There has been an explosion of literature in the field of mesenchymal stem cells (MSCs) in the past 10 years. Many researchers have sought to exploit their potential as a source of reparative cells for clinical use in a variety of contexts. There are, however, several pitfalls that it would be useful to avoid, as MSCs have some awkward properties that may make their use for tissue repair or tissue engineering somewhat risky. In this review, we highlight some recent advances in the understanding of the cell biology of MSCs, and how these may integrate into strategies for more clinical applications.
Adult MSCs are generally thought of as an autologous source of reparative cells, in contrast to the totipotent, and allogeneic embryonic stem cells (ESCs). A major source of adult stem cells is the bone marrow (BM), from which two main populations derive: haematopoietic stem cells (HSCs), which produce the blood-cell lineages, and MSCs, which provide the bone-marrow stromal niche and have the potential to produce several cell lineages, including adipogenic, osteogenic and chondrogenic lineages. BM also contains endothelial precursor cells (EPCs) and probably a common precursor of all three stem-cell types, which remains somewhat less well defined. Recent reports suggest that MSCs may differentiate into endothelial cells [
1,
2], an outcome that may depend on the cytokine context. MSCs can also be obtained from the stromal fraction of lipoaspirates of adipose tissue [
3], and these possess similar properties to BM-derived MSCs. Further sources of human MSCs include the intestinal [
4], limbal [
5], knee-joint [
6‐
9] and prostate [
10] stroma, trachea [
11], nasal mucosa [
12], Wharton's jelly (WJ) [
13,
14], cord blood [
15] and placenta [
16]. MSCs-like cells have also been extracted from tumour stroma [
17], and may have an important role in the fibrotic responses, as reported for the rat kidney [
18] and rat heart [
19]. Thus, these cells are becoming almost ubiquitous residents in many tissues and organs. A very recent finding may even cast doubt on the utility of MSCs themselves; Medici and colleagues [
20] reported that endothelial cells treated with either bone morphogenetic protein (BMP)4 or transforming growth factor (TGF)β2 reverted to a multipotent cell with some characteristics of MSCs, and could be differentiated into several endodermal cell types. The degree to which such cells or 'true' MSCs can be used in tissue repair, gene or cancer therapy may have a bearing on many clinical outcomes.
Characteristics of MSCs from different tissues
The features of several sources of MSCs are shown in Table
1. There is considerable overlap in their gene expression patterns, as expected, but a few notable differences. The general conclusion is that MSCs or their closely related cousins can reside in or pass through most tissues, and that such cells may be isolated and/or cultured by conventional methods such as fluorescence-activated cell sorting (FACS) or plastic adherence, and can be shown to possess multipotency. It is of interest to see that MSCs also express cell-surface epidermal growth factor receptor (EGFR)-1 (Her-1, ErbB-1) and respond to the ligand heparin-binding epidermal growth factor (HB-EGF) with dose-dependent proliferation, which reversibly impairs their trilineage differentiation ability until the stimulus is removed [
21]. As shown in Table
1 there is no consistent subset of surface molecules that are a definition of the MSCs phenotype; thus it is possible that many slightly different subtypes exist, and that their output phenotypes could be conditional upon local and systemic signalling.
Table 1
Human tissue MSCa phenotypes
| CD29, CD44, CD105, CD166 | CD34, CD45 | Yes |
Umbilical-cord perivascular cells [ 118] | CD146, NG2, PDGF-Rb, ALP, SSEA4, Runx1, Oct4 | CD34, CD45, CD144, vWF | Yes |
| CD44, CD90, CD105, HLA G6, IL-1A/B, IL-6, IL-8, IL-14, BMP1, CSF3, FAMC3, GDF15, PDGF-B, TNF-4, TNF-11b, TNF-12, VEGF | CD34, CD40, CD45, CD80, CD86 | Yes |
| CD29, CD44, CD73, CD90, CD105, TLR4 ligand | | Yes |
| CD29, CD44, CD73, CD105, CD117, CD166 | CD14, CD34, CD45 | Yes |
| CD117, plus as listed for stroma | | Yes |
| CD105, CD106, CD54, CD166, CD90, CD29, CD71, pax-6/p75, SSEA1, Tra-1-61, Tra-1-81, CD31, CD34, CD45, CD11a, CD11c, CD14, CD138, Flk1, Flt1 | Vascular endothelial cadherin | Adipo, Osteo |
Osteosarcoma stroma [ 120] | CD44, CD73, CD90, CD105, CD166, HLA class I | CD14, CD19, CD31, CD34, CD45, HLA-DR | Yes |
There are many protocols used to determine the phenotype of MSCs. The essential characteristic is their multipotentiality, which is usually established
in vitro by retrospective analysis of their ability to differentiate into at least three cell types: typically adipocytes, osteocytes and chondrocytes, and these potencies may vary between mouse strains [
22]. Other cell types may be included, depending on the purpose of the study, such as muscle or tendon cells [
23]. A few researchers have used single-cell clones as a source of trilineage-potent MSCs [
24], whereas others have shown their potential by
in vivo analysis in xenografts [
25].
