Cyclin-dependent kinase 4 signaling acts as a molecular switch between syngenic differentiation and neural transdifferentiation in human mesenchymal stem cells

Abstract

Multipotent mesenchymal stem/stromal cells (MSCs) are capable of differentiating into a variety of cell types from different germ layers. However, the molecular and biochemical mechanisms underlying the transdifferentiation of MSCs into specific cell types still need to be elucidated. In this study, we unexpectedly found that treatment of human adipose- and bone marrow-derived MSCs with cyclin-dependent kinase (CDK) inhibitor, in particular CDK4 inhibitor, selectively led to transdifferentiation into neural cells with a high frequency. Specifically, targeted inhibition of CDK4 expression using recombinant adenovial shRNA induced the neural transdifferentiation of human MSCs. However, the inhibition of CDK4 activity attenuated the syngenic differentiation of human adipose-derived MSCs. Importantly, the forced regulation of CDK4 activity showed reciprocal reversibility between neural differentiation and dedifferentiation of human MSCs. Together, these results provide novel molecular evidence underlying the neural transdifferentiation of human MSCs; in addition, CDK4 signaling appears to act as a molecular switch from syngenic differentiation to neural transdifferentiation of human MSCs.

Introduction

In recent years, a growing body of evidence has indicated that under certain experimental conditions, mesenchymal stem/stromal cells (MSCs) can differentiate into mesodermal, ectodermal and endodermal cell types.^1-3 Evidence also suggests that MSCs have the potential to become neural cells, and commitment to these lineages depends on the culture conditions.^1,2,4,5 In addition to neural morphologic differentiation, selected neural induction media functionally convert MSCs to typical neuronal and glial cells,^4,6 indicating the strong possibility that controlled neural differentiation of human MSCs could become an important source of cells for cell therapy of neurodegenerative diseases, because autologous adult MSCs are more easily harvested and effectively expanded than corresponding neural stem cells.

Furthermore, gaining insight into the molecular evidence of neural cell lineage has great potential to improve our understanding of a variety of neurologic disorders in humans. However, the underlying molecular mechanisms of how human MSCs can be converted into neural cells still needs to be determined.

Recent important advances in stem cell biology are the ability to modulate and control stem cell fate and function. Interestingly, cell-permeable small chemical molecules have proven useful for inducing the differentiation of various stem cells.^7,8 Small molecules modulate multiple specific targets within a protein family or across different protein families, allowing the production of a desirable phenotype in a specific lineage differentiation manner. For example, small-molecule inhibitors, such as 5-azacytidine (inhibitor of DNA methyltransferases), suberoylanilide (inhibitor of histone deacetylases), bortezomib (proteasome inhibitor) and imatinib mesylate (kinase inhibitor) induce differentiation of various stem and progenitor cells.⁷ Therefore, small molecule approaches provide new insight into stem cell biology and may contribute to biological therapeutics for tissue repair and regeneration.

Cell cycle regulation, which is mediated by cyclin-dependent kinases (CDKs) in a complex with corresponding cyclins, appears to have a critical role in the initial phases of neuronal development, when neural stem and progenitor cells commit to neuronal fate, withdraw from the cell cycle and begin to differentiate. Several forms of evidence indicate a strong interrelation between cell cycle control and neural differentiation.⁹ Cycling neural progenitor cells generate post-mitotic neurons, in which cell cycle activity is arrested by the increased expression of CDK inhibitory genes. Numerous cell cycle inhibitory genes are expressed in the brain in a ubiquitous fashion (p19^INK4d) or in specific regions as follows: p15^INK4a in the forebrain, p21^Cip1 in the cerebellum and p27^Kip1 in the cerebellum and cortical post-mitotic neurons.^2,10,11 Subsequently, signaling through the CDK-Rb pathway seems to be critical in the regulation of terminal mitosis and commitment to the differentiation of neural progenitor cells.¹² In a murine model, Rb becomes essential immediately following commitment to a neuronal fate, and in the absence of Rb, virtually all neuronal populations undergo apoptotic cell death.¹³ Several groups have recently reported a tight correlation between inhibition of CDK activity and neuronal differentiation, showing, for example, that overexpression of CDK4 delays neurogenesis and inhibition of CDK2 and/or CDK4 expressions induces neuronal differentiation in neural stem cells.^14,15 Together, these findings strongly suggest that cell cycle regulators, such as CDKs and Rb, play substantial roles in the highly regulated processes of self-renewal, asymmetric division and differentiation of neural stem and progenitor cells. We report here that the activity of CDK4 was required for syngenic differentiation, such as adipogenesis and osteogenesis, from proliferation. However, the loss of CDK4 activity or expression selectively led to reversible neural transdifferentiation. The results of this study suggest that the regulation of CDK4 activity is a key signaling device to confer to human MSCs for balancing between syngenic differentiation and reversible neural transdifferentiation.

