Mesenchymal stem/stromal cells (MSC) are currently the best candidate therapeutic cells for regenerative medicine related to osteoarticular, muscular, vascular and inflammatory diseases, although these cells remain heterogeneous and necessitate a better biological characterization. We and others recently described that MSC originate from two types of perivascular cells, namely pericytes and adventitial cells and contain the in situ counterpart of MSC in developing and adult human organs, which can be prospectively purified using well defined cell surface markers. Pericytes encircle endothelial cells of capillaries and microvessels and express the adhesion molecule CD146 and the PDGFRβ, but lack endothelial and haematopoietic markers such as CD34, CD31, vWF (von Willebrand factor), the ligand for Ulex europaeus 1 (UEA1) and CD45 respectively. The proteoglycan NG2 is a pericyte marker exclusively associated with the arterial system. Besides its expression in smooth muscle cells, smooth muscle actin (αSMA) is also detected in subsets of pericytes. Adventitial cells surround the largest vessels and, opposite to pericytes, are not closely associated to endothelial cells. Adventitial cells express CD34 and lack αSMA and all endothelial and haematopoietic cell markers, as for pericytes. Altogether, pericytes and adventitial perivascular cells express in situ and in culture markers of MSC and display capacities to differentiate towards osteogenic, adipogenic and chondrogenic cell lineages. Importantly, adventitial cells can differentiate into pericyte-like cells under inductive conditions in vitro. Altogether, using purified perivascular cells instead of MSC may bring higher benefits to regenerative medicine, including the possibility, for the first time, to use these cells uncultured.
Despite an ever developing interest in the promising properties of mesenchymal stem cells (MSC), the true nature and identity of these cells has always been elusive. MSC, which were originally extracted from adult bone marrow , are now obtained from many tissues and are defined by their ability to adhere to plastic in culture, expression of a set of surface markers and capacity to differentiate into mesodermal cell lineages. Yet, these cells remain heterogeneous and not well characterized [2-4]. Is the heterogeneity of cultured MSC reflective of the existence of distinct MSC progenitors in situ? We have hypothesized that because of their presence in virtually all organs, blood vessels represent a reservoir of these mesodermal stem cells. The presence of related progenitors, named mesoangioblasts because of their ability to differentiate into endothelial cells and other mesodermal lineages, has been already demonstrated in the embryonic dorsal aorta [5, 6]. Moreover, haematopoietic stem cells are generated from a hemogenic endothelium in the embryonic aortic wall . Finally, the identification of mesodermal progenitors in the tunica adventitia suggests that the outmost layer of blood vessels also contributes to tissue homeostasis and repair . These results led us to further investigate the presence of MSC within the vascular wall in various developing and adult human organs. Small blood vessels are formed by endothelial and mural cells. Charles Marie Benjamin Rouget described for the first time mural cells of capillaries as ‘non-pigmented adventitial cells’ or ‘intramural pericytes’, very closely associated with the endothelium and extending numerous ramified protoplasmic processes . His name was given to these cells—Rouget cells—until 1923, when Zimmermann renamed them pericytes, described as being mainly associated with vessel contractile function . Electron microscopy described pericytes as cellular structures encapsulated under the basal membrane of microvessels (diameter: 10–100 μm) and capillaries (diameter <10 μm) where they form a single layer around and in close contact with the endothelium. In this respect, the presence of pericytes in vessels devoid of basement membrane is by definition not expected, as for example in hepatic sinusoids . Thus, pericytes are clearly distinguishable from other perivascular cells, smooth muscle cells and adventitial cells, found around larger blood vessels and outside the basal membrane. Pericyte morphology is also different, these being large, stellate cells. In situ, pericytic cytoplasmic protuberances run parallels to the longitudinal axis of capillaries, but circular around microvessels . Other than pericytes, adventitial cells represent another perivascular cell type. The wall of all blood vessels, but capillaries, is constituted by three layers named tunica intima, tunica media and tunica adventitia. The tunica adventitia is the outmost layer of the vascular wall, originally described as an unorganized structure composed by fibroblasts, collagen and nerves playing a role in maintaining vessel structural integrity. More recently, a number of studies have reported a dynamic role for the tunica adventitia in vascular remodelling [8, 13-16] inflammation and immune response [17-20]. Although pericytes and adventitial perivascular cells have been described for more than a century, it is only recently that the blood vessel wall was demonstrated as a reservoir of progenitor cells. We showed that perivascular cells, i.e. pericytes and adventitial cells, are the in vivo counterpart of MSC obtained in culture from various organs [21, 22]. Pericytes and adventitial cells are two perivascular cell compartments with distinct phenotypes and anatomical locations in situ as well as distinguishable behaviours in culture. These perivascular cells can be prospectively purified by flow cytometry using a well-defined surface marker combination, common in all human organs tested (Table 1). Importantly, pericytes and adventitial cells dissociated from vessel wall contain multipotent precursors with robust regeneration properties similar to those of classic heterogeneous MSC. Some studies show that human MSC may accumulate chromosomal aberrations [23, 24] in contrast to other reports which strongly support their chromosomal stability [25, 26].
