Cholinergic signals preserve haematopoietic stem cell quiescence during regenerative haematopoiesis

The sympathetic nervous system has been evolutionary selected to respond to stress and activates haematopoietic stem cells via noradrenergic signals. However, the pathways preserving haematopoietic stem cell quiescence and maintenance under proliferative stress remain largely unknown. Here we found that cholinergic signals preserve haematopoietic stem cell quiescence in bone-associated (endosteal) bone marrow niches. Bone marrow cholinergic neural signals increase during stress haematopoiesis and are amplified through cholinergic osteoprogenitors. Lack of cholinergic innervation impairs balanced responses to chemotherapy or irradiation and reduces haematopoietic stem cell quiescence and self-renewal. Cholinergic signals activate α7 nicotinic receptor in bone marrow mesenchymal stromal cells leading to increased CXCL12 expression and haematopoietic stem cell quiescence. Consequently, nicotine exposure increases endosteal haematopoietic stem cell quiescence in vivo and impairs hematopoietic regeneration after haematopoietic stem cell transplantation in mice. In humans, smoking history is associated with delayed normalisation of platelet counts after allogeneic haematopoietic stem cell transplantation. These results suggest that cholinergic signals preserve stem cell quiescence under proliferative stress.

H aematopoietic stem cell transplantation (HSCT) is routinely performed to regenerate the haematopoietic and immune systems of patients with cancer, immune or metabolic disorders. However, haematopoietic stem cell (HSC) responses are heterogeneous, but the underlying mechanisms are not yet clear. HSCs have reduced quiescence and cannot self-renew in peripheral blood, compared with bone marrow (BM). Sympathetic noradrenergic fibres innervate the BM 1 and regulate traffic and activation of HSCs and leucocytes [2][3][4][5] . Additionally, stress-induced noradrenergic activity increases HSC proliferation 6 . However, preserving HSC quiescence is critical to prevent HSC attrition, but the mechanisms regulating HSC quiescence under stress are understudied 7,8 . Cholinergic signals cooperate with noradrenergic signals to regulate HSC and leucocyte migration 9 . However, whether cholinergic signals regulate HSC proliferation has not been described. Noradrenaline and acetylcholine (ACh) are generally postganglionic neurotransmitters of (noradrenergic) sympathetic nervous system (SNS) and (cholinergic) parasympathetic nervous system, respectively. However, some sympathetic neurons convert to cholinergic phenotype during postnatal development in sweat glands and the periosteum 10 . Furthermore, the function of sympathetic cholinergic fibres in bone has remained unknown 11 .
Here, we show the source and function of cholinergic signals during regenerative haematopoiesis. Cholinergic-neural signals activate bone-associated nestin + BM mesenchymal stem cells (BMSCs) and regulate HSC quiescence locally in endosteal BM niches. These results illustrate the regulation of stem cell quiescence by the cholinergic system. This mechanism seems to allow stem cells to meet physiological demands and respond to stress, without losing potency.

Results
Cholinergic expansion during stress haematopoiesis. The autonomic nervous system has been evolutionarily selected to efficiently respond to stress. Therefore, we have studied the role of sympathetic cholinergic signals during stress haematopoiesis. WT mice underwent BM transplantation (BMT) following lethal irradiation. BM ACh concentration transiently increased 2 weeks after BMT (Fig. 1a), suggesting a role for cholinergic signals during stress haematopoiesis. To investigate the source of BM ACh, we performed genetic lineage tracing of neuroglial cells in Wnt1-Cre2 mice, and cholinergic cells using ChAT-Ires-Cre mice and ChAT-Gfp mice. Matching the transient ACh increase, cholinergic nerve fibres peaked 2 weeks after transplantation ( Fig. 1b and Supplementary Fig. 1a). Signalling through the GDNF family receptor alpha-2 (GFRα2) promotes the survival of cholinergic neurons 12,13 . Matching the transient increase in cholinergic innervation, BM GFRα2 mRNA expression increased 9-fold 2 weeks after transplantation (Fig. 1c). In sharp contrast, BM tyrosine hydroxylase (TH) + noradrenergic fibres decreased steadily over 4 weeks ( Fig. 1d and Supplementary Fig. 1b). In a separate study, we found that cholinergic neural signals are transmitted in the skeletal system through cholinergic osteoprogenitors. Therefore, we measured ChAT + osteoprogenitors and found them similarly expanded 2 weeks after irradiation ( Fig. 1e-h and Supplementary Fig. 1c-e). These results suggest that both neural and non-neuronal BM cholinergic signals increase during haematopoietic regeneration.