Results

Human MSCs selectively commit to transdifferentiate into neural cells following CDK inhibition

We initially attempted to examine the functional cross-talk between the cell cycle-related kinases and the proliferation or differentiation of human MSCs. We isolated human mesenchymal stem/stromal cells (hMSCs) from the fatty portion of liposuction aspirates and bone marrow (BM). First, the isolated human adipose-derived MSCs (hAD-MSCs) were characterized by specific surface antigen expression, as reported previously,¹⁶ and these cells were positive for known MSC markers (e.g., CD44, CD49d, CD73, CD90 and CD105) but negative for CD34, CD45 and CD106 (Fig. 1A and B). Fluorescence-activated cell sorting analysis using an antibody against the hematopoietic lineage marker CD45 showed that hAD-MSCs cultures were virtually pure and contained < 1% hematopoietic lineage cells (data not shown). The hAD-MSCs are multipotent and differentiated into adipocytes, osteoblasts and neural cells, which were characterized by Oil Red O, von Kossa staining and anti-microtubule-associated protein 2 (MAP2) antibody, respectively (Fig. 1C).

Figure 1. Adipose-derived MSCs selectively commit to transdifferentiate into neural cells following CDK inhibition. (A) Phase contrast image of human adipose-derived MSCs (hAD-MSCs). hAD-MSCs were isolated from the fatty portion of liposuction aspirates. (B) hAD-MSCs were labeled with FITC- or PE-coupled antibodies specific for CD34, CD44, CD45, CD49d, CD73 (not shown), CD90 and CD106 antibodies or immunoglobulin isotype control. The surface phenotype was analyzed by FACS. Colored histograms illustrate the control immunoglobulin and open histograms illustrate the specific antibodies, as indicated. Isolated MSCs are positive for the MSC markers (CD44, CD49d and CD90), whereas they are negative for hematopoietic stem cell markers (CD34, CD45 and CD106). Blue-colored shadows represent the isotype control and green represent the staining against each specified antibodies. (C) Adipogenic differentiation resulted in the formation of lipid vacuoles, which were stained with Oil Red O. Osteogenic differentiation showed a highly enriched extracellular matrix stained with von Kossa. Neurogenic differentiation led to a typical neural cell appearance (i.e., a condensed cell body, spherical and refractile), and were immunostained with anti-MAP2 antibody. (D) hAD-MSCs were isolated from six different donors, and characterized as described in (B and C). hAD-MSCs were treated with 10 μM CDK4i for 24 h and digitally imaged. Treatment with CDK4i induced the morphologic neural transdifferentiation with high efficiency. (E) Quantified comparison of cell death and neural transdifferentiation of hAD-MSCs following CDK inhibitor (PurA, 25 μM and CDK4i, 10 μM) treatment. The rate of differentiating hAD-MSCs was randomly counted from 10 different microscopic images (X200) per sample. Averages were obtained from six different MSCs and donors. (F) hAD-MSCs were treated with Purvalanol A (PurA, 25 μM) and CDK4i (10 μM). Cells were harvested, stained with propidium iodide and analyzed by flow cytometry to determine their DNA contents. (G) Phase contrast images of human fetal white matter progenitor/precursor cells (NPCs), human U251 glioblastoma cells and human AD-MSCs. NPCs were cultured in DMEM (NPC) or DMEM supplemented with ITSFn medium (NPC w/ ITSFn). (H) RT-PCR analysis for neural progenitor/precursor cell markers (NCAM, Msi1, and Sox2), a marker for stem cell multi-potency (Nestin) and mesenchymal stem cell surface markers (CD90 and CD105). GAPDH mRNA was utilized as a control.

hAD-MSCs were expanded and exposed to a series of cell cycle-related kinase inhibitors targeting CDK1, CDK1/2 (Purvalanol A and Roscovitine), CDK4 (Otava Ltd.), Aurora B (ZM447439), Polo-like kinase 1 (BI2536 and PPG), MEK1 (PD98059) and GSK3b (TDZD-8) (Fig. S1). Surprisingly, the exposure of hAD-MSCs to CDK inhibitors (CDK1, CDK1/2 and CDK4) resulted in unexpected changes in morphology, as responsive cells assumed forms typical of neural cells, as indicated by the cell body condensing and becoming increasingly spherical and refractile (Fig. 1D and E; Fig. S1). However, we did not observe significant induction of neural cells from hAD-MSCs that were consecutively cultured throughout several passages without CDK inhibition (data not shown). Among CDK inhibitors, the CDK4 inhibitor has been shown to have high efficiency of transdifferentiation of hAD-MSCs into neural cells (i.e., up to 85% transdifferentiation; Fig. S1). Next, we examined the effect of CDK inhibitor treatment on cell cycle progression in the same experimental condition. As shown in Figure 1F, the treatment of CDK4i showed the similar cell cycle profile compared with the control treatment, whereas those of Purvalanol A treatment reduced G₁ phase but significantly increased S and G₂/M phases. However, treatment with other cell cycle-related kinase inhibitors showed no apparent difference in the proliferation, division and differentiation. In particular, although mitotic kinases Aurora kinase and Polo-like kinase 1 (Plk1) are known to play a crucial role in controlling the mitotic cell cycle and post-mitotic division,¹⁷ selective inhibitions of Aurora kinases and Plk1 did not affect the transdifferentiation of hAD-MSCs. Although the degree of neural transdifferentiation of hAD-MSCs into a neural phenotype was variable, possibly due to donor heterogeneity, we observed sufficient transdifferentiation of all hAD-MSCs that were isolated from six different donors by treatment with CDK4 inhibitor (Fig. 1D).

