|Home | About | Journals | Submit | Contact Us | Français|
Although drugs used to treat several neurological diseases are presumed to target synapses that secrete dopamine (DA), relatively little is known about synaptic vesicle (SV) release mechanisms at single DA synapses. We found that the relative probability of release (Pr) varied between individual DA synapses. Furthermore, DA terminals generally exhibited lower Pr than glutamatergic hippocampal (Hpc) terminals, suggesting that DA release is less reliable than the release of glutamate. Our mathematical model of fluorescence loss shows that Pr is regulated by two independent and heterogeneous elements. Firstly, the size of the recycling SV pool regulates Pr. Secondly, Pr is also independently regulated by additional factors, which are reflected in the time constant of FM 1-43 destaining, τ. We found that the observed difference in Pr between Hpc and DA neurons results because the recycling SV pool is smaller in DA neurons than in Hpc neurons. However, τ does not vary between these two neuron populations. We also identified a population of functional non-synaptic boutons in DA axons, which are not associated with a postsynaptic element and which are not functionally different to boutons that formed conventional synapses. Our work provides a new approach to the study of SV exocytosis in DA neurons and shows that synaptic terminals of DA neurons are functionally heterogeneous and differ from excitatory terminals in mechanisms that regulate Pr.
Neurons that secrete dopamine (DA) are associated with a number of prevalent neurological disorders, including Parkinson's disease (PD). Drugs used in the treatment of PD, as well as antipsychotics and several addictive drugs, are presumed to act by altering the function of DA synapses. Understanding the mechanisms and signalling pathways that regulate SV exocytosis, endocytosis and recycling in DA neurons is critical for understanding the actions of these drugs and for the identification of new therapies. However, relatively little is known regarding cellular mechanisms that regulate the exocytosis of these synaptic vesicles (SVs) in DA neurons at the level of individual synapses.
Synaptic transmission is an innately unreliable process, in that each action potential does not always result in neurotransmitter release (Hessler et al., 1993; Rosenmund et al., 1993). The presynaptic probability of release (Pr) refers to the likelihood that an action potential will cause release of neurotransmitter at a given synapse. Pr is heterogeneous across hippocampal (Hpc) synapses (Rosenmund et al., 1993; Murthy et al., 1997; Ryan et al., 1997). However, mechanisms that govern this functional heterogeneity remain elusive. Moreover, while studies using electrochemical and ultrastructural methods have contributed substantially to our understanding of quantal DA release from SVs (Nirenberg et al., 1997; Pothos et al., 1998; Staal et al., 2004), functional heterogeneity has not been investigated in single DA synapses.
Studies using optical techniques to SV release in Hpc neurons have greatly enhanced our understanding of presynaptic function (for examples see Murthy et al., 1997; Ryan et al. 1997; Smith and Waters, 2002; Virmani et al., 2006; Chen et al., 2008; Tokuoka and Goda, 2008). In this study we provide a description of functional heterogeneity across DA synapses and a comparison of SV exocytosis in DA and glutamatergic Hpc neurons. We find that DA synapses exhibit considerable heterogeneity in Pr. Overall, DA synapses exhibit lower Pr compared to Hpc synapses, suggesting that DA release is even less reliable than the release of glutamate. Furthermore, by modeling our data we find that Pr can be described by two independent factors. Using this analytical approach, we conclude that the lower Pr in DA neurons is due to these synapses containing fewer recycling SVs available for release. We also show that while the kinetics of SV pool release is also heterogeneous at DA synapses there is no detectable difference in release kinetics between Hpc and DA synapses. Finally, in addition to synaptic terminals we show that DA neurons possess a subpopulation of non-synaptic boutons, which appear functionally similar to synaptic terminals of DA neurons. Our work presents a new approach to the study of SV release in DA neurons, with implications for understanding DA secretion at the level of single synapses and examining DA synapses in human disease and therapeutics.
All reagents were from Sigma-Aldrich (St Louis, MO), unless otherwise stated.
Pregnant female mice were either imported from the Animal Resources Centre (Perth, Australia) or bred on-site at the Garvan Institute. Glia were prepared from post-natal day 1 mouse pups. Cortical tissue was collected from pups and dissociated in 0.006% Trypsin (in EBSS, Invitrogen, Carlsbad, CA) at 37 °C for 30 min. Tissue was triturated in culture medium consisting of MEM (Invitrogen), 0.1% Mito+ serum extender (BD Biosciences, Franklin Lakes, NJ), 2.94 g/L glucose (final), and 10% fetal bovine serum (FBS, ThermoTrace, USA). The cell filtered through a 70 um cell strainer (BD Biosciences) and cultured in 75cm2 flasks at 37 °C, 5% CO2. After 10-14 days in culture, glial cells were used to seed 25 mm glass coverslips coated with 0.1 mg/ml poly-D-lysine and 0.5 mg/ml rat tail collagen (BD Biosciences).
Our method for culturing midbrain neurons was based on a previously published method (Rayport et al., 1992). Midbrain regions from postnatal day P0 neonatal mice were collected by microdissection. Tissue chunks were dissociated using papain (20 U/ml) in the presence of 10 μg/ml kynurenic acid. Tissue was triturated and plated at 6 × 104 cells/well in 6 well plates containing 25 mm coverslips that had a monolayer of established glial cells (see above). Cells were plated in 1 ml of basic medium (MEM w/10% heat-inactivated iron-supplemented calf serum). And allowed to settle in an incubator (37 °C, 5% CO2) for 1-2 h. 3 ml of neuronal culture medium was then added, consisting of 46.5% MEM, 40% DMEM, 10% Ham's F12, 0.25% Albumax II BSA (Invitrogen), 0.5 mM glutamine, 5.86 mg/ml D-glucose, 6.16 μg/ml catalase (Worthington, Lakewood, NJ), 10 μg/ml kynurenic acid, 125 nM hydrocortisone, 200 nM progesterone, 25 μg/ml insulin, 100 μg/ml apo-transferrin, 30 nM tri-iodothyronine, 30 nM sodium selenite, 21.4 U/ml superoxide dismutase, 15 μm putrescine, 1% heat-inactivated iron-supplemented calf serum. Cells were treated at the time of plating with glial-derived neurotrophic factor (GDNF, Alomone, Israel) at 10 ng ml−1 to promote midbrain neuron survival. Midbrain neurons were left for up to 24 h before striatal cells were added to each well (see below).
