We used previously described DA differentiation protocol30,31 to derive DA neurons from iPSCs (See Methods). Analysis of the cultures by immunofluorescence showed that 30 days from the induction of differentiation, 76.23% ± 7.8 (mean ± S.D.) of the cells were positive for the neuronal marker βIII-tubulin, and 89.21% ± 6.7 (mean ± S.D.) of the βIII-tubulin-positive neurons were positive for the DA neuron marker tyrosine hydroxylase (TH) (Fig. 1A, B). Western blot analysis confirmed expression of TH, as well as expression of the plasma membrane DA transporter (DAT) in these cells (hence referred to as DA neurons), but not in neurogenin-2 induced iPSC-derived glutamatergic cortical-like neurons (i3Neurons, day 19) used as a control (Fig. 1C).
Conversely, expression of typical SV markers, such as synaptophysin, synapsin, VAMP2 and Rab3, were not present at relevant level in DA neurons at 30 days, but robust expression was observed at 50 days by immunofluorescence, which revealed the expected punctate synaptic pattern (Fig. 1D, E, Supplementary Fig. 1A–C).
Characterization of synaptic function with the calcium indicator Fluo-4 showed basal spontaneous activity within the neuronal population (Fig. 1F, G) while acute depolarization with high K+ stimulation (30 mM KCL) induced a synchronized rise in calcium levels in these neurons (Fig. 1H, I). Collectively, these results indicate that iPSC-derived DA neurons exhibit synaptic properties typical of a neuron from day 50 onwards in culture.
iPSC-derived DA neurons show vesicle pools distinct from those in i3Neurons
We next used transmission electron microscopy (EM) to examine the presence and type of secretory vesicles in neurite varicosities of cortical-like i3Neurons and DA neurons. In i3Neurons, numerous tightly-packed typical presynaptic clusters of ~40 nm vesicles were already clearly visible at day 19 post-differentiation (Fig. 2A–C). In DA neurons, the appearance of clusters of similar vesicles along axonal processes was much delayed relative to i3Neurons (Fig. 2D–F), consistent with the delayed expression of typical SV marker proteins (Fig. 1 and Supplementary Fig. 1). In favorable sections, many such clusters were clearly anchored to a region of the plasma membrane directly opposed to a region of neighboring cells positive for a thick or thin plasma membrane undercoating, confirming that they represent synaptic sites. Moreover, three morphologically distinct populations of such clusters were observed.
In one population, the ~40 nm vesicles were found as in i3Neurons (Fig. 2D, see also Fig. 2B, C), and the post-synaptic membrane had the thick undercoating typical of post-synaptic densities of asymmetric synapses. These synapses are likely glutamatergic, consistent with the reported property of DA neurons in situ to form VGLUT-positive glutamatergic synapses23,32.
In another population, clusters of irregularly shaped vesicles with a clear content and around 60–100 nm in diameter were frequently observed (Fig. 2E). These vesicles were tightly packed, completely segregated from other organelles and generally localized in outpocketing of neuronal processes, reminiscent of bouton-like structures lacking clear postsynaptic densities. In fact, staining for postsynaptic density protein 95 (PSD95), a postsynaptic marker, revealed a marked reduction in PSD95 labeling adjacent to synapsin-positive structures in DA neurons, compared to those found in i3Neurons (Supplementary Fig. 2A, B). Interestingly, appearance of these large vesicles has been observed in DA axonal boutons of the mouse nucleus accumbens13, whose identity has yet to be characterized. Analysis of the vesicle size distribution of both populations in DA neurons showed an average diameter of 54.9 ± 20.5 nm (mean ± S.D.), in comparison to those in i3Neurons (49.3 ± 5.8 nm and mean ± S.D., Fig. 2G, H).
In addition to the two populations of SV clusters described above, another vesicle population observed in DA neurons along cell processes was represented by 60–100 nm round, oval or irregularly shaped larger vesicles with an electron dense content. These vesicles were enriched in neurite varicosities but were much sparser than the other two populations described above. Their appearance suggests that they represent neuro-peptide containing large dense core vesicles (DCVs) (Fig. 2F). The occurrence of DCVs was previously observed in nigrostriatal axons20. Their presence is of special interest, as in cells specialized for amine secretion, such as chromaffin cells of the adrenal gland, amines are stored in DCVs that also contain a variety of peptides.
