Logo of nihpaAbout Author manuscriptsSubmit a manuscriptNIH Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
J Neurosci. Author manuscript; available in PMC 2013 Mar 19.
Published in final edited form as:
PMCID: PMC3489011
NIHMSID: NIHMS409583

Co-release of Dopamine and GABA by a Retinal Dopaminergic Neuron

Abstract

Numerous neurons release two transmitters of low molecular mass, but it is controversial whether they are localized within the same synaptic vesicle, with the single exception of GABA and glycine because they are ferried into the vesicle by the same transporter. Retinal dopaminergic amacrine (DA) cells synthesize both dopamine and GABA. Both transmitters are released over the entire cell surface and act on neighboring and distant neurons by volume transmission, but, in addition, DA cells establish GABAergic synapses onto AII amacrine cells, the neurons that transfer rod signals to cone bipolars. By combining recordings of dopamine and GABA release from isolated, genetically identified perikarya of DA cells from the mouse retina, we observed that a proportion of the events of dopamine and GABA exocytosis were simultaneous, suggesting co-release. Furthermore, a proportion of the secretory organelles in the perikaryon and synaptic endings of DA cells contained both vesicular transporters for dopamine (VMAT2) and GABA (VGAT). Since the majority of the dopamine release events concerned a single transmitter and organelles were present that contained a single transporter, either VMAT2 or VGAT, we conclude that the secretory organelles of DA cells contain variable concentrations of the two transmitters, which are in turn determined by a variable mixture of the two transporter molecules in their limiting membrane. This variability can be explained if the relative numbers of transporter molecules is determined stochastically during the budding of the somatic organelles from the trans-Golgi-network or the retrieval of the vesicular membrane from the plasmalemma after exocytosis.

Introduction

In the central nervous system there is convincing evidence for co-localization of dopamine with other low molecular mass transmitters that act on ionotropic postsynaptic receptors and therefore convey faster excitatory signals to the postsynaptic cell (see Seal and Edwards, 2006; Hnasko and Edwards, 2012). Midbrain dopaminergic neurons make excitatory glutamatergic synapses onto the projection neurons of the accumbens (Sulzer et al., 1998; Joyce and Rayport, 2000; Chuhma et al., 2004; Stuber et al., 2010) and co-transmission of dopamine and GABA was observed in the periglomerular cells of the olfactory bulb (Maher and Westbrook, 2008). In the retina, the dopaminergic amacrine (DA) cells release dopamine (Puopolo et al., 2002) and GABA (Hirasawa et al., 2009) extrasynaptically and establish GABAergic synapses onto AII amacrine cells (Contini and Raviola, 2003), a neuron that transfers rod signals to cone bipolars.

Because of their dual transmitter phenotype, DA cells exert complex functions in the retina: dopamine is released upon illumination, diffuses throughout the retina and acts by volume transmission on most types of retinal neurons, setting their gain for vision in bright light (see Witkovsky, 2004). The physiological significance of the GABA released by DA cells has not been established with certainty: by inhibiting DA cells’ postsynaptic target, GABA may prevent signals of the saturated rods from entering the cone pathway when the dark-adapted retina is suddenly exposed to bright illumination. Thus, inhibition may cause the silent pause and hyperpolarization of the AII amacrine cell when the stimulus reaches photopic range during an intensity series (Xin and Bloomfield, 1999). On the other hand, the GABA released extrasynaptically by DA cells, in concert with other GABAergic amacrines (Hirasawa et al., 2009), may prevent cross-talk between OFF- and ON-pathways in the crowded environment of the inner plexiform layer (vertical inhibition; Roska and Werblin, 2001; Farajian et al., 2011).

In all instances of co-localization or co-release, different vesicular transporters, VMAT2 for dopamine, VGAT for GABA and the various VGLUT isoforms for glutamate, load the transmitters into membrane-bounded secretory organelles (see Seal and Edwards, 2006; Edwards, 2007), either 40–50 nm synaptic vesicles or 80–90 nm large dense core vesicles (LDCV; Nirenberg et al., 1996, 1997). The question therefore arises whether the two transmitters are stored within the same or separate organelles. Unfortunately, in DA cells the metabotropic receptors for dopamine do not co-localize at the postsynaptic active zone with the ionotropic receptors for GABA. It is therefore impossible to deduce the composition of the transmitter contents of individual presynaptic vesicles by teasing out pharmacologically different components of the miniature postsynaptic currents, as previously done for GABA and glycine (Jonas et al., 1998; O’Brien and Berger, 1999) or glutamate and acetylcholine (Li et al., 2004).

In this paper, we tested by simultaneous patch clamp and amperometry whether GABA and dopamine are co-released by genetically identified DA cell bodies and we used immunocytochemistry to investigate the co-localization of VMAT2 and VGAT in the intact retina.

Material and Methods

Animals

We used C57BL/6J mice of either sex both wild type (Jackson Laboratory, Bar Harbor, ME) and transgenic, the latter expressing PLAP under control of a 4.8 kb promoter sequence of the rat TH gene (Gustincich et al., 1997). All procedures involving mice were in accordance with National Institutes of Health guidelines and approved by the Institutional Animal Care and Use Committee of Harvard Medical School.

Electrophysiology

The procedure for dissociation of the retina and short-term culture of the resulting cell suspensions were described in detail previously (Hirasawa et al., 2009). Briefly, retinas of 1- to 3-month-old transgenic mice whose catecholaminergic neurons were labeled by human placental alkaline phosphatase (PLAP) were incubated with papain (5–10 U, Worthington, Lakewood, NJ) and subsequently triturated by squeezing them through the bore of fire-polished Pasteur pipettes. To identify DA cells, the suspension was re-suspended in minimum essential medium (MEM) containing 1:100 monoclonal antibody to PLAP (E6; de Waele et al., 1982) directly conjugated to the fluorescent dye Cy3 (E6-Cy3). Cells were allowed to settle onto concanavalin A-coated glass coverslips glued to the bottom of Petri dishes that were mounted on the stage of an inverted microscope (Nikon Diaphot 300, MVI, Avon, MA). DA cells, stained by E6-Cy3, were identified by scanning the coverslip in epifluorescence. Once the DA cells were identified, the remaining procedures were carried out in visible light with DIC optics.