As a modification to the method of Rayport et al. (1992), we co-cultured midbrain neurons in the presence of primary neurons taken from the striatum, as striatal neurons constitute a major target for DA neurons of the midbrain. To culture striatal neurons, pieces of striatum were dissected from the brains of P0 neonatal mice. The tissue chunks were then dissociated using papain and triturated as described for midbrain neurons. The striatal cells were then added to culture wells containing midbrain neurons at a density of 1.5 × 105 cells/well. This resulted in a midbrain/striatal neuron co-culture. Cells survived without feeding for several weeks, and were left in culture for at least 15 days before they were used in experiments. For a discussion of our choice of age for our cultures, see ST1a.
For culturing Hpc neurons, the dissection of the hippocampal CA1 region was based on a method previously published (Bekkers and Stevens, 1989). The CA1 of P0 neonatal brains were dissected and then dissociated using papain as described above. CA1 neurons were triturated and then plated at a density of 5 × 104 cells/well on coverslips that had an established monolayer of glial cells. Cells were maintained in 3 ml of neuronal culture medium alone (media contents described above). Neurons were fed weekly by exchanging 1 ml of medium in each well. Hpc neurons were used after at least 15 days in vitro (see ST1a).
Cells were first fixed in 4% paraformaldehyde (with 4% sucrose, in PBS) for 2 min at 23 °C, and then in methanol at −20 °C for 10 min. Blocking was carried out for 30 min using 10% bovine serum albumin (BSA, in PBS). Primary antibodies were: mouse anti-bassoon (1:200, Stressgen, Ann Arbor, MI), rat anti-DAT (1:1000), guinea pig anti-VGLUT1 (1:1000), mouse anti-Map2 (1:500), mouse anti-PSD-95 (1:500), rabbit anti-VMAT2 (1:1000, Millipore, Billerica, MA), rabbit anti-Synaptophysin (1:1000, DAKO, Glostrup, Denmark), mouse anti-Synaptophysin (1:1000), and mouse anti-VAMP2 (1:500, Synaptic Systems, Goettingen, Germany). Primary antibodies were diluted in 3% BSA and applied overnight at 4 °C (except anti-VMAT2, which was applied for 48 h). Primary antibodies were detected using secondary antibodies conjugated to AlexaFluor 488, 594, 647, or Pacific Blue (Invitrogen). These fluorophores were stimulated using lasers with emission wavelengths of 488, 594 633 and 405 nm, respectively.
Cells were imaged in an RC21-BRFS chamber (Warner Instruments) and perfused at 0.5 ml/min. FM 1-43 studies were carried out at 23 °C in Tyrode's buffer (136 mM NaCl, 2.5 mM KCl, 2 mM CaCl2, 1.3 mM MgCl2, 10 mM HEPES, 10 mM D-glucose, pH adjusted to 7.35 with NaOH). FM 1-43 (95 μg/ml, Invitrogen) was delivered to cells for 40 s, followed by depolarization for 100 s using buffer containing FM 1-43 and 90 mM K+. This labeling method was chosen because it gave complete loading of the recycling pool of SVs (see Materials and Methods - Relationship between Pr and FM 1-43 destaining, below). Cells were then perfused with FM 1-43 alone for 90 s and washed for 15 min in buffer containing NBQX (10 μM, Alexis Biochemicals) and AP-5 (50 μM). FM 4-64 (100 μg/ml, Invitrogen) labeling was carried out using the same procedure.
Destaining of FM dye was induced by field stimulation using parallel electrodes (1 ms pulses at 10 V, using S48 stimulator, Grass-Telefactor), 60 s after imaging commenced. FM 1-43 was imaged using the 488 line of an ArKr laser (Leica DMIRE2), while FM 4-64 was imaged using the 514 line. Images were acquired every 5 s. Laser power, line averaging (2 line scans were averaged per frame) and frequency of imaging were optimized to ensure data exhibited low noise while also minimizing photobleaching (which amounted to approximately 10% of total fluorescence over 8 min of imaging). Images were acquired using a Leica 63x, NA 1.5 oil-immersion objective.
After each experiment we noted the cell's location in a photo-etched grid on the coverslip (Bellco) and cells were fixed. The same neuron was located after immunolabeling and imaged. The image of the immunolabeled neuron was overlaid with an image showing FM 1-43 or FM4-64 labeling using Adobe Photoshop CS2.
In order to identify whether a given FM 1-43-positive punctum was VMAT2 positive, it was necessary to examine co-labeling between VMAT2 and FM 1-43. Likewise, to determine whether sites were synaptic or non-synaptic, Map2 labeling was examined in these same images. As such, FM 1-43-positive sites were defined as being VMAT2-positive and synaptic/non-synaptic before fluorescence at these sites was quantified as described below.
ImagePro v6 (Media Cybernetics, Bethesda, MD) was used for counting fluorescent puncta and distance measurements. FM 1-43/4-64 intensity was measured using Leica Microsystems software. To study FM 1-43/4-64 fluorescence changes at labeled sites, a region of interest (ROI) was drawn around each punctum. Fluorescent puncta were selected for analysis based on fluorescence loss corresponding to field stimulation. ROIs were circular (1 μm diameter), and positioned with the centre of each ROI occupied by the centre of the punctum. The mean fluorescence intensity (measured in ‘units’) within each ROI was then determined. All suitable puncta were analyzed from each experiment. This varied widely, with each experiment providing between 1 and 22 individual puncta to the final data set. All experimental parameters were kept identical across experiments to minimize inter-experimental variation in the data.
For each presynaptic terminal, a series of 85 separate fluorescence measurements were made at 5 s intervals (7 min total imaging). During this period 10 Hz stimulation mediated a loss of fluorescence. In order to analyze this fluorescence loss, we modeled a first order decaying exponential to fit all 85 data points (see ST2 – Data Modeling and Analysis). This procedure was individually performed for each terminal. The validity of this first order fit is discussed in ST1b. Note that where 2 Hz stimulation was used, 133 data points were acquired over 11 minutes. Modeling was then carried out in the same way as for data acquired during 10 Hz stimulation.
We calculated three key parameters directly from the model curve at each presynaptic site: Fir, ΔF and τ, as described in ST2a. Fir was calculated as the value of the slope of the tangent to the destaining curve at t = 0 s. The validity of using this approach to study the initial rate of destaining is discussed in ST2b. ΔF was calculated from the model curve as the difference between the fluorescence intensity at t = 0 s and t = ∞. Our method of calculating ΔF is not affected by residual fluorescence that is evident after destaining (see ST1c). τ was also calculated directly from the model curve.