The original characteristic of MSCs - their ability to form fibroblastic colonies
in vitro from BM or other tissues plated onto tissue-culture plastic [
26] - has been widely exploited, and there is a vast literature on their many phenotypical characteristics [
27‐
29]. As briefly illustrated in Tables
1 and
2, there are many combinations of cell-surface markers that can be used to select MSCs from mixtures of cells, which are often performed by cell sorting using FACS [
30] or immunomagnetic [
31] methods, although many studies use plastic-adherent stromal cells that are subsequently classified using immunofluorescent phenotyping or flow cytometry [
5,
32]. The exact equivalence of these phenotypes may be questioned, but most studies have shown that the selected cells possess similar multipotentialities. A recent study [
10] found by FACS analysis that some murine prostate cancer xenograft MSCs possessed a Hoechst 33342 'side population (SP)', similar to that observed more commonly for HSCs [
33] and for some epithelial stem cells such as keratinocytes [
34] or tumour cells in the colon [
35], although perhaps not all SP cells from colon-cancer cell lines are stem cells [
36]. It is possible that the murine stromal SP cells resemble those described as human adipose tissue-derived MSC- and EPC-like cells (CD34+, CD90+) that could differentiate into adipocytes and endothelial cells [
37] and that may be more like multipotent adult progenitor cells (MAPCs), which show some overlaps with MSCs and mesangioblasts. However, these are essentially three cell types that differ from each other in their expression profiles, and differ again from ESCs [
2,
38]. It has been reported that murine MSCs and MAPCs possess similar immunomodulatory abilities
in vivo and
in vitro by suppressing alloreactive T-cell proliferation [
39].
Table 2
Recent clinical use of MSCsa in phase I/II trials
Stroke, 16 | Autologous iliac crest | > 90% SH2 | DMEM, 10% FCS, Cryo | 5 × 107 twice | 2 doses, 2-week interval | None | ND | To 5 years | HR for MSCs = 0.344 | Con 34% MSCs 72% | |
MI, 10 | Autologous iliac crest | CD73+ CD90+ CD105+ | DMEM, 10%FCS | 7.5 × 106 MSCs + EPCs | 1 dose | None | ND | 6 months | LVEF 12% up | -- | |
MI, 53 | Allogeneic unmatched 'prochymal' | CD105+ CD166+ CD45- | Cryob | 0.5 to 5 × 106/kg | 1 dose, intravenous | MSCs 5 pt Placebo 7 | ND | 6 months | FEV1 up, LVEF up, arrhythmia down | -- | |
ALS, 10 | Autologous iliac crest | CD29+ CD44+ CD90+ CD105+ CD166+ | Cambrex MSC medium 10% FCS | 11.4 to 120 × 106 | 1 dose, thoracic spine | None | ND | 4 years | MSCs tolerated, SC scars | -- | |
ALS, 19 MS, 15 | Autologous iliac crest | CD29+ CD73+ CD90+ CD105+ CD166+ | DMEM, 10% FCS | ALS 5 × 107 MS 6 × 107 +/- ferumoxide | 1 dose, intrathecal; 1 dose, intravenous | None; CD4/25 Treg cells up | Possible | 6-25 months | ALSFRS stable, EDSS better | -- | |
Refractory Crohn's Disease, 10 | Autologous iliac crest | CD73+ CD90+ CD105+ | DMEM, 10% FCS, Cryo | 1 to 2 × 106/kg [mean wt 57 kg, range 46-113 kg | 2 doses, 1-week interval | Headache (3 patients) Allergy (1 patient) | ND | 14 weeks | CDAI fall (5 patients; > 70 in 3 patients);3 patients worse | -- | |
Paediatric acute leukaemia 8 | Haploid parent BM | Adherent CD phenotype ND | Cryob | 6 × 104 to 107/kg MSCs + UCBT | 1 or 2 doses, 3-week interval | None | None | 6.8 years, no chronic GVHD | All patients PMN+ at 9 to 28 days | 63% | |
Leukaemia 12c | HLA match opposite gender HCT | CD44+ CD73+ CD90+ CD103+ | ND | ND | 1 dose | ND | None | 0.9-138 months | HCT success, no MSCs took | ND | |
Fibroblastic differentiation
Lee and colleagues [
80] reported that human MSCs could differentiate into stromal fibroblasts
