Biomedical Microdevices

, Volume 13, Issue 3, pp 517–526

Evaluation of transdifferentiation from mesenchymal stem cells to neuron-like cells using microfluidic patterned co-culture system


  • De-Yao Wang
    • Institute of Biomedical Engineering, College of Engineering, College of MedicineNational Taiwan University
  • Shinn-Chih Wu
    • Department of Animal Science and Technology, College of Bio-Resources and AgricultureNational Taiwan University
  • Shau-Ping Lin
    • Institute of Biotechnology, College of Bio-Resources and AgricultureNational Taiwan University
  • Shih-Hsiang Hsiao
    • Department of Animal Science and Technology, College of Bio-Resources and AgricultureNational Taiwan University
  • Tze-Wen Chung
    • Department of Chemical and Material EngineeringNational Yunlin University of Science and Technology
    • Institute of Biomedical Engineering, College of Engineering, College of MedicineNational Taiwan University

DOI: 10.1007/s10544-011-9520-z

Cite this article as:
Wang, D., Wu, S., Lin, S. et al. Biomed Microdevices (2011) 13: 517. doi:10.1007/s10544-011-9520-z


We design a microfluidic patterned co-culture system for mouse mesenchymal stem cells (mMSCs) and neural cells to demonstrate the paracrine effects produced by the neural cells in facilitating the transdifferentiation from mMSCs to neuron-like cells. Neural cells and mMSC are orderly patterned in the microfluidic co-culturing system without direct cell contact. This configuration provides us to calculate the percentage of neural marker transdifferentiated by mMSCs easily. We obtain higher transdifferentiated ratio of mMSC in the microfluidic co-culturing system (beta III tubulin: 67%; glial fibrillary acidic protein (GFAP): 86.2%) as compared with the traditional transwell co-culturing system (beta III tubulin: 59.8%; GFAP: 52.0%), which is similar to the spontaneous neural marker expression in the undifferentiated MSCs (beta III tubulin: 47.5%; GFAP: 60.1%). Furthermore, mMSCs expressing green fluorescent protein and neural cells expressing red fluorescent protein were also used in our co-culture system to demonstrate the rarely occurring or observed cell fusion phenomenon. The results show that the co-cultured neural cells increased the transdifferentiation efficiency of mMSCs from soluble factors secreted by neural cells.


Mesenchymal stem cellNeural cellCo-cultureMicropatterningLaser manufacturingPolydimethylsiloxane

1 Introduction

The microenvironment, the niche of cell, plays an important role in cell behavior and function (Fuchs et al. 2004; Arai et al. 2005). Cells dynamically respond to biochemical, mechanical, and topological cues derived from their microenvironment including gradient of soluble cytokines (Yu et al. 2005), interaction with the extracellular matrix (Aigner and Stove 2003; Sands and Mooney 2007), and communication between cells (Bhatia et al. 1999). In addition to its programmed differentiation capacity, one cell lineage is able to adopt other fates via extrinsic induced pathways such as juxtacrine: membrane proteins of one cell surface interact with receptor proteins of an adjacent cell, and paracrine: secreted proteins from one cell diffuse to induce changes of neighboring cells (Alexanian 2005).

Mesenchymal stem cells (MSCs) are programmed for mesodermal lineages differentiation, including osteogenesis, chondrogenesis and adipogenesis in vivo. However, when detached from the physiological niche, the in vitro cultured MSCs also have the plasticity to express the ectoderm neural markers and can adopt morphological characteristics of neurons and astrocytes (Kopen et al. 1999). Secreted soluble factors from neural cells and physical contact with neural cells may cause/facilitate the co-cultured MSCs to cross germinal boundaries, a process called transdifferentiation. However, from a regular co-culture system with randomly mixed two cell types, it is difficult to differentiate whether the neural protein expression of MSCs is induced from the pathway of paracrine, juxtacrine, or both (Alexanian 2005). Although, a semipermeable membrane of transwell co-culture system was employed to separately culture two kinds of cells without contacting each other in the same medium (Wislet-Gendebien et al. 2005), cells are separated far enough and/or the membrane could be an additional diffusion barrier for secreted soluble factors. Furthermore, it is difficult to find out what actually causes MSCs to transdifferentiate into neuron–like cells based on the fragmented and inconsistent results. Therefore, we designed the microfluidic patterned co-culture system, a straightforward and effective method to immobilize cells (Chiu et al. 2000; Park and Shuler 2003; Tan and Desai 2003) and organize them in a controllable and reproducible way, to co-culture green fluorescent protein expressing (GFP+) mouse MSCs (mMSCs) and red fluorescent protein expressing (RFP+) neural cells for characterizing the facilitating mechanism of transdifferentiation and determining the possibility of cell fusion. We used the direct-write CO2 laser microchanneling (Fogarty et al. 2005) to manufacture a microfluidic patterning device and evaluated the cell plasticity with the use of the expression of neural markers, beta III tubulin and glial fibrillary acidic protein (GFAP), in MSCs. The effect of neural cells direct contact on the transdifferentiation of mMSC or neural and mesenchymal stem cells without direct contact but being separated in a micrometer scale distance, orderly patterned were also discussed.