For all data presented, at least 3 separate coverslips were analysed. In general, each coverslip was only used for a single experiment, except on three occasions where a single coverslip was used for two experiments. The maximum number of presynaptic boutons included in a data set from a single coverslip was 23 (out of a total data set of 81 boutons, acquired across 5 separate coverslips in total).
Where we accepted the null hypothesis, we state the P value for the test, and where we rejected the null hypothesis, we indicate that the P value was below the level set for significance (e.g. P < 0.05). The level of significance was set at P < 0.05, except for correlation analyses, where the significance level was set at P < 0.01. Statistical analyses were carried out using SPSS Graduate Pack v.13 (SPSS, Chicago, IL) or R (R Development Core Team, 2007).
Our method for comparing the fluorescence parameters Fir, ΔF and τ took account of two features of the data. First, these parameters were generally not normally distributed, as determined using the Shapiro-Wilk normality test. Second, each data set consisted of groups of observations from a number of different experiments, so that the observations were clustered within experiment. The first feature suggests that a rank-based method is appropriate, whilst the second feature indicates the use of a method that allows for potential within-experiment correlation. We used a method that has been developed to meet these two requirements (Datta and Satten, 2005). The method is a modified version of the Wilcoxon rank-sum test that is valid for clustered data.
Similarly, the method that we used to assess equality of variances was a modified version of Levene's test. Our approach uses a robust (sandwich) variance estimator in the ANOVA on which Levene's test is based, which allows for the clustering of observations within each independent experiment, and hence address potential variability between experiments. This method was developed recently (Iachine et al., 2009).
Correlation analyses were performed using Spearman's ρ test.
In our studies, we analyze the kinetics of destaining of FM dyes from synaptic and non-synaptic sites to study the heterogeneity of Pr. In order to achieve this we developed a rapid approach to compare relative Pr at different release sites, which we describe below.
In order to compare the SV release properties of individual synaptic terminals, we derived a mathematical relationship between FM dye destaining kinetics and Pr. We first assume that if sufficient events could be analyzed, then, for a given stimulation frequency, probability of release (Pr) could then be estimated from the frequency of SV fusion events according to the following:
Where Pr is the probability of release; N is the frequency of SV fusion events (number of fusion events / time); Where N is measured in exocytosis events/s This relationship is based on that previously described by Murthy et al. (1997).
We next assume that the likelihood of an FM dye-labeled SV undergoing exocytosis due to electrical stimulation (and thus detectable loss of fluorescence) is dependent on the proportion of total SVs that are labeled with FM dye. Therefore:
Where Prf is the probability of an FM dye-labeled SV undergoing exocytosis, Pr is the probability of an exocytic event at a given synapse, Vf is the number of SVs that are fluorescently labeled with FM dye, and VT is the total number of recycling SVs.
Our loading protocol employed a single exposure to hyperkalemic buffer for 100 s in the presence of FM dye, which is sufficient to label all recycling SVs in Hpc neuron terminals (Harata et al., 2001; Mozhayeva et al., 2002). We found that increasing the time of exposure to hyperkalemic buffer did not increase the fluorescent labeling of presynaptic terminals in Hpc or midbrain neurons (data not shown). Furthermore, Mozhayeva et al. (2002) found that multiple successive exposures to hyperkalemic conditions did not result in increased FM 1-43 labeling. This demonstrated that a single exposure to hyperkalemic buffer was sufficient to label all recycling SVs. The study by Harata et al. (2001) also found that the degree of presynaptic labeling was the same whether hyperkalemic buffer or 10 Hz electrical stimulation was used to label recycling SVs with FM 1-43. Therefore, our loading protocol labeled all recycling SVs.
Given our assumption that all SVs are labeled at the onset of electrical field stimulation (see above), then, at t = 0, then Vf is very close to VT, and Vf/VT ~ 1. This leads to: Pr = Prf (at t = 0; where t = 0 refers to the onset of FM dye destaining).
The remainder of our derivation therefore only applies at t = 0, i.e. the onset of destaining, when SVs within presynaptic terminals are fully labeled with FM dye. These SVs are then stimulated to undergo exocytosis and recycling by the application of electrical field stimulation. Therefore, frequency of SV fusion events is proportional to the change in fluorescence over time at the onset of FM dye destaining:
Where Floss is fluorescence loss due to an SV fusion event (in fluorescence units); c is the fluorescence intensity per single FM dye-labeled SV (in fluorescence units); t is the time period analyzed.
We assume that FM dye destaining kinetics follow first order exponential decay (see ST1b). In our experiments, we use data modelling to derive a first-order decaying exponential curve that describes the fluorescence data at each synapse. Specifically, we use a mixed linear model to fit a first order decaying exponential curve to the data obtained at each neurotransmitter release site. We then mathematically derive the slope of the tangent to this model destaining curve at t=0 s (the onset of FM dye destaining). The derivation of this tangent is described in ST2a, and the rationale behind using this tangent as an estimate of the rate of fluorescence loss is discussed in ST1d and ST2b. The slope of the tangent to the exponential at t = 0 describes the initial rate of decay of the curve. This initial rate of decay is a measure of the initial rate of fluorescence loss at the onset of field stimulation, and can be written simply as an expression of fluorescence loss over time:
Where Fir is the slope of the tangent to our first-order destaining exponential at t = 0 s (note that the slope value is always negative). If we substitute this expression into (3), we derive an equation that relates Fir and Pr at the onset of FM dye destaining:
This demonstrates the initial rate of fluorescence loss, Fir, is proportional to the Pr at a given synaptic terminal. Based on this mathematical relationship, we used Fir as a relative measure of Pr at synaptic terminals. It should be noted that relative Pr can only be compared across terminals that were all stimulated at the same rate, e.g. 10 Hz. Fir values obtained at different stimulation frequencies cannot be directly compared.
At later time points during destaining, unstained SVs will re-enter the recycling pool, thereby decreasing the proportion of all vesicles that are stained (i.e. Vf/VT moves away from 1). In addition, at later time points Pr could be affected by repeated neuronal stimulation, i.e. plasticity. Therefore, we assume that the rate of fluorescence loss is only an accurate measure of relative Pr at t = 0 (the onset of FM dye destaining). For a discussion of the importance of only examining the initial rate of fluorescence loss in studying Pr, and how our model is able to examine relative Pr even at stimulation rates likely to induce short-term plasticity, see ST1d.