Table 1. Comparison between human pericytes and adventitial perivascular cells [21, 22]
Capillaries and microvessels
Human tissue origin
Adult, foetal and embryonic skeletal muscle and pancreas, adult WAT, foetal skin, small intestine, brain, foetal and embryonic BM, term and mid-term placenta
CD146, NG2, PDGFRβ, α SMA, CD90, CD73, CD105, CD44, ALP
CD34, CD90, CD73, CD105, CD44
Documented differentiation potential
Osteogenic, adipogenic, chondrogenic, myogenic
Osteogenic, adipogenic, chondrogenic, pericytic
Our own experiments demonstrated that cultured human pericytes and adventitial cells injected into immunodeficient mice are not tumorigenic .
Markers of pericytes
In vivo, the main criterion to identify pericytes remains their anatomical localization and morphology. Beside localization, defining the molecular phenotype of pericytes has been a challenge. The 3G5 antigen, first suggested to be expressed by pericytes of the retina and adipose tissue, was later documented as a ubiquitous pericyte marker [27, 28]. Other pericyte markers have been described such as the melanoma-associated antigen, Thy1.1, the ephrin receptor and its ligands, neuropilin-1 and -2 and the Notch receptor and its ligands, Jagged-1 and Jagged-2. Although none of these antigens is present exclusively on pericytes . Vimentin and desmin are expressed by most chick pericytes as well as smooth muscle cells . Alkaline phosphatase (ALP) has been used to identify and isolate pericytes from adult human skeletal muscle [31, 32]. The presence of ALP on pericytes from other tissues was next confirmed . Finally, we have demonstrated that CD146, NG2 and PDGFRβ can be used to purify human pericytes from foetal and adult human tissues .
Perivascular and endothelial cell marker, CD146, aka MUC18, MCAM, Mel-CAM or S-Endo1, is a transmembrane glycoprotein and a member of the immunoglobulin superfamily. Perivascular and endothelial cell marker, CD146 expression is mainly associated with blood vessels, being present on vascular endothelium, pericytes and smooth muscle cells [21, 33-35]. In the embryo, CD146 marks other cell types present in the neural crest, notochord, mesonephros, ectoderm, somites and skeletal muscle rudiment [36, 37].
However, cells in the paraxial mesoderm do not express CD146 before segmentation; only once somites are specified to the muscle cell lineage is CD146 expressed. In humans, CD146 expression is correlated with the development of the trophoblast as early as 12 days post fecundation . The exact role of CD146 is less obvious. In zebrafish, absence of CD146 inhibits lumen formation in intersomitic capillaries preventing establishment of the blood flow and causing defaults in the vascular system . In human adult BM, CD146 is found on adventitial reticular cells identified as sinusoidal pericytes. These osteoprogenitors are able to self renew in serial transplantation and restore the perivascular cell compartment [38-40]. In the absence of CD146, no self-renewal has been observed.
Platelet-derived growth factor, PDGFRβ is expressed on pericytes [21, 31, 35, 41]. In the absence of PDGFβ production by endothelial cells, pericytes of newborn rats detached from blood vessels . Moreover, mouse embryos deficient in PDGFβ and PDGFRβ are viable, but die after birth lacking most pericytes and developing haematological, renal and placental abnormalities at late embryonic stages [43-45]. A compensatory mechanism set to recruit small subsets of pericytes via the PDGFRα subunit may explain the absence of embryonic lethality because complete loss of PDGF signalling is lethal at embryonic day 9.5 .