2 Materials and methods

2.1 Microfluidic patterning device fabrication

A cured Sylgard 184 PDMS (polydimethylsiloxane) (Dow Corning, Midland, MI, USA) about 5 mm thick was prepared and engraved to become the microfluidic chip (Fig. 1(b)) with the computer-aided design (CAD) pattern (Fig. 1(i)) and the PDMS spacer (Fig. 1(d)) using the direct-write CO2 laser micromachining of Venus II System (LaserPro, Taipei, Taiwan). The CAD pattern was output precisely and rapidly onto the surface of PDMS workpieces using laser micromachining. By using plasma cleaner (Harrick, NY, USA, 18W for 10 mins under 1 torr of air), we cleaned up the residual debris generated during engraving and activated the surface of the microfluidc chip to increase its hydrophilicity. We sterilized all the components of microfluidic patterning device including a pre-clean cover glass, the microfluidic chip, the PDMS spacer, two polymethylmethacrylate (PMMA) plates (Hishiron Industries, Tainan, Taiwan) and binder clips. A piece of cover glass was placed in between the plasma activated microfluidic chip and the PDMS spacer and then all of them were pressed to form a temporary seal microfluidic patterning device using two PMMA plates and binder clips (Fig. 1(f)). The morphologies of the microfluidic chip were verified with the scanning electron microscopy (SEM; JEOL JSM-5600, 15 KV, Tokyo, Japan).
Fig. 1

Microfluidic patterning. There are two PMMA plates (a, e), the microfluidic chip (b), a 22 ×22 mm cover glass(c) and the PDMS spacer(d) in the microfluidic patterning device (f). GFP+ MSCs and RFP+ neural cells were injected into the device (g) and co-cultured on the cover glass (h). The CAD pattern of microchannels (i) and SEM results of the microfluidic chip (jq)

2.2 Primary culture of cell

The ICR mice were purchased from the Laboratory Animal Center of National Taiwan University, College of Medicine (Taipei, Taiwan) and kept under standard conditions. All experimental procedures were approved by the Institutional Animal Care and Use Committee. Neurospheres of brain tissue from RFP+ mouse fetus were generated and then differentiated into neural cells; mMSCs were isolated from the femurs and tibias of GFP+ mice MSCs (GFP+ mMSCs). The generation of RFP+ neurosphere was modified from previous studies (Reynolds and Weiss 1992; Johe et al. 1996; Tsai and McKay 2000; Wachs et al. 2003; Ray and Gage 2006). Briefly, brain tissues were dissected from RFP+ ICR mice at embryonic day 14.5 and then transferred into ice-cold PBS. Tissue pieces were physically dissociated using gentle pipette triturating until cell suspension become homogeneous. Cells were filtrated through a 70 micron mesh and centrifuged at 75 g for 10 min to remove debris. We discarded the supernatant and transferred the resuspended cells into a 25 cm2 plastic flask at a cell density of 5 × 104 cells/cm2 using the growth medium including modified Dulbecco’s modified Eagle’s medium/Nutrient Mixture F-12 Ham (DMEM/F12; InVitrogen, Gibco-BRL) without fetal bovine serum (FBS), 100 U/ml of penicillin and 100 μg/ml of streptomycin (InVitrogen, Gibco-BRL), 1 X B27 supplement minus vitamin A (InVitrogen, Gibco-BRL) and basic fibroblast growth factor (bFGF; InVitrogen, Gibco-BRL) at at 20 ng/mL. We replaced 50% of the total volume with fresh growth medium every 2.5 days. To examine their self-renewal abilities, primary neurospheres though first expansion for 5 days were dissociated into single cell using gentle pipetting and then centrifuged at 100 g for 10 min to collect them. We subcultured and resuspend the cells in growth medium to obtain secondary neurospheres through second expansion for further 5 day. To induce the differentiation of neurospheres into RFP+ neural cells, we removed bFGF with the use of differentiation medium including DMEM/F12 with 20% FBS (InVitrogen, Gibco-BRL), 100 U/ml of penicillin and 100 μg/ml of streptomycin, 1 X B27 supplement minus vitamin A for further 4–7 day culture.

