Bone marrow niches for hematopoietic stem cells: life span dynamics and adaptation to acute stress Free

The spatial microenvironment of hematopoietic stem cells (HSCs) is a multiparametric ecosystem composed of multiple hematopoietic and nonhematopoietic populations, their molecular signals, and extracellular matrix components. From these, a functional subunit called the HSC niche provides lifelong support by promoting HSC quiescence, self-renewal, and niche retention.¹ Previous studies used tissue imaging and in vivo cell type–specific ablation approaches to elucidate both the composition and functional roles of putative HSC niches.² As a result, several bone marrow (BM) populations have been reported to support homeostatic HSCs including endothelial cells (ECs) and mesenchymal stromal cells (MSCs) that promote HSC maintenance and niche retention,^3-5 nonmyelinated Schwann cells,⁶ megakaryocytes (MKs)^7,8 and macrophages⁹ supporting HSC quiescence, and osteolineage cells (OLCs) that have been proposed to regulate HSC numbers.^10,11 However, the composition and function of HSC niches exhibit context-dependent adaptation, particularly during aging or response to stress. Here, we review recent findings that have advanced our understanding of the spatial organization and molecular composition of murine and human HSC niches from early development to aging and in response to acute stress.

Embryonic HSCs transition through different organs and are supported by distinct niche cues to fulfill their function.¹² Murine and human definitive HSCs emerge from arterial-like hemogenic precursors at the aorta-gonad-mesonephros region.^12-16 To uncover cell-extrinsic signals that support HSC emergence, Hadland et al compared the transcriptomic profile of murine aorta-gonad-mesonephros–derived ECs with or without hemogenic potential. Cytokine and integrin signaling were enriched in hemogenic ECs, and the respective proteins functionally supported HSCs in vitro (Cxcl12, Notch1/2, Vascular Cell Adhesion Molecule 1 [Vcam1], and Fibronectin).¹⁷ In the human system, Crosse et al investigated niche signals enriched in the ventral side of the dorsal aorta. Endothelin-1 (EDN1), a peptide involved in vasoconstriction, was among the most promising ventrally enriched factors due to its colocalization with Runx1⁺ intra-aortic hematopoietic clusters¹⁸ and contribution to the ventral enrichment of tumor necrosis factor α signaling that supports HSC emergence.¹⁹ Despite failure to detect EDN1 receptor on HSCs, colony-forming assays and transplantation experiments confirmed EDN1’s role in promoting early HSC development.¹⁸ In combination, these studies identified functional cell-extrinsic regulators supporting murine and human HSC emergence.

Next, HSCs migrate to the fetal liver (FL) where they expand in numbers.²⁰ This observation sparked considerable interest in identifying the cell-extrinsic regulators that promoted HSC self-renewal and expansion.^21-27 To profile the FL microenvironment in situ, Lu et al used multiplexed error robust fluorescence in situ hybridization and identified 8 major populations including arteriolar and sinusoidal endothelium, hepatocytes, and 5 hematopoietic populations.²⁸ The frequencies and cellular composition of the HSC microenvironment were similar to average FL locations (50% erythroid, 20% hepatocytes, and 5% arteriolar cells), thus providing no evidence for preferential association of HSCs with specific microenvironment(s).²⁸ However, Notch signaling components (Notch3 and delta like ligand 4 [Dll4]) and the arteriolar marker Ephrin-B2 were enriched within HSC microenvironments at the transcriptome²⁸ and protein level.^29,30 These results suggest spatially resolved signaling activation, although further studies are required to decipher their functional role on HSC function. Besides ECs, macrophages contribute to FL HSC expansion through cytokine secretion.³⁰ Elegant live-cell imaging in zebrafish embryos demonstrated that macrophages surveil the expanding stem cell pool and engulf HSCs with high levels of reactive oxygen species and higher risk to accumulate DNA damage.³¹ These data illustrate that the FL microenvironment not only provides supporting signals but also safeguards developmental hematopoiesis by regulating survival of HSC clones with baseline stress levels.

