Human bone marrow angiogenesis: in vitro modulation by substance P and neurokinin A


Dr Laurent Pelletier, Unité INSERM 318, UJFG–CHU A. Michallon BP 217, 38 043 Grenoble cedex 9, France. E-mail:


Summary. We have previously described a culture system for human bone marrow endothelial cells that organize into capillary tubes associated to pericytes. In the present work, we used this model to assess the angiogenic properties of tachykinins, which have been demonstrated to be involved in neuro–immuno–haematopoietic interactions. The substance P (SP) and neurokinin A (NKA) were similarly potent at increasing in vitro angiogenesis, via NK1 and NK2 receptors respectively. These mediators were not produced by cells in culture, suggesting that in vivo they may be released by nerve fibres in the bone marrow. Therefore, we looked for in situ innervation of the human bone marrow, unknown to date, using immunohistochemistry techniques. As in rodents, arterioles were largely innervated, associated with between one and 10 nerve fibres. Capillary innervation was more restrictive as a unique thin nerve fibre was found in the vicinity of only 6% of these vessels. Finally, no nerve fibres were observed in the vicinity of sinus walls. In conclusion, both in vitro results and the anatomical display of nerve fibres suggest a role in human bone marrow for the vasoactive neuropeptides SP and NKA, which were secreted into a perivascular location. These neural mediators might modulate blood flow in the bone marrow both in the short term by adjusting vascular tone and in the long term by inducing angiogenesis.

The haematopoietic microenvironment is comprised of non-haematopoietic structures found in the haematopoietic cord or marrow logettes. These elements contain the extracellular matrix, adipocytes, bone, vascular network and innervation (Lichtman, 1984). The vascular network of the bone marrow comprises arteries, arterioles, capillaries and venous sinusoids. Luminal endothelial cells are associated with adventitial abluminal cells that contain α-smooth-muscle-type actin. In arteries and arterioles, the media is made of vascular smooth muscle cells. Capillaries are intermittently covered with pericytes while sinuses are lined by abluminal myoid or reticular cells (Galmiche et al, 1993; Charbord et al, 1996). Pericapillary pericytes and parasinusal myoid cells probably fulfil two functions as they are anatomically associated with both endothelial cells and haematopoietic cells. The presence of cells expressing α-smooth muscle actin raised the question of their potential contractile role. Hypothetically, they may respond to endothelium-derived tone regulators such as nitric oxide, prostanoids and endothelin, or to other mediators carried by blood flow. In addition, they could also be stimulated by neuromediators released by nerve fibres located in the vicinity of blood vessels.

Neuromediators could contribute to the regulation of bone marrow blood flow that is instrumental in the control of haematopoiesis. Several observations have thus shown that blood flow in bone marrow is modulated according to the haematopoiesis turnover rate. For example, granulocye-colony stimulating factor (G-CSF) simultaneously increased marrow blood flow and blood content in granulocytes (Iversen et al, 1993). Also, erythropoietin (EPO) or anaemia (either hypo- or normovolumic) caused increases in blood flow (Iversen et al, 1992). A chronic increase in blood flow of bone marrow has also been observed in myelofibrosis (Van Dyke et al, 1970; Lahtinen et al, 1982). Sinuses were distended and a marked proliferation of endothelial cells resulted in a larger vascularity in the bone marrow (Reilly et al, 1985; Wolf & Neiman, 1985). Angiogenesis appeared, therefore, to be at the origin of a long-term increase in bone marrow blood flow.

Vasoactive peptides such as the tachykinin substance P (SP) have angiogenic properties in vitro and in vivo. SP stimulates human or bovine endothelial cell proliferation (Ziche et al, 1990), migration (Ziche et al, 1991), organization into capillary-like tubes in Matrigel (Wiedermann et al, 1996) and angiogenesis in rabbit cornea or rat sponge models (Ziche et al, 1990; Fan et al, 1993). Therefore, neuropeptides such as SP with vasoactive and angiogenic functions could participate in regulation of haematopoiesis by modulation of the microenvironment.

