Brain-derived neurotrophic factor (BDNF) is known to be a crucial regulator of neuronal survival and synaptic plasticity in the mammalian brain. Furthermore, BDNF positively influences differentiation of embryonic neural precursors, as well as that of neural stem cells from adult neurogenic niches. To study the impact of cell-released BDNF on neural differentiation of embryonic stem cells (ESCs), which represent an attractive source for cell transplantation studies, we have generated mouse ESC clones overexpressing BDNF–GFP by use of knock-in technology. After neural differentiation in vitro, we observed that ESC clones overexpressing BDNF–GFP gave rise to an increased number of neurons as compared to control ESCs. Neurons derived from BDNF–GFP-expressing ESCs harbored a more complex dendritic morphology and differentiated into the GABAergic lineage more than controls. Moreover, we show that ESC-derived neurons released BDNF–GFP in an activity-dependent manner and displayed similar electrophysiological properties as cortical neurons. Thus, our study describes the generation of ESCs stably overexpressing BDNF–GFP, which are ideally suited to investigate the ameliorating effects of BDNF in cell transplantation studies of various neuropathological conditions.
Brain-derived neurotrophic factor (BDNF) is a member of the protein family of neurotrophins, promoting survival, differentiation and synaptic plasticity of neurons in the mammalian brain. Furthermore, BDNF is known to support differentiation of embryonic, mesenchymal and neural stem cells into neurons in vitro and in vivo (Kang et al., 2001; Ortega and Alcántara, 2010; Trzaska and Rameshwar, 2011). For example, studies using embryonic and adult neuronal precursor cells from mouse striatum revealed a BDNF-mediated increase in neuron number and neurite outgrowth that was independent from increased cell division and increased survival (Ahmed et al., 1995). BDNF has also been shown to promote the generation of neurons from neural stem cells (NSCs) in rat embryos (Tateno et al., 2004) and from neurospheres obtained from postnatal mouse brain (Silva et al., 2009). The enhancement of neuronal differentiation of stem cells by BDNF suggests a potential beneficial role of this neurotrophin for stem cell-based therapy of, for example, neurodegenerative diseases.
It is known that neurodegenerative diseases, such as Alzheimer's disease, Huntington's disease and mood disorders, such as depression and schizophrenia, are associated with decreased BDNF levels in affected brain regions, which is paralleled by respective changes of BDNF levels in the blood of patients (Castrén and Rantamäki, 2010a; Castrén and Rantamäki, 2010b; Ciammola et al., 2007; Fernandes et al., 2011; Ferrer et al., 2000; Green et al., 2011; Laske et al., 2007; Sen et al., 2008). However, whether reduced serum BDNF levels in patients with Huntington's disease might serve as a reliable biomarker is still under debate (Zuccato et al., 2011). Nevertheless, altered BDNF levels result in changes in neuronal plasticity, neuronal structure and neuronal survival, which probably contribute to outbreak of disease and disease phenotype.
Embryonic stem cells (ESCs) represent an attractive source for cell-transplantation-based curing of diseases, given that their pluripotency enables their differentiation into nearly all cell types, which can then be grafted into diseased tissue. Differentiation of ESCs into neurons and the functional integration into neuronal circuits upon grafting have been studied extensively in order to provide a replacement for degenerating neurons in the treatment of neurological disorders (reviewed by Liu et al., 2013). Moreover, stem cells genetically engineered to produce, for example, neurotrophins, which are then differentiated into neurons, can be used to deliver growth factors into a degenerating brain area to overcome cell death of grafted cells and to possibly enhance integration into the existing niches. Interestingly, mesenchymal stem cells (MSCs) can be epigenetically modified to endogenously express modest levels of BDNF (Somoza et al., 2010). Furthermore, it has been shown that, in a co-culture system with hippocampal neurons, MSCs induce glial-dependent neuronal survival, presumably through the secretion of BDNF (Mauri et al., 2012). Because BDNF enhances differentiation, these stem cells potentially show the capacity to support their own differentiation in an autocrine fashion. Viral transduction of MSCs and adult NSCs in order to produce BDNF has recently been shown to improve neuronal survival of host neurons as well as neuronal differentiation of grafted stem cells (Lim et al., 2011; Ma et al., 2012). Given that ESCs display a higher potential for differentiation than, for example, MSCs, and are easier to access and amplify than adult NSCs, ESCs genetically engineered to stably overexpress BDNF could be a very suitable tool to provide a pluripotent cell source for long-lasting expression and secretion of BDNF after grafting. Furthermore, expression of BDNF by knock-in technology into a defined locus circumvents potential genomic risks, which can occur in the use of viral transduction.
Here, we generated mouse ESCs with a targeted insertion of a BDNF–GFP-encoding cDNA construct under the control of a chicken β-actin promoter into the Rosa26 genomic locus. The C-terminal GFP fusion, which does not affect biological activity of BDNF (Brigadski et al., 2005), enables direct visualization of BDNF transport and secretion in ESCs and differentiated neurons derived from these cells. We show that BDNF–GFP released from these differentiating ESCs greatly enhanced neuronal differentiation compared with control cells. In addition, activity-dependent fusion pore opening and secretion of BDNF–GFP from secretory granules of ESC-derived neurons was observed. Thus, our study describes the generation of ESCs stably overexpressing BDNF–GFP, which are ideally suited to investigate ameliorating effects of BDNF in cell transplantation studies.