To determine whether both small and large vesicles with clear contents are endocytic organelles, cholera toxin-horseradish peroxidase (CTX-HRP) was added to unstimulated, K+ stimulated, and during recovery from K+ stimulation. CTX-HRP labels endocytic organelles from the cell surface and is an efficient marker for tracing recently formed SVs. Incubation with CTX-HRP for 2 h revealed a few vesicles positive for HRP most likely due to spontaneous neuronal activity, and a further increase in the number of HRP-positive vesicles after 30 h (Supplementary Fig. 2C–E). Notably at 30 hours, most of the vesicles including the small and large ones were positive for HRP (Supplementary Fig. 2B). High K+ stimulation for 90 s resulted in massive formation of large HRP-positive vacuoles previously recognized as bulk endosomes (Supplementary Fig. 2D), often detected in the presynapse of mouse primary neuronal cultures after high K+ stimulation33. Following recovery from high K+ stimulation, the majority of SVs were positive for HRP, indicating that the large pool of newly-formed SVs originated from the large vacuoles induced by high K+ stimulation (Supplementary Fig. 2F–H).
Confirming the presence of large vesicles and DCVs in axon terminals
The presence of organelles reminiscent of large vesicles and DCVs which could be storage sites for DA raises the possibility that at least a pool of DA secreted from DA neurons in the striatum may be released from these organelles. As in neurons, these vesicles can be present both in dendrites and axons, we wanted to confirm that the neurites of DA neuron containing them were axons. To address this question, we used a microfluidic compartmentalization device in which neurons are seeded in one chamber but can extend axons to another chamber through long microchannels (640μm in length, Fig. 3A, Supplementary Fig. 3A, B).
DA neurons from a 30-days old culture were replated in one of the chambers and allowed to grow their axons through the microchannels. To facilitate visualization of the axons, DA neurons derived from tdTomato-tagged TH iPSCs were used to track axonal outgrowth in the microchannels (Fig. 3B–D). Moreover, to facilitate axonal outgrowth and formation of synapses, 10 days after plating the DA neurons, iPSC-derived striatal medium spiny neurons (MSNs), which we confirmed to be positive for DARPP32 immunoreactivity by western blotting and immunofluorescence (Supplementary Fig. 3C–E), were seeded in the target chamber. A major synaptic target of DA neurons in the striatal region (nigrostriatal pathway) is the medium spiny neurons (MSNs), which represent 90% of the striatal neuronal population34. Following an additional 10 days, anti-synapsin immunofluorescence of the target chamber yielded abundant puncta revealing presence of axon varicosities (Fig. 3E, F). Importantly, EM revealed the same abundant presence of the organelles reminiscent of irregularly shaped large vesicles and DCVs, revealing that these organelles populate axons (Fig. 3G–J).
Distinct localization of VMAT2 and VGLUT2 in DA neurons
VMAT2 is the vesicular transporter responsible for the loading of dopamine into secretory vesicles of nigrostriatal DA neurons. A non-overlapping localization of VMAT2 and VGLUT2 is in fact supported by previous studies of axons of DA neurons in the ventral striatum and primary cultures of DA neurons23,27,32. To further validate this difference, we assessed the localization of VMAT2 and VGLUT2 in our iPSC-derived DA neurons at 50–55 days of differentiation, i.e., stage when synapses have matured (see Fig. 1). Since individual axons can be easily identified from these neurons by tracing their length, immunofluorescence of these cultures for VGLUT2 and synapsin demonstrated close colocalization of the two proteins in typical synaptic puncta (93.9% ± 7.6; mean ± S.D.), (Fig. 4A, A’, D), confirming presence of glutamate-containing SVs. Since antibodies that yield reliable VMAT2 immunofluorescence were not available, expression of VMAT2 tagged at its C-terminus with either GFP or mCherry was used in this study. VMAT2-GFP colocalized with synapsin (like VGLUT2 vesicles), suggesting that they are indeed SVs and not just endosomes resulting from overexpression (91.8% ± 6.7; mean ± S.D.), Fig. 4B, B’, D). When DA neurons expressing tagged-VMAT2 were stained for VGLUT2, there was indeed only a partial overlap between VGLUT2 and VMAT2 signal (43.5% ± 22.4; mean ± S.D.), Fig. 4C, C’, D). In addition, we further validated the presence of endogenous VMAT2 and VGLUT2 in these neurons by western blot from day 30 and day 60 cultures (Supplementary Fig. 4A). Detection of VMAT2 in striatal mice lysates showed presence of more than one band (Supplementary Fig. 4B, C), which could arise from region-dependent modifications to VMAT235. Of note, VGAT, the GABA vesicular transporter, was also detected in these neurons (Supplementary Fig. 4D), and its presence in mouse DA neurons has also been previously documented14, suggesting strong resemblance of mouse SV pools in iPSC-derived DA neurons. Collectively, our results indicate that the types of transporters, specifically VMAT2 and VGLUT2, do not have the same localization, phenocopying what had been observed in primary neuronal cultures and striatal brain slices23,25.
