Introduction
Differentiation and growth of neurons involve many aspects
including 1) sprouting, extension, and branching of dendrites and axons during development of the nervous system, 2) re-wiring of neurons in hippocampal tissues during learning and memory processes, and 3) regeneration of severed axons. Because of the importance of the
biological aspects of the differentiation and growth of neurons/neurites, a protocol to achieve an appropriate homogenous population of neurons has emerged as key to pursuing many crucial questions.
While many neural studies at the cellular level utilized primary neuron culture or neuronal cell lines (e.g., P19 cells), neurons
differentiated from ESCs have emerged as another meaningful cellular model to investigate neuronal differentiation and growth (Evans and Kaufman, 1981). The main advantage is that neurons derived from ectoderm ES cell development are developmentally and physiologically significant. To date, various protocols for neuronal differentiation of ESCs involve the formation of embryoid bodies (EBs) with or without retinoic acid treatment (Bain et al., 1995; Fraichard et al., 1995; Lee et al., 2000; Wichterle et al., 2002). The hurdle encountered by many studies using ESCs as a model system is that the eventual cellular population differentiated from ESCs is not uniquely neurons (Stavridis and Smith, 2003).
A protocol described by Bardie's group clearly demonstrates that feeder-dependent mESCs can be differentiated into uniform neurons
(Bibel et al., 2004; Bibel et al., 2007). They characterized many aspects of these neurons, such as neuronal markers and electrophysiological properties. They pointed out that many critical steps must be strictly followed in order to acquire optimal results. After all of the crucial factors were deliberately repeated including a step in which cell aggregations of EBs were completed trypsinized, we found that a homogeneous population was still not always obtained. This indicated that some crucial factors and/or details needed to be explored.
Therefore, we set out to systemically investigate many possibilities which might have caused such a discrepancy, including the quality of various reagents and precision of parameters we used. We gradually pinpointed this issue and developed a protocol to consistently
successfully differentiate neurons from mESCs. The approach presented herein has the potential to establish a cellular model for neuronal
research.
Material and methods
Refer to “General Materials and Methods” in chapter 2.
Results
Feeder-independent mESCs were routinely cultivated in ES medium supplemented with LIF. As depicted in Fig. 1, the protocol was
schematically drawn as a time line to indicate the key processes of differentiation. When mESCs were initially cultivated in ES medium without LIF on the bacterial dishes, these cells began to form EBs, and this time was designated day 0. From day 4, these EBs were cultivated in the same medium but supplied with a final concentration of 5 μM
all-trans retinoic acid for 4 more days. On day 8 when the differentiated progenitor cells within EBs were about to be separated and placed on poly-DL-ornithine/laminin-coated plates, these EBs were subjected to trypsinization so as to dissociate the aggregations. The timing was
hereafter designated time after plating due to the many manipulation steps in the following procedure. The culture medium was eventually switched to neurobasal medium, and neurons were maintained for at least 21 days.
In an effort to establish a method to make ESCs uniformly differentiate into neurons, many crucial steps were strictly followed according to Bibel et al. (Bibel et al., 2007). These steps and/or factors included the quality of materials, the number of cells required for seeding,
the complete dissociation of EBs, etc. However, we ultimately pinpointed that the critical step was not to completely dissociate EBs, but was the optimal timing of EBs treated with the freshly made trypsinization solution. As shown in Fig. 2A, neurites were originally grown out in all groups in the initial 24 h, as the initial seeding numbers were all the same.
When time progressed to 48 or 72 h, only EBs treated with an appropriate trypsinization period (i.e., 8 min in this case) continued to grow, maintain neurites, and eventually create neurite networks. The effect of
inappropriate timing on EB trypsinization resulted in cell loss as seen in Fig. 2B.
Meanwhile, over-trypsinization of EBs usually caused the cell
population to consist of different cell types. A heterozygous population of neurons was observed at 24~48 h after plating and become more
significant at 72 h. As seen in Fig. 3 which was taken at 48~72 h, it obviously contained neurons, neurites, and more than 50% of
non-neuronal cell types. One of the cell types was neuronal-like with complicated neurites extending from it which were prone to form aggregations. Another one was non-neuronal demonstrating a flatter morphology, and it was distributed all over the field. Determining exactly what these cells are requires further investigation.
These mESCs were well maintained and the separately formed EBs are shown in Fig. 4A and 4B. Having adopted the above-described modifications, we were consistently able to differentiate homogenous
neurons derived from mESCs. Judging from the appearance of the uniform neuronal morphology in Fig. 4C, this could be distinguished as successful differentiation. We further probed these cells with
neuron-specific class III β-tubulin using immunocytochemistry. In Fig.