Cholinergic regulation of HSC quiescence. Since GFRα2 is required for the survival of cholinergic neurons 12,13 , we used Gfra2 −/− mice as a model to study the cholinergic regulation of . a, c, g, h Each dot is a mouse. Data are mean of biological replicates ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. ANOVA and Tukey's multiple comparisons test.
HSCs. First, we analysed HSC proliferation separately in the endosteal or central BM of Gfra2 −/− mice and WT mice ( Supplementary Fig. 2a). In agreement with a previous study 14 , BM lin − sca1 + ckit + (LSK) CD150 + CD48 − HSCs were >5-fold more abundant in the central WT BM, which contained 7-fold more nucleated cells. However, the frequency of quiescent HSCs was 4-fold higher in the endosteal WT BM. Notably, Gfra2 −/− mice showed 5-fold-reduced frequency of quiescent HSCs in the endosteal BM, whereas HSC proliferation remained unchanged in the central BM (Fig. 2a, b). An identical result was obtained using a different marker combination (LSK Flt3 − ) to label HSCs (Fig. 2c).
We have previously shown that decreased parasympathetic activity in Gfra2 −/− mice derepresses sympathetic activity and causes abnormal BM egress of HSCs and leucocytes via sympathetic activation of the β 3 -adrenergic receptor 9 (encoded by the Adrb3 gene). To investigate the possible contribution of the noradrenergic system to decreased HSC quiescence in the endosteal BM of Gfra2 −/− mice, we intercrossed these mice with Adrb3 −/− mice. Unlike HSC mobilisation 9 , increased endosteal HSC proliferation was not the consequence of derepressed noradrenergic activity in Gfra2 −/− BM because it was not rescued in Gfra2 −/− Adrb3 −/− compound mice (Fig. 2c). These results suggest that cholinergic signals inhibit HSC proliferation locally in the endosteal BM niche.
Niche α7nAChR promotes HSC quiescence. GFRα2 has been shown to promote HSC self-renewal and ex vivo expansion through its co-receptor RET expressed in HSCs 15,16 , but its role during regenerative haematopoiesis is unclear. Furthermore, GFRα2 protein was detected on the membrane surface of some (<5%) mature haematopoietic cells but was not detectable in immunophenotypically defined HSCs ( Supplementary Fig. 2b, c). To clearly dissect haematopoietic-cell-autonomous and HSC-extrinsic regulation in Gfra2 −/− mice, we generated chimeric mice through long-term transplantations of BM cells into lethally irradiated recipients. The decreased quiescence of endosteal HSCs in Gfra2 −/− mice (see Fig. 2c) was not observed in WT mice carrying Gfra2 −/− haematopoietic cells ( Supplementary Fig. 2d, e), suggesting that HSC quiescence is extrinsically regulated by cholinergic signals through the microenvironment.