αSMA and NG2
Alpha smooth muscle actin (αSMA) is a universal marker of smooth muscle cells in large vessels, as seen in the human umbilical cord at term (Fig. 1a). Pericytes of microvessels also express αSMA, in contrast to most pericytes surrounding capillaries [21, 47, 48]. It is suggested that the presence of αSMA in subsets of pericytes, and absence from others, is correlated with one of the main functions of pericytes that is to contract and control blood pressure .
The NG2 proteoglycan (neural glial antigen 2), or chondroitin sulphate proteoglycan 4, was first described in the nervous system  and later discovered on pericytes and smooth muscle cells [21, 51-53]. Not all pericytes express NG2, which is circumscribed to the arterial system [21, 54, 55]. A phenotypic transition from pericytes to smooth muscle cells has been already proposed [56, 57]. Indeed, NG2 expression distinguishes three subsets of human pericytes, associated with: capillaries (NG2+αSMA-), venules (NG2-αSMA+) and arterioles (NG2+αSMA+) [21, 55]. Interestingly, all three phenotypes can be seen simultaneously around big vessels such as the human umbilical cord artery (Fig. 1b), or surrounding large vessels in the developing human placenta . The human umbilical cord at term is composed of three large vessels (two veins and one artery) surrounded by stromal cells and embedded in a mostly acellular substance named Wharton's jelly. The mid-gestation umbilical cord contains an additional vein. Pericyte marker, NG2 is not expressed in either vein of the human umbilical cord, which confirms NG2 as a marker of arterial/venous polarity. Whereas all perivascular cells express αSMA within venules, only rare cells within the venular wall co-express NG2 and αSMA (Fig. 1c). Wharton's jelly does not contain NG2-expressing cells (Fig. 1d). In contrast to the human umbilical cord at term, arteries and veins of the mid-gestation cord do not show a differential distribution of NG2. Mural cells in all large vessels co-express NG2 and αSMA, except for the monolayer immediately in contact with endothelial cells (Fig. 1e). Whether or not the latter can be named pericytes of large vessels is less obvious although their presence in adipose tissue was also proposed . Compared to NG2, CD146 is equally distributed, in the large vessels of the human term umbilical cord, within the arterial wall and has a distribution similar to that of NG2 in the vein [60, 61]. In conclusion, our data define human pericytes as CD146+PDGFRβ+CD34-CD56-CD45- cells (Fig. 2a). Subsets of perivascular cells within this phenotype can be distinguished by NG2 and αSMA differential expression [21, 55].
Pericyte long-term culture
Because of their implications in pathological processes, tissues such as the retina, brain, lung, skin and kidney have been used to establish methods to culture pericytes. These protocols include the growth of pericytes from microvessels of the bovine retina, a method adapted to various tissues such as human placenta and adipose tissue [62-65]. These cultures, albeit enriched in pericytes, remain heterogeneous in term of cell composition. For this reason, we set up to prospectively purify and culture human pericytes (Table 1). When seeded in culture, purified human pericytes do not attach rapidly, but sit for several hours then spread on the pre-coated plastic dish and divide very slowly during the first 2 to 4 passages to eventually proliferate and expand. Foetal muscle derived pericytes for example can be expanded up to 40 population doublings . This high capacity to proliferate is only observed in culture. Indeed, most cultured pericytes tested at passage 5 express Ki67, in contrast to pericytes in situ, of which only a few proliferate (Fig. 2b and c). After around 30 doublings, human pericytes grow more slowly and eventually undergo senescence, similar to conventional MSC [21, 66-68]. In terms of phenotype, long-term cultured pericytes retain expression of discriminating markers (CD146, NG2, PDGFRβ), and, importantly, exhibit MSC phenotype and multipotency . Cultured pericytes express NG2, αSMA, CD44, CD146, PDGFRβ and nestin (Fig. 2d–h), but lack the myogenic and neural cell marker CD56 (Fig. 2i) and the endothelial and haematopoietic cell markers CD34, CD31 and vWF (Fig. 2j–l). As of morphology, cultured pericytes are large, in general more than 50 μm, with an irregular stellate shape. Pericytes do not show contact inhibition; at confluence, they retract and fuse to eventually form ball like structures reminiscent of neurospheres. Nestin expressing mesospheres exhibiting MSC multipotency and myogenic myospheres were similarly isolated from adult murine BM and skeletal muscle respectively [69-72].