5A, the majority of cells showed nuclear staining (blue) and was stained with the neuronal marker (green). The double-staining rate was calculated to be > 93% in any given field. In order to clearly demonstrate the
homogeneity and identify what kinds of cell types these few cells were, differentiated cells from the same experiment were immunostained with both GFAP and DAPI. The results showed that they were non-glial differentiated cells (Fig. 5B). The same GFAP antibody was capable of staining CTX TNA2 cells, an astrocyte cell line (Fig. 5D). Similarly, this glial cell line was barely stained with neuron-specific class III β-tubulin (Fig. 5C). Furthermore, different cell types were blotted with Oct-4 and type III β-tubulin which are respective markers for pluripotent ES cells and differentiated neurons (Fig. 5E). The results demonstrated that only mESCs contained the Oct4 marker, while differentiated neurons did not.
This indicates that mESCs were properly maintained in the culture system.
On the other hand, differentiated mESCs using the current protocol developed herein contained substantial type III β-tubulin. In the same experiment, HEK 293T cells were used as a negative control. In addition, NeuN and MAP2, which are other neuronal markers, were also used on the same blots. The results showed the same pattern as that seen for type III β-tubulin (data not shown). These neurons and neurite networks could be maintained for at least 3 weeks without morphological changes.
In an effort to further characterize these neurons, these differentiated neurons were immunostained with a vesicular glutamate transporter 1/2 (vGLUT1/2) antibody, a marker for glutamatergic neurons. As shown in Fig. 6A of staining with vGLUT1/2 and Fig. 6B of staining with DAPI, >
87% of cells were stained with both as shown in Fig. 6C. Although few cells were stained with DAPI but not vGLUT1/2, they demonstrated the specificity of this vGLUT1/2 antibody. This phenomenon was seen as early as 3 days after cell plating. We further examined synapse formation with this differentiation method. As seen in Fig. 6D, synaptotagmin, a marker of synaptic vesicles, was uniquely expressed in neurons
differentiated from mESCs, but not other cell lines such as mESCs or HEK293T cells. These results further demonstrated that mESCs differentiated using the defined protocol we established uniformly developed into glutamatergic neuron.
We further characterized the neurons derived from mESCs.
Neurons/neurites were confirmed by staining with class III β-tubulin (red), a neuronal marker, as shown in Fig.7. Within the same field,
synaptotagmin (green), which functions as a calcium sensor in the
regulation of neurotransmitter release, was also stained on these neurons three days and nine days after differentiation. We further verified these neurons by western blot analysis. As shown in Fig. 7, AMPA receptor 1 (AMPAR 1), an ionotropic transmembrane receptor for glutamate in the central nervous system, began to express on Day 3 after differentiation, and more on Day 9. PSD-95, a protein that is almost exclusively located
in the post synaptic density of neurons and is involved in anchoring synaptic proteins, also showed a similar pattern to AMPAR 1 or
synaptotagmin. The lysates from HEK 293T cells and primary neuronal cultures of mice cerebral cortex functioned as negative and positive
controls, respectively, for all antibodies used. Within the same blot, Oct-4 and type III β-tubulin, which are respective markers of mESCs and
differentiated neurons, demonstrated that mESCs were properly maintained and neurons were well developed. This also indicated that synapses could potentially be constructed within these neurons.
Discussion
We adopted an approach which contains many crucial steps to differentiate mESCs into neurons (Bibel et al., 2004; Bibel et al., 2007).
Many factors were indeed very crucial in our test as well. However, in almost all the early efforts when we followed the protocol, we obtained neurons and other cellular types. The morphology of which was larger than neurons with no neuronal characteristics such as neurites as seen in Fig. 3. Therefore, many variables were systemically compared in parallel, and trial-and-error experiments were performed. We pinpointed that the critical factor which determined the success of homogenous neuronal differentiation was the dissociation timing of EBs. It was reported that the processes and timing of germ-layer formation derived from EBs in vitro greatly repeated stages of development during the embryonic phase of life in vivo (Leahy et al., 1999; Rohwedel et al., 1999). As to the order and spatial relation existing in these EBs, they developed as endoderm-like cells in the core first, then as ectoderm cells in the rim, followed by mesoderm cells within the entire EB. As we trypsinized the EBs at
different times, the right layers of progenitor cells could be isolated at the appropriate time (as seen in Fig. 2). Conversely, over-trypsinization of the EBs caused a mixed population which resulted from more than one germ-layer of progenitor cells.