in vitro after stimulation by connective tissue growth factor, during which they secreted collagen I and tenascin-C. MSCs were initially α-SMA-negative, but could express this protein on stimulation with TGFβ. Sarraf and co-workers [
24] reported that murine MSCs could differentiate into fibroblasts and myofibroblasts when embedded in a collagen type I matrix and placed under tension, either self-generated or externally applied. The cells secreted both collagenous and elastic fibres. It thus seems likely that BM-derived myofibroblasts and fibroblasts in unsorted BMT experiments have come from the MSC population.
Several publications have reported a flux of BM-derived stromal myofibroblasts and fibroblasts into many tissues [
81,
82], and that damage increases their numbers. Direkze and colleagues [
82] found that lung tissue damaged by paracetamol contained 41% BM cells, compared with 17% in control lungs. A similar pattern was found in full-thickness wounded skin, but with only 4% of BM-derived myofibroblasts. Interestingly, the proportion of kidney-derived myofibroblasts did not rise significantly with injury, but remained at around 20 to 24%. This group also reported BM-derived tumour-associated myofibroblasts and fibroblasts in a mouse model of insulinoma [
83].
Zhao and colleagues reported that the marker CD90 was raised in both MSCs and in prostate cancer stromal fibroblasts, which the authors analysed by quantitative PCR after isolating them from tissues by FACS, using CD90 as a discriminator [
30]. The authors concluded that the CD90-hi cells were not true MSCs, but that they expressed several proteins associated with tumour promotion, such as Sonic hedgehog and TGF-β, as well as pro-angiogenic factors. A further report on lung MSCs suggested that these also respond to TGFβ by differentiating into myofibroblasts [
11].
Xenotransplantation
MSCs are common to all mammals tested to date, and have been viewed as essentially benign because of their frequent lack or suppression of immune effects on hosts. However, an increasing body of literature has reported unfortunate or even malign effects that may result from xenogeneic MSC infusions.
Early work on MSC infusions used human MSCs in foetal sheep [
25,
88], which suggested that this immune-privileged site is permissive for tolerance of foreign cells, which can engraft, differentiate
in situ, and function in a normal way for that cell type for extended periods; in those studies, up to a year. MSCs were detected in cartilage, fat, muscle, heart, BM stroma and thymic stroma.
Much of the clinical literature to date on infusion or injection of MSCs in human disease has failed to show significant adverse effects (AEs) on the recipients [
89], particularly for their use in HSC transplantation [
90]. This is encouraging for possible therapeutic uses of MSCs in tissue repair and regenerative medicine. However, there are reports of human MSCs in xenograft models of disease suggesting that some caution is needed regarding the absolute benevolence of these cells. These reports may reflect differences in the cell biology of the species involved, and need not predict that the same pathologies will occur in people, but nevertheless indicate that the precautionary principle should apply. Some recent examples are presented below.
Teng and colleagues reported that human MSCs could be rendered tumourigenic by the hypermethylation silencing of two tumour suppressor genes that activate p53: Hypermethylated in cancer (HIC)1 and RassF1A [
91]. These cells were transformed, grew as anchorage-independent colonies in agar, and formed sarcoma-like tumours when injected subcutaneously into nude mice.