Styryl dyes such as FM1-43 have been extensively used to identify and analyze rapidly recycling, small neurotransmitter-containing vesicles (Murthy, 1999), which we will refer to as synaptic vesicles (SVs). In this study, we were interested to use FM 1-43 to assess the functional heterogeneity of DA release sites. As our first step, we characterized the basic properties of DA release sites in DA neurons in vitro.
We stained recycling SVs in DA neurons with FM 1-43 as described in the Methods. We then imaged the FM 1-43-labeled neurons (labeling shown in Fig. 1a), and induced destaining of FM 1-43 by electrical stimulation at 10 Hz. The neurons were subsequently fixed and labeled using an antibody directed against the vesicular monoamine transport protein VMAT2, which is localized to vesicles in DA neurons (Fig. 1b, d-f). Antibodies against Map2, dopamine transporter (DAT) and bassoon were also applied to identify dendrites, axons and active zones in DA neurons, respectively.
We then overlaid images of the immunostained neurons on the same neuron previously labeled with FM1-43 (Fig 1d-f). This allowed us to identify which FM1-43 sites were also VMAT2 positive. We classified VMAT2-positive sites as ‘synaptic’ (i.e. belonging to a synapse with pre- and postsynaptic elements) if it was < 0.35 μm from a Map2 positive dendrite (as defined by Krueger et al., 2003). Sites between 0.35 and 2 μm from a dendrite were not analyzed. Not all VMAT2-positive sites exhibited FM 1-43 staining. The majority (64. 4 ± 6.3%) of all VMAT2-positive sites that were defined as synaptic exhibited co-labeling for FM 1-43 (n = 649 sites, 9 experiments), demonstrating that SV recycling was apparent at most synaptic boutons of DA neurons.
We then focused specifically on VMAT2-positive sites that exhibited FM 1-43 labeling. Of all FM 1-43-labeled VMAT2-positive sites, the majority (62.0 ± 4.4%) were synaptic (n = 659 total FM 1-43/VMAT2-labeled sites, 13 experiments) while 19.3 ± 3.2% were defined as non-synaptic (located >2 μm from a dendrite). Thus, DA neurons not only possessed active synaptic terminals but also a subpopulation of non-synaptic boutons, which contained recycling SVs and were therefore presumably capable of neurotransmitter release.
A number of regulatory proteins are localized to SVs in synaptic terminals (Sudhof, 2004). We examined the presence of these regulatory proteins at VMAT2 positive synaptic and non-synaptic sites by immunolabeling of DA neurons. We found that VMAT2-positive sites exhibited immunolabeling for the SV proteins VAMP2 and Synaptophysin (Syp) (Fig. S1). While most VMAT2-positive sites co-labeled with SV proteins VAMP2 (97.3 ± 0.5%) and Syp (88.4 ± 1.7%), only a subpopulation of total VMAT2-positive sites co-labeled for bassoon (39.1 ± 4.1%). These synaptic proteins are characteristically found at mature synapses in Hpc neurons (Krueger et al., 2003). In addition, it was evident from this analysis that not all presynaptic structures in the culture were VMAT2-positive (Fig. S1). These VMAT2-negative sites were presumably formed by non-DA neurons, in particular GABA neurons, which are abundant in the striatum and midbrain. This is also true of sites that were FM 1-43-positive but VMAT2 negative in live cell experiments (see Fig. 1).
It is notable that, although most VMAT2-positive sites did not exhibit bassoon labeling, almost all VMAT2-positive sites that exhibited FM 1-43 labeling also co-labeled for bassoon. 94.5 ± 3.2% of synaptic sites (n = 88 sites from 12 experiments) were immunopositive for bassoon. Furthermore, all FM 1-43/VMAT2-labeled non-synaptic sites (n = 24 sites) were also positive for bassoon. Since bassoon is a marker of active zones, these findings suggested that release of DA primarily occurs at sites that possess active zones in DA neurons.
Conventional synapses contain both a presynaptic element, formed by an axon terminal, and a specialized postsynaptic element, generally formed by a dendrite. To investigate whether any VMAT2-positive presynaptic terminals were associated with postsynaptic specializations, we next examined whether VMAT2-positive puncta exhibited any co-localization with markers of postsynaptic specialization. D1 and D2 receptors are believed to be the primary target for synaptically released DA, however there is evidence that they commonly exhibit extrasynaptic localization (Sesack et al., 1994; Yung et al., 1995). This suggests that these receptors are not a robust marker of postsynaptic specialization. Therefore, to identify postsynaptic specializations we used antibodies directed against postsynaptic proteins known to be involved in the regulation of D1 receptor signalling: the NMDA receptor subunit NR1 and the protein PSD-95 (Harvey and Lacey, 1997; Fiorentini et al., 2003; Zhang et al., 2007). By this approach, we hoped to identify at least some of the postsynaptic sites associated with presynaptic terminals of DA neurons.
Immunolabeling against the postsynaptic proteins PSD-95 and NR1 showed that a subpopulation of VMAT2-positive sites exhibited co-localization against these markers (18.9 ± 3.3% for NR1, n = 1812 sites from 5 neurons; 20.7 ± 3.0% for PSD-95, n = 2186 sites from 8 neurons). This suggested that a subpopulation of all VMAT2-positive puncta were localized to sites that had postsynaptic specializations. However, there was no indication from this data whether the VMAT2-positive sites identified were functional.
In order to examine whether postsynaptic specializations were present at functional VMAT2-positive sites, we next used FM 1-43 labeling and retrospective immunocytochemistry to examine the localization of PSD-95 at sites that labeled both for FM 1-43 and VMAT2-positive (Fig. 2). We found that 46.6 ± 2.4% of all FM 1-43/VMAT2-positive sites (n = 93 sites analysed from 3 experiments) were co-localized with PSD-95-positive postsynaptic specializations, indicating that almost half of the functional presynaptic terminals in our culture system belonged to mature synapses with PSD-95-positive postsynaptic specializations.
Overall, our findings suggested that our DA neuron culture system contained mature presynaptic and postsynaptic structures. Based in part on these findings, we concluded that neuron cultures that were aged at least 15 days in vitro formed mature synapses. The reasons for using neurons of this age are further elaborated in ST1a.