After their FL expansion, HSCs migrate to the fetal spleen and fetal BM (fBM), which becomes the primary hematopoietic organ after birth. HSCs start seeding the fetal spleen by embryonic day 15.5 (E15.5)³²; however, its microenvironment has limited capacity to support their self-renewal and instead promotes myeloid differentiation.³³ HSC migration to the fBM seems to be independent of the fetal spleen, because fetal spleen removal had no impact on fBM HSC numbers.^27,34 Recent studies timed the presence of the first long-term repopulating cell in murine fBM as early as E15.5³⁵ or few days later^32,36 and 12 weeks after conception in humans.³⁷ Unexpectedly, early repopulating activity does not reside within the “classical” HSC compartment, because E16.5 CD150⁺CD48^–cKit⁺Sca1⁺Lineage^– fBM HSCs displayed no measurable long-term reconstitution.³⁵ Gene expression and competitive repopulation assays revealed that fBM HSCs are transcriptionally distinct and functionally inferior from their adult counterparts.³⁵ Instead, embryonic FL-derived multipotent progenitors generated in a HSC-independent manner contribute to both embryonic and adult hematopoiesis by near-lifelong blood production.³⁸ These data revised our view of the fBM as a site with limited capacity to support HSCs.

When does the BM acquire its supportive role? The appearance of repopulating cells coincides with tissue calcification and vascularization. Despite their concomitant progression, osteogenesis is uncoupled from angiogenesis. Deletion of Osx-expressing OLCs did not affect vascular development but significantly impaired homing and engraftment of repopulating cells,³⁶ suggesting a functional role of OLCs in fBM colonization. Transcriptomic analyses revealed significant differences in the composition of fetal vs postnatal/adult BM. Contrary to adult BM where they form an extensive network³⁹ and overlap with leptin receptor⁺ (LepR⁺) cells,^40,41 Cxcl12⁺ MSCs are absent from fBM (Figure 1). Instead, osteochondrocytes, fibroblasts,³⁵ or, an embryonic MSC population (Col3a1^highDecorin^highGsn/Gle3b^high)⁴² are the most abundant fBM populations. The absence of Cxcl12⁺/LepR⁺ MSCs leads to reduced levels of CXCL12, stem cell factor (SCF), and angiopoietin 1 (ANGPT1) in murine⁴² and early-stage human fBM,³⁷ which was not compensated by other MSCs⁴² nor OLCs.³⁶ Expectedly, fBM MSCs displayed inferior capacity to support cocultured HSCs compared with adult MSCs.³⁵ Unlike the MSC compartment, all EC subtypes are present in fBM and share common cytokine profiles with adult BM vasculature (Cxcl12, Scf, and vascular endothelial growth factor C [Vegfc]). However, arteriolar rather than sinusoidal endothelium is the main source of hematopoietic cytokines in fBM (Cxcl12, Jagged-1 [Jag1], and Dll4)^37,42 and promotes HSC proliferation via canonical Wnt/β-catenin signaling.⁴² In summary, murine BM seems to acquire its HSC-supportive role postnatally. Collectively, these studies provide novel insights into the cell-extrinsic signals that support HSCs during their transition through different embryonic stages.

Although HSCs expand in the FL before migrating to the BM, elegant lineage tracing experiments have recently challenged the extent to which FL-HSC expansion contributes to the adult HSC pool and lifelong hematopoiesis.⁴³ Instead, FL expansion was reported to favor hematopoietic differentiation, suggesting that the adult BM HSC compartment is independently established postnatally. Indeed, BM HSCs retain active cell-cycling characteristics 4 weeks after birth, before switching to the adult quiescent phenotype.⁴⁴ Although proposed to be intrinsically determined, this phenotypic HSC switch could be cosupported by the BM niche. In line with this hypothesis, the transcriptional rewiring of BM endothelial and mesenchymal compartments is more pronounced during the first weeks of postnatal life and transition to adulthood than during the 2-year aging period.^45,46 Furthermore, sinusoids and Cxcl12⁺/LepR⁺ MSCs significantly expand (<1% at perinatal day 4 [P4]) and become among the most prevalent nonhematopoietic BM populations by P14⁴⁶ (Figure 1B). Concomitantly, most HSCs spatially associate with sinusoids (but only few with arteries, capillaries, or bone) and rely on EC-derived membrane-bound⁴⁶ and soluble/systemic SCF,⁴⁷ a role later acquired by Cxcl12⁺/LepR⁺ MSCs. Quantitative multiplex 3-dimensional imaging revealed that most adult HSC-niche populations are present in the BM by P21; however, Cxcl12⁺ MSCs and MKs are less frequent compared with adult BM. Although 5-week-old and adult HSCs shared similar niche signatures, 3-week-old HSCs were enriched in direct contact with CXCL12-Discosoma red fluorescent protein (DsRed) expressing MSCs and away from MKs,⁴⁸ a cell type reported to promote HSC quiescence. Notably, the reduced HSC-MK association reflected the twofold reduction in MK abundance in preswitch vs adult BM.⁴⁸ These findings underscore the spatial remodeling of BM tissue as it acquires its HSC-supportive role, as well as its potential contribution to the fetal-to-adult HSC switch.