Adventitial cell characterization
Markers of adventitial cells and long-term culture
The phenotype of progenitors residing in the tunica adventitia of blood vessels was first described in the mouse. Hu et al. identified a population of Sca1+ progenitors abundant in the tunica adventitia, but absent from the tunica media or intima . Upon proper stimulation, purified Sca1+ progenitors differentiated into smooth muscle cells in vitro and in vivo. The authors also demonstrated that Sca1+ adventitial progenitors actively participate in the formation of atherosclerotic lesions by migrating into the tunica media and differentiating in smooth muscle cells. In agreement with these results Passman et al. described a population of Sca1+CD34+ckit-CD140b+ progenitors localized in a subregion of the tunica adventitia and highly active for sonic hedgehog signalling . Sca1+ adventitial progenitors were described to retain the potential to differentiate not only into smooth muscle cells but also into endothelial-like cells and osteoblasts, thus confirming that the tunica adventitia represents a niche for multipotent progenitor cells. Adventitial progenitors with similar phenotype and potential were also described in humans. Mesenchymal stem cell progenitor, CD34+CD31– cells endowed with the ability to give rise to endothelial-like cells were detected in the proximity of the tunica adventitia of adult human thoracic arteries and in the adipose-derived stromal vascular fraction [75-77]. Mesenchymal stem cell progenitor, CD34+CD31– adventitial cells from the human saphenous vein were shown to associate with endothelial cells in vitro and in vivo and to promote neo-angiogenesis through paracrine mechanisms . Altogether, these studies indicate CD34+CD31- cells as a promising cell product for therapeutic angiogenesis, but the potential of adventitial progenitors is not limited to angiogenesis. A number of studies have indeed demonstrated that, in mice and in humans, the tunica adventitia harbours CD34+CD31– mesenchymal stem cell progenitors that may play a role in extravascular tissue homeostasis and repair [22, 77, 79-81]. We have previously reported pericytes and adventitial cells as two anatomically and phenotypically distinct populations of perivascular cells expressing MSC markers in situ and able to generate MSC-like cells in culture [21, 22]. Mesenchymal stem cells, MSC generated from pericytes or adventitial cells were undistinguishable when assessed for expression of conventional MSC surface markers (CD90, CD73, CD105, CD44) and ability to differentiate into bone, fat and cartilage (Table 1). However, similar to murine Sca1+ cells, human adventitia-derived MSC do not express any of the smooth muscle cell markers αSMA, NG2, CD146 and PDGFRβ that are instead homogeneously expressed by pericyte-derived MSC. Moreover, adventitial MSC could acquire a pericyte-like phenotype when stimulated with angiogenic factors . Despite similar morphologies, adventitial cells and pericytes have different abilities to grow in culture. Differently from pericytes, freshly sorted adventitial cells rapidly attach and start proliferating in tissue culture plates without the need for gelatin coating. Adventitial cells retain a proliferative advantage over pericytes over the long term, as showed by a significantly lower population doubling time . Clonal cultures can be grown from both adventitial cells and pericytes [21, 22]. As for other cell types (haematopoietic cells, endothelial cells), expression of CD34 is rapidly down-regulated once cells are cultured in vitro, and CD34 is undetectable after 1 or 2 passages. Finally, similar to pericytes, adventitial cells cultured in a low-attachment plate which maintain cells in a suspended, unattached state, can form spheres that can be serially passage in vitro. Interestingly, similar CD34+CD31- spheroidal colonies have been also isolated from pericyte progenitors of the postnatal rat aorta . Taken together, these studies indicate the co-existence in situ and in culture of distinct perivascular multipotent progenitors able to contribute to distinct subsets of MSC, possibly organized in a hierarchical fashion. Current studies are aimed to investigate the contribution of individual or combined pericytes and adventitial cells to tissue repair.