Based on the formation and development of the embryonic germ layers, they were theoretically more uniform when only the outer layers
of the EBs were trypsinized then seeded onto plates. Nevertheless, we found that successful differentiation greatly depended on an optimal duration of trypsinization, not just on the complete or less-than-complete dissociation of EBs. As long as the trypsinization timing was optimal as seen in those treated for 8 min in Fig. 2, the neurons survived and grew normally. As shown in Fig. 2, neurons/neurites only survived for 24 h in groups of EBs trypsinized for 2, 4, or 6 min. This phenomenon occurred within groups that were inappropriately trypsinized, which could have been due to either apoptosis after EB dissociation (Koyanagi et al., 2008) or cell death because of a lack of sufficient growth cytokines (Cook et al., 2010). Since the initial cell numbers which were dissociated from EBs for plating were the same, the latter factor seems less likely in this case.
Recently, it was shown that control over molecular transport of EBs plays a strategic role in stem cell differentiation (Sachlos and Auguste, 2008).
Whether this is relevant to the phenomenon seen here and how the
appropriate timing of trypsinization creates such a difference remain to be further investigated.
Neurons differentiated from mESCs using this protocol survived for more than 3 weeks with only slight changes in their morphology. Also, the network formed by the neurites gradually grew more complicated and was maintained for the same time. In agreement with Bardie et al.’s results (Bibel et al., 2004; Bibel et al., 2007), the majority of neurons differentiated by the current method were glutamatergic neurons based on the positive staining of vGLUT1/2. Since vGLUT1/2 is a membrane
protein expressed by most pyramidal neurons in the cerebral cortex and hippocampus (Fremeau et al., 2004), the establishment of this method further confirms that this is an excellent cellular model for investigating properties related to the central nervous system (CNS).
Alternatively, there are some other approaches such as gene
manipulation to obtain uniform neuron populations. Utilizing insertion of the sox promoter followed by a green fluorescent protein (GFP) and/or other reporter genes, GFP-positive cells can be sorted into homogenous neuronal populations (Li et al., 1998; Ellis et al., 2004; Plachta et al., 2007). Although such an approach will definitely result in uniform neuronal cells, it could interfere with potential applications in medicine because of the consideration of genomic interference and instability. It might be more ideal and physiologically relevant in terms of obtaining uniform neurons by using the current approach developed in this context.
Glutamatergic, multipolar neurons, and complicated neurite networks began forming at three to nine days post culture of neurons differentiated from mESCs. On the basis of vGLUT1/2, AMPAR, and PSD-95 blots, they indicate synapses could also be constructed within these neurons. Initially, they were immature, with more bipolar neurons at three days after differentiation, as shown in Fig. 7, although already around 5% had begun to show multipolar neurons. As time passed, by the seventh day, the majority showed mature characteristics, such as
multipolar neurons and complicated neuritis. Several papers employing a similar protocol produced glutamatergic neurons like ours. Bibel et al.
developed a method and reported that essentially all neurons were
positive for glutamate vesicular transporter 1/2. By contrast, less than 1%
of the cells were positive when stained with antibodies to other
neurotransmitter systems after one week of culture. Most neurons that they investigated received both majority glutamate and minor GABA (5%) input after three weeks (Bibel et al., 2004). In addition, using a similar method with the modification of maintaining differentiating aggregates on petri dishes with rotation, Hubbard et al. produced highly enriched glutamatergic neurons from suspension-cultured mESCs (Hubbard et al., 2012).
In conclusion, we developed an approach to obtain a homogeneous population of glutamatergic neurons differentiated from mESCs. By trypsinizing EBs in freshly produced trypsin buffer for an appropriate time, uniform neurons could be reproducibly and consistently generated.
Such a method can contribute to building a physiological cellular model for investigating neuronal development/growth, and can also shed light on possible therapeutic applications in regenerative medicine.
Days: 0 4 8 >21 Treatment: EB
formation
neurobasal medium
+RA plating
EB formation
: One million of feeder-independent mES cells were seeded in a 10-cm bacterial dish and cultivated in EB medium which contained ES medium but no LIF.
+RA : A final concentration of 5 μM retinoic acid was added to the EB medium.
plating : Determine the critical timing for EB trypsinization (7-9 min is typical). The optimal density for plating is around 1.2~1.5x105 cells/cm2.
: , 24 h; , 48 h; and , 72 h; time after plating
Fig. 1.
Fig. 1. Flow chart of experimental procedures to differentiate mESCs into neurons. mESCs were routinely cultivated in ES medium which normally contained leukemia inhibitory factor (LIF). On the day when differentiation began, 106 ES cells were cultivated on one 10-cm bacterial culture plate in which the LIF was withdrawn from the ES medium. This was designated day 0. On day 4, a final concentration of retinoic acid of 5 μM was directly added to the medium. Four days later, these embryoid bodies (EBs) were trypsinized and placed on
poly-DL-ornithine/laminin-coated plates. The timing of trypsinization had to be carefully controlled as described in the text. Due to the frequent operating steps from this time point on, the time was designated
hours/days after plating. Therefore, dissociated EBs were seeded onto plates at an optimal density in the range of 1.2~1.5x105 cells/cm2. The medium was replaced with fresh N2 medium at 2 and 24 h after plating.
