Adult Bone Marrow Stromal Cells in the Embryonic Brain: Engraftment, Migration, Differentiation, and Long-Term Survival

Ventricular MSCs invade the brain

To determine whether adult rat bone MSCs assume neuronal functions in vivo, we characterized donor cells after transplantation. We assessed invasion, migration, and localization to definitive sites, phenotypic expression, and long-term survival after infusion of undifferentiated MSCs into E15.5 rat telencephalic ventricles in utero. MSCs were isolated and cultured as described previously (Azizi et al., 1998). Passage 17-21 MSCs were labeled with the DNA synthesis marker BrdU and were treated with bFGF before transplantation (see Materials and Methods). After this in vitro labeling protocol, >90% of donor cells incorporated BrdU. One hundred thousand to 150,000 donor cells suspended in 2-3 μl of DMEM were injected into each E15.5 recipient embryo lateral ventricle (for details, see Materials and Methods). In preliminary studies, we also used green fluorescent protein and cyan fluorescent protein as MSC labels but found that expression in vivo was variable. Consequently, we used BrdU in these studies, which favors detection of cells that divide no more than two times, without dilution of the label.

Briefly, intact uterine horns containing the embryos were exposed, and the embryonic ventricles were illuminated with high-intensity fiber optic light. BrdU-labeled MSCs were infused into the visualized ventricles through the uterine wall. Successful transplantation was confirmed by rapid ventricular diffusion of the tracer fast green, which had been added to the MSC suspension. Surgical wounds were sutured, and the embryos were developed in utero until examination (for details, see Materials and Methods).

We initially characterized short-term survival and engraftment. During the first 24 hr after transplantation, at E16.5, sheets of individual BrdU-labeled MSCs were apparent throughout the ventricular system, including the dorsal, central, and anterior horns of the lateral ventricle, superior tectal aqueduct, and fourth ventricle (Fig. 1). In contrast, 48 hr after transplantation, at E17.5, discrete spherical clusters of engrafted MSCs were observed throughout periventricular germinal zones, such as the telencephalic ventricular (VZ) and subventricular (SVZ) zones (Fig. 2). In summary, MSCs infused into the lateral ventricles diffused freely throughout the ventricular system for 24 hr, engrafting into VZ/SVZ throughout the brain as discrete spherical clusters by 48 hr.

Early phenotypic expression

We next examined whether MSCs engrafted in the embryonic neuroepithelium expressed phenotypic traits associated with neuroprecursor cells. Brains were assayed immunocytochemically for colocalization of BrdU and either nestin or vimentin, intermediate filaments expressed by neuroprecursor or radial glia cells, respectively. (Lendahl et al., 1990; Noctor et al., 2002; Kriegstein and Gotz, 2003; Weissman et al., 2003).

Confocal analysis at E17.5 revealed colocalization of nuclear BrdU and cytoplasmic nestin (Fig. 3) in the forebrain VZ/SVZ. At E19.5, Z-series confocal reconstruction of BrdU/vimentin double-labeled sections revealed a subset of donor cells in the cerebellar peduncle/external germinal layer region extending long vimentin-positive fibers into the parenchyma (Fig. 4). These data suggest that MSCs express neuroprecursor and radial glial phenotypic traits in vivo. As expected, donor cells in germinal zones at E17.5 did not express the mature neuronal markers tau or MAP2 (data not shown). Consistent with this contention, parallel in vitro studies demonstrated that >50% of cultured mouse MSCs express the neuroprecursor/radial glia protein RC2 (Fig. 4).

Migration of MSCs

The widespread distribution of transplanted MSCs raised the possibility that donor cells responded to endogenous guidance cues, engaging host mechanisms of cell migration. We investigated this possibility by double staining the E19.5 and E21.5 neocortical white matter and cerebral cortex for BrdU and for nestin, which is expressed in radial glial fibers in these regions (Noctor et al., 2002; Kriegstein and Gotz, 2003; Weissman et al., 2003). Analysis of Z-series confocal reconstructions revealed BrdU-positive MSCs with elongated, migratory nuclear morphologies closely apposed to radially oriented, nestin-positive fibers. The processes spanned the white matter and cortical plate at E19.5 and E21.5 (Fig. 5). Confocal analysis confirmed differential localization of BrdU-positive nuclei (Fig. 5) and nestin-positive radial glial fibers (Fig. 5) in single optical sections 3, 5, 6, and 8 μm below the section surface. Because germinal zone MSCs extended vimentin-positive fibers (Fig. 4), our observations suggest that MSCs generated both migratory neurons and guiding radial fibers.