Our next aim was to examine SV exocytosis at individual presynaptic sites in cultured DA neurons. In order to achieve this, we first labeled vesicles in midbrain neurons with FM 1-43. We next applied a 10 Hz field stimulation to induce vesicle exocytosis, and hence FM 1-43 destaining. We opted to use 10 Hz stimulation for destaining as the kinetics of FM dye destaining at this frequency has been extensively studied in Hpc neurons (for examples see: Mozhayeva et al., 2002; Krueger et al., 2003; Virmani et al., 2006). During destaining, we acquired 85 separate measurements of average fluorescence intensity at each terminal, taking one image each 5 s for 7 min. This was followed by post-hoc immunostaining for VMAT2, Map2 and DAT and bassoon. We then overlaid the image of the immunostained neuron on the image of the same neuron labeled with FM 1-43. Destaining data was then only analyzed at those sites later defined as dopaminergic by virtue of the fact that there was co-localization of FM1-43 and VMAT2, in a neuron that was also imunolabeled for DAT. Synaptic and non-synaptic sites were defined as described above.
We observed that at each VMAT2-positive synaptic site that exhibited FM 1-43 labeling, fluorescence was initially lost rapidly during stimulation, after which the rate of fluorescence loss gradually decreased over time (Fig. 3a). The FM 1-43 destaining kinetics exhibited considerable heterogeneity across all synaptic DA terminals (see Fig. 3b), suggesting that the amount of FM 1-43 destaining, and the kinetics of destaining, were highly heterogeneous across the entire population of DA terminals. Normalization of fluorescence data against ΔF, a common practice in studies that use optical methods to examine presynaptic function (for examples see Mozhayeva et al., 2002; Krueger et al., 2003; Virmani et al., 2006), decreased the apparent heterogeneity (Fig. 3c) in the destaining profiles. Although normalization against ΔF is commonly carried out, for our purposes normalization actually removed information that was important for the analysis of Pr (see Equation 1 below).
The heterogeneity of destaining kinetics suggested heterogeneity of Pr at DA release sites. We were therefore interested to analyze the extent of the heterogeneity of Pr at these sites. In order to do this, we developed a rapid approach to compare relative Pr at different release sites. In our approach, we take advantage of the fact that Pr is proportional to the initial rate of fluorescence loss at a given presynaptic site (see Methods - Relationship between Pr and FM 1-43 destaining). An example of Fir is shown in Fig. 4a. Since initial rate of fluorescence loss is proportional to initial Pr, it follows that by comparing initial rate of fluorescence loss at different neurotransmitter release sites we can gain insight into the relative initial Pr at these different sites. Simply put, a faster rate of fluorescence loss at t = 0 s reflects a higher initial Pr. For further discussion of the relationship between initial rate of fluorescence loss and Pr; please refer to ST1d and ST2.
In order to measure the initial rate of fluorescence loss, we assume that destaining follows first order kinetics (as discussed in ST1b). We then use a mixed linear model to fit a first order decaying exponential curve to the 85 separate measurements of average fluorescence intensity obtained during destaining for each release site, which provides an estimate of the entire destaining curve for a given presynaptic site. Subsequently, we mathematically derive the slope of the tangent to this model destaining curve at t = 0 s, which we refer to as Fir, for each site (for a discussion of measuring the slope of this tangent as a measure of Fir, see ST1d and ST2b). This value of Fir thus provides a relative measure of Pr at each FM 1-43 labeled site. It is particularly important to note that our method is only designed to examine relative Pr at the onset of destaining (i.e. initial Pr, at t = 0 s), as discussed in ST1d.
Based on the above, we used Fir as a relative measure of Pr at presynaptic sites. In order to assess whether Pr was heterogeneous at DA synapses we first examined Fir across many individual synaptic VMAT2-positive sites. Fir exhibited a broad distribution, suggesting that Fir was heterogeneous across synaptic sites of DA neurons (n = 106 total sites from 11 experiments, Fig. 4b), and thus that Pr was heterogeneous across DA synapses.
We next compared DA terminals with terminals of Hpc neurons. Specifically we were interested to compare Fir at DA boutons with Fir at Hpc boutons, as much research has been carried out previously examining Pr and SV release in Hpc neurons. Since Fir is proportional to Pr, such a comparison provides insight into the relative Pr at these two populations of synaptic terminals.
Synaptic boutons in Hpc cultures were characterized using retrospective immunolabeling with anti-Syp immunolabeling (n = synaptic 81 sites from 7 experiments, Fig. S2a), although to compare DA terminals specifically to glutamatergic sites we used anti-VGLUT1 in some experiments (n = 42 sites from 5 experiments, Fig. S2b). In these experiments, most Syp-positive sites co-labeled for VGLUT1 (66.7 ± 8.8%, n = 55 Syp-positive synaptic sites from 5 experiments analyzed) and were therefore glutamatergic. Synaptic sites were defined based on their proximity (< 0.35 μm) to a Map2-positive structure.
Synaptic sites of DA neurons exhibited a slower initial rate of fluorescence loss compared to Hpc synaptic sites using either Syp (P < 0.01) or VGLUT1 (P < 0.02) to define Hpc terminals ( Fig. 4c). This indicated that DA neurons generally exhibited lower Pr than Hpc neurons at synaptic sites.
It is well characterized that Pr is highly heterogeneous in synapses of Hpc neurons (Hessler et al., 1993; Rosenmund et al., 1993; Murthy et al., 1997). We were therefore interested to compare the degree of heterogeneity in Pr between DA and Hpc synapses. To examine this, we compared the variance in Fir in synaptic sites for these two types of neuron. We observed no significant difference in the variance of Fir between these two types of synapse. This suggests that Pr at DA synapses, like Hpc synapses, is heterogeneous, varying from synapse to synapse.
We next examined factors underlying Pr at DA terminals. In Hpc neurons, Pr is proportional to the number of SVs in the readily releasable pool (Dobrunz and Stevens, 1997; Murthy and Stevens, 1999; Murthy et al., 2001; Waters and Smith, 2002), which is in turn proportional to the total number of recycling SVs, called the recycling pool (Murthy and Stevens, 1999). Thus, the size of the recycling pool indirectly regulates Pr. We were therefore interested to determine if the heterogeneity of Pr we observed in DA neurons could be accounted for by variations in vesicle pool size between synapses. If so, then it would be predicted that a strong correlation would exist between Fir and the total vesicle pool size.