The adult BM is a highly complex tissue composed of multiple hematopoietic and nonhematopoietic populations including sinusoids, arteries, capillaries, MSCs, OLCs, chondrocytes, adipocytes, and neuronal and smooth muscle cells. HSCs depend on reciprocal interactions with their BM niche to fulfill their role in blood cell production.^1,49 In vivo ablation of selected populations and/or factors using reporter mouse models (including those expressing Cre recombinase) has led to the identification of the cellular and molecular BM constituents of the HSC niche (reviewed in detail elsewhere^1,2,50). These experiments showed that ECs promote HSC maintenance via the production of SCF⁵ and CXCL12,⁵¹ whereas MSCs support HSC maintenance³⁹ and niche retention via pleiotrophin,⁵² SCF,⁵ and CXCL12.^3,51 Furthermore, several BM populations regulate HSC quiescence including nonmyelinated Schwann cells (via transforming growth factor β [TGF-β]⁶), MKs (via TGF-β⁸ and CXCL4⁷), macrophages (via atypical chemokine receptor 1 [DARC]⁹), and chondroitin sulfate proteoglycan 4 (NG2) peri-arteriolar cells.⁵³ 

Identifying the anatomical composition of the BM HSC niche in situ is also essential to unravel the cross talk signals supporting HSC function. Surprisingly, homeostatic murine HSCs have been reported in the vicinity of several BM populations but at different frequencies.⁵⁴ Cxcl12⁺/LepR⁺ MSCs^3,48,55 and sinusoids^48,55,56 are the most frequent cellular components of the HSC microenvironment, because 80% and ∼70% of HSCs locate in their immediate proximity, respectively (Figure 2A). MKs^7,8,48 and Nestin-green-fluorescent-protein (GFP) expressing cells^53,57 are the next most frequent populations (20% HSCs), followed by arteriolar endothelium (∼10% HSCs).^48,53,55 Bone, nonmyelinated Schwann cells, and adipocytes are rarely part of the HSC microenvironment, because most HSCs locate several cell-diameters away (<5% HSCs).^6,48,55,56 Notably, the frequency of these associations might marginally vary depending on the utilized HSC identification strategy (Lineage^–CD41^–CD48^–CD150^+6-8,53 or α-catulin-GFP⁺cKit⁺ HSCs^48,55), bone type, and imaged/analyzed volume (single tiles vs volumetric tissue-wide imaging), but have been mostly confirmed by independent studies (reviewed in detail elsewhere⁵⁴). In human BM, most CD34⁺ hematopoietic stem and progenitor cells (HSPCs) associate with CD271⁺ MSCs, ECs, or CD163⁺ macrophages,⁵⁸ whereas CD34⁺CD38^– HSPCs are proximal to MKs, sinusoids,⁵⁹ and, rarely, bone.⁶⁰ However, due to the low purity of functional HSCs within the imaged populations (1 in 617 CD34⁺CD38^– cells are HSCs⁶¹), it is unclear which spatial signatures reflect “true” HSC localization.

It is evident that measuring HSC associations with individual BM populations per sample is insufficient to capture niche complexity. Because several populations functionally support HSCs,¹ they may collectively form an “HSC-niche functional unit.” In fact, most of the arteriolar and bone locations harboring HSCs are also rich in sinusoids and MSCs, because hardly any hematopoietic cell can be found further than 30 μm away from peri-sinusoidal locations due to their frequency and tissue distribution.^41,48,62 In other words, HSCs previously classified as proximal to infrequent BM populations rarely associate with these cell types exclusively. Multiplex visualization of HSCs with multiple BM components demonstrated their simultaneous physical association with different combinations of BM cells, such as perisinusoidal CCXL12⁺ MSCs (∼64%) or perisinusoidal MKs in central BM away from bone surfaces and adipocytes (∼32%; Figure 2B).⁴⁸ Notably, the frequency of murine HSC-niche associations correlated with the abundance of homeostatic BM niches, providing no evidence for active HSC migration toward imaged BM populations or their combinations. These data demonstrate that not all HSCs physically associate with the same BM population(s); however, even higher multiplexing is required to profile these complex spatial interactions.