Bone marrow innervation has been documented in laboratory animals and particularly in rats where numerous myelinated and non-myelinated nerve fibres were observed (Calvo, 1968). The nerve fibres were found to line the vascular network from the entry of the nutrient artery into the haematopoietic cord to some capillaries and sinuses, while some fibres were found in the parenchyma. Ultrastructural studies have further shown that efferent nerve endings came into contact with stromal elements (Yamazaki & Allen, 1990). This raised the hypothesis of a neural control of haematopoiesis by microenvironmental cells. Following the anatomical description of bone marrow innervation in rat and mouse, functional experiments were conducted in vivo using chemical or surgical denervation and stimulation of nerve cells. This established that nerve fibres were instrumental not only in the release of mature and progenitor cells into the bloodstream but also in the amount of progenitor cells in the bone marrow (Webber et al, 1970; Maestroni et al, 1992; Afan et al, 1997).

Neuropeptide-containing nerve fibres have been extensively studied in primary and secondary rodent lymphoid organs (for review see Stevens-Felten & Bellinger, 1997). The presence of SP-, neurokinin A (NKA)-, calcitonin gene-related peptide (CGRP)- or vasoactive instestinal peptide (VIP)-containing nerve fibres in lymphoid organs supported the hypothesis of their role as neuro-immunological links. SP-, NKA-, CGRP- and VIP-containing nerve fibres have also been detected in rodent bone marrow, both in a perivascular location and in the parenchyma (Bjurholm et al, 1988; Weihe et al, 1990; Hukkanen et al, 1992). In vivo deletion of SP and CGRP abrogated normal blood cell production (Broome et al, 2000). These experiments demonstrated the involvement of neuropeptides in bone marrow physiology. Moreover, tachykinins contribute to human haematopoiesis regulation (for review see Rameshwar & Gascon, 1997). In vitro, SP stimulated and NKA inhibited granulocyte output, whereas both mediators stimulated the release of cells of the erythroid lineage. In both cases, tachykinins interacted with specific receptors expressed at the surface of stromal cells that, in turn, produce stimulatory [interleukin 3 (IL-3)] or inhibitory (transforming growth factor-β, macrophage inhibitory protein-1α) cytokines. Therefore, an increasing number of observations support a function for neuropeptides in human bone marrow. However, to our knowledge, neuropeptide-containing fibres have not been detected in human bone marrow to date.

In this study, we first used a culture model to examine whether tachykinins contribute to human bone marrow angiogenesis as they do in other tissues. The finding that SP and NKA stimulated the growth of endothelial tubes via NK1 and NK2 receptors, respectively, persuaded us to look for bone marrow innervation. Using trephine iliac crest biopsies, we showed that the vascular network in human bone marrow is innervated. Although we did not focus on neuropeptide-containing nerve fibres, these observations strengthen the hypothesis of a role for tachykinins of bone marrow origin in microenvironment function.

Materials and methods

Bone marrow endothelial cells isolation and culture. Sternal bone marrow aspirates were obtained from patients undergoing sternotomy for cardiac surgery and were collected into bottles containing heparin (100 I.U./bottle). Endothelial cell isolation and culture was performed as previously described (Rafii et al, 1994; Pelletier et al, 2000). After centrifugation, bone marrow supernatants were digested in collagenase (Type I collagenase 0·1%; Sigma, St Louis, MO, USA). Cells were resuspended and selected using magnetic beads (M280 tosylactivated dynabeads; Dynal, Oslo, Norway) coated with Ulex Europaeus Agglutinin-1 (UEA-1; Vector Laboratories, Burlingame, CA, USA) added at a ratio of 1·5 beads per cell. The targeted cells were recovered on a magnetic particle concentrator (MPC; Dynal) and resuspended in E-STIM medium (Becton Dickinson, Bedford, MA, USA), containing 1 ng/ml recombinant human vascular endothelial growth factor (Becton Dickinson).

Selected cells were cultured on rat tail type I collagen (12 µg/cm2; Becton Dickinson) and human fibronectin-coated (50 µg/cm2; Becton Dickinson) 25 cm2 flasks or Lab Tek (8 well permanox slides; Nunc, Naperville, IL, USA).

In vitro angiogenesis assays. Culture treatments with neuropeptides were performed daily from d 7 (75% confluency) to d 12. Neuromediators, antagonists or antibodies were added as 1/100 of the culture medium volume in vehicle: phosphate-buffered saline (PBS) with 0·5% (w/v) bovine serum albumin (BSA; Sigma). Controls contained isotype or normal serum for immunoneutralization assays. Acetylcholine was obtained from Fluka (Buchs, Switzerland), and SP, NKA, VIP and CGRP were obtained from Sigma.