In order to determine whether this correlation exists, we needed to first obtain a measure of the relative recycling pool size at each neurotransmitter release site. To achieve this, we measured the total fluorescence loss during destaining (ΔF). ΔF was calculated from the first order model curve at each site as the difference between the fluorescence intensity at the final fluorescence intensity after destaining is complete (see ST2a for calculation method), such that:
Where Finit = the initial FM 1-43 fluorescence intensity and Fresidual is the residual fluorescence observed after destaining is complete. ΔF is therefore a measure of the total vesicle pool size, and is independent of variation in the residual (unreleasable) fluorescence at synaptic terminals (see ST1c for further discussion regarding residual fluorescence and the calculation of ΔF).
We investigated whether the heterogeneity of Fir, and thus Pr, could be explained by heterogeneity in ΔF. We determined ΔF from the model curve at each synaptic DA site (shown in Fig. 4d, calculation method described in ST2a). ΔF, which reflects recycling pool size, was heterogeneous at DA synaptic sites (Fig. 4e). Furthermore, the variance in ΔF was not significantly different between DA and Hpc synapses, suggesting that the degree of heterogeneity in SV pool size across presynaptic terminals was similar between these two types of neuron.
When values of ΔF were compared between DA and Hpc neurons, we observed that synaptic sites of DA neurons exhibited a lower ΔF than Hpc synaptic sites (Fig. 4f), using either Syp (P < 0.01) or VGLUT1 (P < 0.05) to define Hpc terminals, indicating that synaptic DA terminals contained fewer recycling SVs than Hpc glutamatergic terminals.
According to our decaying exponential model (See ST2a), Fir is related to ΔF by the equation:
where τ is the time constant of fluorescence decay for a given synapse.
Equation 1 shows that Fir is directly related to ΔF. Indeed, as predicted from this equation, there is a correlation between Fir and ΔF at individual synaptic DA sites (Spearman correlation co-efficient, ρ = −0.41, Fig. 5a). This relationship between ΔF and Fir was consistent with previous observations that recycling SV pool size was correlated with Pr (Murthy et al., 1997; Murthy et al., 2001). Our findings suggest that the heterogeneity of Pr was at least partly due to heterogeneity of recycling SV pool size in DA neurons. We also found a correlation between Fir and ΔF at Hpc synaptic sites (ρ = −0.43). This is consistent with previous findings in Hpc neurons (Dobrunz and Stevens, 1997; Murthy and Stevens, 1999; Murthy et al., 2001; Waters and Smith, 2002).
However, given the direct relationship between ΔF and Fir (Equation 1), the observed correlations were not as strong as anticipated, suggesting an additional mechanism regulating Pr. According to Equation 1, Fir is also determined by τ. If τ were constant between all terminals, then Fir would be solely determined by ΔF, with an anticipated correlation co-efficient of −1. However, if τ were variable the correlation between ΔF and Fir would be weaker. We therefore examined whether variation in τ accounted for the relatively weak correlations observed.
Values of Fir and ΔF from synaptic DA sites were grouped according to their value of τ (Fig. 5b). We then examined whether sites with similar τ values exhibited a stronger correlation between ΔF and Fir than the overall ungrouped correlation (ρ = −0.41). When sites were grouped according to percentiles of τ, within each percentile the correlation co-efficient, ρ, was very high, generally close to −1 (individual values shown in Fig. 5b). Therefore, the relatively weak correlation we observed between ΔF and Fir, evident in the scatter of data in Fig. 5a, was attributable to variation in τ across presynaptic terminals, as illustrated in Fig. 5b. Similar results were also obtained in Hpc terminals (Fig. 5c). This observation firstly demonstrated the direct relationship between ΔF and Fir, as defined by our first-order model. Secondly, our results showed that τ was not constant, but rather was variable across DA and Hpc terminals. According to Equation 1, the value of τ plays a direct role in determining Fir. Therefore our data suggests that τ reflects a heterogeneous factor that contributes to Pr, and that Pr could not be attributed solely to variations in SV pool size.
An example of τ is shown in Fig. 5d. When we plotted the τ values from a population of synaptic terminals of DA neurons, τ exhibited a broad, non-normal distribution with a skew towards 0 (Fig 5e), similar to the distribution observed for Fir. We propose that the heterogeneity of τ reflects variability of SV pool release kinetics between individual boutons (see ST1e).
When we examined the variance of τ between DA and Hpc synaptic boutons we observed that there was no significant difference between these populations. This suggests that although the kinetics of SV pool release is heterogeneous across both DA and Hpc synapses, the degree of heterogeneity is similar in these two types of neurons. The heterogeneity of release kinetics, of which τ is a measure, has been documented previously in Hpc neurons (Smith and Waters, 2002).
Previous research has assumed a relationship between τ and Pr in which τ was simply determined by Pr (Krueger et al., 2003). According to these previous assumptions, the SV pool size would influence Pr, which in turn would determine τ. This idea makes some intuitive sense, in that synapses with higher Pr may be expected to release SVs at a faster rate, therefore exhibiting a smaller value of τ.
However, the model presented by Equation 1 suggests that this previous assumption that Pr determines τ is not correct. According to Equation 1, Fir (which is a measure of Pr) is determined by both ΔF and τ. To clarify this relationship between τ and Fir, we reasoned that if SV pool size determines Pr, then in our model ΔF must determine Fir. Following on, if ΔF determines Fir, and Fir determines τ, then ΔF must correlate with τ. However, we found no significant correlation between ΔF and τ at DA terminals (ρ = 0.161, P = 0.10, Fig. 5f). This demonstrated firstly that τ could not be determined by Fir. Rather, this lack of correlation supported our model, in which Fir is determined by both ΔF and τ. This suggests that τ reflects a distinct parameter that regulates Pr. Interestingly, this finding indicates that the size of the recycling SV pool and the kinetics of SV pool release (as measured by τ) were not related, and furthermore that the factors reflected by ΔF and τ regulate Pr independently of each other. We will address this point further, below (Fig. 7).
According to our model of fluorescence loss at synaptic terminals, Fir is determined both by the value of ΔF and the value of τ at the level of individual terminals. Having observed that values of ΔF were significantly different between DA and Hpc synapses, we next examined whether values of τ were different in these two populations of synapses. We detected no significant difference between values of τ in these two types of neuron (Fig. 5g), using either Syp or VGLUT1 to define Hpc terminals. This demonstrated that although τ is highly heterogeneous, there is no significant difference in the kinetics with which SV pool release occurs in DA and Hpc neurons.