To exclude the possibility that neural progenitor/precursor cells are contaminated in the isolated hAD-MSCs populations, we obtained human neural progenitor/precursor cells (hNPCs) from fetus subcortical white matter and examined neural progenitor/precursor cell markers (NCAM, Msi1 and Sox2), markers for stem cell multi-potency (Nestin) and mesenchymal stem cell surface markers (CD90 and CD105) (Fig. 1G and H). Importantly, RT-PCR analyses revealed that neural progenitor/precursor cells were not contaminated in the isolated hAD-MSCs populations. Together, this unexpected observation indicates that CDK inhibition morphologically and selectively induces the neural transdifferentiation of hAD-MSCs.

Inhibition of CDK4 activity induces neural transdifferentiation in both human adipose- and bone marrow-derived mesenchymal stem/stromal cells

hAD-MSCs efficiently acquired the typical appearance of neural cells as early as 12 h post-treatment with CDK4 inhibitor (Fig. 2A). We also examined the neural transdifferentiation of hAD-MSCs by CDK inhibitor using time-lapse microscopy. hAD-MSCs were transduced with rAd-H2B-RFP by HP4-PTD co-treatment to visualize DNA then treated with CDK4 inhibitor 24 h post-transduction (Fig. 2B). As expected, treatment with CDK4 inhibitor transformed hAD-MSCs into neural cells based on the condensed, spherical and refractile appearances of the cells.

Figure 2. Inhibition of CDK4 activity induces neural transdifferentiation in both human adipose- and bone marrow-derived mesenchymal stem/stromal cells. (A) Human AD-MSCs were treated with the control DMSO or inhibitors selective for CDK1 and CDK2 (data not shown) or CDK4, as indicated, and digitally imaged. Treatment with CDK4 inhibitor (CDK4i, Otava Ltd.) induced the morphologic neural transdifferentiation with high efficiency. (B) Time-lapse microscopic image of neural transdifferentiation of hAD-MSCs by CDK4i treatment. Isolated MSCs were infected with a recombinant adenovirus encoding H2B-RFP-fused protein (rAd-H2B-RFP) to visualize DNA, treated with CDK4i and digitally imaged. The arrow indicates the induced neural transdifferentiation of hAD-MSCs. (C) hAD-MSCs were treated with CDK4 inhibitor or control DMSO, while the 24 h post-treatment neural subtype markers, neural precursor (Nestin), neuron markers (β-tubulin III and MAP2) and glial markers (GFAP and NL3) were amplified by RT-PCR. (D) Human bone marrow-derived MSCs (hBM-MSCs) were labeled with FITC- or PE-coupled antibodies specific for CD34, CD44, CD45, CD49d or CD70 antibodies, and immunoglobulin isotype control. Open histograms illustrate the control immunoglobulin, and colored histograms illustrate the specific antibodies, as indicated. Isolated hBM-MSCs were positive for CD44, CD49d and CD70, but negative for CD34 and CD45. hBM-MSCs were cultured for 2–3 wk in the medium for adipogenic or osteogenic differentiation. Differentiation into adipocytes and osteoblasts was confirmed by staining using Oil Red O and alkaline phosphatase, respectively. (E) Time-lapse microscopic image of neural transdifferentiation of hBM-MSCs by CDK4i treatment. Isolated hBM-MSCs were infected with rAd-H2B-RFP, treated with CDK4i and digitally imaged. The arrow indicates the induced neural transdifferentiation of hBM-MSCs. (F) hBM-MSCs were treated with CDK4 inhibitor or control DMSO, while the 24 h post-treatment neural subtype markers, neural precursor (Nestin), neuron markers (β-tubulin III and MAP2) and glial markers (GFAP and NL3) were amplified by RT-PCR.