Generation of genetically engineered BDNF–GFP-expressing ESCs
In order to stably express BDNF–GFP in ESCs and cells derived thereof, a knock-in targeting strategy into the Rosa26 locus was chosen (Fig. 1A). After homologous recombination, the stop BDNF–GFP allele was obtained (Fig. 1B), containing the coding sequence of mouse BDNF with a C-terminal in-frame fusion with GFP (Kolarow et al., 2007) and the long 3′UTR including the proximal and distal polyadenylation sites of the endogenous mouse Bdnf gene. Using the long BDNF 3′UTR should allow the correct processing and subcellular targeting of BDNF–GFP-encoding mRNA (see, for example, Tongiorgi and Baj, 2008). The Rosa26 locus had been modified with the early cytomegalovirus enhancer element and the chicken β-actin promoter (CAG), to ensure ubiquitous expression. The insertion of a loxP-flanked transcriptional stop cassette (floxed neo-stop) led to the suppression of transcription of BDNF–GFP. Left (1.2 kb) and right (4.3 kb) arms of the Rosa26 locus were included in the targeting construct to allow for homologous recombination with the Rosa26 locus on chromosome 6 of mouse ESCs (V6.5). The neomycin resistance gene, together with the floxed transcriptional stop cassette, was inserted between CAG promoter and the coding sequence for BDNF–GFP, to allow for selection of cell clones carrying the integrated construct with G418. By using 5′, 3′, and neo probes (Fig. 1A,B), Southern blot analysis revealed that the correct clones were formed (stop; Fig. 1D). Using the Cre-recombinase-expressing plasmid pCrepac and subsequent selection with puromycin (Taniguchi et al., 1998), Cre-recombinase-mediated excision of the floxed neo-stop cassette led to the BDNF–GFP-encoding allele becoming transcriptionally activated (Fig. 1C). Several cell clones were isolated, and the expected activation of the BDNF–GFP-encoding allele was verified by Southern blot analysis in two independent clones, named E2 and H1 (Fig. 1D).
Expression of BDNF–GFP in ESCs and ESC-derived neurons
Neural differentiation of ESCs via formation of embryoid bodies (EBs), dissociation of the derived cells, and plating on astrocytes or laminin-coated coverslips was performed using standard procedures (see Materials and Methods). In all experiments using BDNF–GFP-expressing ESCs (BDNF–GFP ESCs), two independently derived BDNF–GFP ESC clones were investigated (i.e. E2 and H1), and both clones showed very similar results. Therefore, the pooled results from both clones are shown for some experiments.
Anti-GFP immunohistochemistry revealed that the BDNF–GFP protein was already expressed in undifferentiated ESCs (Fig. 2A). Following in vitro differentiation, both BDNF–GFP ESC clones exhibited typical neuronal morphology 10 days after dissociation of EBs (Fig. 2B,C). Subcellular localization of BDNF–GFP was analyzed using live-cell imaging of GFP epi-fluorescence (Fig. 2C–F). At early stages of differentiation, BDNF–GFP fluorescence in ESC-derived neurons was mainly in cytosolic compartments, excluding the nucleus, as expected for pro-BDNF–GFP, which is directed to endoplasmic reticulum and to the Golgi apparatus (compare with Brigadski et al., 2005; Haubensak et al., 1998). Additional targeting of BDNF–GFP into secretory granules of the regulated secretion pathway was observed in neuronal processes and became more pronounced at later stages of neuronal differentiation (Fig. 2B,C). The subcellular distribution of BDNF–GFP in neurons derived from BDNF–GFP ESCs was comparable to the expression pattern observed in primary hippocampal neurons transfected with BDNF–GFP-encoding cDNA (Fig. 2E). The BDNF–GFP fluorescence intensity in the ESC-derived neurons was ∼10 times lower than in transfected hippocampal neurons, suggesting a much lower expression level of the GFP-tagged BDNF in the ESC-derived neurons compared to transfected neurons. Differentiated V6.5 ESCs (control; Fig. 2D), and untransfected hippocampal neurons (Fig. 2F) exhibited only background levels of fluorescence under identical recording conditions.
For further expression analysis, cell lysates of purified BDNF–GFP ESC-derived neurons, undifferentiated BDNF–GFP ESCs and undifferentiated control cells (V6.5, Ctrl) were subjected to western blot analysis and probed with an anti-GFP antibody. The non-differentiated BDNF–GFP ESCs revealed a single GFP-reactive band with a molecular mass of ∼58 kDa, representing the uncleaved precursor pro-BDNF–GFP (Haubensak et al., 1998). Upon neuronal differentiation, the BDNF–GFP ESCs gained the capacity to cleave the pro-protein, thus allowing detection of an additional GFP reactive band at ∼43 kDa, representing mature BDNF–GFP (Fig. 3A). Thus, the ESCs are capable of cleaving pro-BDNF only during neuronal differentiation, enabling the secretion of mature BDNF.
The relative expression levels of endogenous BDNF and BDNF–GFP, respectively, were investigated at the mRNA and at the protein level. Bdnf mRNA and BDNF–GFP-encoding mRNA content in purified neurons derived from BDNF–GFP ESCs was assessed by quantitative PCR (qPCR). These experiments revealed a 35-fold higher expression of BDNF–GFP-encoding mRNA as compared to endogenous Bdnf mRNA levels in the BDNF–GFP ESC-derived neurons (Fig. 3B). However, anti-BDNF western blot analysis (Fig. 3C) revealed that the overexpression of BDNF–GFP protein in ESC-derived neurons was much lower than that for its mRNA. Analysis of cell lysates obtained from purified BDNF–GFP ESC-derived neurons showed an approximate 8-fold overexpression of pro-BDNF-GFP over endogenous pro-BDNF levels. The ratio of pro-BDNF–GFP to mature BDNF–GFP was ∼4∶1. Likewise, pro-BDNF was also the dominant molecular species of endogenously expressed BDNF in the same cells, with mature BDNF being barely detectable.