Adipose tissue is an abundant source of perivascular MSC
Although pericytes and adventitial cells are ubiquitous multipotent progenitors, isolation of these cells for clinical purposes is not feasible from all human tissues. Based on availability, dispensability, harvesting procedure and progenitor frequency, human adipose tissue emerges as an abundant and convenient source of therapeutic perivascular progenitors. Indeed, the adipose stromal vascular fraction (SVF) is highly enriched in blood vessels and SVF-derived perivascular cells give rise in vitro to MSC, previously derived indirectly in primary culture of the unsorted SVF as ADSC (adipose-derived stem cells) . In 2001, Gronthos et al. demonstrated that ADSC express all MSC markers: CD105, CD106, CD166, CD44 and importantly the CD146, perivascular and endothelial cell marker . At the same time, Zuk et al. demonstrated that ADSC have similar capacity as MSC to differentiate into adipogenic, osteogenic, chondrogenic and myogenic cell lineages, at a clonal level [84, 85]. CD146+CD34-CD31-CD45 –pericytes and CD34+CD146-CD31-CD45– adventitial cells represent around 15% and 20% of the stromal vascular fraction respectively [21, 22] whereas perivascular cells in adult bone marrow are less than 0.5% of total mononuclear cells. Identification of two frequent multipotent progenitor cell subsets in human adipose tissue represents a critical step towards the clinical use of autologous stem cells. Purification and combination of pericytes and adventitial cells from lipoaspirates would yield clinically relevant numbers of progenitor cells devoid of bystander cells or negative regulators (endothelial cells), that could be in some indications directly transplanted without ex vivo expansion. Such an approach would significantly improve the efficacy and safety of current cell therapy strategies making use of cultured total stromal cells (James et al., 2012, In press).
Perivascular cells as stem/progenitor cells for regenerative medicine
Perivascular cells for cardiovascular repair and regeneration
Despite the multiple roles of pericytes in the pathophysiology of the cardiovascular system, their application in cardiovascular regenerative medicine remains to be tested. We recently investigated the therapeutic potential of pericytes in ischaemic heart repair. Upon transplantation into acutely infarcted hearts of NOD/SCID mice, human muscle-derived and cultured pericytes significantly improved cardiac function when compared to control injections. Pericytes exhibited cardio-reparative effects such as promotion of angiogenesis, reduction of scar and inhibition of chronic inflammation, likely because of their secretion of a variety of trophic factors (Chen et al., submitted). Dar et al. very recently described the production of pericytes from spontaneously differentiating embryoid bodies derived from human pluripotent stem cells (hPSC) . These CD105+CD90+CD73+CD31- multipotent mesodermal precursors express the pericyte markers CD146, NG2 and PDGFRβ but not the smooth muscle cell marker, αSMA. hPSC-derived pericytes transplanted into immunodeficient mice with ligature induced limb ischaemia not only induced vascular regeneration but also promoted muscle repair, by incorporating into the damaged muscle and vasculature . Pericytes have been also used to engineer vascular grafts: a cylindrical synthetic scaffold seeded with human muscle pericytes and transplanted into the sectioned rat aorta supported the development of a structurally and functionally normal blood vessel . These results suggest that pericytes can serve as a cell source for cardiovascular therapy. The use of adventitial cells in cardiovascular repair and regeneration has been also investigated. Campagnolo et al. showed that CD34+CD31- adventitial cells interact with endothelial cells and promote the formation and stabilization of capillary-like structures. A significant pro-angiogenic effect of adventitial cells was observed after administration in mouse ischaemic limbs, as shown by full recovery of blood flow as early as 7 days post treatment.
These results indicate the therapeutic capacity of adventitial cells in post injury angiogenesis/vasculogenesis . Very recently, the same group reported that transplantation of adventitial cells improves repair of the mouse infarcted heart through angiogenesis involving microRNA-132 (miR-132). Adventitial cells treatment enhanced cardiac repair in multiple aspects, augmenting cardiac contractility, attenuating LV dilatation, reducing cardiomyocyte apoptosis and interstitial fibrosis, increasing myocardial blood flow and neovascularization and decreasing vascular permeability. The paracrine function of adventitial cells activates endogenous repair responses, including pro-angiogenic and pro-survival Akt/eNOS/Bcl-2 signalling. Furthermore, adventitial cells constitutively express and secrete miR-132 and markedly up-regulate its expression upon stimulation, which in turn acts as a paracrine activator of cardiac healing . Together these data indicate the therapeutic potential of pericytes and adventitial cells in ischaemic tissue repair.