In order to delineate the neural transdifferentiation of hAD-MSCs by CDK inhibition, we then examined the expression of neural marker transcripts, neuronal markers [β-tubulin III (βIII Tub), microtubule-associated protein 2 (MAP2)], glial markers [glial fibrillary acidic protein (GFAP), neuroligin 3 (NL3)] and a broad marker for neural precursor cells (Nestin). mRNAs were amplified by RT-PCR from hAD-MSCs 24 h post-treatment with CDK4 inhibitor. hAD-MSCs treated with CDK4 inhibitor expressed significantly higher levels of the expected neural marker genes, βIII Tub, MAP2, GFAP and NL3 mRNAs, with the exception of Nestin, compared with the control-treated MSCs and hNPCs (Fig. 2C). In addition, we observed similar results from human bone marrow-derived MSCs (hBM-MSCs), which showed sufficient transdifferentiation into neural cells by treatment with CDK inhibitors (Fig. 2D–F). hBM-MSCs treated with CDK4 inhibitor also showed the typical morphology of neural cells (Fig. 2E) and expressed significantly higher levels of the neural marker genes, βIII Tub, MAP2, GFAP and NL3 mRNAs, compared with the control-treated hBM-MSCs and hNPCs (Fig. 2F). However, the induction of these neural marker proteins in CDK4 inhibitor-treated MSCs was not likely due to the increased confluence of cell proliferation, because CDK inhibition mostly decreases the proliferation of somatic and stem cells (data not shown). These data indicate that neural-like morphology induced by CDK4 inhibition corresponds to the expression of neural marker genes.

Targeted inhibition of CDK4 expression induces the neural transdifferentiation of MSCs

To explore the direct involvement of CDKs in adipose-derived MSC transdifferentiation into neural cells, we generated recombinant adenovirus (rAd) that selectively depleted CDK1, CDK2 and CDK4 (rAd-GFP-shCDK1, -shCDK2 and -shCDK4, respectively). However, the transduction efficiency of rAd is extremely low in stem cells that express very low levels of the primary rAd receptor.¹⁸ The discovery of protein transduction domains (PTDs) enabled the transduction of rAd into adult stem cells.¹⁹ Thus, the co-treatment of HP4-PTD derived from herring protamine (HP) with rAd-expressing green fluorescence protein (rAd-GFP) dramatically enhanced in vitro transduction of rAd-GFP into hAD-MSCs.¹⁹ Under similar experimental conditions, we transduced rAd-GFP (a negative control), rAd-GFP-shCDK1, rAd-GFP-shCDK2 or rAd-GFP-shCDK4 into ADSCs, and thereby significantly depleted endogenous CDK1, CDK2 and CDK4, respectively (Fig. 3A). hAD-MSCs were then transduced with rAd-GFP or rAd-GFP-shCDKs and digitally monitored by time-lapse microscopy. Importantly, hAD-MSCs transduced with recombinant adenoviruses expressing shCDK1 or shCDK2 did not exhibit the clear appearance of neural morphology (Fig. 3B). However, transduction of rAd-GFP-shCDK4s into hAD-MSCs resulted in the obvious induction of neural morphology, which was distinguished from the normal cell morphology by several criteria, including the spherical and retractile appearance of the cells and the presence of bipolar or multipolar neurite-like projections (Fig. 3B). However, we showed that depletion of CDK1 or CDK2 responded very slowly and had a weak, neural-like morphology at a later time point (data not shown). In summary, although the treatment of CDK1 and CDK2 chemical inhibitors slightly induced the neural transdifferentiation of hAD-MSCs in normal culture medium (Fig. S1), inhibition of CDK4 rather than CDK1 and CDK2 appeared to be the major responder with the highest frequency of transdifferentiation of hAD-MSCs into neural cells. This effect of neural transdifferentiation of hAD-MSCs by CDK1 and CDK2 inhibitors appeared to be related to the broad spectrum of CDK inhibition in a concentration-dependent manner. Therefore, two chemical compounds with primary activity against CDK1 and CDK2 provided additional activity toward CDK4.²⁰

Figure 3. Targeted inhibition of CDK4 expression induces the neural transdifferentiation. (A) To deplete the endogenous CDK1, CDK2 and CDK4, ADSCs were transduced with recombinant adenovirus expressing GFP fused-shCDK1 (rAd-GFP-shCDK1), -shCDK2 (rAd-GFP-shCDK2) or -shCDK4 (rAd-GFP-shCDK4), respectively, in conjugation with HP4-PTD. Forty-eight hours post-transduction, cell lysates were prepared and immunoblotted with anti-CDK1, anti-CDK2, anti-CDK4 and anti-actin antibodies. (B) hAD-MSCs were transduced with rAd-GFP [as a control (data not shown)], rAd-GFP-shCDK1, rAd-GFP-shCDK2 or rAd-GFP-shCDK4 and rAd-H2B-RFP to visualize DNA. Transduced hAD-MSCs were cultured in normal culture medium and digitally monitored by time-lapse microscopy. Images were taken at 6 min intervals, showing the false-colored GFP and RFP emissions. The red-colored arrows indicate the neural cells induced by CDK4 depletion. (C) Neural sub-type marker transcripts (Nestin, S100, NL3 and GFAP) were amplified by RT-PCR from hAD-MSCs transduced with rAd-GFP or rAd-GFP-shCDK4. GAPDH mRNA was utilized as an internal control for RT-PCR.