Overall, our findings demonstrate that presynaptic terminals of DA neurons are functionally heterogeneous, exhibiting considerable variation in Pr, SV pool size and SV pool release kinetics, as measured by Fir, ΔF and τ respectively, at the level of individual synapses. Our findings suggested that the degree of variation was similar to that observed in Hpc neurons, in which functional heterogeneity has been well documented. In addition, DA synapses exhibit lower Pr than Hpc synapses, attributable to their lower number of recycling SVs, suggesting that DA synapses are more reluctant to release neurotransmitter than Hpc synapses.
When examining VMAT2-positive sites that exhibited FM 1-43 labeling, we observed that DA neurons formed not only synapses but also a subpopulation of non-synaptic release sites (refer to Fig. 1). In order to further examine these structures, we studied FM 1-43 fluorescence loss at non-synaptic sites in DA neurons (Fig. 6). These non-synaptic boutons were functional structures, containing active zones and recycling SVs that label with FM 1-43, but with no apparent postsynaptic dendrite. Destaining at non-synaptic sites exhibited first order kinetics.
We examined Fir at non-synaptic boutons (n = 31 sites from 11 experiments, Fig. 6). We found that the values of Fir were not significantly different between synaptic and non-synaptic boutons. Therefore, there was no detectable difference between Pr at these two types of structure. Similar results were observed when we stimulated DA neurons at 2 Hz to destain synapses (see Fig. S3), suggesting that our observations using 10 Hz stimulation were not stimulation dependent (see ST1d). Overall, these findings suggest that synaptic and non-synaptic sites are functionally similar in terms of their capacity for SV exocytosis.
Decreased extracellular Ca2+ has been shown to decrease Pr, as measured using FM 1-43, in Hpc neurons (Murthy et al., 1997). To confirm that our method of analyzing relative Pr is sensitive to conditions known to alter Pr, we analyzed Fir in the presence of decreased extracellular Ca2+. We found that decreasing buffer Ca2+ from 2 mM to 0.5 mM resulted in a significant decrease in the initial rate of fluorescence loss at synaptic sites of FM 1-43 labelling in Hpc neurons (Fig. 7), reflecting a reduction in Pr. This finding confirms that our method is sensitive to detecting alterations in Pr.
We also observed that decreased extracellular Ca2+ resulted in an increase in the value of τ at Hpc boutons in 0.5 mM Ca2+ (Fig. 7). Decreased Ca2+ did not alter ΔF, however. This would be expected as ΔF reflects the size of the recycling pool and would not necessarily be expected to change when Ca2+ concentrations are altered.
Our data obtained using decreased extracellular Ca2+ helps to illustrate how our model of destaining works. By decreasing the Ca2+ level from 2 mM to 0.5 mM, the τ value at presynaptic sites was increased. In accordance with Equation 1, this increase is τ resulted in a decrease in the initial slope of the destaining curve (Fir). Given that Fir reflects Pr, our findings suggest that τ reflects a cellular factor, or factors, that regulate Pr and that these factors are sensitive to changes in Ca2+ concentrations. These observations also emphasize that SV pool release kinetics reflect yet-to-be defined factors that regulate Pr independently of SV pool size.
We analysed presynaptic function in DA neurons using FM 1-43. FM 1-43 labeling was evident at over 60% of VMAT2-positive synaptic boutons. However, we observed that FM 1-43 labeling was evident at an even higher proportion of synaptic sites in Hpc neurons (82.8 ± 2.3%). This may suggest that a relatively high proportion of sites containing SVs in DA neurons are not release sites. Our findings also indicate that active zones are necessary for DA release, as they are in other neuron types (Rosenmund et al., 2003). VMAT2 co-localized with SV proteins VAMP2 and Syp, and almost half of all functional VMAT2 puncta co-localized with postsynaptic specializations, suggesting that DA neurons form mature synapses under our culture conditions (see ST1a).
DA axons possessed non-synaptic boutons containing recycling SVs and active zones. Synaptic and non-synaptic boutons exhibited similar FM 1-43 destaining properties, and thus appear equally capable of DA release. Previous research in Hpc neurons suggests that although dendrites can play a role in presynaptic maturation during synaptogenesis (Ahmari et al., 2000), neurons also possess non-synaptic sites that are functional and mobile (Krueger et al., 2003). While their physiological relevance is unknown, non-synaptic sites in our cultured DA neurons are reminiscent of non-synaptic varicosities in DA axons in the brain (Descarries et al., 1996), whose role in neurotransmission remains unclear. This is a highly interesting area for future investigation.
Our modeling and analysis of styryl dye destaining showed mathematically that the rate of fluorescence loss at t = 0 s (Fir) is proportional to initial Pr. As long as destaining data fits a first order model, an assumption we have justified (see ST1b), then we can obtain a measure of Fir and hence of relative Pr, by taking the slope of the tangent to the model curve at t = 0 s. However, our approach does not exclude alternative methods of data modeling, so long as the initial rate of fluorescence loss is estimated accurately. This method is only designed to examine relative Pr at the onset of destaining (t = 0 s), and it is only possible to compare Fir for a specified stimulation frequency. Furthermore, providing the first order fit is valid, our analysis of relative Pr is not affected by subsequent short-term plasticity (see ST1d).
By analysing Fir, we observed that initial Pr was heterogeneous in DA neurons, consistent with previous observations of Pr in Hpc neurons (Rosenmund et al., 1993; Dobrunz and Stevens, 1997; Ryan et al., 1997). Although Pr was heterogeneous, DA terminals generally exhibited lower Pr than Hpc terminals. Furthermore, the degree of Pr heterogeneity that we observed in Hpc neurons (coefficient of Fir variation = 0.66) fell within the range observed in Hpc neurons by Murthy et al. (1997), suggesting that our method exhibits similar sensitivity and robustness in analyzing Pr heterogeneity to previous methods (see ST2b). Interestingly, when pooled Fir values from DA synapses are plotted as a single histogram (Fig. S4), the shape of the plot, with its broad distribution and skew towards 0, resembles the distribution of Hpc Pr published by Murthy et al. (1997).