Enhanced neuronal differentiation of BDNF–GFP ESC lines
Following the induction of neural differentiation, EBs were dissociated and plated on laminin-coated coverslips. Two days after plating, MAP2-positive neurons were quantified as a percentage of the overall number of cells, yielding 66%±3 and 69%±3 (±s.e.m.) neuronal cells for BDNF–GFP ESCs for clone E2 and H1, respectively (control V6.5 cells: 33%±4, n = 4; significantly different with P<0.001 from E2 and H1; Fig. 4A,B). Notably, the increased neuronal differentiation of BDNF–GFP ESCs was completely inhibited when BDNF-scavenging TrkB-Fc receptor bodies (1 µg/ml; see also Walz et al., 2006) were present in the culture medium, thus indicating that the enhanced neuronal differentiation was attributable to released BDNF–GFP produced by the BDNF–GFP ESCs. In contrast, TrkB-Fc receptor bodies had no effect on the differentiation of V6.5 control ESCs, indicating that endogenous BDNF in these cells did not affect basal neuronal differentiation. Glial cell differentiation from the differentiated BDNF–GFP ESCs was not significantly changed when compared to control cells, as evaluated by immunostaining against the astrocytic marker GFAP and the oligodendrocyte marker O4 (Fig. 4B). Furthermore, 5 days after plating, the percentage of GAD67-positive neurons relative to all neurons was almost doubled (E2, 38%±1.4; H1, 42%±0.74) as compared to V6.5 control cells (24%±3; Fig. 4C,D). This relative increase in GABAergic neurons was also completely abolished by scavenging BDNF with TrkB-Fc receptor bodies, suggesting a shift of neuronal differentiation in BDNF–GFP-overexpressing ESCs towards the GABAergic lineage.
To validate whether the enhanced neuronal differentiation of BDNF–GFP ESCs compared to control V6.5 ESCs is also detectable at the mRNA level, we performed qPCR analysis of different neuronal and glial markers (Fig. 5). These experiments allowed us to further quantify the enhanced differentiation, revealing a ∼3-fold increased mRNA encoding of MAP2 and a ∼2-fold of β3-tubulin, which was inhibited by scavenging BDNF–GFP with TrkB-Fc receptor bodies. In contrast, the levels of the mRNAs encoding glial markers (GFAP and Olig2) were not significantly changed in both conditions, confirming the immunofluorescence quantification (Fig. 4B). qPCR analysis of the GABAergic markers GAD67 and GAD65 (also known as Gad1 and Gad2) already showed a ∼3-fold enhanced mRNA expression by 2 days after EB dissociation (as compared to protein expression on day 5; Fig. 4C,D). This effect was inhibited in the presence of TrkB-Fc receptor bodies. These data further support the enhanced GABAergic differentiation of the BDNF–GFP-overexpressing cells as compared to control cells.
Furthermore, neurons derived from BDNF–GFP ESCs showed increased dendritic complexity compared to control cells (Fig. 6). This was analyzed on day three after plating, when neuritic morphology became prominent. Thus, the average number of primary dendrites per neuron was 5.7±0.4 in the E2, and 6.0±0.2 in the H1 clone, compared to an average number of 3.9±0.1 primary dendrites in V6.5 control neurons (Fig. 6B). Again, incubation with TrkB-Fc receptor bodies completely abolished this effect on dendrite outgrowth in BDNF–GFP-overexpressing cells, whereas neurons differentiated from V6.5 control ESCs remained unaffected under these conditions. Sholl analysis (Fig. 6C) confirmed this result for primary dendrites, yielding 6.5±0.4 dendritic intersections for BDNF–GFP-expressing neurons (E2 and H1) and 4.5±0.6 intersections for control neurons in the first Sholl segment (starting radius of 6 µm). In addition, Sholl analysis revealed that neurons derived from BDNF–GFP-overexpressing ESCs showed increased dendritic branching as evaluated by an increase in intersection numbers over the measured distance of 26 µm as compared to neurons from control ESCs. As for the number of primary dendrites (Fig. 6B), treatment with TrkB-Fc receptor bodies completely abrogated the BDNF-mediated effect on dendritic branching.
Analysis of BDNF–GFP secretion from BDNF–GFP ESC-derived neurons
The BDNF-dependent enhanced differentiation of ESC-derived neurons indirectly suggested secretion of BDNF–GFP from these cells. To directly prove activity-dependent release of BDNF–GFP from ESC-derived neurons and to allow comparison with the respective release from primary neurons, we analyzed depolarization-induced secretion of BDNF–GFP. To ensure intact excitability of ESC-derived neurons, we verified the electrophysiological properties using whole-cell patch-clamp recordings. We observed firing of repetitive action potentials in response to current injection in BDNF–GFP ESC-derived neurons (supplementary material Fig. S1). Action potential properties were similar to the firing patterns observed in primary hippocampal and cortical neurons, and in control ESC-derived neurons (Jüngling et al., 2003). Altogether, these results revealed complete neuronal differentiation of the BDNF–GFP ESCs.