Patch pipettes with tip resistance of 3–5 MΩ for whole-cell recording were constructed from borosilicate glass. Electrodes were connected to an Axopatch 200B amplifier (Axon Instruments Inc., Foster City, CA) controlled by pCLAMP 8.0 analysis software (Axon Instruments) and current or voltage output was viewed directly on the screen of a computer attached to the amplifier via a DigiData 1200 Interface (Axon Instruments). The digitized data were stored on a PC and analyzed with Origin 6.1J (OriginLab, Northampton, MA). Frequency of the low pass Bessel filter in the amplifier was set at 2 kHz. Sampling frequency was 6.7 kHz. Bath temperature was routinely kept at 30–31°C with a heater controller (TC-344B, Warner Instruments, Hamden, CT).

Membrane currents of DA cells were recorded by using the whole-cell configuration of the patch-clamp technique as previously described (Feigenspan et al. 1998). The standard extracellular solution contained (in mM): 137 NaCl, 5.4 KCl, 2 CaCl2, 1 MgCl2, 5 HEPES and 10 Glucose (pH 7.4). The pipette solution contained (in mM): 111–115 cesium methanesulfonate (CsMeSO3), 10 NaCl, 10 TEA-Cl, 10 HEPES, 0.05 EGTA, 1 MgCl2, 14 Phosphocreatine, 3 ATP-Mg, 0.3 GTP-Na (pH 7.2). The liquid junction potential of the pipette solution (~ −10mV, measured by using a 3 M KCl agar bridge) was corrected. Series resistance was maintained at <20 MΩ and uncompensated.

Amperometric recordings were obtained with polyethylene-insulated, carbon fibers 5 μm in diameter (T650, Cytec Carbon Fibers LLC, Greenville, SC). Immediately prior to use, the tip of the electrode was cut with a razor blade, beveled at a 45° angle with a diamond coated polishing wheel (Sutter Instrument Co., Novato, CA) and cleaned by ultrasonication in isopropanol for 3 s. Electrodes were backfilled with a 4 M potassium acetate, 150 mM KCl solution and electrical connection to the headstage (CV201A/Axopatch 200A, Axon Instruments) was made by a chlorided silver wire. An electrode holding potential of +650 mV versus an Ag/AgCl reference electrode was applied by the Axopatch 200A after appropriate modification to the holding potential circuitry (Axon Instruments Hardware Modification Note #22). Sensitivity of the electrode was checked with current measurement in response to 50 μM dopamine reconstituted from 10 mM stock solution in 100 mM perchloric acid. The output of the Axopatch 200A was filtered at 2 kHz by an internal 8-pole low-pass Bessel filter. Sampling frequency was 6.7 kHz. The amount of dopamine was estimated to be equivalent to two electrons per oxidized dopamine molecule.

To stimulate release of GABA and dopamine, ECl was set near −60 mV, the holding potential was −90 mV and 3 s depolarizing pulses to −10 mV were delivered to the DA cell bodies. To rule out the possibility that the coincidence of GABA and amperometric events was due to chance, we carried out a Monte Carlo simulation under the null hypothesis of independence between the timing of GABA current events and that of the amperometric current events, since the release of dopamine and GABA were both based on a Poisson process (see Results). For each depolarization trial, we measured the times of the last events of GABA release (t=Tg) and dopamine release (t=Td) after the onset of depolarization (t=0) and counted the number of GABA (Ng) and dopamine (Nd) events. The times of the Ng GABA events were uniformly distributed on (0,Tg) and the times of the Nd dopamine events were uniformly distributed on (0,Td). So, for each trial, under the null hypothesis of independence of GABA and dopamine events, we simulated the times of the events and computed the number of events that coincided within 4 ms (actual time interval 3.37 ms). We then summed the coincidences over the 43 trials. This computation was repeated 10,000 times.

For analysis of the time course of GABA and dopamine events, data were subsequently low-pass filtered at 100 Hz by digital Gaussian filter (Clampfit 8.1, Axon Instruments), which distorted the 10–90% step rise time to 3.4 ms. After this filtration, threshold for detection of GABA events became 50 pS amplitude, which corresponds to ~5,000 molecules based on a mathematical model of GABAA receptor-mediated Cl currents over the area surrounding the release site (Hirasawa et al., 2009). Both types of current events were detected by eye with the threshold defined as three times the value of the standard deviation (SD) of baseline noise. Statistical values are given as mean ± SD; confidence limits were determined by Student’s t tests.

The sensitivity of the techniques used to detect dopamine and GABA was estimated in the following way: (i) Amperometry: oxidizing reaction of dopamine on the carbon surface occurs after one-dimensional diffusion from the releasing pore. Since the time course of dopamine oxidation is almost instantaneous at a very positive potential (+650mV), the shapes of oxidative current spikes observed should closely resemble those predicted for instantaneous point source release. The time integral Q of the current transient I can be directly related to the amount of oxidized transmitter by Faraday’s law:Q=∫Idt=Men

Where M is the number of molecules, e is the elementary charge[double contour integral operator] and n is the number of moles of electrons transferred per mole of transmitter oxidized. Threshold for the detection of dopamine events was 1fQ in charge, corresponding to ~3,000 in molecules. (ii) When GABA molecules released from the fusion pore arrives at GABA receptors surrounding the release site, the elementary GABA current ΔI through a membrane unit surface ΔS is determined by the local concentration of GABA, the density of the GABAA receptor Cl channels, the single channel conductance, and the receptor binding affinity. The three-dimensional diffusion of GABA molecules is assumed to occur on the plane. The total current I over the area of the plasma membrane surrounding the release site is thus given by the following form: I=γVν[double contour integral operator](c)dS where γ is the single channel conductance, V is the driving force of Cl, ν is the surface density of the GABAA receptors, and p(c) is the dose–response function that represents the open probabilities of GABAA receptor channels at the GABA concentration c (Hirasawa et al., 2009). Threshold for the detection of GABA events was 50pS in amplitude corresponding to ~5,000 in molecules.

Immunocytochemistry

To investigate the localization of VMAT2 and VGAT in DA cell bodies, the wild type mice were anesthetized by IP injection of a 0.1 ml solution containing equal parts of 5% ketamine (Ketaset, Bristol-Myers Co., Syracuse, NY) and 1% xylazine (Rompun, Bayer Co., Shawnee Mission, KS). The eyes were enucleated, opened at the equator and their posterior segments were immersed in 2% formaldehyde (Tousimis, Rockville, MD) in 0.15 M Sörenson phosphate buffer (pH 7.4). Neural retinas were separated from the pigment epithelium and the outer ocular tunics and kept in the fixative fluid for 2 hr at room temperature.