ΔF, a relative measure of recycling SV pool size, was also heterogeneous across synaptic terminals of DA neurons, similar to previous studies in Hpc neurons (Ryan et al., 1997; Moulder et al., 2007). DA terminals generally exhibited lower ΔF than Hpc neurons, suggesting a smaller pool of recycling SVs, to which we attribute the lower Pr at DA terminals. Previous observations in Hpc neurons have shown that SV pool size correlates with Pr (Murthy et al., 1997; Murthy et al., 2001), which is consistent with our findings. The recycling pool in Hpc neurons is estimated between 20 and 45 SVs using FM 1-43-based techniques (Murthy et al., 1997; Murthy and Stevens, 1999; Harata et al., 2001). We infer from our data that DA terminals in vitro contain between 15 and 34 recycling SVs on average, since ΔF in DA terminals (mean ΔF = 39.97 ± 1.62 units) is about 77% that of glutamatergic synapses (mean ΔF = 52.33 ±3.57 units). In drawing this conclusion, we assume that SVs are the same size in DA and Hpc neurons (see ST1f), and that large dense-core vesicles (LDCVs) do not contribute significantly to FM 1-43 loss at DA terminals (see ST1c).
τ was also heterogeneous at DA terminals, similar to observations at Hpc synapses (Smith and Waters, 2002). We found that in a first order model of fluorescence loss, the initial rate of fluorescence loss, Fir, was defined by both ΔF and τ. Hence, we propose that τ measures factors regulating Pr independently of SV pool size, and that heterogeneity of τ across synapses reflects innate variability in these factors (ST1e). A discussion of our measurement of τ compared to previous values is provided in ST1g.
We found no difference in τ between DA and Hpc synapses, demonstrating that although DA synapses contain a smaller recycling SV pool, this pool undergoes release with similar kinetics to Hpc synapses. This may suggest that there is no difference in the cellular factors that underlie τ between these neuron types.
Our findings further demonstrated that ΔF and τ are independent factors. This is shown firstly through the lack of correlation between these parameters, and secondly through our finding that τ, but not ΔF, was altered when extracellular Ca2+ was decreased. Waters and Smith (2002) demonstrated that phorbol dibutyrate treatment increases Pr without altering recycling pool size, which, interpreted in the context of our findings, supports our conclusion that Pr can be regulated independently of SV pool size. Furthermore, our observations suggest that the kinetics of SV release, reflected by τ, are independent of the number of releasable SVs present at a presynaptic terminal.
Our method provides an approach to identify factors that regulate Pr independently of vesicle pool size, i.e. factors that regulate τ. At this stage, the factors underlying τ are somewhat unclear. We show that τ is influenced by extracellular Ca2+, and that this change in τ in turn alters Fir. This suggests that τ, in part, reflects molecular factors that are sensitive to Ca2+ and regulate Pr. This could include Ca2+-sensitive molecules such as synaptotagmin I (Fernandez-Chacon et al., 2001) and synapsin I (Chi et al., 2003). In Hpc neurons, it is unlikely that τ reflects the probability of SV fusion (Pv), since in these neurons Pv is thought to be a constant (see ST1h). However, it is unknown whether Pv is variable across DA synapses. Thus, in DA neurons Pv may contribute to variation in τ, and hence to variation in Pr. It has also been suggested that τ reflects the proportion of SVs localized in the RRP (Waters and Smith, 2002).
Plasticity occurs at Hpc synapses during 10 Hz stimulation, the predominant form of which is depletion, in which Pr declines as the RRP becomes depleted of SVs (Dobrunz and Stevens, 1997). A recent study suggested that synapses that exhibit greater depletion at 10 Hz stimulation subsequently exhibit slower FM dye destaining kinetics (Virmani et al., 2006), suggesting that τ may incorporate a factor defining plasticity during intense stimulation (discussed in ST1d).
GABA terminals exhibit less heterogeneity in SV pool size when compared against glutamatergic terminals (Moulder et al., 2007). The authors suggested that GABA terminals may therefore be less susceptible to synaptic plasticity. When viewed in the context of the conclusions of Moulder et al. (2007), our findings that the degree of functional heterogeneity was similar between DA and Hpc neurons may suggest that these neuron types are similarly susceptible to plasticity. However, Hpc terminals with relatively low Pr tend to undergo greater short-term facilitation in response to paired-pulse stimulation (Murthy et al., 1997). Assuming that this phenomenon holds true at DA terminals, then DA synapses, which have relatively low Pr, may be susceptible to short term facilitation. DA synapses may also be susceptible to RRP depletion, resulting in a decrease in neurotransmitter secretion during intense stimulation, assuming that Pr and rate of depletion are correlated in DA neurons, as has been shown at Hpc synapses (Dobrunz and Stevens, 1997). Thus, further investigation is necessary to examine these possibilities.
In conclusion, we describe a new approach to the study of relative Pr across a population of presynaptic terminals. We find that DA synapses are functionally heterogeneous. We describe previously uncharacterized differences in the regulation of SV exocytosis between DA neurons and glutamatergic Hpc neurons, while by contrast synaptic and non-synaptic boutons in DA neurons are not different.
Drugs acting at DA synapses are used to treat neuropathologies such as PD and schizophrenia. There is also emerging evidence that synaptic function itself may be altered in PD (Serulle et al., 2007). Our approach to studying DA synapses allows further examination of signalling mechanisms that regulate DA secretion, the role of presynaptic function in disease and examination of the effects of therapeutic drugs on synaptic function.
This work would not have happened without Amadeus Energy Ltd, Mr Bill Gruy and Mr Geoffrey Towner. We thank Nick Kell and Joanna Knott for their support, Dr. T. Lewis, Dr. M. Christie and Dr. J. Clements for their insightful comments regarding data analysis, and Dr. S. Janmaat and Dr. J. Bekkers for their generous overall assistance. Dr. D. Sulzer provided the methods for culturing dopamine neurons. Dr. J. Bekkers provided the methods for culturing hippocampal neurons. We also thank the following people for their helpful comments in the preparation of this manuscript: Dr. S. Janmaat, A. Daniel, Dr. E. Vissel, Dr. D. Ryugo, Dr. R. Callister, and Dr. S. Oleskevich. Our work was supported by NHMRC Australia Grant #188819 (BV), a NSW State Government Spinal Cord Injury & Related Neurological Conditions Research Grant administered by the Office for Science and Medical Research (BV), The NSW State Government's BioFirst Award (BV), the Australian Postgraduate Award (JAD), the Baxter Family Scholarship (JAD), the Gowrie Trust Scholarship (JAD), The Baxter Trust, The Lyndsay and Heather Payne Medical Research Foundation, Perpetual Philanthropic Foundation, NIH Grant NS32519-14 (LI), Amadeus Energy Ltd., Mr Bill Gruy, Mr Geoff Dixon and Mrs Dawn Dixon, and the Henry H Roth Charitable Foundation.