The fact that fractions of the HSC compartment associate with distinct BM populations may reflect the reported HSC heterogeneity⁶³ and the unique requirements of HSC subsets for cell-extrinsic signals (Figure 3). Along these lines, lineage-biased HSCs have been reported to occupy distinct BM locations with von Willebrand factor+ (VWF⁺) MK-biased HSCs being close to MKs, whereas VWF^– HSCs displaying peri-arteriolar localization.⁶⁴ MKs are important regulators of HSC quiescence via CXCL4 or TGF-β signals^7,8; however, whether quiescent and cycling HSCs occupy distinct BM locations remains controversial.^53,55 Of note, dormant long-term label-retaining HSCs that contain most of the transplanted stem cell activity^65,66 displayed identical BM localization to sinusoidal, arteriolar, MK niches, and their combinations compared with proliferating HSPCs.⁴⁸ These data challenged the existence of spatially resolved predefined BM niches that differentially support dormant and activated HSCs. In line with these findings, HSCs and multipotent progenitors displayed similar localization in relation to ECs,⁵⁶ MSCs,⁶⁷ and bone surfaces (<5 μm difference), although more systematic analyses are required. Are HSCs migratory within BM tissue and therefore able to receive signals from spatially distinct BM niches in steady state? In vivo live-cell imaging of highly purified homeostatic Mds1^GFP/+Flt3^Cre HSCs revealed baseline motility compared with downstream progenitors in their native BM microenvironment (<10 μm in 150 minutes).⁵⁶ These results suggest stable HSC-niche interactions, whereas a tamoxifen-induced labeling system reported heterogeneous motility behaviors (Pdzk1ip1-CreER R26^LSL-Tom).⁶⁸ 

Another possibility for the nonuniform HSC distribution is their association with distinct, possibly unknown, subsets of BM cell types, which have been traditionally viewed as homogeneous. Single-cell profiling revealed extensive heterogeneity within BM populations at the transcriptome and protein level.^69-72 Recent studies revealed that some of these BM subpopulations are spatially and functionally distinct including the adipo- and osteo-primed Cxcl12⁺/LepR⁺ MSCs,⁷¹ the osteo-primed osteolectin+ peri-arteriolar LepR⁺ MSCs that support lymphopoiesis,⁷³ the metaphyseal apelin+ arteriolar ECs supporting HSC maintenance,⁷⁴ the rare type S ECs regulating epiphyseal osteogenesis and supporting transplanted HSCs,⁷⁵ as well as the metaphyseal Lin^–CD31^–Sca1^–CD51⁺PDGFRa⁺PDGFRb⁺ OLC and IL7⁺CXCL12⁺ MSC subsets supporting lymphopoiesis.^67,76 Along the same lines, multiplex imaging identified sinusoids as the primary site for myelopoiesis, with potentially distinct sinusoidal subsets supporting maturation of granulocytes, monocytes, or dendritic progenitors.⁷⁷ Notably, OLCs display distinct transcriptional profiles depending on their proximity to HSCs.⁷⁸ Whether HSCs actively migrate toward predefined, spatially and functionally distinct BM niches, or whether specific matrix-immobilized signals suffice, or whether HSCs induce such changes on their neighboring cells requires further investigation (Figure 3). Collectively, these studies highlight the multifaceted heterogeneity of the BM microenvironment at the transcriptional, protein, and spatial level and the need for deeper characterization of functional subpopulations beyond conventional cellular compartmentalization.