For confocal microscopy, fixed retinas were washed in PBS, cryoprotected in 20% sucrose in PBS, frozen in the liquid phase of partially solidified monochlorodifluoromethane and finally cut in a cryostat at a thickness of 10 μm. Sections were preincubated for 2 hr in blocking solution, containing 2% bovine serum albumin (BSA, Sigma-Aldrich, St. Louis, MO), 10% normal goat serum (Cat. # S-1000, Vector Labs, Burlingame, CA), 2% cold water fish gelatin (Ted Pella Inc., Redding, CA), and incubated overnight at room temperature in a mixture of rabbit polyclonal α-VMAT2, guinea pig polyclonal α-VGAT and sheep polyclonal α-tyrosine hydroxylase (TH) antibodies (see Table 1 for source and characterization). Slides were washed in several changes of block solution, followed by 3 hr incubation in a mixture of donkey α-rabbit, goat α-guinea pig and donkey α-sheep secondary antibodies (all from Invitrogen-Molecular Probes, Carlsbad, CA), resp. conjugated to Alexafluor 568 (Cat. # A10042), Alexafluor 488 (Cat. # A11073) and Alexafluor 680 (Cat. # A21102). All secondary antibodies were diluted 1:500 with block solution. The retinas were finally rinsed in several changes of PBS and mounted in Vectashield (Vector Labs).

Table 1
Properties of the antibodies used for immunocytochemistry

Fluorescence was detected in a Zeiss LSM 510 Meta confocal imaging system (Carl Zeiss MicroImaging Inc., Thornwood, NY), equipped with three visible wavelength lasers, META spectral emission detectors and a Zeiss Axioplan 2 microscope. We used a 100X Plan-Apochromatic objective 1.4 NA and set the pinhole at 1 Airy unit for each of the three wavelenghts (543, 488 and 633 nm). Eighteen micrographs were obtained sequentially from the three channels by averaging 16 scans (1024 × 1024 pixels) and stored as TIFF files.

Using the ImageJ 1.40g morphometric analysis tool, in each confocal micrograph of the DA cell bodies the green (VGAT), red (VMAT2) and blue (TH) channels were split, thus converting them into greyscale images. By drawing the contours of the cell and nucleus in the TH image, the cytoplasm was selected and the intensity of all its pixels was obtained in both green and red images. The images were thresholded on the basis of the distribution profile of the pixel intensities; then, pixels whose intensity was above that of the background were assigned value 1 (top 55% of the green pixels and 43% of the red pixels, a difference caused by the higher background of the staining with the VGAT antibody) and the remaining pixels were given value 0. It must be noted, that our approach was biased on the side of a false null hypothesis (i.e. concluding independence between green and red, see below).

Since the largest diameter of the organelles stained green, red or yellow was 5 pixels, we considered independent of one another any two pixels separated from each other by a distance (vertical or horizontal) of 5 + 1 pixels. Therefore, in the arrays of digitized pixel intensity values of each red and green image, we selected every 6th pixel in both rows and columns (subset 1). We then shifted by (1, 1) the selected pixels 5 times in the rows and 5 times in the columns, creating a total of 25 subsets of data for each red and green image. We repeated this procedure for all 18 micrographs of DA cell bodies.

In each subset, the overlap of green and red was then calculated by multiplying the two digitized values for each pixel, i.e. only pixels with value 1 in both the red and the green arrays were assigned value 1. The numbers of double positive (GR), green only (G0), red only (R0) and double negative (00) pixels were counted in each of the subsets and the data were arranged into 450 contingency tables.

We calculated the odds ratio (OR) in each contingency table and tested the null hypothesis that the true OR was equal to 1, i.e. green in a red pixel had the same probability as green in a pixel that was not red. OR provides a reasonable estimate of the strength of the association when the number of positive pixels is small compared with the number of negative pixels. We modeled the binary outcome of green using logistic regression, with adjustment for the binary outcome of red. We estimated standard errors and confidence intervals using the method of Generalized Estimating Equations (Liang and Zeger, 1986) to adjust for dependence among the 25 subsets within each of the 18 micrographs.

Electron microscopy

The technique used to identify with the electron microscope the DA cell bodies in PLAP-expressing mice was described in detail previously (Gustincich et al., 1997). Briefly, after aldehyde fixation, retinas were incubated in a β-glycerophosphate, alkaline lead citrate solution. They were subsequently postfixed in osmium ferrocyanide and stained en bloc with uranyl acetate. After embedding in low viscosity Epon and thin sectioning, micrographs were obtained with a Jeol 1200 EX electron microscope.

For EM immunocytochemistry, retinas fixed in formaldehyde as described above were washed in phosphate buffered saline (PBS) containing 0.2 M glycine to quench free aldehyde groups, infiltrated with 2.3 M sucrose in PBS for 15 min and frozen in liquid nitrogen. Cryosections were obtained at − 120° C with a Reichert Ultracut S microtome equipped with a FC S cryo attachment (Leica Microsystems, Bannockburn, IL) and transferred to formvar-carbon coated grids, that were subsequently floated on drops of 1% BSA in PBS (PBS/BSA) on parafilm (10 min). Both the antibodies and the Protein A-gold (Cell Microscopy Center, Utrecht, The Netherlands) were diluted in 1% PBS/BSA and the diluted antibodies were centrifuged at 15,000 rpm for 1 min. Grids were transferred sequentially to: rabbit polyclonal α-VMAT2 (1:50, 30 min), PBS/BSA (4x), Protein A-gold 15 nm in diameter (1:75, 20 min), PBS (4x), 0.1 % glutaraldehyde in PBS (5 min) to denature the α-VMAT2 antibody, PBS/BSA (4x), rabbit polyclonal α-VGAT (1:50, 30 min, see Table 1), PBS/BSA (4x), Protein A-gold 10 nm in diameter (1:75, 20 min), PBS (4x), H20 (double distilled, 6x). Staining with 0.3% uranyl acetate in 2% methylcellulose (Sigma) in H2O for 10 min was carried out on ice. Grids were picked up with metal loops and excess fluid was removed by streaking on filter paper, leaving a thin coat of methylcellulose (blue interference color when dry). Micrographs were obtained in a Technai G2 Spirit BioTWIN transmission electron microscope equipped with an AMT 2k CCD camera.