Longer-term distribution of MSCs

We characterized the overall distribution of transplanted MSCs in recipient brains at E19.5, E21.5, and postnatal day 3 (P3). Widespread distribution of MSCs was detected in distant regions of the telencephalon, including the following: E19.5 neocortex and white matter (Fig. 6); E21.5 dorsal and ventral hippocampus, subventricular zone, white matter, and cortical plate (Fig. 6); and P3 subventricular zone, rostral migratory stream (RMS), and olfactory bulb (Fig. 6).

Expression of immature neuronal markers in diverse areas

The widespread migration of donor cells to distant CNS areas may expose different subpopulations of MSCs to different microenvironments. In turn, environmental signals may induce different MSC fates. Consequently, we examined area-specific MSC phenotypic expression using confocal analysis.

MSCs confined to the subventricular zone continued to express the neuroepithelial marker nestin at E19.5 and E21.5 (Fig. 7). In contrast, another subset of donor cells in the pons and basal ganglia exhibited astrocytic morphologies and expressed the astrocyte marker GFAP (Fig. 8).

Four days after transplantation, MSCs in the midbrain exhibited multipolar neuronal morphologies and expressed the immature neuronal protein β-III-tubulin (Fig. 9), consistent with immature neuronal phenotypic expression.

Maturation of phenotypic expression: temporal and area specificity

To assess ongoing development, we examined E19.5 and E21.5 sections through the neocortex for the mature neuronal proteins tau and MAP2. MSCs with elongated nuclear morphologies in the white matter at E19.5 did not express tau, as indicated by rotating the Z-series reconstruction 360° (Fig. 10). In contrast, MSCs in cortical layers at E21.5 expressed mature neuronal tau and MAP2 (Fig. 10). BrdU/tau and BrdU/MAP2 cellular colocalization was determined confocally by analysis of Z-series reconstructions and single optical sections in corresponding yz and xz orthogonal planes. In summary, different subsets of MSCs expressed different gene products and morphologies in a temporal and area-specific manner, suggesting that MSCs respond to specific signals in different microenvironments.

Expression of region-specific phenotypes: calbindin and tyrosine hydroxylase

To determine whether MSC-derived neuronal subtypes were regionally specified, we examined the olfactory bulb as a prototype. The bulb was selected because we detected an extensive population of NeuN⁺ donor cells in the periglomerular region at 1 and 2 months (see next section, Quantitative analysis of MSC distribution and survival) and because different gene products have been localized to distinct subpopulations of periglomerular cells (Halász et al., 1977, 1985). For example, the periglomerular layer contains calbindin- and TH-immunoreactive neurons in distinct, non-overlapping subregions.

In fact, we detected calbindin expression in BrdU⁺ MSCs within the external plexiform layer of the periglomerular region (Fig. 11), consistent with previous reports of the distribution of these neurons in the normal mature brain (Halász et al., 1985). Rotational analysis of 360° around the y-axis definitively colocalized BrdU and calbindin (Fig. 11).

Conversely, TH was colocalized to a distinct BrdU⁺ subpopulation in a separate subregion of the periglomerular region (Fig. 12), as described previously (Halász et al., 1985). In summary, these observations are consistent with the regional specification of differentiation of neuronal subtypes from the MSC donor population.

Quantitative analysis of MSC distribution and survival

To obtain a quantitative estimate of the distribution and survival of transplanted MSCs, we used unbiased stereology, using a modification of Abercrombie's method (see Materials and Methods) (Kramer et al., 1999). We examined multiple brain areas at 1 and 2 months postnatally, after transplantation of MSCs at E15.5. The donor cells were widely distributed, consistent with the embryonic pattern. At 1 and 2 months postpartum, abundant BrdU-positive cells were apparent in the following: olfactory bulb and rostral migratory stream; the frontal, parietal, and occipital cortices (predominantly in the upper laminas); the hippocampus and dentate gyrus; and the VZ and SVZ. MSCs were so densely packed in the SVZ (Figs. 2, 3) that it was often difficult to resolve individuals, and we used other areas to perform reliable morphometry. We chose the 1 month frontal cortex and olfactory bulb as representative areas for detailed analysis and scored multiple sections from each of three rats (Table 1). Abundant cells were apparent, although numbers varied from animal to animal (Table 1). In the frontal cortex, BrdU-positive cells varied from 118 ± 22 to 501 ± 88 per 16 μm section. To obtain an estimate of donor cell number per unit volume, we counted cells in the area circumscribed by the rostral extreme of the ventricle to the frontal pole and by the dorsal extreme of the ventricle to the dorsal extent of the cortex. We summed the volume over the total number of sections counted for each rat (for details, see Materials and Methods). We calculated that the frontal cortices of the three individual rats contained 4792, 2152, and 1124 BrdU⁺ cells/mm³, respectively, 1 month after transplantation.