We then performed RT-PCR analysis to determine whether the observed neural morphology was accompanied by the expression of neural marker genes. As shown in Figure 3C, hAD-MSCs’ selectively depleting CDK4 expression showed significant upregulation of neural marker gene expression, such as S100, NL3 and GFAP mRNA, compared with the control MSCs or MSCs depleting CDK1 or CDK2 (data not shown). Together, the targeted inhibition of CDK4 expression efficiently and selectively induced the neural transdifferentiation of hAD-MSCs.

Inhibition of CDK4 activity attenuated the syngenic differentiation of hAD-MSCs

To explore the effect of CDK4 inhibition on the adipogenic and osteogenic differentiations of hAD-MSCs, the MSCs were treated with CDK4 inhibitor and further cultured in adipogenic and osteogenic media for 2 wk. The hAD-MSCs were efficiently differentiated into adipocytes (Fig. 4A) and osteoblasts (Fig. S2), which were characterized by Oil Red O and von Kossa stainings, respectively, in the absence of CDK4 inhibitor treatment. The treatment of high concentration of CDK4 inhibitor efficiently induced the neural transdifferentiation of hAD-MSCs within 24 h post-treatment, but the cells mostly underwent apoptotic cell death following 3 d post-treatment (Fig. 4A and B). However, the treatment of a low concentration of CDK4 inhibitor, even after 2 wk of treatment, was viable but very weakly induced the neural transdifferentiation. In these conditions, the treatment of a low concentration of CDK4 inhibitor sharply reduced the adipogenic (Fig. 4A and C) and osteogenic (Fig. S2) differentiations of hAD-MSCs. We further examined the adipogenic expression marker transcripts, Adiponectin, Lipoprotein lipase and PPARγ2 in the MSCs cultured in the absence or presence of CDK4 inhibitor (Fig. 4D). Similarly, the treatment of CDK4 inhibitor during the adipogenic differentiation significantly reduced the expressions of these adipogenic markers compared with the control treatment. Taken together, these data suggest that the inhibition of CDK4 activity may attenuate the mesodermal syngenic differentiation of hAD-MSCs.

Figure 4. Inhibition of CDK4 activity reduces the differentiation potential of adipose-derived MSCs into syngenic lineages. (A) hAD-MSCs were treated with CDK4 inhibitor (2 μM, Otava Ltd.) or control DMSO and then cultured in the adipogenic differentiation media. At 2 wk post-treatment, the MSCs showed the formation of lipid vacuoles, which were stained with Oil Red O. (B) hAD-MSCs were treated with CDK4 inhibitor (2 μM and 10 μM) or control DMSO for 24 h. The hAD-MSCs treated with the high concentration of CDK4 inhibitor showed a typical neural cell appearance. (C and D) hAD-MSCs were treated with CDK4 inhibitor (2 μM) or control DMSO and further cultured in the absence or presence of adipogenic differentiation media for 2 wk. The relative rate of adipogenic differentiation, the lipid vacuoles of which were stained with Oil Red O, was measured at OD500 nm (C). Adipogenic marker transcripts (Adiponectin, Lipoprotein lipase and PPARγ2) were amplified by RT-PCR from the MSCs as above. GAPDH mRNA was utilized as an internal control for RT-PCR (D).

Reversibility between the differentiation and dedifferentiation of adipose-derived MSCs by forced regulation of CDK4 activity

Interestingly, we observed the potential dedifferentiation of hAD-MSCs by the removal of CDK4 inhibitor. Thus, we determined whether or not the regulation of CDK activity was reciprocally reversible between the transdifferentiation and dedifferentiation of human MSCs. We transduced rAd-H2B-RFP into MSCs by co-treatment with HP4-PTD and then replaced the tissue culture medium 24 h post-transduction. We periodically managed the activity of CDK4 by the treatment or removal of CDK4 inhibitor after an interval of 12 h. Inhibition of CDK4 activity clearly induced the neural morphology of MSCs < 10 h post-treatment with CDK4 inhibitor (Fig. 5A). Unexpectedly, the removal of CDK4 inhibitor from the MSCs, which were treated with CDK4 inhibitor for 12 h, clearly abolished the neural morphology of the differentiated MSCs and transformed it into the normal morphology of MSCs (Fig. 5A). The neural transdifferentiation of MSCs by CDK4 inhibitor treatment was confirmed by the upregulation of neural marker gene expression, whereas removal of CDK4 inhibitor significantly downregulated the expression of neural marker genes (Fig. 5B). Further treatment of CDK4 inhibitor, thus hypophosphorylating Rb, induced the re-differentiation of MSCs into neural cells (Fig. 5C). We observed the repeated morphologic and biochemical changes of MSCs into neural transdifferentiation and dedifferentiation in responding to CDK4 inhibitor and, in turn, the control treatment. This is an unusual observation, because the regulation of CDK4 activity was sufficient to drive into the reversible neural transdifferentiation, even in the absence of known extra- or intra-cellular neurogenesis signals, such as sonic hedgehog (SHH), 3-isobutyl-1-methylxanthine (IBMX), retinoic acid (RA), nerve growth factor-β (NGF-β) and basic fibroblast growth factor (bFGF).²¹ It appears that differentiated MSCs are not mature neural cells, which still bear stem cell properties. Although at this stage we do not understand how the CDK4 activity mechanistically regulates the neural transdifferentiation of MSCs, it is possible that either CDK4 signaling (or the hypophosphorylation of unidentified substrates of CDK4s) drives the neural transdifferentiation of MSCs, or CDK4 signaling instructs MSCs to a neural fate (Fig. 5D).