Using time-lapse video microscopy, we analyzed the fusion pore opening of BDNF–GFP-containing secretory granules in neurons differentiated from BDNF–GFP ESCs. Depolarization of neurons was induced by superfusion with extracellular solution containing elevated K+ (50 mM). To facilitate detection of depolarization-induced fusion pore opening, 0.3 mM of the low-molecular-mass dye Bromophenol Blue (BPB) was added to the extracellular solution, which efficiently quenched the green fluorescence of BDNF–GFP vesicles following instantaneous diffusion of BPB through the opened fusion pore (Kolarow et al., 2007). Under these conditions, elevated-K+-induced depolarization allowed us to observe fusion pore opening of BDNF–GFP-containing vesicles as a sudden decrease in vesicular fluorescence intensity (Fig. 7). BDNF–GFP ESC-derived neurons were analyzed 7–14 days after dissociation of EBs. Analysis was restricted to clearly discernible single vesicles in neurites. Approximately 10% of these vesicles (79 out of 713 vesicles analyzed in six cells) disappeared instantaneously, which represents fusion pore opening upon depolarization. The remaining vesicles in the same neurites that were analyzed did not show any decrease in fluorescence intensity in this assay, as indicated by stable fluorescence throughout the analysis. These results were comparable to the respective fusion pore opening events that were detected in BDNF–GFP-overexpressing primary hippocampal neurons (Fig. 7C,E).
Whereas the BPB assay is an unequivocal proof of fusion pore opening in secretory granules containing BDNF–GFP, actual release needs to be quantified as a decrease in BDNF–GFP fluorescence intensity of vesicles in the absence of BPB, thus representing diffusion of BDNF–GFP out of the vesicles. Analysis of release in these experiments was again focused on clearly discernible single vesicles in neurites of BDNF–GFP ESC-derived neurons. Because the intravesicular pH is acidic before fusion pore opening, which quenches the pH-sensitive intravesicular GFP fluorescence, fusion pore opening in this experiment is evident from the initial fluorescence increase of vesicles undergoing exocytosis owing to pH equilibration with the extracellular solution (Brigadski et al., 2005). This increase was followed by a subsequent decrease of BDNF–GFP fluorescence intensity due to diffusion of the neurotrophin out of the vesicles (Fig. 8). On average (nine vesicles from three cells), these vesicles released ∼20% of their BDNF–GFP content [the remaining fluorescence intensity at the end of stimulation (300 seconds) was 79.8%±3.5], which was again indistinguishable from the respective release data obtained from BDNF–GFP-overexpressing primary hippocampal neurons (Fig. 8C,E). Interestingly, we also observed a population of vesicles showing only fusion pore opening, as described previously (Matsuda et al., 2009; de Wit et al., 2009); this reflects a typical feature of neurosecretory vesicles (for a review, see Lessmann and Brigadski, 2009). Importantly, numerous single discernible vesicles in neurites showed stable fluorescence throughout the depolarization, reflecting the large number of secretory vesicles that are presumably not in the readily releasable pool (see Lessmann and Brigadski, 2009).
BDNF is critically involved in morphological differentiation and synaptic plasticity of central nervous system (CNS) neurons. Furthermore, there is evidence for reduced brain BDNF levels being associated with different neurodegenerative disorders, e.g. Alzheimer's and Huntington's disease. Thus, supplying BDNF via neurons derived from ESCs in affected brain areas is a promising therapy model for these diseases. Here, we generated ESCs stably overexpressing GFP-labeled BDNF and analyzed expression levels, localization and release of BDNF–GFP from neurons differentiated from ESCs. Our results showed an ∼8-fold overexpression of BDNF protein in ESC-derived neurons. We observed vesicular expression and secretion of BDNF–GFP, similar to previous observations in primary neurons. Furthermore, the released BDNF–GFP retained the biological activity of BDNF, thereby enhancing the efficacy of neuronal differentiation of ESCs. Our study suggests that these genetically engineered BDNF–GFP-expressing ESCs are ideally suited for delivery of visible, GFP-labeled biologically active BDNF in vivo. Transplantation of BDNF–GFP-overexpressing ESC progenitors into diseased tissue would enable the investigation of beneficial BDNF-mediated effects on terminal ESC differentiation and on damaged host tissue with direct tracking of grafted BDNF–GFP-expressing cells.
Neuronal differentiation of BDNF–GFP ESCs
To our knowledge, we have generated for the first time an ESC line overexpressing a visible and biologically active BDNF–GFP. Following appropriate differentiation of these BDNF–GFP ESCs and transplantation into target tissue, BDNF–GFP can be tracked by fluorescence microscopy, enabling the analysis of transport, secretion, re-uptake and biological effects of BDNF. Successful differentiation of BDNF–GFP ESCs into functional neurons, as judged by morphology, resting membrane potential and firing of Na+-channel-dependent action potentials was observed, as described previously (Fig. 2; supplementary material Fig. S1; see Jüngling et al., 2003).
Expression level of BDNF–GFP in ESCs
We estimated that the expression of BDNF–GFP in hippocampal neurons was ∼10 times higher than in neurons derived from BDNF–GFP ESCs. This can be explained by the fact that expression in hippocampal neurons was driven by an expression plasmid, which contains a strong immediate early CMV promoter (pCMV IE; see Haubensak et al., 1998), whereas the expression of BDNF–GFP in ESC-derived cells is controlled by the Rosa26 locus, which contains a chicken β-actin promoter element combined with the cytomegalovirus (CMV) early enhancer element (Fig. 1, CAG). It is generally accepted that expression from the Rosa26 locus containing the CAG promoter element can give considerable expression levels. However, these expression levels cannot be compared with the levels reached by using expression plasmids and transfection procedures.