GFP+ mMSCs were isolated from bone marrow of the femurs and tibias of GFP+ mice at 6 to 8-weeks-old and cultured at a cell density of 2 × 105 cells/cm2 with complete medium consisting of MEM alpha supplemented with 20% FBS with 2 mM L-glutamine (InVitrogen, Gibco-BRL), 100 U/ml penicillin, and 100 μg/ml streptomycin (Hsiao et al. 2010). After incubation for 72 h, the non-adherent cells were removed by refreshing the medium. Plastic-adherent GFP+ mMSCs from mouse marrow were purified by immunodepletion with magnetic microbeads conjugated to either anti-CD11b or anti-CD45 antibody (Miltenyi Biotec Paris, France).

2.3 Microfluidic patterning

The cell suspension, GFP+ mMSCs in MEM alpha and RFP+ neural cells in DMEM/F12, was delivered individually into the microchannels of the microfluidic patterning device by simple injection (Fig. 1(g)). Cells attached to the substrate exposed to the microflow of cell suspension and were incubated at seeded densities of 3 × 106 cells/cm2 for 24 h. After that, GFP+ mMSCs and RFP+ neural cells were patterned separately on the cover glass and followed by culturing at 37°C in differentiation medium (Fig. 1(h)). Laser scanning microscope (LSM; Leica, Bensheim, Germany) and Axiovert 100TV fluorescence microscope (Zeiss, Germany) were used to observe the cells’ behavior. Four different culturing conditions were used to investigate the transdifferentiation mechanism: undifferentiated MSCs, microfluidic patterned co-cultured MSCs and neural cells, randomly mixed co-cultured MSCs and neural cells, and transwell co-cultured MSCs and neural cells. In the randomly mixed co-culture, GFP+ mMSCs/neural cells (1:1) were mixed and cultured at seeded densities of 1 × 105 cells/cm2 on a cover glass. In the transwell co-culture system, GFP+ mMSCs were seeded at densities of 2 × 105 cells/cm2 and separately co-culture with neuron seeded at densities of 4 × 105 cells/cm2 in a 0.4 μm pore size transwell device (BD, CA, USA) in the same medium.

2.4 Immuno-histochemistry analysis

Before fixation, cells were cultured with 2.5 μg/ml Hoechst 33342 (InVitrogen, Gibco-BRL) for 30 min to stain nuclei. According to a standard immunohistochemistry analysis procedure, we used 3% paraformaldehyde (Wako) in phosphate buffer solution (PBS) to fix samples at 37°C for 30 min. After that, we used 5% FBS in PBS for 30 min at room temperature to block samples. The primary antibodies we used were rabbit anti-mouse beta tubulin IgG (Epitomics, Burlingame, CA), rabbit anti-mouse GFAP IgG (Santa Cruz, CA, USA) (Anjos-Afonso et al. 2004; Isakova et al. 2006) and mouse anti-mouse O4 (R&D Systems, USA) diluted 1:100 in 5% FBS. All primary antibodies were applied to stain samples overnight at 4°C and then rinsed five times each with PBS. The secondary antibodies, Alexa-Fluor647 conjugated donkey anti-rabbit IgG (InVitrogen, Gibco-BRL) and Alexa-Fluor647 conjugated goat anti-mouse IgM (InVitrogen, Gibco-BRL), were applied at 1:200 diluted in 5% FBS for 3 h. After five more rinses in PBS and mounting with glycerol/water (1:1), stained samples were inspected by LSM.