Figure 5. The reversibility between transdifferentiation and dedifferentiation of adipose-derived MSCs by forced regulation of CDK4 activity. (A and C) The hAD-MSCs were transduced with rAd-H2B-RFP by co-treatment with HP4-PTD, and replaced medium 24 h post-infection. Transduced hAD-MSCs were treated with CDK4 inhibitor (CDK4i) for 12 h, and replaced with normal culture medium. This cycle was repeated twice. Time-lapse microscopy images showing the repeated morphologic changes of hAD-MSCs into neural transdifferentiation and dedifferentiation (A). mRNAs were isolated from hAD-MSCs, and neural sub-type markers transcripts were amplified by RT-PCR (B). Protein extracts were prepared from the above hAD-MSCs, and analyzed by immunoblotting using anti-Rb, anti-phospho-Rb S807/811 (phosphorylated-Rb at residues serine 807 and 811), anti-E2F1, anti-Nestin and anti-actin antibodies (C). (D) A schematic model of neural transdifferentiation of mesodermal hMSCs by CDK4 inhibition. The inhibition of CDK4 activity selectively led to transdifferentiation into neural cells, but attenuated the syngenic differentiation of hMSCs.

Interestingly, it is widely recognized that the terminally dysregulated CDK activity has been linked to the pathogenesis of neurodegenerative diseases such as Alzheimer disease (AD), amyotrophic lateral sclerosis (ALS) and stroke.²² Specifically, aberrant activation of CDKs is also associated with apoptosis and neuronal dysfunction in response to various neuronal stressors.^23,24 These findings strongly indicated that the regulation of CDK activity is a critical process for the neurogenesis of neural stem and precursor cells. However, the question remains regarding what would happen if the CDK activity or expression was lost or upregulated during neurogenesis. Importantly, recent evidence has revealed that hMSCs have the potential to trans-differentiate into neural cells, and that commitment to this lineage depends on the culture conditions.^1,2,4,5 However, there is no molecular evidence on how the neural transdifferentiation of hMSCs could be induced and regulated. This study has provided evidence that CDK4 suppression is required for the neural transdifferentiation of hMSCs. In addition, the transdifferentiation of hAD-MSCs by CDK4 inhibition induces functional neural cells in vivo (Fig. S3). Therefore, CDK signaling, in particular CDK4 signaling, will offer new insight into how human MSCs can be integrated into the neural cell lineage and may provide a new target for the therapeutic application of neurologic diseases.

Materials and Methods

Isolation of human adipose-derived mesenchymal stem/stromal cells (hAD-MSCs) and bone marrow-derived MSCs

This study was approved by the Institutional Review Board (IRB) at Kangbuk Samsung Hospital of the Sungkyunkwan University School of Medicine. For all liposuctioned tissues used herein, we obtained written informed consent from all donors prior to surgery. The following information from all donors represents gender, age and surgical site, respectively; donor #1 (F, 52, abdomen), donor #2 (F, 33, arm), donor #3 (F, 20, thigh/hip), donor #4 (F, 27, thigh), donor #5 (F, 37, flank) and donor #6 (F, 39, abdomen). hAD-MSCs were isolated from adipose tissue using methods previously described,²⁵ with minor modifications. Briefly, the adipose tissue was washed with phosphate-buffered saline (PBS) and centrifuged at 1,200 g for 5 min to remove the red blood cells. Washed aspirates were treated with 0.0075% collagenase (Type XI; Sigma-Aldrich) PBS for 1 h on a shaker at 37°C. The enzymatic digestion was stopped by the addition of an equal volume of growth medium (high glucose DMEM, 10% FBS and 1% penicillin/streptomycin) and centrifuged at 1,200 g for 10 min. The pellet was reconstituted with normal growth medium, plated in a T-75 tissue culture flask and incubated at 37°C in 5% CO₂.