To further understand the type of vesicles VMAT2 is localized on in these neurons, we first performed correlative light and electron microscopy (CLEM) for VMAT2 in DA neurons. Towards this aim, VMAT2-GFP was transfected into DA neurons at day 50 on gridded glass dishes and visualized for VMAT2 fluorescence at day 53. VMAT2-GFP-positive structures were then identified on the gridded dish, and selected regions were processed for EM analysis (Fig. 4E). EM observation revealed that clusters of VMAT2-GFP fluorescence in control DA neurons corresponded to a mixture of predominantly larger vesicles (>60 nm, Fig. 4E’, F), which are mostly empty and occasionally electron-dense, indicating that VMAT2 proteins are indeed present on larger-sized secretory vesicles.
Differential sorting of VMAT2 and synaptophysin when expressed in an ectopic system
To further assess differences in the intracellular sorting of VMAT2 relative to other SV proteins, we capitalized on a recent demonstration that clusters of SV-like organelles – condensates that appear as large droplets in fluorescence microscopy – can be generated in fibroblastic cells (COS7 cells) by exogenous expression of synapsin and synaptophysin36,37. We had found that when additional bona fide SV proteins, such as VAMP2, SCAMP5, synaptotagmin 1, VGLUT1, VGAT1 and Rab3A are expressed together with synaptophysin and synapsin in these cells, they co-assemble with synaptophysin in the same vesicles. Thus, we examined whether also VMAT2, like these other proteins, can assemble into vesicles generated by synaptophysin expression.
First, we expressed VMAT2-GFP alone in COS7 cells and found that it localizes to the cis- and trans- Golgi complex, in addition to scattered vesicular puncta throughout the cytoplasm (Supplementary Fig. 5A). When co-expressed together with mCherry-synapsin, VMAT2 co-assembled with synapsin into droplet-like condensates (Fig. 5A, B, Supplementary Fig. 5B, C) reminiscent of those generated by synaptophysin and synapsin (Fig. 5C, D). However, these condensates were composed of larger and irregularly shaped vesicles (81.1 ± 34.9 nm; mean ± S.D.), clearly different from those found in the synaptophysin-synapsin condensates (43.3 ± 8.2 nm ; mean ± S.D.) (Fig. 5B, D, E, F). Strikingly, when VMAT2-GFP, synaptophysin and mCherry-synapsin were co-expressed together, VMAT2 vesicles and synaptophysin vesicles segregated from each other and assembled into distinct phases within the mCherry-synapsin phase, with synapsin being more concentrated (based on a higher fluorescence intensity) in the synaptophysin subphase (Fig. 5G–J). Most interestingly, CLEM of cells co-expressing synaptophysin, synapsin and VMAT2 revealed that the two phases detectable by fluorescence correlated with two classes of vesicles: small SV-like vesicles in the synaptophysin phase and larger vesicles in the VMAT2 phase (Fig. 5K). Synaptophysin and synapsin condensates have previously been shown to have liquid-like property36. To test whether VMAT2-synapsin droplets are phase separated vesicles, both synaptophysin (untagged) and VMAT2 (VMAT2-GFP) were co-expressed with synapsin (mCherry-synapsin) separately and photobleached for measurement of fluorescence recovery kinetics (Fig. 5L, M). Both types of condensates showed a similar fast fluorescence recovery after photobleaching for mCherry-synapsin, supporting the role of phase separation mechanisms mediated by synapsin in the formation of both VMAT2 and synaptophysin condensates. Furthermore, other SV proteins, such as VAMP2, Rab3, synaptotagmin-1, SCAMP5 or SV2C (an SV protein highly expressed by nigrostriatal DA neurons), were all positively colocalized with the VMAT2-synapsin clusters (Fig. 6A–E). Moreover, either VGLUT2 (VGLUT2-GFP, Fig. 7A, B) or VGLUT1 (VGLUT1-GFP, Fig. 