2.5 Stereological cell count

We used an unbiased stereological estimation of the total number of MSCs expressing neural markers(Gundersen and Jensen 1987; West et al. 1991) MSCs expressing neural biomarkers were identified by staining for both beta III tubulin and GFAP. Neural expression of MSCs were counted in randomly selected areas (375 × 375 μm) using a 40× objective. At least thirty separate regions of each sample were photographed and 100 cells were scored. The neural expression of MSCs was determined based on the percentage of beta III tubulin and GFAP producing GFP+ mMSCs of the total number of MSCs calculated by stereology. The results were compared using the Student’s t-test for pairs and unpairs analysis. Statistical significance was deemed at P < 0.05(*), P < 0.01(**), and P < 0.001(****). Statistical analyses were performed using SigmaPlot (Jandel, Corte Madera,USA)

3 Results

3.1 Microfluidic device

The microfluidic patterning device was composed of the microfluidic chip (Fig. 1(b)), the PDMS spacer (Fig. 1(d)), a cover glass (Fig. 1(c)), and two PMMA plates (Fig. 1(a) and (e)). Microfluidic channels were formed via that the microfluidic chip sealed with the cover glass using binder clips (Fig. 1(f)). The CAD pattern (Fig. 1(i)) consisted of 11 500 μm width microchannels separated by 600 μm and was etched into the surface of PDMS workpieces using the CO2 laser micromachining (Fig. 1(j–q)). Due to the location of focused laser spot and the output of laser power, the channel depth was about 1 mm and the channel cross section (Fig. 1(q)) exhibited a Gaussian-like profile, which can be attributed to the decay of light transmission and the distribution of the laser energy (Bowden et al. 2003; Cheng et al. 2005; Yen et al. 2006; Yuan and Das 2007).

3.2 Differentiated neurons from neurosphere

Primary RFP+ neurospheres were formed during first expansion for 4 days in growth medium. The second expansion is used for experiments that stem cells in neurospheres are able to present their self-renewal of sphere-forming. After dissociating into a single cell using gentle pipetting, neural stem cells were subcultured for further 4 days in growth medium to generate secondary neurospheres (Fig. 2). Neurospheres were induced to differentiate with the addition of 20% FBS and the absence of bFGF at day 11. Neurospheres have abilities to not only subclone into new formed spheres but also undergo neural differentiation. Differentiated cells migrated from neurosphere and expressed neural surface markers, beta III tubulin for neuron (Fig. 3(a–c)), GFAP for glia cell (Fig. 3(d–f)), and O4 for oligodendrocyte (Fig. 3(g–i)), after 3–4 day serum induction.
Fig. 2

Expansion and differentiation of RFP+ neurospheres. Morphologies of RFP+ cells from neurosphere to differentiated neural cells in bright field and in red fluorescence were observed during expansion and differentiation. All scale bars are 100 μm
Fig. 3

Serum induced differentiation of neurospheres. Examination of RFP+ neurospheres for morphologies under DIC (a, d, g), for RFP expression as the red signals in (b, e, h), and for the expression of neural surface markers as the purple signals in beta III tubulin (c), GFAP (f), and O4 (i). All scale bars are 50 μm

3.3 Co-cultured MSCs and neural cells

We used GFP+ mMSCs and RFP+ neural cells for co-culturing experiments. Cells were patterned separately and co-cultured without cell-cell contact in the same medium. Cells were delivered individually into the microfluidic patterning device by sample injection and cultured within the device for 24 h. These two cells were seeded in a pattern format (Fig. 1(h)) on the same cover glass with organized patterns (Fig. 4(a–c)). After cells adhered to the culture plate, we uncovered the device and cultured these cells (neural cells and mMSCs) for 1–3 days in differentiation medium. During co-culture, patterned cells proliferated, extended and migrated crossing over 600 μ width gap space between GFP+ mMSCs and RFP+ neurons to contact each other (Fig. 4(d–f)). Green and red fluorescent signals indicated the location of GFP+ mMSCs and RFP+ neural cells and revealed organized border between them. The well organized boundary between GFP + mMSCs and RFP + neural cells give us a very easy way to evaluate the behavior of these cells. The microfluidic patterned co-culturing system could also provide cells with more orderly border than random mixed co-culture did (Fig. 5).
Fig. 4