Furthermore, in BDNF–GFP ESC-derived cells, we observed a 35-fold elevated mRNA level, but only an 8-fold elevated protein level of BDNF–GFP in relation to endogenous BDNF. This might be explained by differences in mRNA stability, translational efficacy and/or protein half-life. In addition, it is possible that the non-spliced BDNF–GFP-encoding allele in the Rosa26 locus, which lacks the 5′UTR and other 5′ non-coding BDNF exons, is transcribed faster than the endogenous Bdnf gene. However, the missing 5′UTR, which is important for translational initiation and regulation, might also mean that the translation of BDNF–GFP-encoding mRNA is decreased. Finally, the molecular mass of the fusion protein BDNF–GFP is approximately double (58 kDa) that of endogenous BDNF (32 kDa), which might also lead to a reduction in translational efficacy.
The moderately overexpressed BDNF–GFP was targeted to secretory granules and was found in all neurites of differentiated neurons, as previously shown for BDNF–GFP and endogenous BDNF in primary hippocampal and cortical neurons (Brigadski et al., 2005; Haubensak et al., 1998; Kohara et al., 2007; Matsuda et al., 2009; Swanwick et al., 2004). To put the present level of overexpression into context, relatively high levels of virus-mediated BDNF expression in bone marrow stromal cells, MSCs and adult NSCs have been previously shown to exert beneficial effects, for example after grafting into lesioned tissue (see e.g. Lim et al., 2011; Ma et al., 2012; Makar et al., 2008; Makar et al., 2009). Furthermore, modest endogenous levels of BDNF protein in mesenchymal and neural stem cells (Chekhonin et al., 2011; Li et al., 2006; Mauri et al., 2012) have also been shown to be beneficial on co-cultured neurons (Mauri et al., 2012). Hence, the 8-fold overexpression of functional BDNF–GFP we achieved here is of physiological relevance.
Enhanced neuronal differentiation of BDNF–GFP ESCs
As described previously for MSCs or adult NSCs overexpressing untagged BDNF (Lim et al., 2011; Ma et al., 2012), or as revealed by application of BDNF to EBs (Kang et al., 2001) or to NSCs (Silva et al., 2009), neuronal differentiation of our BDNF–GFP-overexpressing ESCs was strongly enhanced (Figs 4, 5). Given that this enhancement was critically dependent on the extracellular availability of BDNF (see inhibition by TrkB-Fc receptor bodies; Fig. 4), this result proves a causal connection between release of BDNF–GFP and enhanced neuronal differentiation. These results further prove that GFP-tagged BDNF is released from ESCs and developing neurons and that it retains the biological activity to activate TrkB receptors, which are expressed in NSCs (Ahmed et al., 1995). As an additional indicator of the impact of BDNF on neuronal differentiation in the context of neurite outgrowth, we detected a higher number of primary dendrites and increased dendritic branching in neurons differentiated from BDNF–GFP-overexpressing ESCs, an effect that was completely abolished in the presence of TrkB-Fc receptor bodies. These data are consistent with several reports stressing the positive effect of BDNF on neurite and dendritic outgrowth (Baj et al., 2011; Cohen-Cory et al., 2010; McAllister, 2002).
Furthermore, we detected an enhanced GABAergic differentiation of ESC-derived neurons in the BDNF–GFP-overexpressing ESCs, indicating that BDNF is strongly involved in the differentiation towards the GABAergic lineage. These data are in accordance with the previously reported effects of BDNF on GABAergic differentiation of NSCs from the postnatal mouse forebrain (Silva et al., 2009) and the strong BDNF-dependent support of differentiation of GABAergic neurons (Huang et al., 1999; Kohara et al., 2007; for a review, see Gottmann et al., 2009). Furthermore, these data are in agreement with the role of BDNF in neuroprotection of GABAergic neurons (Canals et al., 2001; Pérez-Navarro et al., 2000; Petersén et al., 2001).
Analysis of BDNF–GFP secretion from BDNF–GFP ESC-derived neurons
Fusion pore and release measurements of BDNF–GFP allowed us, for the first time, to visualize in real time the secretion process of a neurotrophin from ESC-derived neurons (Figs 7, 8). The observed kinetics and the degree of BDNF–GFP release were similar to the comparable data obtained for BDNF–GFP overexpressed in primary hippocampal neurons (Brigadski et al., 2005; Kolarow et al., 2007). In addition, the relative number of fusion events was similar, indicating that the ready releasable pool of BDNF vesicles is not different between both experimental situations. When comparing the fluorescence intensity of BDNF-GFP vesicles prior to release, BDNF-–GFP expression in differentiated ESCs was roughly 5–10-fold lower than that of BDNF–GFP-transfected primary hippocampal neurons (R.E. and P.L., data not shown). Although these data suggest that release properties of BDNF–GFP were not affected by the extent of overexpression, it remains to be shown whether time-resolved release of BDNF at the – even lower – physiological expression level proceeds similarly.
The similar properties of BDNF–GFP release between the differentiated BDNF-GFP ESCs and BDNF–GFP-transfected neurons also provide evidence that the previously described release properties are not a result of changes in the physiological characteristics of neurons following transfection. For example, fusion pore opening and dilatation (with dilatation being critical for proper diffusion of the large BDNF–GFP molecule out of a vesicle) proceed such that BDNF–GFP can also be released from untransfected neurons.