In the 1 month olfactory bulb, MSCs varied from 49 ± 9 to 239 ± 32 cells per 16 μm section among individual animals (Table 1). To estimate the density of MSCs, we used an area circumscribed by the rostral extreme of the bulb to 500 μm caudally and a height spanning the dorsoventral extent of the bulb. The depth was estimated from section thickness, as above. For the three rats, MSC density measured 10,864, 10,091, and 2229 cells/mm³. Consequently, the MSCs displayed extensive long-term survival from the embryonic to the young adult host brain.

To determine whether the surviving MSCs exhibited neuronal traits, we examined BrdU⁺ cells for NeuN, a neuron-specific marker expressed in postmitotic cells (Sarnat et al., 1998). Colocalization of BrdU and NeuN was assessed by confocal microscopy in the frontal cortex and olfactory bulb (see Materials and Methods). In the frontal cortex, virtually all of the BrdU⁺ cells stained for NeuN, and 99, 100, and 97% of the MSCs were NeuN positive in the three animals, respectively (Table 2, Fig. 13). In the olfactory bulb, 70, 83, and 79% of the MSCs were NeuN⁺, respectively (Table 2). In summary, MSCs in widely dispersed areas that differ functionally and anatomically expressed this common neuronal trait.

Overview

The present study was designed to determine whether MSCs, which can be differentiated into presumptive neurons in vitro, exhibit neuronal characteristics in the brain in vivo. We chose the neurogenetic embryonic rat brain as a fertile substrate to assess neuronal potentials. In vitro analysis confirmed that the isolated MSCs used in these studies exhibited the ability to differentiate into osteoblasts and adipocytes, indicating the multipotentiality of the starting population. This represents the first demonstration that MSCs engraft, migrate, differentiate, and survive long term in the embryonic brain in vivo.

Ventricular MSCs enter the brain

Infusion of MSCs into the ventricles at E15.5 resulted in the homogeneous distribution of sheets of cells in the forebrain ventricular system within 24 hr, indicating free diffusion within CSF pathways (Fig. 1). Forty-eight hours later, at E17.5, discrete spherical clusters of MSCs were detected throughout periventricular germinal zones (Fig. 2), reminiscent of the pluripotent neurospheres that form in NSC cultures (Reynolds and Weiss, 1992; Reynolds et al., 1992). The MSC spheres observed in vivo may exhibit the multipotentiality of in vitro neurospheres, a contention we are now examining.

Widespread distribution and phenotypic expression of MSCs

The evidence for extensive MSC migration implies that the cells exhibited another characteristic of normally developing brain neurons, the localization to multiple, distant definitive sites. Abundant donor cells were detected in the neocortex, hippocampi, cerebellar peduncles, colliculi, thalamus, midbrain, forebrain germinal zones, rostral migratory stream, and olfactory bulbs. The MSCs are apparently equipped to follow diverse migratory paths, through different microenvironments to multiple target areas with extensive long-term survival (see below).

The cells expressed different gene products and morphologies in a region-specific manner, apparently exhibiting the plastic ability to respond specifically to different microenvironments. Donor cells confined to the SVZ continued to express neuroepithelial nestin at E19.5 and E21.5. MSCs in the midbrain exhibited multipolar neuronal morphologies and expressed β-III-tubulin, a neuronal protein, at E17.5. Neocortical cells, on the other hand, expressed mature neuronal tau and MAP2 at E21.5. Donor cells in the pons and basal ganglia exhibited astrocytic morphologies and expressed GFAP. MSC GFAP was not detected in the neocortex at this time. We are presently examining GFAP expression in MSCs transplanted at different ages after different postoperative intervals. Indeed, in previous studies, MSC GFAP has been detected in multiple forebrain regions after transplantation to the mouse at P3 (Kopen et al., 1999).

In addition to expressing mature, generic neuronal traits in an area-specific manner, the MSCs exhibited subregion-specific differentiation. Appropriate subsets of MSCs expressed the specific trait calbindin within the external plexiform layer of the olfactory bulb periglomerular region (Fig. 11), consistent with previous reports of neurons in the mature brain (Halász et al., 1985).

MSCs in non-overlapping periglomerular subregions expressed TH, characteristic of bulb dopaminergic neurons in this subregion (Fig. 12) (Halász et al., 1985).

In summary, donor MSCs exhibited subregion-specific differentiation, consistent with plasticity of this adult population.

