Several recent studies have suggested that stromal cells from bone marrow may be capable of generating either neurons or glia, both in vivo [4, 5] and in vitro .
In this study we induced stromal cells from human bone marrow to differentiate into cells with a neural phenotype, comparing two different protocols. Furthermore, we explored for the first time the involvement of the transcription factor Hes1 in inducing neural phenotype in BMSC. Both treatments led to a neural phenotype in a similar but not identical manner. Indeed, at the end of both treatments, BMSC expressed neural markers such as NF, TUJ-1 and GFAP, whereas nestin and vimentin progressively decreased with differentiation. These results were associated with different morphological modifications and required times of exposure ranging from one day to one week, depending on the treatment. The first induction media (a cocktail of PKC and PKA activators and FGF-1) rapidly determined dramatic modifications of cell shape and induced the expression of several neural markers in less than 24 hours. Differentiation was transient and cells reverted to their original phenotype when the inducing factors were removed. On treating BMSC with retinoic acid and BME biochemical markers consistent with differentiation towards a neural phenotype were not expressed before seven days. Moreover, morphological changes were not as drastic as in the other treatment, although differentiation persisted even when inducing agents were removed.
It is well known that an increase of intracellular cAMP and consequent PKA activation represents a critical regulator for differentiation of neurons and glia . Moreover, agents that increase intracellular cAMP levels induce neuroendocrine differentiation in human prostate carcinoma cells , neuronal differentiation in C6 glioma cells [19, 20] and processes elongation in MCD-1 medulloblastoma cell line . In the same manner, PKC is involved in neural differentiation regulating neurite outgrowth and branching [22, 23]. Iacovitti et al.  showed that FGF-1, together with co-activator molecules such as dopamine, PKA and PKC activators, induced differentiation of an embryonic carcinomal stem cell line into dopamine neurons. Deng et al.  found that human BMSC can be differentiated into early neural progenitors by conditions that increase intracellular cAMP. After six days of treatment these cells expressed NSE and vimentin but not markers of mature neurons, such as NF-M or MAP-2. In our experiments, treatment 1 induced BMSC to differentiate into more mature neuron-like cells within 24 hours.
Retinoic acid (RA), a derivative of vitamin A, is essential in maintaining normal cellular growth and development. In fact, RA is present in various tissues of both embryonic and adult animals, in particular in the nervous system [25–30], where it promotes neuronal differentiation . Previous studies have demonstrated that RA induces both a greater number of neurites as well as increased neurite length in cultured neurons [for a review, see ]. Retinoic acid has been used in combination with other factors to induce differentiation of BMSC into neural cells [12, 33–35]. Since it has been suggested that BME is capable of supporting the viability and differentiation of fetal mouse brain neurons , we used low concentrations of this factor in combination with retinoic acid (treatment 2). With this treatment, BMSC slowly differentiated into neuron-like cells and after 7 days they expressed NF-M. Cells survived in induction media for up to at least 14 days.
These results confirmed data reported by other authors suggesting that BMSC are able to differentiate in cell types of varying embryonic origin. In particular, using two different treatments we induced BMSC to differentiate into cells with neuronal phenotype. At the end of both treatments cells expressing neural markers such as neurofilament and GFAP were obtained, whereas vimentin and nestin decreased with differentiation. The presence of GFAP in treated BMSC may be ascribed not only to differentiation into astrocytes, but also to neural precursors, that have been demonstrated to express this protein .
Prior to inducing differentiation, BMSC expressed nestin, a commonly used marker of neural precursor. A transient expression of this intermediate filament in non-neural cells, such as hepatic stellate cells, myogenic and epithelial cells has recently been observed [38, 39]. Moreover, in our experiments the patterns of mRNA and protein expression did not exactly match. For example, mRNA for neurofilaments was present in BMSC and was not modified during differentiation, whereas the protein was absent in untreated cells and its expression was induced by treatments. These data suggest that the expression of neural markers (such as neurofilaments, nestin, GFAP, NSE) is controlled at a translational rather than a transcriptional level. In fact, the constitutive expression of these proteins by BMSC confirms the hypothesis that these cells are "multidifferentiated" cells and thus can retain the ability for neuronal differentiation, as already suggested by Tondreau et al. .
Microscopic examination of the cultures revealed that, using either of the two protocols, only a small percentage of cells (15–20%) showed immunoreactivity for neural markers, suggesting that only a subset of BMSC can differentiate into cells with a neural phenotype. We can therefore hypothesize that this subfraction of BMSC may be constituted by tissue-specific progenitor cells with restricted differentiation potential, capable of giving rise, under specific experimental conditions, to cells characteristically found in other tissues. In fact, morphological and phenotypic heterogeneity of BMSC has been demonstrated: BMSC include small round cells and large polygonal-shaped cells with different multilineage potentials [41, 42]. On the other hand, our findings may indicate the possibility that a population of multipotential stem cells resides in adult bone marrow.
The results reported demonstrate how the microenvironment may be capable of affecting cellular differentiation, since activation of PKs alone leads to neuron-like cell morphology, but the cells express neural markers after both treatment protocols. This observation suggests that specific marker expression is not always accompanied by typical morphological modifications [5, 43]. In contrast, Black and Woodbury  have demonstrated that BME induces stromal rat cells to differentiate into neuronal phenotype cells, but at concentrations 10–100 times higher than those used in the present study. In our experiments, RA and BME concentrations were sufficient to affect the expression of neural markers but not to induce morphological modifications. The differences observed between the two treatments respect to morphological changes and to time required for differentiation may indicate the activation of different intracellular molecular mechanisms in these processes. It has been suggested that the molecular mechanisms involved in the action of RA during embryogenesis may occur via a complex signaling pathway [for review see ]; moreover, it has been hypothesized that both in vitro and in vivo, the number of neural-specific genes encoding transcription factors and signaling molecules induced by RA during neural differentiation are comparable .
In a recent report Lu et al.  observed a neuron-like morphology in BMSC exposed to various stressors. An increase in NSE immunoreactivity was also observed, but it could not be confirmed by RT-PCR, suggesting that these modifications can be the result of cell shrinkage. On the contrary, our results showed the increase of neural markers not only by immunocytochemistry, but also by RT-PCR and western blot.
Hes1 is a transcription factor that plays a key role in neurogenesis regulation maintaining cortical progenitors by inhibiting neurogenesis . Our results show that after 5 hours of treatment 1, the expression of Hes1 transiently increases. We could speculate that the upregulation of Hes1 expression induces differentiation of BMSC into neural precursors and its inhibition leads these cells to differentiate into more mature cells, as observed in vivo during development . On the other hand, we observed an increase of Hes1 expression after 7 days of treatment 2. This result may indicate that retinoic acid treatment leads cells to acquire a less mature phenotype as compared to treatment 1. This hypothesis is confirmed by nestin expression that decreases after treatment 1, but that is not modified by treatment 2.