Overall, our ESCs that show an ∼8-fold overexpression of BDNF–GFP will be a valuable tool in future studies investigating BDNF targeting, secretion and its action after proper release in real time. Our results strongly support the application of knock-in ESCs overexpressing BDNF–GFP in cell therapy approaches for various neuropathological conditions. It would be valuable to test whether the enhanced neuronal differentiation of BDNF–GFP-overexpressing ESCs is also observed in vivo, leading to a long-lasting functional improvement, especially in mouse models of neurodegenerative diseases. As we observed an increase in GABAergic neuronal differentiation from BDNF–GFP ESCs, a cell transplantation of early neural progenitors might become particularly attractive to pathologies where GABAergic neurons are degenerating, as in, for example, Huntington's disease.
Materials and Methods
Generation of targeted BDNF–GFP ESC clones by insertion into the Rosa26 locus
The coding sequence of mouse pre-pro-BDNF–GFP was fused (compare Brigadski et al., 2005; Kolarow et al., 2007) to the long 3′UTR of mouse BDNF (NCBI RefSeq ID: NW_001030694.1, corresponding region: 70780131–70783754) comprising the two BDNF-3′UTR polyadenylation (pA) sites (corresponding regions: proximal pA site 70780414–70781175; distal pA site 70783428–70783668). The respective DNA was obtained by PCR from BAC clone RP23-390F19 (Bacpac Resources Center Children Hospital, Oakland Research Institute, CA). Then, the pre-pro-BDNF–GFP-3′UTR sequence was cloned into a Rosa26-targeting vector (Remedi et al., 2009) with deletion of the IRES-GFP sequence. This vector contained the left and right homology arm of the Rosa26 locus (Soriano, 1999), and included a 5′ sequence of a ubiquitous CAG promoter [combination of the cytomegalovirus (CMV) early enhancer element and chicken β-actin promoter] and a loxP-flanked stop cassette, also harboring the neomycin resistance gene. The targeting vector was constructed so that an EcoRI digest could serve for the subsequent analysis of ESC clones by Southern blotting. Linearization of the targeting construct for electroporation into ESCs (V6.5) (Eggan et al., 2001; Rideout et al., 2000) was achieved by AsisI digestion. The targeting vector was completely sequenced to ensure the functional integrity of the construct.
Southern blotting of selected ESC clones was performed as described previously (Sambrook and Russel, 2001) after EcoRI restriction digest of genomic DNA. Hybridization probes were generated by PCR with appropriate DNA templates and the following primer pairs: 5′probe fw: 5′-CAGGCAAAAAGGGGAGACCA-3′, 5′probe rev: 5′-CGTTGGGCCTAACTCGAGTC-3′; 3′probe fw: 5′-CATAGTGAACACTGAATGGC-3′, 3′probe rev: 5′-TATGTGTACCAGTGCAGGTG-3′; Neo probe fw: 5′-TGCTCGACGTTGTCACTGAAGC-3′ and Neo probe rev: 5′-TACCGTAAAGCACGAGGAAGC-3′ (see Fig. 1A,B). Electroporation of a recombined ESC clone harboring the heterozygous stop BDNF–GFP allele with the Cre-recombinase-expressing plasmid pCrepac and subsequent selection with puromycin (Taniguchi et al., 1998) led to Cre-recombinase-mediated excision of the floxed neo-stop cassette and thereby to the transcriptionally activated BDNF–GFP allele. Puromycin-selected ESC clones were verified for correct recombination events by Southern blotting with the same hybridization probes as for the ESC clones with the heterozygous stop BDNF–GFP allele (see above).
ES cell culture and in vitro differentiation of EBs
Mouse ESCs (V6.5 line) were cultured on mouse embryonic feeder cells (inactivated with mitomycin-C) in the presence of leukemia inhibitory factor (LIF) according to standard protocols (Hogan et al., 1994). For embryoid body (EB) formation, 2×106 ESCs/10 ml differentiation medium [DMEM high glucose + 15% knockout serum replacement, including 2 mM glutamine, 1 mM sodium pyruvate, 0.1 mM MEM non-essential amino acids, 50 U/ml penicillin/streptomycin (all from Invitrogen, Carlsbad, CA) and 0.1 mM β-mercaptoethanol] were incubated on non-adherent bacterial plastic dishes (100 mm) for 4 days. In vitro differentiation of ESCs into neural cells was promoted by the addition of 5 µM all-trans-retinoic acid for an additional 2 days. EBs were further cultivated for 1 week on polyornithine-coated dishes in neurobasal (NB) medium with addition of B27 supplement, glutamax and penicillin/streptomycin (all from Invitrogen). EB dissociation was carried out as described previously (Jüngling et al., 2003) and, thereafter, dissociated cells were plated on glass coverslips coated with polyornithine (0.5 mg/ml) and laminin (5 µg/ml in NB medium), or tissue culture dishes, at a density of 3.8×104 cells/cm2. In experiments where BDNF signaling was blocked, recombinant human TrkB-Fc receptor bodies (R&D systems) were added at a concentration of 1 µg/ml in the respective medium. TrkB-Fc receptor bodies were applied during the whole differentiation process, starting with the production of EBs until the final analysis of cells.
When quantifying neural differentiation by immunofluorescent staining or qPCR, no purification of ESC-derived neurons by immunoisolation was performed. However, to determine the expression level of exogenous versus endogenous BDNF by qPCR and western blotting (see Fig. 3), an anti-L1-immunopanning was performed after EB dissociation as described previously (Jüngling et al., 2003), in order to specifically select for neuronal cells.