The NK1 receptor antagonist SR140333 and the NK2 receptor antagonist SR48968 (both provided by Professor C. Advenier) were used at a 10−6 mol/l final concentration and added 2 h before the neuropeptides.

Rabbit antiserum to SP was used at 1/200 final concentration. Measures were performed in triplicate on at least four different cultures. Effects of the treatments were assessed after immunolabelling the endothelial cells by comparing the number of endothelial tubes between treated and control wells.

Immunofluorescence studies on endothelial cell cultures. Endothelial cells were labelled with anti-von Willebrand Factor (VWF) antibodies for measuring tube development.

The culture medium was removed from Lab Tek culture chambers and the monolayer was washed once with PBS prior to fixation. Cells were fixed in cold methanol for 30 min at 4°C. Incubations with primary and secondary antibodies were performed at room temperature for 30 min; each step was followed by two washes in PBS. Primary polyclonal rabbit anti-VWF antibody (Dako, Glostrup, Denmark) and secondary tetramethylrhodamine B isothiocyanate (TRITC)-conjugated mouse anti-rabbit antibody (Sigma) were used at 1/225. An Aristoplan microscope (Leitz Wetzlar, Wetzlar, Germany) was used for observation.

Endothelial tubes were counted on the entire surface of each well (0·8 cm2/well). Where these structures formed networks, each branch was considered as one tube (Pelletier et al, 2000). Values found for treated culture numbers are expressed as a percentage of values found for control conditions.

Immunochemistry on bone marrow. Fourteen bone marrow trephines without histological inflammatory or tumoral changes were obtained from Department of Pathology (J. Minjoz Hospital, Besançon). They were fixed in Shaffer solution, decalcified in 7% EDTA and embedded in paraffin.

Double-labelling immunohistochemistry (CD34/synaptophysin) was performed on sample sections as follows: 5 µm sections were dewaxed in toluene and acetone, and rehydrated in Tris-buffered saline. After microwave heating (3 × 5 min in pH 6 citrate buffer), sections were incubated with primary monoclonal antibody to CD34 for endothelial cell labelling (Immunotech, Westbrook, ME, USA, 7860, dilution 1/50, 30 min). Staining was performed using the avidin–biotin–peroxidase technique (Kit HRP 2391; Immunotech) and diaminobenzidine substrate.

After a complementary treatment with trypsin, nerve fibres were labelled when sections were incubated with the antibody to synaptophysin (A010, dilution 1/50; Dako) and stained with avidin biotin–alcalin phosphatase technique (Kit AP 2392; Immunotech) and Fast Red substrate (Sigma). Anti-PS100 (directed against a protein expressed by Schwann's cells of the peripheral nervous system; Z0311, dilution 1/1500; Dako) and anti-neurofilaments (directed against the intermediate filament protein present in neuronal processes and peripheral nerves; M762, dilution 1/400; Dako) were also used. Labelling specificity of CD34 and nerve fibre antigens was assessed using control (irrelevant) isotypic antibodies.

Sinuses were identified as enlarged vessels delimited by a unique CD34-positive endothelial layer. Vascular structures were considered as capillaries when they consisted of only endothelial cells associated or not with a discontinuous abluminal pericyte layer. Arterioles and arteries were conversely surrounded by one or several layers containing smooth muscle cells.

The percentage of innervated vessels was determined as the ratio of vessels associated with synaptophysin-positive nerve fibre(s) on total vessel number.

Statistical analysis. Values are expressed as mean ± standard error of the mean (SEM). Statistically significant differences between series were assessed by analysis of variance (anova) using statview software (Abacus Concepts, Berkeley, USA). Mean differences were considered as significant when P < 0·05.


In vitro angiogenesis assays

The angiogenic effect of neuromediators was assessed by counting capillary tubes in endothelial cell cultures in the presence or absence of mediators. The neuromediators tested were acetylcholine, SP, NKA tachykinins plus the neuropeptides VIP and CGRP.