Aging induces functional alterations to the hematopoietic compartment and BM microenvironment, which are already evident in middle-aged animals (12 months of age).^79,80 Despite the marked increase in the frequency of immunophenotypic HSCs in aged BM tissue, old HSCs display reduced repopulation potential, homing efficiency, and B-/T-lineage output compared with their young counterparts.^81-83 Aging-associated effects on hematopoiesis are evolutionary conserved as human HSCs from old individuals also exhibit increased cell frequencies, cycling activity, and myeloid bias.⁸⁴ Mechanistically, the functional impairment of old HSCs is caused by alterations in their genomic stability, epigenetic profile, and defective DNA repair machinery.^80,85,86 Furthermore, data from heterochronic transplantations demonstrated that the aged BM microenvironment also contributes to the aging-associated phenotype of old HSCs. Young HSCs display reduced repopulation efficiency and increased myeloid and impaired lymphoid output upon transplantation in aged mice.^87,88 Strikingly, old HSCs retain their myeloid-biased lineage output when transplanted in lethally irradiated young recipients with damaged BM niches,⁸⁰ but not in sublethally irradiated⁸⁰ or nonconditioned young BM niches, which seem to restore myeloid skewing.⁸⁹ Furthermore, young unperturbed BM niches restored the transcriptional but not DNA methylation profile of transplanted aged HSCs. Taken together, these findings indicate that cell-intrinsic and environmental factors collectively contribute to the aging-associated HSC phenotype.

Single-cell profiling and functional assays revealed that several components of the BM microenvironment undergo significant remodeling during aging and instigate aging-related phenotypes on young HSCs (Figure 1). Old MSCs expand, display elevated levels of reactive oxygen species and inflammatory cytokines (C-X-C motif chemokine ligand 2 [Cxcl2] and C-X-C motif chemokine ligand 5 [Cxcl5]) known to impair HSC self-renewal (interleukin-1α/β [IL-1α/β] and IL-6),^45,90,91 and induce myeloid skewing on cocultured young HSPCs.⁹² The differentiation trajectory of old MSCs is imbalanced toward the adipocytic lineage resulting in reduced OLC numbers and aging-related bone loss.^88,93 As a result, aged murine^94,95 and human⁹⁶ BM accumulate adipocytes, which are known negative regulators of steady-state hematopoiesis.⁹⁷ Beyond their decreased frequency, aged OLCs display a proinflammatory signature mediated by colony-stimulating factor 1 (CSF1), receptor activator of nuclear factor kappa B ligand (RANKL), and C-C motif chemokine ligand 5 (CCL5), which further contribute to the myeloid skewing of old HSCs.^87,98 

Aging also remodels the BM vasculature. Sinusoids become discontinuous and leaky,^57,99 whereas the frequency and size of arterioles^95,100 and type H vessels⁹³ are decreased. Coculture with old ECs induces aging-associated phenotypes on young HSPCs, but young ECs cannot revert the myeloid bias of old HSCs.⁹⁹ Sympathetic innervation is also affected by aging, demonstrated by the reduction of tyrosine hydroxylase–expressing nerve fibers in murine BM.⁹⁵ Recent studies revealed that BM nerves are important mediators of aging phenotypes on the hematopoietic system. Unilateral surgical denervation of femoral and sciatic nerves in young mice induced premature aging via adrenergic receptor β3 signaling (ADRβ3) on both HSCs and BM niches (decreased arteriolar volume and mild expansion of dysfunctional MSCs), respectively.⁹⁵ Expectedly, the age-related remodeling of the BM microenvironment affects spatial interactions of HSCs with BM niches in both mice and humans. Significantly less old murine HSCs physically associate with arterioles and MKs,^57,95 whereas their perisinusoidal localization is maintained (Figure 1A). The reduced association between old HSCs and MKs was recently also observed in human BM,⁶⁰ suggesting conserved aging-associated phenotypes. However, further studies are required to systematically unravel the exact localization of old HSCs in relation to reported BM niches in mouse models and human samples. Taken together, these data illustrate a functional interconnection between biological aging, BM niche remodeling, and functional defects on HSCs.