To further investigate how cholinergic signals regulate CXCL12, we treated WT mice with cholinergic antagonists selective for nicotinic or muscarinic receptors. Only the nicotinic antagonists decreased BM CXCL12 content by 40% (Fig. 3d), suggesting that cholinergic signals are transduced by nicotinic receptors in the BM microenvironment. Therefore, we treated different BM stromal cells with acetylcholine or nicotine ( Supplementary Fig. 3a). CXCL12 production was highest in MS-5 cells (resembling nestin + BMSCs) and was equally induced by ACh and nicotine, but was not affected by the muscarinic receptor antagonist atropine ( Fig. 3e; Supplementary Fig. 3b). These results are consistent with more abundant expression of nicotinic receptors than muscarinic receptors in Nes-GFP + BMSCs 19 . Among nicotinic receptors, mRNA expression of α 7 (Chrna7) was high in Nes-GFP + cells ( Supplementary Fig. 3c) 19 . Moreover, CXCL12 secretion was 5-fold lower in primary BM cultures from Gfra2 −/− mice or Chrna7 −/− mice (Fig. 3f), suggesting that cholinergic signals regulate BM CXCL12 expression through the α 7 nicotinic ACh receptor (α7nAChR). Supporting this possibility, CXCL12 was 3-fold lower in the BM extracellular fluid from Chrna7 −/− mice (Fig. 3g). Together, these results suggest that cholinergic signals induce CXCL12 expression via α7nAChR in niche cells. Preserved HSC quiescence after chemotherapy. Since CXCL12 is essential to preserve HSC quiescence 17,18 , we asked whether its regulation by cholinergic signals might affect HSC quiescence and activation during emergency haematopoiesis. 5-Fluorouracil (5-FU) is commonly used to treat cancer since it triggers apoptosis of proliferative cells but spares quiescent cells 21 . BM Gfra2 and its ligand neurturin (Nrtn) mRNA expression increased 7 days after 5-FU (Fig. 3h, i), resembling the NRTN-GFRα2 induction after BMT. Gfra2 −/− mice and Chrna7 −/− mice exhibited a decreased ability to regulate haematopoietic recovery and fine-tune immune responses acutely after 5-FU (14-18d); in contrast, haematopoiesis normalised in these mice after 20d, when stress haematopoiesis reverted to homeostasis (Fig. 3j), underscoring the role of cholinergic signals during stress haematopoiesis.
To confirm that cholinergic niche regulation is required to preserve HSC quiescence under proliferative stress, Gfra2 −/− or control donor/recipient mice were transplanted with 500 endosteal HSCs and 10 5 congenic BM helper cells; 4 weeks after transplantation, HSC quiescence was compromised only in the cholinergicneural-deficient endosteal BM ( Fig. 3k and Supplementary Fig. 4).
HSCs cycling in cholinergic-deficient niche. To evaluate the functional consequences of decreased HSC quiescence in the cholinergic-neural-deficient niche, we performed HSC long-term competitive repopulation assays using Gfra2 −/− or control mice as donors or recipients. Consistent with increased WT HSC proliferation in cholinergic-deficient niches and with its normalisation in WT niches (see Fig. 3k), multilineage haematopoietic reconstitution from Gfra2 −/− HSCs was unchanged, and that from WT HSCs remained high, 4-16 weeks after transplantation into Gfra2 −/− mice ( Fig. 4 and Supplementary Fig. 6). This was opposite to, and not explained by, the competitive disadvantage of CD45.1 + (compared with CD45.2 + ) cells 25 (Fig. 4, blue lines).
Reduced HSC self-renewal in cholinergic-deficient niche. To determine whether decreased HSC quiescence in Gfra2 −/− mice compromises the HSC self-renewing programme, we compared the gene expression profiles (RNA-Seq) of endosteal and central HSCs isolated from WT mice and Gfra2 −/− mice. Notably, gene set enrichment analysis (GSEA) showed that central (compared with endosteal) BM WT HSCs exhibited increased ribosomal, mitochondrial and GFRα1-related pathways (Fig. 5a, Supplementary  Fig. 7a-d and Supplementary Data 1), suggesting a higher activation and different response to GFR signalling. Furthermore, the expression of target genes of the GFRα1/2 co-receptor RET, which promotes HSC self-renewal and expansion 15,16 , was reduced in HSCs from Gfra2 −/− mice (Fig. 5b). In contrast, myc-related glycolysis and Notch1-dependent pathways, which regulate HSC maintenance 26,27 and GFRα1/2-dependent, but RET-independent, maintenance of cardiac progenitors 28  To investigate the impact of the cholinergic regulation of HSCs on their self-renewal ability, 500 donor-derived HSCs were isolated from the primary recipient mice and were retransplanted (together with congenic helper BM cells) into lethally irradiated secondary recipients of the same genetic background as the primary recipient mice (Fig. 6a). Contrasting their normal reconstitution capacity in primary recipients (see Fig. 4b-g) and consistent with their reduced self-renewing gene programme (see Fig. 5d), endosteal HSCs from Gfra2 −/− mice failed to reconstitute secondary recipients ( Fig. 6b-g). Notably, WT HSCs continued to yield increased reconstitution shortly (4w) after secondary transplantation into Gfra2 −/− mice; however, by 8 weeks their reconstitution capacity dropped sharply as a sign of decreased self-renewal or premature exhaustion ( Fig. 6b-g). Altogether, these results suggest that cholinergic signals preserve HSC quiescence and self-renewal in the endosteal BM niche under proliferative stress.