Co-culture for 1 and 3 day. Microfluidic patterned GFP+ mMSCs and RFP+ neurons co-cultured for 1 day before contact (ac). After 3 day, two cell types contacted with each other (df) on the same cover glass. Green signals represent the location of GFP+ MSCs in (b, e) and red sinals represent the location of RFP + neural cells in (c, f). Scale bars are all 500 μm
Fig. 5

Random mixed co-cultured GFP+ mMSCs and RFP+ neurons. Signals of GFP+ mMSCs (green), RFP+ neurons (red), Hoechst stained nucleus (blue), and beta III tubulin expression (purple) are merged. The scale bar is 100 μm

3.4 Neural markers expressed in mMSCs with and without cell-cell contact induction

Results of immunocytochemistry analysis were shown with the stained nucleus using Hoechst 33342 in blue signals, GFP+ mMSCs in green signals, RFP+ neurons in red signals, and Alexa-Fluor647, a dye spans near-IR spectra, conjugated antibody for beta III tubulin and GFAP in purple signals (Figs. 6, 7). GFP+ mMSCs expressed higher level of beta III tubulin (67.1 ± 14.5%, t-test, p < 0.01) and GFAP (86.2% ± 5.9%, t-test, p < 0.001) without contacting with any neural cells in the first day of microfluidic co-culture (Fig. 8). However, when used a traditional transwell co-culturing system to separately GFP+ mMSCs with RFP+ neural cells, we obtained lower transdifferential ratio. mMSCs expressed beta III tubulin (59.8% ± 14.2%) at almost the same level as their naturally expression of beta III tubulin (47.5% ± 9.5%) and at lower expression level of GFAP (52% ± 5.4%; t-test, p < 0.05).
Fig. 6

GFP+ mMSCs expressed beta III tubulin (*)without neuron contact. Green, red, blue, purple signals represent GFP+ mMSCs, RFP+ neurons, Hoechst stained nucleus, and beta III tubulin expression individually. The sacle bar is 100 μm
Fig. 7

GFP+ mMSCs expressed GFAP (*) without neuron contact. Green, red, blue, purple signals represent GFP+ mMSCs, RFP+ neurons, Hoechst stained nucleus, and GFAP expression individually. The sacle bar is 100 μm
Fig. 8

Beta III tubulin and GFAP expression in different culture conditions. The comparison of beta III tubulin and GFAP expression between MSCs cultured on a cover glass, co-cultured with and without neural cell contact, and co-cultured with neural cells using a transwell device. (*: p < 0.05, **:p < 0.01,***:P < 0.001, t-test.)

During microfluidic co-culture, although dynamic cell migration might mingle the border between two cell types, most GFP+ mMSCs and RFP+ neural cells still can be recognized into two groups creating co-culture condition consisted of physical contact and soluble factors induction (Fig. 9). LSM was used to confirm if the neural biomarkers in purple signals were expressed within the GFP+ mMSCs (data not shown).
Fig. 9

Boundary formation between GFP+ mMSCs and GFP+ neurons. Expression of neural markers (purple) on GFP+ MSCs (green) induced by cell contact and secreted factors from neural cells (red) during co-culturing from day 1 (ad),2 (eh) to 3 (il). All scale bars are 100 μm

3.5 Cell fusion

When GFP+ and RFP+ signal colocolizating inside the same cell, it indicates the happening of cell fusion during co-culture. However, only a very few case of fused cell was observed (Fig. 10). Although neural biomarkers expressing MSCs could be the result of a cell fusion process between MSC and neural cells, this phenomenon can also coexist during co-culture, they might not be the major cause of GFP+ mMSCs differentiating towards neural markers expressing neural-like cells after co-culture.
Fig. 10

Cell fusion happened during co-culture (white arrow). Green, red, blue,purple signals represent GFP+ mMSCs, RFP+ neurons, Hoechst stained nucleus, and GFAP expression individually. All scale bars are 100 μm