To isolate human bone marrow mesenchymal stem cells, human bone marrow aspirates were obtained from healthy donors. Briefly, human bone marrow mononuclear cell (hBM-MNC) fraction was isolated by Ficoll-hypaque (density 1.077 g per litter, Sigma-Aldrich) density gradient centrifugation (400 g for 50 min). hBM-MNCs were washed with PBS and centrifuged at 100 g for 10 min. The pellet was reconstituted with growth medium (DMEM, 10% FBS and 1% penicillin/streptomycin) and plated in a T-75 tissue culture flask.

Differentiation assay

For adipogenesis, the cells were plated onto multi-well plates in a growth medium and maintained until they reached 100% confluence (2~3 wk). The cells were cultured in adipogenic medium (low glucose DMEM, 10% FBS and 1% penicillin/streptomycin supplemented with 1 μM dexamethasone, 1 μM indomethacin, 10 μg/ml insulin and 500 μM IBMX) for 3 d, then transferred to an adipocyte maintenance medium (low glucose DMEM, 10% FBS and 1% penicillin/streptomycin supplemented with 10 μg/ml insulin) for 1 d. For osteogenesis, the cells were plated onto multi-well plates for 24 h before induction of differentiation. The cells were cultured for 4 wk in an osteogenic medium (low glucose DMEM, 10% FBS and 1% penicillin/streptomycin supplemented with 0.1 μM dexamethasone, 50 μM L-ascorbate-2-phosphate and 10 mM β-glycerophsphate disodium). The osteogenenic medium was changed every 3 d.²⁶

Histologic assays

Adipogenic differentiation was determined by Oil Red O staining to detect lipid droplet formation. Cells were fixed in 10% formalin solution for 30 min, washed with deionized water to remove formalin, then were fixed with 60% isopropanol for another 5 min. Oil Red O working solution was added, and the cells were incubated for 30 min. After staining, the cells were washed with deionized water until the fluid ran clear. Osteogenic differentiation was determined by von Kossa staining. The cells were fixed in 10% formalin solution for 30 min, washed with deionized water, then were fixed with 5% silver nitrate solution for another 30–60 min with exposure to UV. Then, 5% sodium thiosulfate was added for 5 min.²⁷

Immunoblotting, immunofluorescence and FACS analyses

Cells for the immunoblots were lysed in NE buffer [20 mM HEPES (pH 7.6), 250 mM NaCl, 1.5 mM MgCl₂, 20% glycerol, 0.1% Triton X-100, 1 mM DTT, 1 mM PMSF and protease inhibitor cocktail], and the lysates were clarified by centrifugation at 13,200 rpm for 30 min. For immunofluorescence staining, the hMSCs were grown on coverslips (1 × 10⁴ cells/each) for 2 d. The cells were fixed with 5% formaldehyde and incubated with antibodies in 3% skim milk solution. DNA was stained with Hoechst dye. The cells were harvested and analyzed by flow cytometry as follows: FITC- or PE-conjugated antibodies against CD34, CD44, CD45, CD49d, CD73, CD90 and CD106; and isotype controls (BD Biosciences). For cell surface staining, the cells were incubated with antibodies in a FACS buffer (0.5% FBS and 0.09% sodium azide in PBS) for 30 min on ice. The cells were then washed three times and fixed with 1% formaldehyde. For intracellular staining, the cells were first fixed with a Cytofix/Cytoperm buffer (BD Biosciences), followed by washing with a Perm/Wash buffer (BD Biosciences) and incubated with antibodies in the Perm/Wash buffer for 30 min on ice. After washing with Perm/Wash, the cells were incubated with FITC- or PE-conjugated corresponding secondary antibodies for 30 min on ice and analyzed with a Becton Dickinson FACScan cytometer using the CellQuest (Clontech) and WinMDI 2.8 (www.facs.scripps.edu/software.html) software packages.

Quantitative real-time RT-PCR

Total RNA was extracted with the RNeasy Mini Kit (Qiagen) and quantified by UV spectroscopy. To prepare RNA for PCR analysis, 1.2 μg of RNA were converted to cDNA using the ImProm-II Reverse Transcription System (Promega). The qRT-PCR reactions were performed with the SYBR Green PCR Master Mix on a MicroAmp optical 96-well reaction plate (Applied Biosystems). The mRNA expression of candidate genes was determined by real-time quantitative PCR using the AMI PRISM 7000 SDS (v1.1) instrument, according to the manufacturer's instructions (Applied Biosystems). The following specific primers were used: GAPDH (5′-CAA GGT CAT CCA TGA CAA CTT TG-3′/5′-GTC CAC CAC CCT GTT GCT GTA G-3′); βIII Tub (5′-CTA CGA CAT CTG CTT CCG CA-3′/5′CTT GCG CCG GAA CAT GG-3′); MAP2 (5′-TCA GAG GCA ATG ACC TTA CC-3′/5′-GTG GTA GGC TCT TGG TCT TT-3′); S100 (5′- ATG TCT GAG CTG GAG AAG GC-3′/5′-TCA CTC ATG TTC AAA GAA CTC GTG-3′); NL3 (5′-AGA GGG ACT TTT CCA GAG AG-3′/5′- GAC GGA ATA GTC AAA GTC AG-3′); GFAP (5′-TCG CTG GAG GAG GAG ATC C-3′/5′-GCG ATC TCG ATG TCC AGG G-3′); Sox2 (5′-AAC ATG ATG GAG ACG GAG-3′/5′-TGC TGG GAC ATG TGA AGT-3′); Msi1 (5′-GTT TCG GCT TCG TCA CTT-3′/5′-AGT GGT ACC CAT TGG TGA A-3′); NCAM (5′-AGG ATC TCA TCT GGA CTT TG-3′/5′-CCA TCC AGA GTC TTT TCT TC-3′); and Nestin (5′-CGC TGG CGG GAG AAG C-3′/5′-GCC ACG CGC TCC TGG TA-3′).