Nicotine impairs hematopoietic regeneration. Since the cholinergic HSC regulation is transduced by nicotinic receptors in the niche ( Fig. 3d-m and Supplementary Fig. 5), to complement the loss-of-function models, we treated mice with nicotine, as gainof-function. A 3-day nicotine treatment increased quiescent endosteal BM HSCs ( Supplementary Fig. 8). Therefore, transplanted mice were treated with nicotine or vehicle over 7 weeks (Fig. 7a). Decreased blood multilineage reconstitution in nicotinetreated mice (Supplementary Fig. 9a-h) was explained by decreased donor-derived chimerism in peripheral blood, persisting 7 weeks after transplantation ( Fig. 7b-d). This was likely caused by increased quiescence of transplanted HSCs ( Fig. 7e and Supplementary Fig. 9i). As a complementary loss-of-function model, transplanted mice were treated with a nicotinic receptor antagonist (Fig. 8a). Nicotinic receptor blockade increased BM HSC proliferation after transplantation (Fig. 8b, c). These results suggest that nicotinic signalling reduces HSC proliferation and haematopoietic reconstitution after transplantation.
Smoking delays recovery after transplantation. To investigate the possible association of nicotinic signalling with human HSC proliferation after transplantation, we retrospectively analysed the platelet recovery of 248 patients following allogeneic HSC transplantation, taking into consideration their smoking history before transplantation (Supplementary Data 3). Multivariate analysis indicated that normalisation of platelet counts was significantly delayed in previous smokers ( Fig. 9), independently of the type of donor (matched family vs. alternative donor) and the number of CD34 + cells infused, which also influenced platelet reconstitution kinetics, as expected 32,33 . Since increased transplant-related mortality has been noted in smokers treated for chronic myeloid leukaemia 34 , these results suggest that cholinergic nicotinic signalling might impact clinical haematopoietic recovery after transplantation.

Discussion
This study shows that cholinergic signals contribute to preserve HSC quiescence in endosteal BM niches. Cholinergic signalling increases in these niches upon haematopoietic stress (myeloablation, irradiation) and preserves HSC quiescence under proliferative stress, thereby helping protect HSCs from exhaustion.
Daily release of noradrenaline by sympathetic fibres innervating the BM activates the β 3 -adrenergic receptor and reduces CXCL12 expression, permitting HSC egress to the circulation 2 . Moreover, the SNS-β 3 -adrenergic-receptor-CXCL12 axis causes HSC mobilisation and correlates with HSC hyperproliferation under chronic stress 6 . However, preserving HSC quiescence under stress is critical to prevent HSC attrition, but the underlying mechanisms remain largely unknown 7,8 . Besides noradrenergic fibres, cholinergic nerve fibres have been found in periosteal regions and inside the BM during postnatal stages 35,36 , but their function had remained elusive. We recently demonstrated that dual cholinergic signals (acting centrally and peripherally) regulate HSC traffic between the BM and the bloodstream 9 , but whether cholinergic signals regulate HSC quiescence, which is essential to preserve HSC self-renewal, was unknown. Non-myelinating Schwann cells associated with peripheral nerves reduce HSC proliferation 37 , but it remained unclear whether and how different neurotransmitters released by these nerve fibres, and particularly acetylcholine (ACh), regulate HSC quiescence 11 .