4 Discussion

Cell-cell interaction is a complex process. Secreted factors from adjacent neural cells do affect MSC behavior and function. In 2000, Sanchez-Ramos assessed the influence of cell-cell interactions and the factors released from developing neural tissue. They found that the interactions played an important role in neural differentiation of MSCs (Sanchez-Ramos et al. 2007). Though several studies have reported that MSCs transdifferentiated into neuronal-like cells when co-culturing with neural cells, the findings have been challenged for the lack of neural specific markers, the heterogeneous nature of the MSCs, and the random distribution of the co-cultured cell population (Song and Sanzhez-Ramos 2003; Krabbe et al. 2005; Phinney and Prockop 2007). Our study demonstrated a novel design of microfluidic pattern GFP+ mMSCs and RFP+ neural cells co-culturing system in order to evaluate that the neural expression of MSCs is facilitated by not only the juxtacrine via cell-cell contact but also the soluble factors mediated paracrine signaling. In the microfluidic patterned cell culture device, patterning cell is easy and straightforward without the need of complicated fabricating process or complicated pre-modified procedures for the culture plates or substrates as shown in literature (Bhatia et al. 1999). Cells in our system were patterned with well-defined boundaries; besides, highly organized cell pattern provides us a better way to evaluate the cell migration, immune-cytochemistrical characters and transdifferentiation than randomly displayed cell pattern. Targeted cells could also be harvested from specific locations on the surface using laser microdissection and used for molecular level analysis (Lee et al. 2006). Moreover, with a defined cell border created by the microfluidic patterning, we could further identify the evident of cell fusion during co-culture. The fused cell reported both GFP+ and RFP+ signals indicated the possibility of cell fusion.

What is more, we emphasized the influence of soluble factors derived from neural cells also play an important role to cause paracrine induction of neural expression of MSCs in our system. Though the transwell co-culture system could provide to separate two cell type culturing simultaneously, the permeable cell insert membrane is a significant diffusion barrier followed by the establishment of concentration gradient. That might affect the soluble factors concentration and paracrine effect from neural cells. Therefore, except the effect of juxtacrine from cell-cell interaction, we used microfluidic patterned cells without contact each other and co-cultured them under the same medium conditions to demonstrate a free diffusion environment for secreted factors from paracrine induction. During microfluidic patterned co-culture, mMSCs express higher neural markers in contrast to the cells in the transwell co-culture system. This data suggests that, barrier-free environment created in microfluidic patterned co-culture system emphasizes the importance of soluble factors in transdifferentiated induction and helps us to clarify the influence of concentration gradient of soluble factors in co-culture condition.

We observed that mMSCs have abilities to express neural markers, beta III tubulin and GFAP. The expression of neural biomarkers in MSCs might be caused by not only transdifferentiation of the neuron marker expression but also the clonal selection from the original MSCs populations. We felt the necessity to note that there is the clonal amplification theory as well as transdifferentiation possibility. Besides, the way we used to isolate and purify mecenchymal stem cells from mice may affect the expression level of neural markers. Varied approaches used to culture-expand and select for MSCs make it difficult to directly compare experimental results. Moreover, spontaneously expressed neural markers, beta III tubulin and GFAP, from MSCs are not efficient to represent the fates of neural cells specifically (Egerbacher et al. 1995; Hainfellner et al. 2001). This reflects the fact that intermediate filaments form dynamic networks whose composition changes in a stage-specific manner during cellular differentiation. For further investigation, we will use more specific markers or different levels of expression/epigenetic analysis in the future studies.

5 Conclusions

We designed a microfluidic patterned co-culturing system by using PDMS spacer engraved by the direct-write CO2 laser micromachining to separate and pattern GFP+ mMSCs and RFP+ neural cells. The system demonstrated that co-cultured mMSCs and neural cells, even without neural cell contact, can significantly increase neural marker, beta III tubulin and GFAP, producing mMSCs. Paracrine effects from neural cells are required for an efficient transdifferentiation from mMSCs to neuron-like cells. Furthermore, the two reporters, GFP and RFP, provide us a tool to assay the few cases of fused cell happened. Microfluidic patterned co-culture could facilitate to evaluate the plasticity and behavior of cells and dynamic cross-talks between cells on micrometer scale.


We acknowledged the financial support by National Science Council of Taiwan for this research (NSC98-2221-E-002-045-MY3). This research was also partially financial supported by NTUH Grants VN97-100, National Taiwan University Hospital, Taipei, Taiwan.

Copyright information

© Springer Science+Business Media, LLC 2011