As shown in Fig 1A, the daily addition of SP induced a significant increase in tube number. The dose–response curve was bell shaped, with a maximum significant (P < 0·01) stimulation obtained at 10−8 mol/l (+39·5 ± 13·8%). At 10−9 mol/l, the development of tubes was also significantly increased (+27·6 ± 9·0%). As 10−8 mol/l was the concentration that produced the greatest tube development, it was subsequently used for all studies. SP neutralization with a specific rabbit immune serum abolished the SP-induced increase in tube formation (Fig 1B), while addition of neutralizing antiserum had no effect on tube development in untreated cultures. As the biological effects of SP have been reported to be mediated by binding to NK1 receptors (Ziche et al, 1990, 1991; Parenti et al, 1996; Rameshwar et al, 1997), we blocked the NK1 receptor by adding the specific inhibitor SR140333. SR140333 abolished the stimulating effect of SP (P < 0·01; Fig 1B), while addition of SR140333 had no effect in the absence of SP.

Figure 1.

Angiogenic effect of SP and NKA in vitro. SP (A) and NKA (B) stimulated the growth of capillary-like tubes with a maximal effect at 10−8 mol/l and 10−9 mol/l respectively. (C) SR140333 or SP immuno-neutralization abolished the stimulation by SP (10−8 mol/l), while SR48968 (10−6 mol/l) reduced the effect of the NKA (10−9 mol/l) (D). Significant differences are indicated by * for P < 0·05 and ** for P < 0·01.

NKA induced a stimulating effect similar to that of SP, although NKA significantly increased tube development at lower concentrations. As observed for SP, the NKA dose–response curve was bell shaped (Fig 1C). NKA increased the tube number at 10−9 mol/l and 10−10 mol/l, by 36·6 ± 20·5% (P < 0·01) and 23·8 ± 18·3% (P < 0·01) respectively. As NKA was effective at 10−9 mol/l, we used this concentration in further experiments.

While SP binds preferentially to NK1 receptors, NKA binds to NK2 receptors (Rameshwar et al, 1997). The addition of 10−6 mol/l of the NK2 receptor antagonist SR48968 reduced the stimulatory effect of NKA (Fig 1D) by 76·1% (P < 0·01). SR48968 alone did not affect the number of tubes.

In situ staining of bone marrow innervation

In preliminary studies (data not shown), nerve elements were stained on marrow biopsies using three antibodies: anti-PS 100, anti-neurofilaments and anti-synaptophysin. Anti-synaptophysin staining detected the greatest number of nerve fibres. In addition, there was no obvious difference in the nature and localization of stained elements when the three antibodies were compared. Therefore, subsequent observations were made using anti-synaptophysin antibody. The study of vascular and nervous elements was thus performed on 14 human trephine bone marrow biopsies. Vascular structures were also counted after staining with anti-CD34 antibody. Nerve fibres were observed in contact with blood vessels or in a close perivascular vicinity. Arteries and arterioles were richly innervated as neural fibres were found alongside 72·5 ± 6·3% of these vessels (Fig 2A and B). The number of nerve fibres in the adventitia of arteries increased parallel to their size, from one to 10 fibres, and sometimes associated in bundles (Fig 2A). Only a small proportion of capillary sections (6·1 ± 1·5%) were associated with one nerve fibre (Fig 2C and D). In rare instances where arterioles or capillaries were longitudinally cut, nerve fibres lined the wall of the vascular structures (Fig 2D). The diameter of synaptophysin-positive fibres was also greater in periarteriolar locations compared with the capillary-associated nerve fibres. In these specimens, no nerve fibres were observed in a parasinusal location or in the parenchyma.

Figure 2.

Innervation of vascular structures in human bone marrow. Double labelling of endothelial cells with anti-CD34 antibody (brown) and of nerve fibres with anti-synaptophysin (red). Arteries and arterioles are lined on 72% of their length by nerve fibres grouped in bundles for arteries (A; ×100) or isolated for arterioles (B; ×250). Capillaries are lined on 6% of their length by synaptophysin-positive nerve fibres. Longitudinal section of a capillary (C; ×250): a nerve fibre is located between pericytes which form an intermittent covering of the abluminal side of CD34+ endothelium. Transverse section of a capillary (D; ×200): nerve fibres can also be seen in contact with endothelial cells.


Using a culture system that enabled the study of angiogenesis in human bone marrow, we assessed the angiogenic activity of several neuropeptides.