Total RNA from differentiated ESCs was extracted, and reverse transcriptase reactions with random primers using 0.2–1 µg RNA were performed. Taqman qPCR gene expression assays (Applied Biosystems) were carried out with Taqman probes (Assay ID) detecting mRNA encoding MAP2 (Mm00485230_m1), GAD67 (Mm00725661_s1), GAD65 (Mm00484623_m1), β3-tubulin (Mm00727586_s1), GFAP (Mm01253033_m1) and Olig2 (Mm01210556_m1). A custom Taqman gene expression assay was specifically generated, comprising the junction of the endogenous BDNF coding sequence and the adjacent 3′UTR in order to distinguish it from BDNF–GFP transcripts with the following sequences: Forward primer, 5′-GCTGGCGATTCATAAGGATAGAC-3′; reverse primer, 5′-TATACAACATAAATCCACTATCTTCCCCT-3′; probe sequence, 5′-TATGTACACTGACCATTAAA-3′. Furthermore, a custom Taqman gene expression assay was specifically generated for GFP with the sequences: Forward primer, 5′-GAGCGCACCATCTTCTTCAAG-3′; reverse primer, 5′-TGTCGCCCTCGAACTTCAC-3′; probe sequence, 5′-ACGACGGCAACTACA-3′. As reference genes, either Gapdh (Mm99999915_g1) (Murphy and Polak, 2002) or Gusb (Mm00446956_m1) was used. qPCR reactions were performed with the 2×TaqMan gene expression master mix (Applied Biosystems) and analyzed with an ABI 7300 qPCR machine (Applied Biosystems).
Lysates of undifferentiated or differentiated ESCs were prepared using RIPA buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.1% SDS, 0.1% sodium deoxycholate, 1% NP-40) including protease and phosphatase inhibitors (Halt Protease and Phosphatase Inhibitor Cocktail, Thermo Scientific). Cellular lysis was performed for 45 minutes at 4°C, followed by sonication with five pulses at 50% force with the Ultra Turax sonicator IKA R104. After centrifugation to sediment cellular debris, 25 µg of protein was separated on a 10% or 12% SDS-PAGE gel, blotted onto a nitrocellulose membrane and probed with a mouse monoclonal anti-GFP (Roche) or a rabbit ployclonal anti-BDNF antibody (Santa Cruz Biotechnology) and a secondary HRP-conjugated goat anti-mouse or -rabbit antibody (Dianova). Signal detection was achieved by ECL Prime western blotting detection reagent (GE healthcare) and the Peqlab FUSION-SL Advance 4.2 MP analyzer. Quantification was performed by densitometric analysis using the Bio1D 15.02 software (Vilber Lourmat).
Immunofluorescence and quantification of neuronal differentiation
For BDNF–GFP expression analysis, undifferentiated BDNF–GFP ESCs or neurons derived from BDNF–GFP ESCs were fixed with 4% paraformaldehyde followed by immunohistochemical staining with an anti-GFP rabbit polyclonal serum (kindly provided by Dr Matthias Klugmann, Sydney) and a goat anti-rabbit Alexa-Fluor-488-coupled antibody (Invitrogen). For quantification of neuronal differentiation, control and BDNF–GFP-overexpressing differentiated ESCs were fixed with 4% paraformaldehyde 2 or 5 days after plating and stained with immunofluorescence antibodies against MAP2 (mouse monoclonal or rabbit polyclonal, both Millipore), GFAP (rabbit polyclonal, Abcam), O4 (mouse monoclonal, Millipore) and GAD67 (mouse monoclonal, Millipore; secondary goat or donkey anti-mouse or -rabbit Alexa-Fluor-488 or -546-coupled antibodies, Invitrogen) and analyzed with a Leica DM5500 fluorescence microscope (Leica microsystems, Wetzlar, Germany). Note that Alexa-Fluor-488-coupled secondary antibodies could be used here because endogenous BDNF–GFP fluorescence was too low to be detected by the Leica DM5500 fluorescence microscope. The percentage of positive cells was assessed relative to all cells (DAPI staining) or all neurons (MAP2-postive cells) by counting five independent microscopic fields with a total of 300 cells per condition and experiment.
Assessment of dendritic outgrowth
For the analysis of dendrite outgrowth, cells were fixed 3 days after plating and stained by immunofluorescence with the monoclonal MAP2 antibody and analyzed with the Zeiss Axiovert laserscanning microscope 710 (Carl Zeiss, Oberkochen, Germany). Pictures of five independent fields per condition were made, and subsequently, primary dendrites of 50–100 neurons were counted per condition and experiment.
Sholl analysis was used to investigate dendritic complexity in relation to the distance from the soma [starting radius, 6 µm; ending radius, 26 µm; at further distances (28–40 µm), dendrites overlapped with neurites from neighboring cells making clear analysis of numbers of intersections impossible]. Concentric Sholl segments (concentric radial interval of 2 µm) were generated starting at a distance of 6 µm from the center of the soma. The number of dendritic intersections was analyzed per Sholl segment using the Fiji Image J software (Schindelin et al., 2012). The axon was excluded based on the absence of MAP2 staining.