The current study expands the cholinergic regulation of HSC and leucocyte traffic 9 to the preservation of HSC quiescence and self-renewal under proliferative stress. The results suggest that the sympathetic nervous system (SNS) can simultaneously promote two opposing processes (stem cell activation/migration and induction of quiescence) through different neurotransmitters in separate BM niches. In this model, a stress-response system (SNS) could simultaneously trigger migration of activated stem cells in one niche (through noradrenaline), but at the same time protect the stem pool from exhaustion by promoting quiescence in another niche (through acetylcholine). This data illustrates a gatekeeper mechanism that can simultaneously trigger migration of activated stem cells in one niche, but at the same time protect the stem pool from exhaustion by promoting quiescence in another niche (Fig. 10).
Importantly, the expression of Gfra2 and its ligand Nrtn, and the presence of cholinergic (but not noradrenergic) fibres increases in the BM under haematopoietic stress induced by myeloablation with 5-FU or after irradiation. Therefore, increased cholinergic nicotinic signalling in endosteal BM niches preserves HSC quiescence during emergency haematopoiesis.
Three independent experimental paradigms of our study (acute myelosuppression in two independent cholinergic-deficient lines, Gfra2 −/− and Chrna7 −/− mice, and BM reconstitution following lethal irradiation) convey the same message-increased HSC proliferation when cholinergic signals are impaired in the BM niche. However, this decreased HSC quiescence is associated with reduced self-renewing gene programme and diminished self-renewal in serial HSC transplantations. Decreased HSC quiescence in Gfra2 −/− mice can be transiently reverted when the haematopoietic cells are transplanted in a WT microenvironment, although HSCs from Gfra2 −/− mice fail to reconstitute secondary recipients. Conversely, the Gfra2 −/− microenvironment decreases WT HSC quiescence and self-renewal in serial HSC transplantations. Altogether, these data points towards the cholinergic regulation of the HSC niche-rather than HSCs themselves-as a gatekeeper mechanism to protect HSC quiescence under stress.
The observed endosteal and central HSC distribution is in agreement with previous studies showing that the majority of HSCs are located in the perisinusoidal (central) niche 14,38 . However, consistently with our work, many studies highlight the presence of a smaller subset of HSCs associated with bone surfaces (endosteal niches), which were initially proposed to harbour quiescent HSCs 39-45 that expand to regenerate the damaged BM 46,47 . Our results support these contentions and more recent descriptions of low-permeability blood vessels (including endosteal arterioles and capillaries named "transition zone vessels") that serve as specialised niches for quiescent HSCs, whilst highly permeable sinusoids allow for HSC activation and trafficking 20,[48][49][50] . We speculate that dynamic regulations in these two niches might regulate the reversible switch between dormant and activated HSCs based on physiological requirements (homoeostasis vs stress) 51,52 .
Our findings reveal Nes-GFP + LepR-Cre-targeted BMSCs as a target for cholinergic HSC regulation in the endosteal BM niche. CXCL12 levels are reduced in Nes-GFP hi cells associated with decreased HSC quiescence in the endosteal BM of cholinergicneural-deficient mice. Since CXCL12 is essential to preserve HSC quiescence 17,18 and our data shows that nicotine induces CXCL12 secretion by BM stromal cells, this appears to be one mechanism of cholinergic regulation of HSC quiescence through the niche. Nes-GFP hi cells have been previously associated with HSC quiescence in arteriolar niches 20 and this might also be the case for endosteal transition zone vessels 48 . It is possible that other stromal cells expressing cholinergic receptors, such as endothelial cells 53 and monocytes/macrophages 54-56 , might be targeted by this cholinergic regulation. Therefore, future studies should elucidate the possible regulation of other HSC niche cells by cholinergic signals.