Results

Electrophysiology

To establish whether dopamine and GABA were released simultaneously and were therefore contained within the same secretory organelles, we exploited the property of DA cell bodies to release extrasynaptically each of the transmitters (Puopolo et al., 2001; Hirasawa et al., 2009). Retinas of transgenic mice that expressed human placental alkaline phosphatase (PLAP) under control of a promoter sequence of the gene for TH were dissociated by enzymatic digestion and mechanical trituration. Since PLAP resides on the outer surface of the cell membrane, we could identify DA cells bodies by staining the cell suspension with the monoclonal antibody to PLAP E6-Cy3. We made use of the presence of GABAA receptors over the entire surface of DA cell bodies to measure by whole-cell patch-clamp recording the Cl current activated by the exocytosis of their own GABA (Hirasawa et al., 2009). Simultaneously, we measured the oxidation current of dopamine by carbon fiber amperometry (Wightman et al., 1991); the inset in Fig. 1 illustrates the experimental setup. Beveled carbon fiber electrodes were approximated to the cell surface until contact was established, as indicated by dimpling of the plasma membrane. The recording area, measured electrochemically, was ~40μm2 or 6% of the cell surface. Oxidation currents were recorded by applying to the carbon fiber electrode +650 mV versus an Ag/AgCl reference electrode, a potential sufficient to oxidize dopamine (Hochstetler et al., 2000).

Figure 1
Co-release of GABA and dopamine by isolated DA cell bodies

The release of dopamine and GABA elicited by depolarization of DA cell bodies was characterized previously (Puopolo et al., 2001; Hirasawa et al., 2009). Figures 1A and B show that when (i) ECl was set near −60 mV, (ii) the holding potential was −90 mV and (iii) intracellular EGTA was 0.05 mM, a 3 s depolarizing pulse to −10 mV elicited both transient Cl current events and amperometric events. The Cl current events were mediated by GABAA receptors, because they were abolished by the antagonists bicuculline and SR95531 (resp.100 μM and 20 μM, data not shown; see Hirasawa et al., 2009) and the amperometric spikes disappeared when the carbon fiber electrode was held at potentials too low to cause dopamine oxidation (<100 mV versus the Ag/AgCl reference electrode; see Puopolo et al., 2001). We measured 145 GABA and 106 amperometric events in 29 cells, including all events that appeared within 9 s after onset of depolarization (Fig. 1C). Latencies of both types of events, as measured from onset of the membrane depolarization to the occurrence of the first GABA or amperometric current transient, exhibited considerable variability (GABA events, 723 ± 834 ms; amperometric events, 1,188 ± 1,265 ms). In addition, the duration of the discharge of both types of events was very long (up to almost 9 s) and extremely variable in different trials, as illustrated in Figures 1C and D. As reported previously (Puopolo et al., 2001, Hirasawa et al., 2009), this considerable variability in latency and duration of the discharge is a property typical of extrasynaptic release as compared with synaptic vesicle exocytosis at the presynaptic active zone.

As illustrated in Figure 1B, inspection of the traces revealed that the majority of the GABA and amperometric events were not associated with one another, but a proportion of them appeared to coincide in time. From the raster plot of Figure 1C, we generated a perievent time histogram (Fig. 2A top) by measuring the time intervals between GABAergic events and the immediately preceding and following dopaminergic events throughout all 43 trials (n=402 event pairs, bin width 4 ms). The histogram peaked at a ±4ms time window (Fig. 2A bottom, actual interval −1.14 to +2.23 ms). We thus concluded that 16% of the amperometric events (17 out of 106 events) coincided with GABAergic events, because no peak would have occurred if association between the two types of currents had been completely random. Interestingly, this coincidence rate was close to that observed between serotonin and GABA release events in pancreatic beta cells (~6.3%; Braun et al., 2007) and between catecholamine-and neuropeptide release events in chromaffin cells (~14%; Whim, 2006).

Figure 2
Coincidence of GABAergic and amperometric events

The coincidence of GABA and amperometric events, however, could have been due to chance. To rule out this possibility, we first observed that the histograms of inter-event time intervals for dopamine and GABA release could be described with a single exponential fit (Fig. 2B). Therefore, the release of both transmitters was based on a Poisson process, as is the case for the secretion of catecholamines by adrenal chromaffin cells (Chow et al., 1992) and serotonin by leech Retzius cells (Bruns et al., 2000). Then, under the null hypothesis of independence between the two types of events, we carried out a Monte Carlo simulation. In 104 trials, the average number of coincidences of GABA and amperometric events was 1.34 and the maximum number of coincidences in any one trial was 8. In no instance we were able to reproduce the 17 coincidences observed experimentally. Therefore, with P<10−4, there was strong evidence against the null hypothesis of independence of GABAergic and amperometric events. We therefore concluded that the coincidence of dopaminergic and GABAergic events was not due to chance and reflected release of the two types of transmitters from the same organelle.

When we examined coincident GABAergic and amperometric events on an expanded time scale (Fig. 3), we observed that, in approximately half of the pairs (8 out of 17), the peak of the amperometric events preceded that of the GABAergic events by 1.14 ± 0.86 ms (Fig. 3A). This was an expected result, bacause oxidation of the dopamine released in the area of contact between cell surface and carbon electrode is instantaneous at the very positive potential used (+650mV), whereas activation of the GABAA receptor channels surrounding the site of exocytosis takes place after radial diffusion of the released transmitter. In the remaining pairs (9 out of 17), however, the opposite was true, with the GABAergic events preceding the amperometric ones by 2.23 ± 1.47 ms (Fig. 3B). To compare the rising phases of the two types of events, the currents were averaged and their times of onset were made to coincide (Fig. 3C top). In the coincident pairs in which amperometric events preceded the GABAergic events, the rise times of the two currents were remarkably similar after normalization and, at half-maximal amplitude, the amperometric event preceded by 0.7 ms the GABAergic event (Fig. 3C bottom). In contrast, when amperometric events followed the GABAergic events, their waveform was quite variable and both their 10–90% rise time and half width were slower (Fig. 3D top and Table 2). After averaging, the resulting amperometric trace appeared flattened and irregular in shape (Fig. 3D bottom). No significant differences in charge and amplitude were observed between the coincident amperometric events that preceded and those that followed the GABAergic events (Table 2).