Several reports now suggest a strong possibility that MSCs may be permissive for the proliferation and dissemination of breast-cancer stem cells. For example, Yan and colleagues extracted MSCs from breast tumours that possessed trilineage potential and augmented the growth of mammary tumours when co-infused into animals. The MSCs also stimulated 'mammosphere' formation
in vitro, which was EGF-dependent [
92]. Similar effects on mammosphere formation were found by Klopp
et al[
93], who found reduced E-cadherin expression in normal and breast-cancer epithelial cells, and the MSCs augmented tumour development when co-injected into mice.
It may be that MSCs can differentiate into fibroblastic cells by the action of breast tumour-secreted osteopontin, and by so doing enhance the tumourigenic and metastatic potential of the MDA-MB231 cancer cell ine [
94]. Human MSC cultures possess a subpopulation of ALDH-positive cells that stimulate the induction of cancer stem cells (also ALDH+) in the human breast line SUM159
in vitro[
95]. When non-obese diabetic/severe combined immunodeficiency (NOD/SCID) mice bearing such tumour xenografts were given intratibial injections of human MSCs, the MSCs homed to the tumours, which grew faster and possessed more cancer stem cells than did control tumours. It was inferred from microarray analyses of tumour cell-MSC co-cultures that several cytokines may be expressed in the tumour cells that act to promote these effects: CXCL1, CXCL5 and CXCL6, IL-6 and IL-8. Antibodies to CXCL7, itself a product of MSCs, blocked the secretion of these cytokines. The regulation was further controlled by breast-tumour cell-derived IL-6 stimulating both the secretion of CXCR7 and the chemotaxis of MSCs to the tumour cells. The MSCs then increased the population of ALDH+ epithelial cancer stem cells. These authors also reported similar juxtapositions of ALDH+ MSCs and breast-cancer cells in sections of human breast tumours.
In a similar study, Zimmerlin and colleagues studied adipose-derived MSCs, and reported the ability of these cells to enhance the growth of breast-tumour pleural effusions from patients in both
in vitro and
in vivo settings [
96]. The authors separated the metastatic breast-tumour cells into active and resting phenotypes based on CD90 positivity, with high or low scatter respectively. When co-injected with the adipose-derived MSCs into mice, only the active effusion cells were tumourigenic.
Park and co-workers reported the migration of human umbilical-cord MSCs towards human glioma cells
in vitro, and that overexpression of CXCR4 increased this trait. Further, in a xenograft model of glioma in nude mice, these cells displayed enhanced migration into the tumours [
97]. In an experiment in mice using transplantation of GFP-tagged BM, GFP-positive MSCs migrated into the prostate of castrated mice, and these cells were increased by testosterone in a Wnt-dependent manner. These findings were also seen in a human prostate tumour xenograft, in which MSCs expressing an exogenous Wnt antagonist, secreted Frizzled-related protein (SFRP)-2, induced tumour shrinkage by necrosis and apoptosis [
98].
Kucerova and colleagues reported that adipose-tissue MSCs could promote growth in nude mice of tumours of the xenografted human melanoma cell line A375 [
99]. This was achieved by suppression of apoptosis and an increase in proliferation. Another melanoma line, 8MGBA, did not share this attribute; instead, MSCs were inhibitory.
Two recent reports suggest that MSCs may give rise directly to mesenchymal tumours. Using a comparison of infused normal MSCs,
in vitro spontaneously transformed MSCs, and osteosarcoma murine cells, Mohseny and co-workers [
100] concluded that aneuploidy, chromosomal translocations and the homozygous loss of the Cdkn2A (p16) locus on chromosome 4 were implicated in tumour progression. The genetic changes seemed to occur around MSC passages 5 to 9 in culture, during which time the cells went into 'crisis', and thereafter they possessed the ability to grow in soft agar independently of substrate. The authors showed a series of 88 human osteosarcomas that possessed similar defects in the homologous cyclin-dependent kinase inhibitor (CDKN)2A locus on chromosome 9. Kaplan-Meier analyses of these patients with osteosarcoma showed very poor survival for patients negative for this locus (zero survival at 50 months follow-up versus 70% survival if positive for more than one allele). Although this study did formally prove the origins of these human osteosarcomas to be MSCs, it warrants a cautious approach when using these cells in the clinic.