Previous studies have suggested the expression of cholinergic nicotinic receptors in HSPCs and their possible activation by nicotine 57 . Specifically, α7nAChR has been proposed to mark HSPCs with a prominent role during inflammatory responses 58 . We have shown that cholinergic signals activate α7nAChR and induce CXCL12 expression in endosteal BMSCs as one mechanism promoting HSC quiescence. Indeed, our results show that α7nAChR deletion in LepR-Cre-targeted SSC-enriched cells 22 , which overlap with Nes-GFP + cells 23,24 , or in Nes-Cre ERT2 -targeted HSC niche cells both increase endosteal (not central) BM HSC proliferation. Consequently, nicotine exposure increases endosteal HSC quiescence in vivo and impairs haematopoietic regeneration after HSC transplantation in mice. Our retrospective analysis of platelet recovery after clinical allogeneic HSC transplantation shows a significant delay in the normalisation of platelets in smokers, compared with non-smokers, independently of the stem cell dose and type of donor. Although other factors could contribute to delayed haematopoietic reconstitution in smokers, our data suggest that the cholinergic nicotinic signalling might affect clinical haematopoietic recovery after transplantation. These data are consistent with the reportedly inhibitory effects of nicotine during in vivo haematopoiesis 59 and in vitro megakaryopoiesis 60 , and suggest that the cholinergic nicotinic signalling might affect clinical haematopoietic recovery after transplantation. Our data might relate to the increased transplant-related mortality previously noted in high-dose smokers who received HSC transplantation for the treatment of chronic myeloid leukaemia 34 .
In summary, these results suggest that the cholinergic nicotinic regulation of the endosteal HSC niche preserves HSC quiescence under haematopoietic proliferative stress. They also reveal how adult stem cells can be regulated by different signals from the autonomic nervous system to meet physiological demands at various sites and time, and to efficiently respond to stress without being exhausted.  66 , were kindly provided by Prof. Lynda Bonewald and seeded in collagen I pre-coated-6 well dishes and grown with α-MEM medium supplemented with 5% FBS and 5% iron-supplemented calf serum (Invitrogen). All cultures were maintained with 1% penicillin-streptomycin (Invitrogen) at 37°C in a water-jacketed incubator with 5% CO 2 and 1:5 split with 0.05% trypsin-EDTA (Invitrogen) every three or 4 days, when cells reached about 80% confluence. Routine tests confirmed the absence of mycoplasma contamination in the cultures.  We isolated HSPCs from bone-associated and non-associated marrow fractions as previously described 41 . Briefly, long bones were flushed gently to obtain hematopoietic cells less tightly associated with the bone and the flushed-bones were then crushed in a mortar to obtain hematopoietic cells of the endosteal compartment. Cells were stained with the above-detailed antibodies and analysed by flow cytometry or sorted (FACS Aria cell sorter, BD Bioscience). HSCs were immunophenotypically defined as lin − CD48 − sca-1 + c-kit + CD150 + cells.
To isolate nestin + cells, bones were cleaned off surrounding tissue, crushed in a mortar with a pestle, and digested with collagenase (catalogue number C2674, Sigma; 0.25% collagenase in PBS supplemented with 20% foetal bovine serum) in water bath at 37°C for 30 min with agitation. Cells were filtered through a 40-μm strainer and erythrocytes were lysed as described above. The resulting BM-enriched cell suspensions were pelleted, washed and resuspended in PBS containing 2% FCS for further analyses. BM stromal CD45 − (30-F11) CD31 − (MEC 13.3) Ter119 − cells were further purified according to GFP fluorescence using an LSRFortessa flow cytometer (BD Biosciences) for immunophenotypic analysis, or a BD FACS Aria or BD Influx Sorter (BD Biosciences) for cell sorting.
To separate endosteal and non-endosteal nestin + cells, we gently flushed the long bones as described above. We digested both the flushed fraction and the remaining bone samples with collagenase (catalogue number C2674, Sigma; 0.25% collagenase in PBS supplemented with 20% foetal bovine serum) in a water bath at 37°C for 30 min with agitation.
Choline acetyltransferase + cell detection. To assess which cell populations express choline acetyltransferase (ChAT) under steady state and after stress, transplantations were performed. To assess the stromal ChAT + cells, ChAT-GFP and ChAT-IRES-Cre;Ai14D (CD45.2) recipient mice were split dose irradiated with  In vivo treatments. Haematopoietic recovery in mice was assayed administering a single dose of 5-FU (150 mg/kg, i.p., diluted in PBS), and bleeding mice for the first time before the 5-FU injection and then every 3-4 days for haematological analysis.