Figure 3
In each pair of coincident current events, the peak of the amperometric spikes either preceded or followed the peak the GABAergic current
Table 2
Properties of the amperometric events that coincided with the GABA current events

In agreement with a previous study (Puopolo et al., 2001), the charge distribution of 89 single and 17 paired amperometric events exhibited a relatively broad-spectrum of sizes, ranging from 1.2 to 39.6 fC or 103 to 105 molecules, assuming that two electrons were transferred for every molecule of dopamine oxidized (Fig. 4A). The mean charge of coincident amperometric events [10.2 ± 7.5 Q(fC), corresponding to (3.19 ± 2.34)×104 molecules] and that of single amperometric events [7.3 ± 6.3 Q(fC), corresponding to (2.28 ±1.97) ×104 molecules] were not significantly different from one another (P<0.1). On the other hand, when we estimated the number of GABA molecules released during the 147 GABAergic events (see Material and Methods), they ranged from 2.0×103 to 3.5×106 molecules and therefore exhibited a spectrum of sizes much broader than the amperometric events (Fig. 4B). In contrast with the amperometric events, however, the mean number of molecules of coincident GABAergic events [(6.29 ± 9.86)×105] was significantly larger than that of single events [(1.64 ± 3.63)×105 molecules, P<0.01]. This result was due to the fact that a sizable fraction of the GABAergic events exceeding ~5×105 molecules in size (38%) coincided with amperometric events.

Figure 4
Quantum size distribution of the dopaminergic and GABAergic events

Comparison of the distributions of the quantal sizes of dopamine and GABA (Fig. 4) shows that both populations were skewed toward smaller events, which is the expected result for the distribution of transmitter contents in populations of cytoplasmic organelles whose diameters are normally distributed. Indeed, all secretory cells that undergo exocytosis, including dopaminergic neurons, exhibit a similar, normal distribution of organelle diameters and skewed distribution of organelles’ volumes. In addition, however, the population of GABAergic events included a relatively small number of very large transmitter quanta that were probably released upon exocytosis of secretory organelles of larger diameter. Indeed, the bodies of DA cells, when examined with the electron microscope, contained a small number of moderately dense granules up to 0.3 to 0.5 μm in diameter (Fig. 5). When the released dopamine was associated with the large quanta of GABA, the number of monoamine molecules was one order of magnitude smaller that that of the bolus of GABA. Furthermore, the dopaminergic event was often polymorphic and its peak followed that of the GABAergic event. This distortion of the amperometric events was either caused by nonuniform contents of the large secretory organelles or “diffusional filtering”, i.e. radial diffusion of the dopamine from a site of exocytosis situated at some distance from the carbon electrode.

Fig. 5
The cell body of DA cells contains a polymorphic population of secretory organelles

Immunocytochemistry

To identify GABAergic and dopaminergic organelles in DA cells, we obtained confocal images of cryostat sections of formaldehyde-fixed retinas from wt C57BL/6J mice after triple staining with antibodies to TH, VGAT and VMAT2 (Fig. 6A). We focused our attention to the perikaryon of DA cells, because their synaptic endings were densely stained by both antibodies (Contini and Raviola, 2003) and individual organelles could not be resolved. All cell bodies stained by α-VMAT2 antibody also contained TH and therefore were DA cells, for in the C57BL/6J mouse retina type 2 catecholaminergic amacrines are TH-negative (Contini et al., 2010). Organelles stained by the transporter antibodies were scattered throughout the cytoplasm of DA cells: after image thresholding, they measured from 1 to 5 pixels in diameter (equivalent to 0.3 to 1.4 μm) and, when the blue channel (TH) was eliminated, their staining was green (VGAT), red (VMAT2) or varying combinations of the two colors (Fig. 6B–D). This seemed to indicate that this heterogeneous population of organelles had bound varying amounts of the two antibodies and that a proportion of them contained both transporters.

Figure 6
DA cell bodies contain organelles stained by antibodies to VMAT2 and VGAT

We digitized the intensity of all pixels in the green and red images of the cytoplasm of 18 DA cell optical sections and used odds ratio (OR) statistics to measure the strength of the association between the two transporters. In this way, we forfeited the chance to count stained organelles in order (i) to avoid subjective bias in estimating the relative green and red staining intensity of the organelles that bound both antibodies, and (ii) rule out the possibility that co-localization of the two transporters was due to the superimposition of different organelles within the thickness of the optical section. The null hypothesis was that the true OR was equal to 1, i.e. green in a red pixel had the same probability as green in a pixel that was not red. OR provides a reasonable estimate of the strength of the association between green and red when the number of positive pixels is small compared with the number of negative pixels. Hypothesis testing was done by using two different methods based on logistic regression models, both adjusting for dependence among the pixels within each organelle (see Material and Methods) and calculated the analytical confidence interval. The lnOR was estimated to be 1.09 with a 95% confidence interval (0.82, 1.4). This means that there was dependence between green and red, since under independence the lnOR would have been 0. The probability of (green in a red+ pixel) divided by the probability of (green in a red− pixel) was 2.74, i.e. there was 2.74 times the likelihood for a green pixel of being red+ than being red−. Under independence, the ratio would have been 1. We could therefore rule out that the co-localization of VGAT and VMAT2 in a number of the cytoplasmic organelles of DA cells was not due to biased sampling or chance superimposition. Statistical simulation showed that inclusion of additional micrographs was unnecessary.

Electron microscopy

To resolve individual organelles in the axonal endings of DA cells, formaldehyde-fixed retinas were treated sequentially with α-VMAT2 and α-VGAT antibodies, followed respectively by Protein A conjugated to 15 nm and 10 nm gold particles (Fig. 7). Both antibodies were rabbit polyclonals, because they strongly bind Protein A: we therefore crosslinked with glutaraldeyde the residual epitopes of the first antibody before the application of the second one. We focused our attention to the most scleral stratum of the inner plexiform layer, where synaptic endings of DA cells are very frequent and can be readily identified because they are the only processes in which synaptic vesicles are labeled with antibodies to VMAT2. In the clusters of negatively stained synaptic vesicles, the majority of 15 and 10 nm gold particles were scattered individually. Less frequently, however, particles of different sizes occurred in pairs, very close to one another, and either superimposed over or apparently associated with adjacent agranular vesicles (Fig. 7A). The crowding, however, of the vesicle clusters did not permit to rule out completely the possibility that the members of the pairs were bound to separate vesicles that overlapped within the thickness of the cryosection. LDCV were also occasionally positive for both VMAT2 and VGAT (Fig. 7B).