Are the functional consequences of organismal aging on the HSC compartment reversible? Although calorie restriction, exercise, or systemic exposure to bloodborne factors failed to rejuvenate old HSCs,^80,89 several studies reported that targeting specific niche-derived signals reverted HSC function and aging-associated phenotypes. Because the expression of several BM-derived supportive signals decreases with age, their exogenous administration has been used to rejuvenate old HSCs (Table 1). Ex vivo treatment of old HSCs with thrombin-cleaved osteopontin⁸⁸ or insulin-like growth factor 1⁷⁹ increased their T- and B-cell lineage output at the expense of myeloid cell production, respectively. Furthermore, infusion of an ADRβ3 agonist (BRL37344) reverted donor chimerism and multilineage reconstitution of old to young HSC levels.⁹⁵ Along the same lines, administration of netrin-1, a secreted protein mainly produced by ECs and MSCs, partially restored DNA damage response on BM niche cells.¹⁰¹ Although netrin-1 treatment did not alter the frequency of immunophenotypic HSPCs, it led to a fourfold increase of functional HSCs and reverted their engraftment deficiency and their myeloid skewing.¹⁰¹ In addition to supplementation of downregulated niche-derived supportive factors, blocking elevated negative regulators of HSC function has also beneficial effects on reverting aging-associated phenotypes. Administration of TGF-βR1 inhibitor reverted the aging-associated expansion of immunophenotypic HSCs and restored their myeloid lineage output after transplantation.¹⁰² Furthermore, counteracting the upregulation of IL-1α/β inflammatory signaling significantly improved lymphopoiesis, restored lymphoid/myeloid differentiation output,¹⁰³ and improved hematopoietic regeneration in chemotherapy-treated old mice.⁹¹ Interestingly, similar results were obtained by modulating the microbiome of aged mice either by 8-week antibiotic treatment¹⁰³ or fecal microbiota transplants from young mice.¹⁰⁴ These results demonstrate reversibility of functional defects of aged HSCs by manipulating niche-derived signals. Further studies are required to develop combinatorial intervention strategies for robust rejuvenation, potentially by simultaneous targeting HSC intrinsic programs and BM niche fitness to efficiently overcome crossrestricting phenotypes.

Before the onset of aging-associated phenotypes, systemic exposure to acute stress elicits damage responses on adult BM and hematopoietic¹¹³ tissues that are reminiscent of aging. BM sinusoids and MSCs from adult mice exposed to infection-mimicking agents exhibited significant overlap with those from untreated old mice, particularly in inflammatory and stress-associated transcriptional programs.⁴⁵ In the context of nonlethal dose of medically relevant cytotoxic conditioning (150 mg/kg 5-fluorouracil [5-FU] or <6Gy total body irradiation [TBI]), HSC-driven hematopoietic recovery is regulated by cell-intrinsic programs (reviewed in detail elsewhere¹¹⁴) and cell-extrinsic signals. Below, we focus on how BM niches adapt and collectively interact to support BM recovery in response to cytotoxic conditioning.

Although the role of several populations is context dependent, BM ECs are critical for both steady-state and emergency hematopoiesis. Ubiquitous vascular endothelial growth factor receptor 2 (VEGFR2) deletion required for sinusoidal regeneration¹¹⁵ or EC-specific Jag1 deletion¹¹⁶ compromised hematopoietic recovery in irradiated mice. Conversely, genetically induced rescue of EC survival significantly improved HSC numbers in chemotherapy-treated mice lacking the preapoptotic factor Bax.¹⁰⁵ Mechanistically, EC recovery relies on vascular endothelial growth factor A (VEGF-A) from hematopoietic progenitors^74,115 and ECs,¹¹⁷ tumor necrosis factor α from granulocytes,¹¹⁸ as well as VEGF-C from ECs and MSCs¹¹⁹ (Figure 4). The functional relevance of synergistic interactions between ECs and hematopoietic cells is further highlighted by the fact that blood cell ablation phenocopies stress-induced vascular defects.⁷⁴ During the ablation phase, ECs undergo p53/CDK5-mediated apoptosis by upregulating SEMA3A (semaphorin 3A) and its receptor neuropilin 1 (NRP1).¹¹¹ Systemic NRP1 inhibition or EC-specific NRP1/SEMA3A deletion accelerated vascular remodeling and hematopoietic reconstitution after cytotoxic conditioning (Table 1) but did not prevent early EC damage.¹¹¹ During the recovery phase, upregulation of VEGF-A, angiopoietin 1/Tie2, and pleiotrophin signaling facilitates both EC and blood cell regeneration.^52,107 Notably, cytotoxic conditioning differentially affects EC populations causing dose-dependent alterations on sinusoidal vessels,¹¹⁵ whereas arteriolar endothelium is mostly preserved.⁷⁴ After irradiation injury, a rare apelin+ EC subpopulation expands and functionally contributes to murine EC and hematopoietic recovery.⁷⁴ Similar emergency mechanisms exist in humans as a rare CD105⁺CD31⁺ EC subpopulation significantly expands after chemotherapy and promotes hematopoietic and EC regeneration via IL-33.¹²⁰ Collectively, these data highlight that EC regeneration is a prerequisite for hematopoietic recovery after myeloablation.