Primary hippocampal microcultures
Dissociated mouse hippocampal microcultures were prepared as described previously with minor modifications (Brigadski et al., 2005): Postnatal rat neocortical astrocytes (P0–P2) were isolated and cultured for 2–4 weeks in BME medium containing 10% FCS until expansion to confluence. Astrocytes were passaged and seeded on glass coverslips at a density of 5×104 cells per 3.5 cm dish in BME/10% FCS, to yield astrocyte islands of 100–300 µm in diameter after 18–24 days in vitro (DIV). Five µM ARAC was added 3 days after seeding of astrocytes to avoid further division of astrocyte islands. Dissociated postnatal mouse (P0–P2) hippocampal neurons were plated in BME/10% FCS at a density of 1–10 neurons per astrocyte island onto the coverslips. After 20 hours, the plating medium was exchanged to serum-free medium (neurobasal medium with 2% B27 supplement, Invitrogen). All animal experiments were performed according to approved guidelines.
Transfection of hippocampal microcultures
Mouse hippocampal microcultures were transfected with expression vectors coding for C-terminal GFP-fused mouse BDNF. The plasmid (4.5 µg DNA per 3.5 cm dish) was transfected at 4–5 DIV using the Ca2+ phosphate precipitation method as described previously (Haubensak et al., 1998). During incubation (2.5 hours), 5 µM DNQX and 50 µM D,L-APV were added to reduce excitotoxicity. The transfection medium was replaced, cultures were washed twice with warmed HEPES-buffered saline (HBS), and thereafter the cultures were kept in neuron-conditioned neurobasal/B27 medium. The cells were used for time lapse imaging experiments 4–9 days after transfection.
Coverslips with differentiated ESCs were transferred to a submerged recording chamber mounted on the stage of an upright microscope (Axio Examiner A1, Zeiss, Germany) equipped with a 63× water immersion objective (Zeiss, Germany). Coverslips were placed in the submersion chamber, filled with HBS at room temperature (23–25°C). Whole-cell patch-clamp recordings were performed on cells with neuronal morphology by use of infrared DIC videomicroscopy (Dodt and Zieglgänsberger, 1990). Patch pipettes were pulled from borosilicate glass (GB150F-8P, Science Products, Hofheim, Germany) to resistances of 7–10 MΩ. The standard pipette solution used for current-clamp recordings contained (in mM): 80 potassium gluconate, 17.5 KCl, 9 NaCl, 1 MgCl, 10 HEPES, 0.2 EGTA, 2 ATP, 0.2 GTP, pH adjusted to 7.2 with KOH (Bi and Poo, 1998; Gärtner and Staiger, 2002). A liquid junction potential of +10 mV of the pipette solution was corrected for. Resting membrane potential, and active and passive membrane properties were measured in the current-clamp mode of the patch amplifier. Patch-clamp recordings were performed using an EPC8 amplifier operating with PatchMaster software (HEKA Elektronik, Lamprecht, Germany). Data were analyzed off-line with FitMaster (HEKA Elektronik, Lamprecht, Germany).
Fusion pore opening assay and release measurements
Visualization of epifluorescence signals was performed as described previously (Kolarow et al., 2007). Briefly, coverslips with cells were transferred into a bath chamber (Luigs & Neumann) and inspected with an upright fluorescence microscope (Olympus BX51W) using a 60× water immersion objective (LUMFI, Olympus, Hamburg, Germany, NA: 1.1). Image capture was performed using a cooled CCD camera (CoolSnap HQ2, Photometrics, Huntington Beach, CA, 14-bit dynamic range), controlled by VisiView software (Visitron Systems, Puchheim, Germany). The exposure times for time-lapse recordings (between 0.3 and 2.0 seconds) were adjusted for every cell such that vesicles were clearly distinguishable from the background without driving the CCD chip into saturation. Processing of images was performed by MetaMorph (Molecular Devices, Downingtown, PA) and CorelDraw X4 software without compromising the evident primary image information. However, in the figures, fluorescence in the soma and the proximal dendrites was often enhanced close to saturation in order to make the single vesicles in distal neurites clearly visible. Only cells showing fusion pore opening after stimulation with elevated K+-containing solution were analyzed.
Data are presented as mean±s.e.m. Differences were tested using two-way analysis of variance (ANOVA) and post hoc Tukey test (Fig. 4; Fig. 6B). ANOVA with repeated measures was performed for Sholl analysis (Fig. 6C), testing initially with a two-way ANOVA with repeated measures (factors: ESC type, TrkB-Fc treatment, distance). As this gave a significant interaction for ESC type × TrkB-Fc treatment with P<0.001, we investigated group differences further with consecutive one-way ANOVA with repeated measures for the four groups (E2/H1 and Ctrl, each ±TrkB-Fc treatment) and post-hoc Tukey's test. To have paired testing in qPCR experiments, the paired Student's t-test was applied (Fig. 5). Unpaired Student's t-test was applied when only two groups were compared (Fig. 3B). Differences were assumed to be significant if P<0.05. Unless otherwise stated, non-significant differences are not indicated.
We would like to thank Danka Dormann, Ruth Jelinek, Regina Ziegler, Sabine Mücke and Martina Bohndorf for excellent technical support.
Experiments were performed by J.L., R.E. and P.L. with help and advice by K.N. and T.B. The manuscript was written by J.L., V.L. and B.L. with help from K.G. and T.B. The study was designed by V.L., B.L., K.G. and J.L. The study was jointly supervised by B.L. and V.L.
This work was supported by the German Research Foundation DFG (to V.L. and B.L.) [grant numbers LE 1020/2-1 to V.L., LU 775/5-1 to B.L.].