Figure 7
EM-Immunocytochemistry on retinal cryosections

Discussion

Combined amperometric and patch clamp recordings from acutely isolated perikarya of DA cells detected events of GABA and dopamine release upon membrane depolarization. Previous evidence had shown that this release was due to exocytosis, because the events were transient on the millisecond time scale and were Ca2+-dependent (Puopolo et al., 2001, Hirasawa et al., 2009). The majority of the release events of one transmitter were unaccompanied by the release of the other, but a significant number of them were simultaneous. Confocal microscopy of intact retinas showed that the perikarya of DA cells contained organelles that bound antibodies to both vesicular transporters: some organelles were immunoreactive for VGAT or VMAT2 alone and others for variable mixtures of both transporters, which suggests variable contents of dopamine and GABA. Taken together, these data strongly suggest that a proportion of the secretory organelles in the cytoplasm of DA cells bodies contain both GABA and dopamine and release them simultaneously by exocytosis upon depolarization of the cell membrane.

We cannot provide a precise estimate of the extent of transmitters’ co-localization within the population of somatic secretory organelles: in amperometry, the contact area between carbon electrode and cell membrane was ~1/17th of the total body surface, whereas the patch electrode measured GABA release over the entire surface. Amperometry indicates that a proportion of the secretory organelles contain dopamine in absence of detectable amounts of GABA. They likely correspond to the cytoplasmic organelles that are exclusively VMAT2 positive. Because the carbon fiber only contacts a fraction of the cell surface, we cannot prove that all GABAergic events that occur in absence of a simultaneous dopamine event are caused by exocytosis of organelles exclusively filled with GABA. They must however exist, for immunocytochemistry demonstrates the presence of organelles stained by VGAT alone. If we extrapolate the amperometric data to the entire cell surface, the total number of dopaminergic events would be one order of magnitude larger than the number of events of GABA release and, paradoxically, the coincident events over the entire cell surface would be twice as many as the recorded GABAergic events. Thus, either amperometry measured dopamine release from a surface larger than the tip of the carbon fiber or the release of the smallest packets of GABA molecules was lost in the noise of the patch clamp recordings. Indeed, threshold for detection of the GABAergic currents was an amplitude of 50 pS, corresponding to a GABA quantum of ~5,000 molecules, whereas threshold for the detection of dopamine was a charge of 1 fC or ~3,000 molecule (Puopolo et al., 2001). Noteworthy is also the fact that when dopamine release was associated with a huge GABA bolus, the monoamine quantum was one order of magnitude smaller than the concomitant GABA quantum. We have no explanation for this finding: coumpound exocytosis such as that seen in pancreatic s-cells (Eliasson et al., 2008) is unlikely, because in DA cells secretory granules are few in number and far apart.

What is the identity of the somatic organelles that release GABA and dopamine? DA cells are a typical example of a neuron that synthesizes, in addition to dopamine and GABA, a large repertory of molecules: these include the neuropeptide CART, the hormone insulin, the cytokine IFN-α and the chemokine MCP-1 (Gustincich et al., 2004). In fact, electron micrographs show that the DA cell cytoplasm contains secretory granules, 0.3 to 0.5 μm in diameter and 100 nm LDCV, together with a plethora of polymorphic membranous compartments, that can be described as vesicles, tubules and cisterns (Fig. 5). Our confocal images showed that the organelles that contained VMAT2, VGAT or varying amounts of both transporters exhibited apparent sizes of 0.3 to 1.4 μm and, therefore, probably corresponded to the granules and LDCVs present in the electron micrographs. It must emphasized, however, that our identification of the immunoreactive organelles was biased toward structures larger that a single pixel. Indeed, the broad spectrum of sizes of the transmitter quanta (103 to 106) and the fact that both their GABAergic and dopaminergic populations were skewed toward smaller events indicate that the secretory organelles were highly heterogeneous and had to include a proportion of the small vesicles visible with the electron microscope.

Little is known of the global traffic of all these somatic, membrane-bounded organelles, that contain either or both transmitter transporters in their limiting membrane, accumulate dopamine and/or GABA in their interior and participate in regulated exo- and endocytosis. Apparently, they have acquired at the trans-Golgi network variable mixtures of the two types of transporter molecules and therefore accumulate varying amounts of the two transmitters. In addition, as previously noted (Puopolo et al., 2001), small vesicular profiles may be temporarily docked at the somatic cell surface and undergo multiple cycles of exo- endocytosis, thus recycling their complement of transporter molecules with the plasmalemma.

There is evidence that also synaptic vesicles in DA cells may contain both dopamine and GABA: confocal microscopy had shown co-localization of VMAT2, VGAT and GABA in the axonal varicosities of DA cells (Contini and Raviola, 2003) and we present here electron micrographs of retinal cryosections suggesting that synaptic vesicles and LDCVs are labeled by antibodies to both transporters. It is interesting that, at DA-to-AII amacrine cell synapses, GABAA but not dopamine receptors are clustered in postsynaptic active zone (Contini and Raviola, 2003), an indication that dopamine, irrespective of its site of release –synaptic or extrasynaptic– acts by volume transmission on receptors diffusely distributed over the surface of DA cells’ postsynaptic targets.

Co-localization and co-release of two low molecular mass transmitters is common in both central and peripheral neurons (see Hnasco and Edwards, 2012), but it is still unclear whether they are contained within the same synaptic vesicle, except for GABA and glycine, that are both co-transported by VGAT (Woicik et al. 2006). In interneurons of the tadpole spinal cord, pharmacological dissection of miniature excitatory postsynaptic currents showed that part of the synaptic vesicles co-released glutamate and acetylcholine (Ach) and the remaining ones released a single transmitter (Li et al., 2004). In contrast, in retinal starbust cells, Lee et al. (2010) suggested that GABA and Ach were contained in separate vesicles. The case of glutamate and GABA is controversial. Two studies, both based on light and electron microscope immunocytochemistry of vesicular transporters reached opposite conclusions: in the hippocampus, terminals of axons that made both asymmetric and symmetric synapses onto dentate granule cells contained two distinct populations of synaptic vesicles, one positive for VGAT and the other for the glutamate transporter VGLUT2 (Boulland et al., 2009). In contrast, in the cerebral cortex of adult rats, VGAT and VGLUT1 sorted to the same population of vesicles in a subset of axon terminals that formed both symmetric and asymmetric synapses (Fattorini et al., 2009). These discrepancies may be reconciled if one postulates that synaptic vesicles contain variable mixtures of both transmitters. When both transmitters act on ionotropic receptors, the physiological action at the synapse is determined by the nature of the receptors that accumulate in the postsynaptic active zone. Therefore, it is not surprising that a single ending could establish both asymmetric and symmetric contacts on the same postsynaptic cell, as observed in the chick ciliary ganglion (Tsen et al., 2000), hippocampus (Boulland et al. 2009) and cerebral cortex (Fattorini et al., 2009).