MSCs are critical for steady-state hematopoiesis^4,5; however, their role in blood cell recovery has been less understood until recently. Human MSCs are considered radio resistant,¹²¹ whereas murine BM MSCs decreased after TBI.⁷² TBI induces metabolic dysfunction on PDGFRa⁺Sca1^– MSCs, which is, at least partially, repaired by mitochondrial transfer from neighboring HSPCs.¹²² LepR⁺ MSCs produce VEGF-C and restore sinusoidal ECs and hematopoiesis after TBI.¹¹⁹ Furthermore, MSCs support sympathetic nerve recovery and hematopoietic regeneration via secretion of nerve growth factor¹²³ counteracting restricting parasympathetic signals.¹²⁴ In addition to their homeostatic role in regulating HSC quiescence⁶ and circadian mobilization,¹²⁵ sympathetic BM nerves also support hematopoietic recovery. Repetitive cycles of neurotoxic chemotherapy (cisplatin or vincristine) caused sympathetic nerve damage and impaired hematopoietic recovery in mice, similar to iatrogenic neuropathy observed in patients who survive cancer.¹²⁶ Conversely, HSC frequencies and hematopoietic regeneration were restored after genetic or chemically mediated neuroprotection via conditional Trp53 deletion in sympathetic neurons or 4-methylcatechol administration in vivo.¹⁰⁸ To exert their supportive role, BM nerves activate ADRβ2/β3 on LepR⁺ MSCs¹²³ or a7 nicotinic receptor signaling on Nestin-GFP cells¹²⁷ to increase the expression of key hematopoietic cytokines (SCF, VEGF, angiopoietin 2 [ANG2], and CXCL12). In summary, these findings illustrate that ECs, MSCs, and neurons retain their supportive homeostatic roles and collectively promote BM tissue recovery and hematopoietic regeneration after cytotoxic conditioning.

Several BM niche populations adapt in response to cytotoxic stress and regulate HSC behavior in a distinct manner from their steady-state role. In homeostasis, MKs promote HSC quiescence via TGF-β1/SMAD⁸ and CXCL4.⁷ However, after 5-FU administration, MKs transiently support HSC proliferation via fibroblast growth factor 1 signaling,⁸ before reinstating HSC quiescence.^110,106 This stage-resolved regulation is pivotal for robust hematopoietic regeneration because MK ablation compromises HSC function and emergency myelopoiesis.¹⁰⁶ Conversely, transient TGF-β inhibition prolongs HSC cycling and enhances hematopoietic regeneration.¹¹⁰ In addition to MKs, BM-resident group 2 innate lymphoid cells (ILC2) transiently support hematopoietic proliferation after 5-FU. During the ablation phase, dying B-cell progenitors,¹⁰⁹ and sinusoids¹²⁰ release IL-33, which activates ILC2s. Activated ILC2s produce granulocyte-macrophage colony-stimulating factor that supports hematopoietic recovery,¹⁰⁹ despite being dispensable for steady-state hematopoiesis.¹²⁸ Similarly, lymphatic vessels expand in a VEGFR3-/IL-6–dependent manner and support hematopoietic regeneration after stress, although not required for homeostatic HSCs.¹²⁹ 

Despite being the first population reported to regulate HSCs,¹ OLCs do not seem to directly support homeostatic HSCs.^5,49 OLCs are resistant to cytotoxic injury,¹¹⁵ although later studies confirmed radio-resistance of osteoprogenitors rather than mature osteoblasts.⁷² OLCs were thought to support hematopoietic recovery via thrombopoietin/thrombopoietin receptor (MPL) signaling¹³⁰; however, recent data identified hepatocytes as the main functional thrombopoietin source after myeloablation.¹³¹ Nevertheless, a radio-resistant OLC subpopulation (CD73⁺ nerve growth factor receptor^high [NGFR^high] Runx2⁺) has emerged as a top hematopoietic cytokine producer that supports myeloid cell recovery after irradiation and hematopoietic transplantation.⁷² Furthermore, N-Cadherin+ OLCs are reported as functional sources of SCF because their conditional deletion in chemotherapy-treated mice led to reduced HSPC numbers but not BM cellularity.¹³² Nevertheless, further studies are required to determine the magnitude of OLC's contribution to hematopoietic recovery after myeloablation.