To assess the role of Chrnα7 signalling in stromal cells, Chrnα7-floxed;LepR-Cre (CD45.2) recipient mice were split dose irradiated with 12 Gy and transplanted with 2 million CD45.1 bone marrow nucleated cells. Mice were culled at 4 weeks post-transplantation and cell cycle status was analysed.
To generate recipients to assess the role of Chrnα7 signalling in stromal cells, Chrnα7-floxed;Nes-CreERT2 (CD45.2) mice, pregnant dams were treated with 4 mg tamoxifen on days 1 and 3 via oral gavage. Induced recipients mice were split dose irradiated with 12 Gy and transplanted with 2 million CD45.1 bone marrow nucleated cells. Mice were culled at 4 weeks post-transplantation and cell cycle status was analysed.
To assess the role of Chrnα7 signalling in HSCs, CD45.1 recipient mice were split dose irradiated with 12 Gy and transplanted with 2 million Chrnα7-floxed;Vav-Cre (CD45.2) bone marrow nucleated cells. Mice were culled at 4 weeks post-transplantation and cell cycle status was analysed.
For the nicotinic agonist and antagonist experiments, CD45.2 C57BL/6 J mice were treated with 100 μg/ml nicotine (Sigma, N3876-5ML) in 1.5% saccharin (Sigma) via their drinking water. To assess in the transplantation setting, CD45.2 C57BL/6 J recipient mice were split dose irradiated with 9.5 Gy and transplanted with 2 million CD45.1 bone marrow nucleated cells. Mice were administered with 65ug/ml nicotine 24 h later via their drinking water for the remainder of the study, changed every 2-3 days. For the antagonist experiments, CD45.2 C57BL/6 J recipient mice were split dose irradiated with 12 Gy and transplanted with 2 million CD45.1 bone marrow nucleated cells. Mice were treated with 20 mg/kg/day Hexamethonium bromide (i.p.) daily for 30 days.
Immunofluorescence. Immunofluorescence staining of cryostat sections was performed as previously described 72 . Briefly, tissues were permeabilized with 0.1% Triton X-100 (Sigma) for 10 min at RT and blocked with TNB buffer (0.1 M Tris-HCl, pH 7.5, 0.15 M NaCl, 0.5% blocking reagent, Perkin Elmer) for 1 h at RT. Primary antibody incubations were conducted overnight at 4°C. Secondary antibody incubations were conducted for 2 h at RT. Repetitive washes were performed with PBS + 0.05% Triton X-100. Stained tissue sections were counterstained for 5 min with 5 μM DAPI and rinsed with PBS. Slides were mounted in Vectashield Hardset mounting medium (Vector Labs). ChAT-Ires-Cre-positive nerve fibre staining was performed on half-bones according to previous reports (Acar et al., 2015), whereby bones were longitudinally bisected using a cryostat, blocked O/N in staining buffer (5% donkey serum, 0.5% IgePal, 10% DMSO) supplemented with 1% BlokHen (Aves Labs, Cat. No. BH-1001) and stained with primary and secondary antibodies for 3 days in staining buffer with daily intervening washes in PBS at RT. We used the following primary antibodies: tyrosine hydroxylase (1:1000, rabbit polyclonal antibody, Millipore), Gfra2 ELISA. CXCL12 protein levels were measured by conventional ELISA. Briefly, 96well plates were coated overnight at 4°C with 2 µg/ml of monoclonal CXCL12/ SDF-1 antibody (MAB350, R&D Systems). After blocking, BM extracellular fluids were incubated with the antibody for 2 h at room temperature, followed by addition of biotinylated anti-human and mouse CXCL12/SDF-1 antibody (BAF310, RD). Streptavidin-horseradish peroxidase conjugate (RPN1231V, Dako) was used to detect the signal, and the reaction was developed with horseradish peroxidase substrate (TMB, ES001-500ML, Chemicon, Millipore). Standard curve was performed with recombinant SDF-1 alpha (350-NS, R&D).