A key to explain these results and other apparent discrepancies in the literature, is that the synaptic vesicle proteins are continuously exchanged with the presynaptic membrane during multiple events of fusion and recycling. Mechanisms exist to ensure that synaptic vesicles acquire upon recycling the entire complement of proteins that are critical to their functional performance. To this purpose, specific motifs in the amino acid sequence of several vesicular proteins, including the transporters VMAT2 and VGLUT1, recruit specific adaptor molecules that shepherd them into the clathrin coat of coated vesicles (see Saheki and De Camilli, 2012). Furthermore, specific adaptors appear to address VGLUT1 and VMAT2 to different pools of vesicles in the presynaptic ending (Tan et al., 1998; Voglmaier et al., 2006; Onoa et al., 2010; Hua et al., 2011). It is yet unclear, however, whether such mechanisms are stringent enough to ensure identical complement of transporter molecules to each synaptic vesicle in the axonal endings of neurons that synthesize multiple low molecular weight transmitters. In fact, (i) each vesicle can accommodate 9 to 14 transporter molecules in their membrane (Takamori et al., 2006); (ii) transporters are interchangeable; and (iii) the presence of a vesicular transporter is sufficient to cause the accumulation of the pertinent transmitter within the vesicle (Takamori et al., 2000, 2001). This suggests that the presence of variable numbers of VMAT2 and VGAT molecules within the membrane of the same vesicle may result from stochastic sorting of the two transporters during recycling from the cell membrane. In a similar fashion, the sorting of both VMAT2 and VGAT to the various secretory organelles in the cell body may be determined stochastically at the moment of their budding from the trans-Golgi network or the somatic cell membrane.

Acknowledgments

We thank M Ericsson for help with EM immunocytochemistry, Dr. RH Edwards for his generous gift of the VGAT antibody and Dr. B Bean for discussions. This work was supported by the National Institutes of Health, Grant number: EY01344 (E.R.), the Harvard Neurodiscovery Center and Harvard Catalyst (R.A.B). H.H.

Footnotes

Conflict of interest: none

Author contributions: HH and ER designed and performed research, analyzed data and wrote the paper. RAB designed, carried out and analyzed statistics.