Finally, some BM populations switch from suppressive to supportive roles. Adipocytes are negative regulators of steady-state hematopoiesis, with their abundance inversely correlating with HSC frequencies.^97,112 In contrast, they transiently expand after myeloablation and promote hematopoietic recovery through SCF and adiponectin production,¹³³ before transdifferentiating toward MSCs.¹³⁴ It is possible that adipocytes act as transient mediators of hematopoietic recovery until (peri)vascular BM niches are restored. These data highlight the plasticity of several BM populations and the ability to adapt to extrinsic stress and fulfill hematopoietic demands.

How does acute stress affect the interaction between HSCs and their BM niches? Despite growing evidence for the role of BM niche populations in supporting hematopoietic regeneration, little is known on how myeloablation affects HSC localization. Recent studies reported that expression of CD49b separates chemotherapy-sensitive from chemotherapy-resistant HSCs (CD49b^–CD48^–CD34^–Flk2^–cKit⁺Sca1⁺Lineage^–). Three days after 5-FU, chemotherapy-resistant HSCs are reported to associate with endosteal N-Cadherin+ OLCs, whereas no enrichment was observed toward vessels nor MKs.¹³² Interestingly, whether HSC localization is re-established after chemotherapy administration seems to be age dependent. Although young HSCs seem to re-establish spatial associations with sinusoids 30 days after 5-FU administration, aged HSCs are reported to migrate from sinusoidal toward arteriolar locations in 18-month-old mice.⁵⁷ 

Taken together, these data provide evidence for the dynamic role of BM niches in differentially regulating HSC fate decisions in homeostasis and response to stress, while highlighting the need for deeper longitudinal analysis of HSC localization.

The application of single-cell and spatial technologies has considerably improved our understanding of BM complexity and its remarkable ability to adapt to organismal demands and extrinsic challenges. It becomes increasingly evident that HSC function is constantly regulated by a complex integrated framework of cell-intrinsic programs and cell-extrinsic signals from multicellular BM niches. Leveraging our mechanistic understanding of the underlying programs toward developing potent intervention strategies could improve the clinical outcome of several conditions. For example, developing combinatorial approaches to simultaneously revert cell-intrinsic vulnerabilities of aged HSCs and rejuvenate their BM niche(s) could significantly enhance regeneration of aged BM tissue. Such findings have the potential to revolutionize treatment of diseases with unfavorable outcome in older patients, such as hematologic malignancies, in which robust hematopoietic function requires both efficient elimination of malignant clones and lifelong healthy hematopoiesis. To achieve clinical translation, thorough preclinical analyses of human HSCs in large cohorts of BM biopsies are essential. The low purity of currently used markers for in situ identification of human HSCs and the need for large-tissue volumetric imaging to unravel 3-dimensional BM niches remain major challenges. To address those, establishing more elegant marker-combination strategies and imaging modalities to perform volumetric imaging of human BM samples is pivotal. Finally, the development of human BM organoids¹³⁵ provides a unique platform to functionally assess intervention strategies aiming to enhance human hematopoiesis in the context of aging or myeloablation in a controlled system.

The authors apologize to all researchers whose work could not be cited because of space limitations. The authors thank all members of the Kokkaliaris lab for their feedback on the text and the figures. All figures were created with BioRender.com.

K.D.K. is funded by the Mildred Scheel Career Center Frankfurt (Deutsche Krebshilfe), the Else Kröner-Fresenius-Stiftung, and the European Hematology Association. J.H. is supported by the Else Kröner-Fresenius-Stiftung.

Contribution: J.H. and K.D.K. wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Konstantinos D. Kokkaliaris, University Hospital and Goethe University Frankfurt, Theodor-Stern-Kai 7, 60596 Frankfurt am Main, Germany; email: kokkaliaris@med.uni-frankfurt.de.