7C, D), when co-expressed with VMAT2 (VMAT2-FLAG), synaptophysin (untagged) and synapsin (mCherry-synapsin) showed selective preference of the glutamate transporters for the synaptophysin-synapsin condensates over the VMAT2-synapsin clusters. In addition, since VGAT, the GABA transporter was also detected by western blot in DA neurons (Supplementary Fig. 4D), we co-express VGAT (VGAT-FLAG) with VMAT2 (VMAT2-GFP), synaptophysin (untagged) and synapsin (mCherry-synapsin). Like the VGLUTs, VGAT preferred the synaptophysin-synapsin (Fig. 7E, F) over the VMAT2-synapsin condensates. The localization of VGAT to synaptophysin vesicle condensates is consistent with previous findings37, suggesting that VGAT-positive vesicles are most likely comprised of SSVs.
Knockdown (KD) of AP3, an adaptor protein known to be important for the intracellular sorting of VMAT223, prevented the formation of VMAT2-synapsin condensates (Supplementary Fig. 6A–E) but did not impact the formation of synaptophysin-synapsin condensates (Supplementary Fig. 6F, G), pointing to different sorting pathways for the two proteins. Interestingly, knockdown of AP2, an adaptor protein essential for clathrin-mediated endocytosis, did not impede formation of both VMAT2 and synaptophysin vesicle clusters (Supplementary Fig. 7), suggesting Golgi-dependent mechanisms for these vesicles, in addition to endocytic origins37. The separation of VMAT2-positive vesicles from vesicles that share greater similarity to bona fide SVs is consistent with the differential localization of VGLUT2 and VMAT2-postive vesicles in mouse DA neurons of the ventral striatum23,25,32 and in iPSC-derived DA neurons as shown above. Note that while DA neurons have both large vesicles and DCVs, lack of a dense core in these vesicles is not unexpected, as fibroblastic cells do not express the molecular machinery to generate the peptide containing secretory granules of the classical regulated secretory pathway38. Hence it remains possible that VMAT2 is localized to both large clear and DCVs in DA neurons.
Presence of small and large vesicles in axonal varicosities of mouse striata
To gain direct insight on the structural differences in SV populations of DA axonal terminals in situ, EM of striatal homogenate from brains of adult mice (3–6 months old) (Fig. 8A) was performed, where the somatodendritic regions of DA neurons are absent in this region. Examination of the crude striatal homogenate embedded in agarose showed preservation of individual synaptosomes, with dense appearance in the cytoplasm. In some nerve terminals, tightly-packed SSVs with clear post-synaptic densities (PSDs) were observed (Fig. 8B), however, the presence of varicosities lacking clear PSDs (bouton-like) comprising of pleiomorphic vesicle pools were frequently observed (Fig. 8C). Since available antibodies to VMAT2 are not specific, we immunolabel striatal homogenate isolated from HA-tagged VMAT2 transgenic mice with HA (to label VMAT2) and synaptophysin antibodies in the agarose-embedded synaptosomes (Fig. 8D, E). Indeed, the sizes of vesicles labeled by HA were much larger than those labeled with synaptophysin (Fig. 8G, H). Immunogold antibody staining for synapsin showed labeling of both SSVs and large vesicles in the agarose-embedded striatal synaptosomes (Fig. 8F, G, H), consistent with those visualized in axonal varicosities of iPSC-derived DA neurons (Fig. 5). These data suggest that DA neurons are comprised of different SV populations with distinct vesicle size identities in vivo.…Read more by Tübingen, USA, Interfaculty Institute of Biochemistry, Rafiq, Nisha Mohd, Martin Shaun, Yale University School of Medicine, Germany, New Haven, University of Tübingen, Department of Neuroscience, Jaya, Kenshiro, Rosenfeld, Mishra, Fujise