References

  • Boulland JL, Jenstad M, Boekel AJ, Wouterlood FG, Edwards RH, Storm-Mathisen J, Chaudhry FA. Vesicular glutamate and GABA transporters sort to distinct sets of vesicles in a population of presynaptic terminals. Cerebral Cortex. 2009;19:241–248. [PMC free article] [PubMed]
  • Braun M, Wendt A, Karanauskaite J, Galvanovskis J, Clark A, MacDonald PE, Rorsman P. Corelease and differential exit via the fusion pore of GABA, serotonin, and ATP from LDCV in rat pancreatic beta cells. J Gen Physiol. 2007;129:221–231. [PMC free article] [PubMed]
  • Bruns D, Riedel D, Klingauf J, Jahn R. Quantal release of serotonin. Neuron. 2000;28:205–220. [PubMed]
  • Chow RH, von Rüden L, Neher E. Delay in vesicle fusion revealed by electrochemical monitoring of single secretory events in adrenal chromaffin cells. Nature. 1992;356:60–63. [PubMed]
  • Chuhma N, Zhang H, Masson J, Zhuang X, Sulzer D, Hen R, Rayport S. Dopamine neurons mediate a fast excitatory signal via their glutamatergic synapses. J Neurosci. 2004;24:972–981. [PubMed]
  • Contini M, Raviola E. GABAergic synapses made by a retinal dopaminergic neuron. Proc Natl Acad Sci USA. 2003;100:1358–1363. [PMC free article] [PubMed]
  • Contini M, Lin B, Kobayashi K, Okano H, Masland RH, Raviola E. Synaptic input of ON-bipolar cells onto the dopaminergic neurons of the mouse retina. J Comp Neurol. 2010;518:2035–2050. [PMC free article] [PubMed]
  • de Waele P, De Groote G, Van De Voorde A, Fiers W, Franssen J-D, Herion P, Urbain J. Isolation and identification of monoclonal antibodies directed against human placental alkaline phosphatase. Arch Int Physiol Biochim. 1982;90:B21.
  • Edwards RH. The neurotransmitter cycle and quantal size. Neuron. 2007;55:835–858. [PubMed]
  • Eliasson L, Abdulkader F, Braun M, Galvanovskis J, Hoppa MB, Rorsman P. Novel aspects of the molecular mechanisms controlling insulin secretion. J Physiol. 2008;586:3313–3324. [PMC free article] [PubMed]
  • Fattorini G, Verderio C, Melone M, Giovedì S, Benfenati F, Matteoli M, Conti F. VGLUT1 and VGAT are sorted to the same population of synaptic vesicles in subsets of cortical axon terminals. J Neurochemistry. 2009;110:1538–1546. [PubMed]
  • Gustincich S, Contini M, Gariboldi M, Puopolo M, Kadota K, Bono H, LeMieux J, Walsh P, Carninci P, Hayashizaki Y, Okazaki Y, Raviola E. Gene discovery in genetically labeled singe dopaminergic neurons of the retina. Proc Natl Acad Sci USA. 2004;101:5069–5074. [PMC free article] [PubMed]
  • Gustincich S, Feigenspan A, Wu D-K, Koopman LJ, Raviola E. Control of dopamine release in the retina: a transgenic approach to neural networks. Neuron. 1997;18:723–736. [PubMed]
  • Hirasawa H, Puopolo M, Raviola E. Extrasynaptic release of GABA by retinal dopaminergic neurons. J Neurophysiol. 2009;102:146–158. [PMC free article] [PubMed]
  • Hnasko TS, Edwards RH. Neurotransmitter corelease: mechanism and physiological role. Annual Rev Physiol. 2012;74:225–243. [PMC free article] [PubMed]
  • Hochstetler SE, Puopolo M, Gustincich S, Raviola E, Wightman RM. Real-time amperometric measurements of zeptomole quantities of dopamine released from neurons. Anal Chem. 2000;72:489–96. [PubMed]
  • Hua Z, Leal-Ortiz S, Foss SM, Waites CL, Garner CC, Voglmaier SM, Edwards RH. v-SNARE composition distinguishes synaptic vesicle pools. Neuron. 2011;71:474–487. [PMC free article] [PubMed]
  • Jonas P, Bischofberger J, Sandkuhler J. Corelease of two fast neurotransmitters at a central synapse. Science. 1998;281:419–424. [PubMed]
  • Joyce MP, Rayport S. Mesoaccumbens dopamine neuron synapses reconstructed in vitro are glutamatergic. Neuroscience. 2000;99:445–456. [PubMed]
  • Li W-C, Soffe SR, Roberts A. Glutamate and acetylcholine corelease at developing synapses. Proc Natl Acad Sci USA. 2004;101:15488–15493. [PMC free article] [PubMed]
  • Liang K-Y, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika. 1986;73:13–22.
  • Maher BJ, Westbrook GL. Co-transmission of dopamine and GABA in periglomerular cells. J Neurophysiol. 2008;99:1559–1564. [PubMed]
  • Nirenberg MJ, Chan J, Liu Y, Edwards RH, Pickel VM. Ultrastructural localization of the vesicular monoamine transporter-2 in midbrain dopaminergic neurons: potential sites for somatodendritic storage and release of dopamine. J Neurosci. 1996;16:4135–4145. [PubMed]
  • Nirenberg MJ, Chan J, Liu Y, Edwards RH, Pickel VM. Vesicular monoamine transporter-2: immunogold localization in striatal axons and terminals. Synapse. 1997;26:194–198. [PubMed]
  • O’Brien JA, Berger AJ. Cotransmission of GABA and glycine to brain stem motoneurons. J Neurophysiol. 1999;82:1638–1641. [PubMed]
  • Onoa B, Li H, Gagnon-Bartsch JA, Elias LAB, Robert H, Edwards RH. Vesicular Monoamine and Glutamate Transporters Select Distinct Synaptic Vesicle Recycling Pathways. J Neurosci. 2010;30:7917–7927. [PMC free article] [PubMed]
  • Puopolo M, Hochstetler SE, Gustincich S, Wightman RM, Raviola E. Extrasynaptic release of dopamine in a retinal neuron: Activity dependence and transmitter modulation. Neuron. 2001;30:211–225. [PubMed]
  • Roska B, Werblin F. Vertical interactions across ten parallel, stacked representations in the mammalian retina. Nature. 2001;410:583–587. [PubMed]
  • Saheki Y, De Camilli P. Synaptic Vesicle Endocytosis. Cold Spring Harbor Perspect Biol. 2012 doi: 10.1101/cshperspect. a0905645. [PMC free article] [PubMed] [Cross Ref]
  • Seal RP, Edwards RH. Functional implications of neurotransmitter co-release: glutamate and GABA share the load. Current Opinion in Pharmacology. 2006;6:114–119. [PubMed]
  • Stuber GD, Hnasko TS, Britt JP, Edwards RH, Bonci A. Dopaminergic terminals in the nucleus accumbens but not the dorsal striatum corelease glutamate. J Neurosci. 2010;30:8229–8233. [PMC free article] [PubMed]
  • Sulzer D, Joyce MP, Lin L, Geldwert D, Haber SN, Hattori T, Rayport S. Dopamine neurons make glutamatergic synapses in vitro. J Neurosci. 1998;18:4588–4602. [PubMed]
  • Takamori S, Rhee JS, Rosenmund C, Jahn R. Identification of a vesicular glutamate transporter that defines a glutamatergic phenotype in neurons. Nature. 2000;407:189–194. [PubMed]
  • Takamori S, Rhee JS, Rosenmund C, Jahn R. Identification of differentiation-associated brain-specific phosphate transporter as a second vesicular glutamate transporter (VGLUT2) J Neurosci. 2001;21:RC182. [PubMed]
  • Takamori S, Holt M, Stenius K, Lemke EA, Grønborg M, Riedel D, Urlaub H, Schenck S, Brügger B, Ringler P, Müller SA, Rammner B, Gräter F, Hub JS, De Groot BL, Mieskes G, Moriyama Y, Klingauf J, Grubmüller H, Heuser J, Wieland F, Jahn R. Molecular anatomy of a trafficking organelle. Cell. 2006;12:831–846. [PubMed]
  • Tan PK, Waites C, Liu Y, Krantz DE, Edwards RH. A Leucine-based Motif Mediates the Endocytosis of Vesicular Monoamine and Acetylcholine Transporters. J Biol Chem. 1998;273:17351–17360. [PubMed]
  • Tsen G, Williams B, Allaire P, Zhou Y-D, Ikonomov O, Kondova I, Jacob MH. Receptors with opposing functions are in postsynaptic microdomains under one presynaptic terminal. Nature Neurosci. 2000;3:126–132. [PubMed]
  • Voglmaier SM, Kam K, Yang H, Fortin DL, Hua Z, Nicoll RA, Edwards RH. Distinct endocytic pathways control the rate and extent of synaptic vesicle protein recycling. Neuron. 2006;51:71–84. [PubMed]
  • Whim MD. Near simultaneous release of classical and peptide cotransmitters from chromaffin cells. J Neurosci. 2006;26:6637–6642. [PubMed]
  • Witkovsky P. Dopamine and retinal function. Doc Ophthalmol. 2004;108:17–40. [PubMed]
  • Wightman RM, Jankowski JA, Kennedy RT, Kawagoe KT, Schroeder TJ, Leszczyszyn DJ, Near JA, Diliberto EJ, Jr, Viveros OH. Temporally resolved catecholamine spikes correspond to single vesicle release from individual chromaffin cells. Proc Natl Acad Sci USA. 1991;88:10754–10758. [PMC free article] [PubMed]
  • Xin D, Bloomfield SA. Comparison of the responses of AII amacrine cells in the dark- and light-adapted rabbit retina. Vis Neurosci. 1999;16:653–665. [PubMed]
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...