A further report of tumours arising from genetically defective MSCs has recently appeared [
101]. These authors deleted p53, Rb or both genes in adipose tissue-derived murine MSCs that underwent Cre-LoxP excision. Wild-type and Rb-negative MSCs were phenotypically normal, whereas the p53-negative and p53-negative/Rb-negative MSCs were transformed, and could initiate leiomyosarcomas in half the animals when injected into the flanks of NOD-SCID/IL-2Rγ
-/- mice. The transformed MSCs approached tetraploidy, and were deficient in the ability to differentiate into adipocytes, yet had increased ability to become osteocytic. The authors noted that human leiomyosarcomas frequently display loss of p53 or Rb.
These examples indicate the possibility that MSCs could be involved in the growth of carcinomas, melanomas and sarcomas, and thus their use as repair agents for normal tissues or organs needs to be seen in this light. In addition, MSCs may also be exploited precisely for their homing attribute; by modifying them appropriately, infused MSCs may home in on tumours and deliver therapeutic reagents. Such experiments have been reported for an anti-tumour viral vector Delta-24-RGD transfected into MSCs, which homed to breast and ovarian tumours in mice and reduced systemic viral toxicity to negligible levels compared with virus-alone infusions [
102]. A different strategy was used by Sato and colleagues, who transfected MSCs with EGFR; these cells homed to both B16 murine melanoma and GL261 glioma tumours [
103]. When the MSCs were co-transfected with IFN-α, there was significantly increased survival of GL261-bearing mice. Secchiero and colleagues [
104] reported recently that BM MSCs could affect the outcome of Epstein-Barr virus (EBV)-positive or EBV-negative metastatic non-Hodgkin's lymphomas in nude-SCID mice. Mice receiving MSCs survived for longer periods than those without (40 and 59 d median survival, respectively).
Human MSCs have been used as a model for tumour therapy after transduction with IFN-β [
105]. Two murine pulmonary metastatic xenograft models were used: A375SM melanoma and MDA 231 breast carcinoma. In both models, the IFN-β MSC infusions resulted in MSC engraftment within the tumour stroma, and significantly prolonged survival of the mice compared with IFN-β injections alone. These results encourage research into the personalising of such treatments for suitable patients.
Clinical use of MSCs
Clinical trials using MSCs are being carried out for a variety of important diseases such as stroke, MI, multiple sclerosis (MS), amyotrophic lateral sclerosis (ALS), and leukaemia (Table
2). In general, MSCs appear to be well-tolerated, with most trials reporting lack of AEs in the medium term, although a few showed mild and transient peri-injection effects. There are no agreed phenotypic MSC markers that should be used, so the exact clinical effects of such sorted cells may be uncertain, as the cell populations used could differ. In addition, clinical outcomes are variable, and generally show small improvements, but to date few studies have reported either a long period of observation, the outcomes of more than one MSC infusion, or whether MSCs survive engraftment. That human MSCs circulate in the bloodstream has been reported [
106,
107], and immortalized MSC lines have been produced from similar cells [
108]. There may be a direct effect of the infused cells, but long-term clinical MSC engraftment is not yet routinely investigated by methods such as paramagnetic iron nanoparticles. Where this test has been performed, it has shown only possible survival of MSCs [
32], or none at all [
109]. It was found that the BM engraftment of donor MSCs in patients with leukaemia receiving whole-body irradiation and HSC rescue did not occur, but this did not influence the HSC engraftment [
110,
111]. However, those data are in contrast to studies in which sex-mismatched BMT resulted in donor-derived stromal cells in several organs including liver [
112], and endothelial cells in the BM of patients with chronic myeloid leukaemia [
113]. It is possible that some major influences of transplanted MSCs are systemic or paracrine via the release of cytokines or other molecules that affect responses in the target organ. Such influences are exemplified by a recent study on a rat model of hepatic failure, in which anti-apoptotic effects were seen after infusion of cultured MSC-conditioned medium [
47].
Autologous MSC infusions were performed in 16 patients with severe middle cerebral artery stroke, who were successfully followed up for up to 5 years, during which 58% of controls but only 25% of MSC-infused patients died [
114]. All patient MSCs were cultured in the presence of 10% FCS, and harvested to achieve 10
8 cells/person, then delivered in two intravenous infusions of 5 × 10
7 cells, 2 weeks apart. No side effects were noted, and similar levels of other disease parameters (vascular problem, seizure) were seen in both groups of patients. There was an association between MSC infusion and the levels of serum stromal cell-derived factor-1.
