population highlights the applicability of the findings to a wide array of transplant settings, although it also limits the potential for identifying more subtle differences in IR patterns and for analyzing correlations with clinical outcomes, which can already be underpowered in a cohort this size. Larger studies will be essential to identify specific IR patterns and their relationship to clinical associations for distinct transplant approaches. Incorporation of machine learning approaches may further expand the depth of these analyses.³¹ 

Taken together, the findings from this study elucidate differences in IR in the bone marrow compared with the PB and identify a distinct abundance of NK cells in the marrow early after transplant. Furthermore, the site-specific lymphocyte profiles identified here and their potential distinct correlations with relapse illustrate the importance of tissue-based analyses for studying transplant biology and clinical outcomes.

This study was made possible thanks to the generous participation of our patients.

This research was supported by a grant from the Sawiris Foundation. It was also supported by the Memorial Sloan Kettering Cancer Center (MSKCC) core grant from the National Institutes of Health (NIH) P30-CA008748, NIH P01 CA23766, the Susan and Peter Solomon Divisional Genomics Program, and the Parker Institute for Cancer Immunotherapy. S.D. reports research funding from an NIH/National Cancer Institute (NCI) KO8 (1K08CA29327-01), the Memorial Sloan Kettering (MSK) Leukemia Specialized Program of Research Excellence Career Enhancement Program (NIH/NCI P50 CA254838-01), the Parker Institute for Cancer Immunotherapy, the MSK Center for Tumor-Immune Systems Biology Pilot, and the MSK Gerstner Physician Scholar Program. V.Z. reports research funding from NIH/NCI R25CA272282. J.U.P. reports funding from the National Heart, Lung, and Blood Institute/NIH Award K08HL143189 and the MSKCC Core Grant NCI P30 CA008748. The authors also acknowledge the use of the Integrated Genomics Operation Core, funded by the NCI Cancer Center Support Grant (P30 CA08748), Cycle for Survival, and the Marie-Josée and Henry R. Kravis Center for Molecular Oncology. Images in the Visual Abstract were created in BioRender. Lorenc, R. (2025) https://BioRender.com/f79z291.

Contribution: S.D., J.K., and P.V. designed and performed experiments, analyzed data, and wrote the manuscript; J.S. and W.D. performed computational analyses; P.G., Z.K., and L.M. performed sample coordination and collected clinical data; V.Z. and T.F. performed statistical analyses; J.U.P. and K.H. provided analytic guidance and assisted with manuscript preparation; B.Gipson. and R.Lorenc. performed experiments; J.A.F. analyzed single-cell data; G.S. collected clinical data; R.Lin., E.P., B.Gyurkocza., B.S., M.T., I.P., S.A.G., J.B., M.-A.P., R.T., and C.C. assisted with study design, clinical sample collection, and data interpretation; O.A.-W. assisted with manuscript preparation; M.R.M.v.d.B. supervised analyses and interpreted data; and A.M.H. designed the study, supervised analyses, interpreted data, and cowrote the manuscript.

Conflict-of-interest disclosure: I.P. has received research funding from Merck and serves as a data and safety monitoring board member for ExCellThera. M.T. receives research funding from AbbVie, BioSight, Orsenix, GlycoMimetics, Rafael Pharmaceuticals, and Amgen; is on the advisory board of AbbVie, Daiichi Sankyo, Orsenix, KAHR, Jazz Pharmaceuticals, Roche, BioSight, Novartis, Innate Pharmaceuticals, Kura, Syros Pharmaceuticals, and Ipsen Biopharmaceuticals; and received royalties from UpToDate. B.S. reports consulting and research support from Hansa Biopharma and Gamida Cell. S.A.G. receives research funding from Miltenyi Biotec, Takeda Pharmaceuticals, Celgene Corporation, Amgen, Sanofi, Johnson & Johnson, and Actinium Pharmaceuticals, Inc; and is on the advisory boards of Kite Pharma, Inc, Celgene Corporation, Sanofi, Novartis, Johnson & Johnson, Amgen, Takeda Pharmaceuticals, Jazz Pharmaceuticals, and Actinium Pharmaceuticals, Inc. M.-A.P. reports honoraria from Allogene, Caribou Biosciences, Celgene, Bristol Myers Squibb, Equillium, ExeVir, ImmPACT Bio, Incyte, Kite/Gilead, Merck, Miltenyi Biotec, MorphoSys, Nektar Therapeutics, Novartis, Omeros, OrcaBio, Pierre Fabre, Sanofi, Syncopation, Takeda, VectivBio AG, and Vor Biopharma; serves on data and safety monitoring boards for Cidara Therapeutics and Sellas Life Sciences; has ownership interests in Omeros and Orca Bio; and received institutional research support for clinical trials from Allogene, Genmab, Incyte, Kite/Gilead, Miltenyi Biotec, Novartis, and Tr1x. M.R.M.v.d.B. has received research support from Seres Therapeutics and stock options from Seres Therapeutics and ThymoFox; has received royalties from Wolters Kluwer and Juno; has consulted, received honorarium from, or participated in advisory boards for Seres Therapeutics, Thymofox, Garuda, Novartis (spouse), Bristol Myer Squibb (spouse), Galapagos (spouse), and BeOne (spouse); has intellectual property licensing with Juno Therapeutics; holds a fuduciary role on the board of DKMS (a nonprofit organization); and is the chairman of the scientific advisory board for Smart Immune. J.U.P. reports research funding, intellectual property fees, and travel reimbursement from Seres Therapeutics and consulting fees from Da Volterra, CSL Behring, Crestone Inc, MaaT Pharma, Canaccord Genuity, Inc, and RA Capital; serves on an advisory board of and holds equity in Postbiotics Plus Research; serves on an advisory board of and holds equity in Prodigy Biosciences; and has filed intellectual property applications related to the microbiome (reference numbers 62/843,849, 62/977,908, and 15/756,845). Memorial Sloan Kettering Cancer Center has financial interests relative to Seres Therapeutics. S.D. has received research support from Tigen Pharma. A.M.H. holds intellectual property related to interleukin-22 and graft-versus-host disease; has a research collaboration with Evive Biotech that provided funding for an unrelated clinical trial; and serves in a volunteer capacity for the American Society for Transplantation and Cellular Therapy. The remaining authors declare no competing financial interests.

Correspondence: Alan M. Hanash, Memorial Sloan Kettering Cancer Center, 1275 New York Ave, New York, NY 10065; email: hanasha@mskcc.org.