At 48 h after plating, the culture medium was switched to
neural-basal-B27 medium. For the long-term culture of neuronal cells, it was necessary to partially replace the neuro-basal-B27 medium every other day. These neurons survived for at least 21 days.
24 h 48~72 h
Ratio of neurites length/neuronal cell numbers (Arbitrary Unit)
0
Fig. 2. Effects of different timing of EBs trypsinization on neuronal differentiation. A pool of EBs on day 8 was divided into four groups and subsequently subjected to 2, 4, 6, and 8 min of trypsinization. After
seeding on poly-DL-ornithine/laminin-coated plates, morphological
changes were recorded at 24, 48, and 72 h after plating. It was noteworthy that all plates contained the same initial cell density of 1.35 x 105/cm2 which was also a crucial factor in successful neuronal differentiation according to the literature and our experience. These pictures were taken with inverted phase-contrast microscopy. Each photo shown is
representative of one experiment that was repeated 3 times. The scale bar indicates 40 μm.
Field 1Field 2
48~72 h after plating
Fig. 3.
Fig. 3. Heterogeneous neuronal differentiation after trypsinization of whole EBs. The EBs were trypsinized and completely dissociated. This typically took 12~14 min. Cells were then counted at an optimal density (1.35 x 105/cm2) for seeding onto poly-DL-ornithine/laminin-coated plates. Morphological changes were recorded daily. The pictures shown here were taken 48~72 h after plating with two different fields. Each photo shown is representative of one experiment that was repeated five times. The scale bar indicates 40 μm.
A B
C
Fig. 4.
Fig. 4. Uniform neurons differentiated from mESCs. A. Morphology of feeder-independent mESCs cultivated on gelatin-coated plates. B. EBs were observed as early as day 1 or 2 when they were cultivated as a suspension on bacterial plates. The morphology of these EBs was recorded on day 4. The inset is a magnification of one of the EBs. C.
After trypsinization of the EBs with optimal timing, these isolated cells were plated onto poly-DL-ornithine- and laminin-coated plates. A population of uniform neurons was obtained and observed. The picture shown here was taken 48 h after plating. Each photo shown is
representative of one experiment that was repeated four times. The scale bar indicates 40 μm.
α Oct-4 α Type III
β tubulin
α Actin
1 2 3
A B
C D
E
Fig. 5.
Fig. 5. Morphological and biochemical evidence showing that uniform neurons had differentiated from mESCs. A. Neurons were fixed on coverslips using paraformaldehyde. Immunocytochemistry was performed using type III β-tubulin as the first antibody and Alexa
488-conjugated anti-mouse immunoglobulin G (IgG) as the secondary antibody. A superimposed picture of green (type III β-tubulin) and blue (DAPI) staining is shown. White triangles indicate cells stained with DAPI only, but not type III β-tubulin. B. Neurons differentiated from mESCs in the same experiment on the other coverslip were blotted with GFAP as the first antibody and rhodamine-conjugated anti-mouse IgG as the secondary antibody. C and D. CTX TNA2 cells, an astrocyte cell line, were used to examine the specificity and efficacy of antibodies. In a paralleled experiment, these cells were stained with the same antibodies as the ones used in A and B. Therefore, type III β-tubulin staining and GFAP staining were shown in C and D, respectively. Each photo shown is representative of one experiment that was repeated three times. The scale bar indicates 40 μm. E. Undifferentiated mESCs along with the neurons differentiated from mESCs were homogenized, run onto the gel, and blotted with Oct4 and class III β-tubulin antibodies. HEK 293T cells were used as a control of the cell type. The blot is representative of one experiment that was repeated three times. An actin blot served as the loading control.
C
A B
α Synaptotagmin α Actin
1 2 3 4
D
Fig. 6.
Fig. 6. Glutamatergic neurons were differentiated from mESCs. A.
Cells differentiated at 72 h after plating were fixed in paraformaldehyde, and the immunocytochemistry protocol was performed as described in
“General Materials and Methods”. Anti-vGLUT1/2 was used as the primary antibody. B. The same field was stained with DAPI. C. Merged image of A and B demonstrating that more than 95% of cells were
“General Materials and Methods”. Anti-vGLUT1/2 was used as the primary antibody. B. The same field was stained with DAPI. C. Merged image of A and B demonstrating that more than 95% of cells were