In a phase I trial, Lasala and co-workers [
115] infused a mixture of fresh peripheral mononuclear cells (a source of EPCs) and cultured BM MSCs (7.5 × 10
6 each) into the ischaemic myocardium of patients with angina pectoris who had over 70% stenosis in one or both coronary arteries. Left ventricular ejection fraction was increased by 12% at 1 month, and remained with an 11% increase at 6 months after the infusions, and cardiac ischaemia was decreased by 1.8-fold at 6 months only. The patients reported increased quality of life, and no AEs were seen. This is encouraging because the MSCs were cultured in bovine serum during their expansion
in vitro.
Another phase I trial used cultured autologous MSCs in patients with ALS [
116]. The cells were infused in cerebrospinal fluid into the thoracic spinal canal, and patients were monitored by MRI for 4 years, during which no AEs were noted locally or systemically. No attempt was made to track the survival of the MSCs. There was little change in the disease progression.
In a similar study, Karusis
et al[
32] studied patients with MS and patients with ALS, who also had no AEs after a single intrathecal infusion of autologous MSCs. In some patients, the MSCs were labelled with superparamagnetic iron-oxide nanoparticles, and there was evidence of their retention in the occipital horns of the ventricles, the meninges of the spinal cord, the parenchyma and the nerve roots for up to 3 months. In short-term (24 hours) immunological analyses of circulating blood cells, there were 72% more regulator T cells (CD4/CD25+) in both patient groups, and reductions of 30 to 60% in the proportions of myeloid dendritic cells positive for CD83, CD86 and human leukocyte antigen (HLA)-DR. Phytohaemagglutinin-stimulated lymphocytes were also 63% less reactive than before infusions. These data strongly suggest a decrease in the activation status of the host lymphocytic cells and antigen-presenting cells after MSC treatment.
A phase I trial of autologous MSCs (1 to 2 × 10
6 cells/kg twice) for refractory Crohn's disease (CD) has been reported [
117]. CD MSCs were similar to normal MSCs in immunomodulatory effects and phenotype, and did not cause side effects. Five common drugs used in the treatment of CD (adalimumab, methotrexate, azathioprine, dexamethasone, 6-mercaptopurine) all allowed MSCs to inhibit the proliferation of peripheral bone-marrow cells in
in vitro tests, whereas infliximab had a similar but non-significant trend. Three of the ten patients improved their Crohn's Disease Activity Index, but three were worse by 6 weeks after treatment, and required surgery. The authors concluded that the methods were safe, but warranted further longer-term investigations.
A report was recently published which detailed a study involving patients with paediatric leukaemia who received a transplant of unmatched umbilical-cord blood with parental haploidentical BM-derived cultured MSCs on one or two occasions [
109]. Any graft-versus-host disease (GVHD) reaction that occurred was acute and responded to steroid therapy, and no episodes of chronic GVHD were noted in the 6.8 year study. Using blood analyses, all patients were found to be chimaeric with regard to all HSC-derived blood-cell lineages within 3 months of transplant (positive for polymorphonuclear leukocytes by 9 to 28 days and platelets by 36 to 98 days), which was comparable with historical control patients who had received no MSC treatment. All patients were analysed by BM sampling for surviving donor MSCs and the degree of BM white blood-cell chimaerism at varying times after engraftment. At no time were any donor MSCs found, despite full haematopoietic chimaerism. This is of considerable interest because many of the putative benefits of MSCs treatment presume long-term engraftment of the cells, which may not have occurred here. It remains possible that some MSCs had engrafted into unsampled organs and exerted a systemic effect from those locations. This possibility would apply to another leukaemia study [
110], in which 12 patients with leukaemia were given sex-mismatched but HLA-matched BMT, and their BM was analysed up to 11 years later for evidence of donor-derived MSCs. In all cases, the BMT resulted in full blood-cell chimaerism, but in no case was there any evidence of donor BM MSC survival. This was true for each of three different conditioning regimens before BMT. Again, other body sites were unsampled. It is possible that the haplotyping was not sufficiently close to prevent a host response against the infused MSCs.