RNA isolation, reverse transcription and quantitative real-time PCR (qPCR). Total RNA extraction was performed with TRIzol (Invitrogen), followed by treatment with DNase to eliminate contaminating genomic DNA with RNase-free DNase Set (Qiagen) and RNA clean-up with RNeasy mini kit (Qiagen). Alternatively, for small cell numbers, RNA was isolated using the Dynabeads® mRNA DIRECT™ Micro Kit (Invitrogen). Reverse transcription was performed using the Reverse Transcription System (Promega), following the manufacturer's recommendations. The expression level of each gene was determined by the relative standard curve method, using a standard curve prepared from serial dilutions of a mouse or human reference total RNA (Clontech). The expression level of each gene was calculated by interpolation from the standard curve. All values were normalised to Gapdh as the endogenous control. The sequence of oligonucleotides used for quantitative real-time RT-PCR are available in Supplementary Data 4.
RNAseq sample preparation. Pools of 30 viable LSK CD48 − CD150 + cells were sorted (BD Influx™ cell sorter) into 4 µl lysis buffer (0.5 U/µl SUPERase In RNase Inhibitor in 0.2 % (v/v) Triton X-100) containing 12.5 mM DTT and 2.5 mM dNTP. RNAseq was performed following Smart-seq2 protocol 73 . Briefly, RNA was primed with oligo-dT primers and reverse transcribed using Superscript II Reverse Transcriptase (200 U/μl, Thermo Fisher 18064071). KAPA HiFi Hotstart  ReadyMix (KAPA Biosystems, KK2601) and IS PCR primers were used to amplify cDNA, after which cDNA was purified with Agencourt AMPure XP beads (Beckman Coulter, A63881). cDNA quality was checked with Agilent Bioanalyser 2100 using Agilent High Sensitivity DNA chip and quantity was measured with Quant-iT™ Picogreen double-stranded DNA assay kit (Thermo Fisher P7589). Pooled libraries of four replicates per sample were prepared using the Illumina Nextera XT DNA preparation kit. Amplified libraries were purified using Agencourt AMPure XP beads, quality-checked using Agilent high-sensitivity DNA chip on Agilent Bioanalyser 2100 and quantified using KAPA qPCR quantification kit (KAPA Biosystems, KK4824). Sequencing of three replicates per condition were performed on the Illumina Hi-Seq 4000 (single end, 50 bp read length) by the CRUK Cambridge Institute Genomics Core facility.
Bioinformatic analysis of RNAseq. RNAseq reads were aligned to Mus musculus genome (Ensembl version 38.81) using GSNAP (version 2015-09-29) with parameters (-B 5 -t 24 -n 1 -Q -N 1). Reads in features were counted with htseq-count (HTSeq version 0.5.3p3) with the parameter (-s no). Quality control was performed with the following cut-offs: more than one and a half million uniquely mappable reads, <20% of reads mapping to mitochondrial genes over mitochondrial + nuclear genes and >8500 high coverage genes identified. Counts were normalized using size factors as calculated by DESeq2 using a 10% FDR, and then log10 transformed. Highly variable genes were selected using the method described by Brennecke et al. PCA was then performed in R using the prcomp function.
Gene set enrichment analyses were performed as described 74 (http:// www.broadinstitute.org/gsea/index.jsp), using a weighted statistic, ranking by signal to noise ratio, 1000 gene-set permutations, and a custom gene set database gene lists manually compiled from the literature.
GEO accession number-GSE94078 Statistical analysis and reproducibility. We used similar response variables for the calculation of sample size. For each hypothesis, the total cell number and composition of the different cell populations was determined by flow cytometry, in peripheral blood (after each treatment) and in the bone marrow (at the end of the experiment). The compared groups were set similarly in all procedures. For each scenario, we established the number of conditions in which to test the hypothesis, and two groups were randomly assigned to each condition (control group and experimental group). Results were scored blindly. Data shown in figures are means ± SEM; n and p-values are indicated in every single figure.
One-way ANOVA and Bonferroni comparison were used for multiple group comparisons, and unpaired two-tailed t-tests for two-group comparisons. The data met the assumptions of the tests. Significant statistical differences between groups were indicated as: *P < 0.05; **P < 0.01; ***P < 0.001. Statistical analyses and graphics were carried out with GraphPad Prism software and Microsoft Excel.