We found that SP had angiogenic activity on human bone marrow endothelial cells and that this effect was mediated by NK1 receptors. NK1 receptors are expressed by cultured cells, as confirmed by reverse-transcription polymerase chain reaction (data not shown). These results are in keeping with other reports using different models. SP was angiogenic in vivo in rabbit cornea and rat sponge systems (Ziche et al, 1990; Fan et al, 1993). In vitro, it stimulated human or bovine endothelial cell proliferation (Ziche et al, 1990), migration (Ziche et al, 1991) and organization into capillary-like tubes in Matrigel (Wiedermann et al, 1996). These effects were transduced via the NK1 receptor.

In our culture model, endothelial and mesenchymal cells are spatially and functionally associated (Pelletier et al, 2000). Endothelial cells are organized in branching tubes, among stromal cells that differentiate into pericytes expressing α-SM-type actin along the tubes. This model combined, therefore, two major cellular components of the bone marrow microenvironment. We observed that angiogenic effects can be the result of the stromal cell responsiveness to mediators. An indirect action of SP has been described for the growth of human marrow haematopoietic progenitors. This involved the release of IL-1, IL3, IL6, stem cell factor (SCF), G-CSF and granulocyte-monocyte colony-stimulating factor (GM-CSF) by stromal cells (Rameshwar et al, 1993, 1994; Rameshwar & Gascon, 1995; Hiramoto et al, 1998). We have previously shown that SCF, G-CSF and GM-CSF were potent angiogenic factors in our culture system (Pelletier et al, 2000). SP might, therefore, stimulate the development of tubes through both a direct action on endothelial cells and/or indirectly by prompting stromal cells to produce haematopoietic cytokines with angiogenic properties.

We also observed that NKA was angiogenic to human marrow microvascular endothelial cells, acting via NK2 receptors. SP and NKA thus have similar angiogenic properties, although NKA seemed to be more efficient than SP at lower concentrations.

Other neuromediators, such as VIP and CGRP, also showed weak but significant angiogenic properties in this culture system (data not shown).

To our knowledge, the angiogenic activity of NKA has not yet been described. Moreover, it has been shown that both NKA or NK2 agonists were devoid of growth promoting activity on human or bovine endothelial cells (Haegerstrand et al, 1990; Ziche et al, 1990; Villablanca et al, 1994). Therefore, it should be considered that NKA either is specifically angiogenic in the marrow microvascular endothelial cells or that this NKA angiogenic effect is mediated by stromal cells.

Tachykinins are produced by nerve cells, but SP can also be synthesized and released by endothelial cells. Therefore, they may act as an autocrine/paracrine mediator (Ralevic et al, 1990). However, neutralization of endogenous SP or blockade of NK1 receptors did not modify basal development of tubes, indicating that SP was not secreted by cultured cells in the supernatant. Therefore, NKA and SP could act in a paracrine way in vivo. As a consequence, we studied human bone marrow innervation by immunohistochemistry. Nerve fibres were always found in perivascular locations. While 75% of arteries or arterioles were associated with one to 10 nerve fibres, only 6% of capillaries were in contact with usually only one thin nerve fibre. Conversely, no nerve fibres were observed along sinuses. Our results extend to humans, data previously obtained only in laboratory animals showing nerve fibres all along the medullary vascular network from nutrient artery to small capillaries (Calvo, 1968; Yamazaki & Allen, 1990). However, whereas in rodents a small number of nerve fibres were found close to the sinusal wall or in the parenchyma, we did not observe similar innervation in human marrow. Further studies should determine the nature of these nerve fibres and assess, in particular, the innervation of human bone marrow by neuropeptide-containing fibres.

In conclusion, our data indicate that SP is angiogenic in human bone marrow, as already described in other systems. NKA is similarly angiogenic in the human model, which was an unknown activity for this tachykinin. Our results reinforce reports that tachykinins have a function in human bone marrow. Finally, we have shown that human bone marrow is innervated and that nerve fibres are interwoven with the vascular network. The release of tachykinins in such locations would modulate haematopoiesis by both stimulating angiogenesis and controlling the production of stromal cell cytokines.


We are grateful to Professor C. Advenier for the gift of NK1 and NK2 antagonists. This work was supported by grants from Institut National de la Santé et de la Recherche Médicale (CRI N°950401) and from Fonds d'Organization et de Recherche en Transfusion Sanguine. L. Pelletier was supported by a grant from the Ministère de l'Enseignement Supérieur et de la Recherche.