Transdifferentiation and cell fusion

The diverse fates of the MSCs argue against a single intrinsic mechanism governing differentiation. MSCs express heterogeneous traits characteristic of (1) radial glia, (2) SVZ progenitors, (3) migratory cells, (4) parenchymal neurons, and (5) glia. MSC fate, consequently, appears to be regulated by multiple influences, presumably including different microenvironments. This may be relevant to several recent reports suggesting that marrow cells can assume different fates through fusion with other cell types. The sheer diversity and widespread distribution of distinct MSC-derived neural types in our studies argues against fusion as a dominant mechanism. Additionally, we never observed multiple nuclei in extensive confocal searches. Furthermore, previous studies have suggested that fusion is a rare event. For example, genetically altered MSCs cultured with embryonic stem (ES) cells yielded only 2-11 hybrid clones per 10⁶ marrow cells (Terada et al., 2002), and 1 in 10⁵ when genetically altered neural cells were exposed to ES cells (Terada et al., 2002; Ying et al., 2002). In contrast, cultured marrow cells that differentiate into neurons in the absence of other cell types express a variety of neuroglial characters (Sanchez-Ramos et al., 2000; Woodbury et al., 2000; Lee et al., 2003), and marrow multipotent adult progenitor cells differentiate into endothelium, neuroectoderm, and endoderm, as well as mesenchymal derivatives (Jiang et al., 2002a,b, 2003). Furthermore, pancreatic cells differentiate into hepatocytes in vitro and in vivo in the absence of fusion (Krakowski et al., 1999; Shen et al., 2000). In vivo, male donor peripheral blood stem cells transdifferentiate into epidermal, hepatic, and gastric mucosal cells in human females, and the cells contained only one X and one Y chromosome, excluding fusion as a mechanism (Korbling et al., 2002). Similarly, differentiation of male marrow cells into mucosal cells throughout the gastrointestinal tract of human female recipients occurred in the absence of evidence of fusion (Okamoto et al., 2002). Nevertheless, in some systems, fusion may play a role: donor hematopoietic stem cells rescued mutant mice with hepatic deficiency by fusing with endogenous hepatic cells (Wang et al., 2002). Moreover, Alvarez-Dolado et al. (2003) recently found that whole bone marrow cells grafted into lethally irradiated mice were capable of fusing with some Purkinje neurons, cardiomyocytes, and hepatocytes. However, the frequency of fusion with Purkinje cells was exceedingly low, 2 of 1.5 × 10⁶ at 2 months and 5 of 1.5 × 10⁶ at 4 months. In contrast, we detected thousands of MSCs throughout the brain. This provocative work, however, may not be directly relevant to the present study. First, their experiments demonstrated that hematopoietic marrow cells fuse in vivo. However, our studies were performed with marrow stromal cells, an entirely different population. Second, they used irradiated adult mouse hosts, whereas we examined normal embryonic rats. Finally, we found presumptive neurons in virtually all parenchymal and germinal areas, whereas they only described fusion in rare Purkinje cells. In summary, the relevance of their important work with hematopoietic cells grafted to lethally irradiated mice, yielding fusion of undetermined frequency, must be documented in additional studies. Furthermore, in a counter example, transplanted marrow cells differentiated into human buccal mucosa in vivo, of which 12.7% were donor derived, although only 0.1% exhibited evidence of fusion (Tran et al., 2003). It is particularly difficult to envision mechanisms through which cellular fusion could generate the multiplicity of cell types from MSCs in the present study. Consideration of the quantitative aspects of the present results emphasizes this point.

Long-term survival of transplanted cells: quantitative analysis

MSC-derived cells exhibited widespread survival at 1 and 2 postnatal months, the longest times examined. Cells were abundant in the germinal VZ/SVZ and RMS, as well as widespread parenchymal areas. Widespread survival indicates adaptation of the cells to diverse brain niches that differ anatomically, physiologically, and functionally, substantiating plasticity of the MSCs in vivo.

Morphometry was performed at 1 month to estimate long-term survival and plasticity. MSCs were so densely packed in the VZ/SVZ that resolution of individuals and accurate counting were not possible. We chose the frontal cortex and olfactory bulb as representative of functionally critical parenchymal and germinal areas that require migration. Although sections from both areas exhibited abundant BrdU⁺ cells, numbers varied from animal to animal, presumably reflecting variable access from different ventricular locations, differences in migratory pathways, altered mitotic rates, and variable survival itself. In the frontal cortex, there were 4792, 2152, and 1124 donor cells/mm³ in individual rats. Clearly, thousands of cells survive into young adulthood.

Abundant cells were also detected in the olfactory bulb. In the 1 month bulb, there were 10,864, 10,091, and 2229 MSCs/mm³ in individual rats. In summary, long-term survival of donor cells in diverse, developmentally changing brain environments indicates the viability and plasticity of MSC derivatives.

To begin identifying the phenotype of the MSCs at 1 month, we stained sections for NeuN, a postmitotic, neuron-specific marker (Sarnat et al., 1998). Approximately 99% of the frontal cortex MSCs expressed NeuN, whereas 70, 83, and 79% were NeuN⁺ in the olfactory bulbs of three separate animals. It should be reiterated that donor cells in the SVZ did not express NeuN at this age, indicating that expression is region appropriate and not a function of intrinsic MSC “aging.” We conclude that transplanted MSCs exhibit long-term, persistent plasticity in vivo, extending at least into young adulthood.