Perivascular cells for the regeneration of other tissues
The effect of pericytes on muscle regeneration was examined in immunodeficient mouse models of skeletal muscle injury and dystrophy . Pericytes (CD146+CD34-CD45-CD56-) purified using flow cytometry from human skeletal muscle biopsies were injected into the hind-limb muscles of SCID-non-obese diabetic (NOD/SCID) mice after injury by intramuscular injection of cardiotoxin. Pericyte-derived muscle fibres were detected by the presence of human-specific spectrin and centrally located human nuclei . Both freshly sorted and long-term cultured pericytes generated more human myofibres than human CD56+ skeletal myoblasts and total unsorted muscle cells. This ruled out the possibility that the myogenic potential observed in pericytes results from a contamination by myoblasts in culture. Interestingly, myogenic potential also exists in pericytes residing in other human organs, including placenta, white adipose tissue and pancreas.
Pericytes purified from those non-muscular organs not only exhibited myogenic potential in culture but also regenerated human dystrophin- or spectrin-positive myofibres upon injection into mdx/SCID or cardiotoxin-treated NOD/SCID mouse muscles, in which they also promoted angiogenesis [21, 64]. These results confirmed that pericytes sorted from healthy and dystrophic human skeletal muscle biopsies by ALP expression regenerate human myofibres in the muscles of dystrophic immunodeficient mice . Lately, using a transgenic labelling of alkaline phosphatase in the cre/lox inducible expression system, the same group demonstrated that pericytes residing in the postnatal skeletal muscle naturally participate in the regeneration of the injured/dystrophic skeletal muscle . New results further suggest a connection between muscle-residing pericytes and NF-κB activation in human muscle regenerative response following eccentric contractions . These results document the role of pericytes in muscle regeneration and suggest their future applications in skeletal muscle therapy. Human pericytes can also make bone. When cultured in standard osteogenic medium, pericytes exhibited alkaline phosphatase expression and mineral deposition. Pericytes seeded onto Gelfoam scaffolds and implanted into skeletal muscle pockets in immunodeficient mice developed into bony nodules . Furthermore, when cultured in the presence of an osteoinductive growth factor, Nell-1, pericytes exhibited robust osteogenic differentiation on either culture plastic ware or human cancellous bone chip (hCBC) scaffold. Upon implantation into a muscle pouch in the nude mouse, pericytes seeded on hCBC formed significantly more new bone than hCBC scaffold alone. Nell-1 significantly increased pericyte proliferation as well as osteogenic differentiation in vitro and in vivo . These and other more recent results suggest pericytes as novel therapeutic cells for skeletal regenerative medicine (James et al. 2012, In press). Overall, the superior regenerative capacity of pericytes can be attributed to multiple factors, including intrinsic multilineage developmental potential, robust paracrine function and efficient migration in response to stimuli [21, 60, 64, 91-93].
We and others have demonstrated that adventitial cells, regardless of their tissue of origin also display developmental features typical of MSC [17, 22, 73, 78, 79, 88, 94]. Although myogenic potential remains to be determined, the ability of adventitial cells to differentiate into major mesodermal cell lineages, including bone and cartilage, suggests a contribution of these cells in mesodermal organ development and post injury regeneration. Altogether, these reports suggest that beyond a mere structural constituent of the vascular wall, the adventitia is a dynamic reservoir of stem/progenitor cells that participate in vascular remodelling and regeneration of surrounding tissues.
Mesenchymal stem cells are one of the most promising stem/progenitor cell populations for regenerative medicine. Despite this potential, MSC are heterogeneous and their origin in situ has long remained unknown. We and others recently demonstrated that two main types of perivascular cells, pericytes and adventitial cells, are the native counterparts of MSC in developing and adult human organs. Studies on prospectively purified subsets of MSC will reveal important facts regarding their stemness, developmental biology, migratory capacity and regenerative potential. This will provide a better understanding of perivascular/MSC behaviour in health and disease and allow their optimal utilization in regenerative medicine.
Mihaela Crisan is grateful for the award of an EMBO Long term Fellowship and NWO Veni grants. Mirko Corselli is supported by a California Institute for Regenerative Medicine (CIRM) training grant. W.C. was supported in part by a Predoctoral Fellowship from the American Heart Association. Experiments described have been in part supported by funds from the University of Pittsburgh and University of California at Los Angeles and by grants from the National Institute of Health, Commonwealth of Pennsylvania, Pittsburgh Foundation, CIRM and Medical Research Council to B.P.
Conflict of interest
The authors confirm that there are no conflicts of interest.