Generation of recombinant adenoviruses

Oigonucleotides encoding shRNAs against CDK1, CDK2 and CDK4 were synthesized and inserted into the pSuper.puro vectors, which contain a H1 promoter and T5 terminal sequences (Oligoengine). The specific shRNA sequences used were as follows: CDK1 (CDK1 shRNA 5′- GAT CCC CGG GGA TTC AGA AAT TGA TCT TCA AGA GAG ATC AAT TTC TGA ATC CCC TTT TTA -3′/5′- AGC TTA AAA AGG GGA TTC AGA AAT TGA TCT CTC TTG AAG ATC AAT TTC TGA ATC CCC GGG-3′); CDK2 (CDK2 shRNA 5′-GAT CCC GCC TGG AGA TTC TGA GAT TGT TCA AGA GAC AAT CTC AGA ATC TCC AGG TTT TTA-3′/5′-AGC TTA AAA ACC TGG AGA TTC TGA GAT TGT CTC TTG AAC AAT CTC AGA ATC TCC AGG CGG-3′); and CDK4 (CDK4 shRNA 5′-GAT CCC CCA GTT CGT GAG GTG GCT TTA CTT CAA GAG AGT AAA GCC ACC TCA-3′/5′-AGC TTA AAA AAC AGT TCG TGA GGT GGC TTT ACT CTC TTG AAG TAA AGC CAC-3′).

For adenoviral vector preparation, DNA fragments containing the H1 promoter, target-specific shRNA and a T5 sequence were excised from the pSuper-based plasmids and inserted into pShuttle-GFP vectors (Quantum). Adenoviral vectors were subsequently generated by recombination of the pShuttle plasmids encoding shRNAs with pAdeasy 1 in Escherichia coli BJ5183. The resulting recombinant DNAs were extracted with the DNA-spin purification kit (Intron) and digested with PacI (NEB). Recombinant adenoviruses (rAd-GFP-shCDK1, -shCDK2 and -shCDK4) were recovered from 293A cells that were calcium phosphate precipitate-transfected with the linearized recombinant adenoviral DNA.⁶ HP4, a type of protein transduction domain (PTD), was synthesized by PEPTRON, Inc. A mixture of rAds at a multiplicity of infection (MOI) of 100 plaque-forming units (PFU) per cell, and HP4 (100 nM) was incubated in a serum-free medium for 30 min at room temperature. The cells were washed and incubated with the rAds and HP4 preparation. After 2 h, the cells were washed and incubated with a serum-containing medium.¹⁹

Time-lapse microscopic analysis

The cells were infected with an expression adenovirus encoding H2B-RFP and GFP-shCDK1/-shCDK2/-shCDK4. The infected cells, maintained in DMEM containing 10% FBS and 1% penicillin/streptomycin, were imaged in XT 0.15 mm dishes. Analyses were based on 50-ms exposures taken every 6 min using a LSM500 META confocal microscope (Carl Zeiss) with a Χ40 NA0.75 objective, or using an Axiovert 200M microscope containing a Zeiss AxioCam HRm (Carl Zeiss).

Human neuroglial progenitor/precursor cells

Human neuroglial progenitor/precursor cells (hNPCs) were obtained from fetus subcortical white matter, as described previously.²⁸ The cells were suspended in DMEM/F12 supplemented with 20% FBS or ITSFn medium (DMEM/F12, 5 μg/ml of insulin, 50 μg/ml of transferrin, 30 nM sodium selenium and 5 μg/ml of fibronectin) supplemented with 20 ng/ml of FGF2 (Sigma-Aldrich) and 20 ng/ml of EGF.

Glossary

Abbreviations:

hAD-MSCs

human adipose-derived mesenchymal stem/stromal cells

hBM-MSCs

human bone marrow-derived mesenchymal stem/stromal cells

CDK

cyclin-dependent kinase

βIII Tub

β-tubulin III

MAP2

microtubule-associated protein 2

GFAP

glial fibrillary acidic protein

rAd

recombinant adenovirus

NPCs

neuroglial progenitor/precursor cells

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.