Confocal laser scanning microscopy and ...

R E S E A RC H AR T I C L E

Confocal Laser Scanning Microscopy and Ultrastructural Study of VGLUT2 Thalamic Input to Striatal Projection Neurons in Rats Wanlong Lei,1* Yunping Deng,2 Bingbing Liu,1 Shuhua Mu,1 Natalie M. Guley,2 Ting Wong,2 and Anton Reiner2* 1 2

Department of Anatomy, Zhongshan Medical School of Sun Yat-Sen University, Guangzhou, 510080, PR China Department of Anatomy & Neurobiology, University of Tennessee Health Science Center, Memphis, Tennessee 38163

ABSTRACT We examined thalamic input to striatum in rats using immunolabeling for the vesicular glutamate transporter (VGLUT2). Double immunofluorescence viewed with confocal laser scanning microscopy (CLSM) revealed that VGLUT2þ terminals are distinct from VGLUT1þ terminals. CLSM of Phaseolus vulgaris-leucoagglutinin (PHAL)-labeled cortical or thalamic terminals revealed that VGLUT2 is rare in corticostriatal terminals but nearly always present in thalamostriatal terminals. Electron microscopy revealed that VGLUT2þ terminals made up 39.4% of excitatory terminals in striatum (with VGLUT1þ corticostriatal terminals constituting the rest), and 66.8% of VGLUT2þ terminals synapsed on spines and the remainder on dendrites. VGLUT2þ axospinous terminals had a mean diameter of 0.624 lm, while VGLUT2þ axodendritic terminals a mean diameter of 0.698 lm. In tissue in which we simultaneously immunolabeled thalamostriatal terminals for VGLUT2 and striatal neurons for D1 (with about half of spines

immunolabeled for D1), 54.6% of VGLUT2þ terminals targeted D1þ spines (i.e., direct pathway striatal neurons), and 37.3% of D1þ spines received VGLUT2þ synaptic contacts. By contrast, 45.4% of VGLUT2þ terminals targeted D1-negative spines (i.e., indirect pathway striatal neurons), and only 25.8% of D1-negative spines received VGLUT2þ synaptic contacts. Similarly, among VGLUT2þ axodendritic synaptic terminals, 59.1% contacted D1þ dendrites, and 40.9% contacted D1-negative dendrites. VGLUT2þ terminals on D1þ spines and dendrites tended to be slightly smaller than those on D1-negative spines and dendrites. Thus, thalamostriatal terminals contact both direct and indirect pathway striatal neurons, with a slight preference for direct. These results are consistent with physiological studies indicating slightly different effects of thalamic input on the two types of striatal projection neurons. J. Comp. Neurol. 521:1354–1377, 2013. C 2012 Wiley Periodicals, Inc. V

INDEXING TERMS: basal ganglia; striatum; thalamostriatal; VGLUT2; intralaminar thalamus; parafascicular nucleus

The cerebral cortex gives rise to a major excitatory input to the striatum that provides it with an instructive signal critical for its role in motor control (Gerfen, 1992; Wilson, 1992). The cortical input mainly ends as terminals that make asymmetric synaptic contact with dendritic spines of striatal projection neurons, which make up the vast majority of striatal neurons (Albin et al., 1989; Reiner and Anderson, 1990; Gerfen. 1992). The corticostriatal input arises from two neuron types, an intratelencephalically projecting (IT) type found predominantly in layer III and upper layer V, and a pyramidal tract (PT) type found primarily in lower layer V (Wilson, 1987; Cowan and Wilson, 1994; Levesque et al., 1996a,b; Levesque and Par-

C 2012 Wiley Periodicals, Inc. V

1354

ent, 1998; Wright et al., 1999, 2001; Reiner et al., 2003; Parent and Parent, 2006). PT-type corticostriatal neurons preferentially contact striatal neurons projecting to the Grant sponsor: National Institutes of Health; Grant numbers: NS-19620, NS-28721 and NS-57722 (to A.R.); Grant sponsor: National Science Foundation of China; Grant numbers: 31070941, 30770679, 20831006; Grant sponsor: Major State Basic Research Development Program of China; Grant number: 973 Program, No. 2010CB530004 (to W.L.). *CORRESPONDENCE TO: Dr. Anton Reiner, Department of Anatomy & Neurobiology, University of Tennessee Health Science Center, 855 Monroe Ave., Memphis, TN 38163. E-mail: [email protected] or Dr. Wanlong Lei, Department of Anatomy, Zhongshan Medical School of Sun Yat-Sen University, 74 Zhongshan Rd 2, Guangzhou, 510080, PR China. E-mail: [email protected] Received May 23, 2012; Revised August 31, 2012; Accepted October 2, 2012 DOI 10.1002/cne.23235 Published online October 10, 2012 in Wiley Online Library (wileyonlinelibrary.com)

The Journal of Comparative Neurology | Research in Systems Neuroscience

521:1354–1377 (2013)

Thalamic input to rat striatum

external segment of globus pallidus (GPe), while IT-type cortical neurons preferentially target striatal neurons projecting to the internal pallidal segment (GPi) or the substantia nigra pars reticulata (SNr) (Lei et al., 2004; Cepeda et al., 2008; Reiner et al., 2010). The striatum also receives a substantial excitatory input from the thalamus, which ends in large part on the spines and dendrites of striatal projection neurons (Wilson et al., 1982; Smith et al., 2004). The thalamic projection is topographically organized and arises heavily from intralaminar, mediodorsal, and midline thalamic nuclei (IMMC) (Berendse and Groenewegen, 1990; Groenewegen and Berendse, 1994), but also from specific sensory nuclei of the thalamus. The IMMC thalamic regions projecting to striatum receive polysensory cortical and brainstem input and a feedback projection from the internal segment of the globus pallidus (GPi). Although the precise role of this input is uncertain, it is thought to play a role in attentional mechanisms concerning motor planning and preparedness (Smith et al., 2004, 2009, 2011; Kato et al., 2011). To further characterize the role of this input, we examined the thalamic input to striatum, with a particular interest in determining the relative abundance of axospinous versus axodendritic contacts by thalamostriatal terminals, in comparison to corticostriatal terminals, and in assessing if thalamostriatal terminals differ in their targeting of direct and indirect pathway striatal neurons. Prior studies report that such a difference may exist, but the data are conflicting (Sidibe and Smith, 1996; Salin and Kachidian, 1998; Giorgi et al., 2001; Bacci et al., 2004). Excitatory thalamic projection neurons use the vesicular glutamate transporter VGLUT2 for packaging glutamate in synaptic vesicles, while excitatory cortical neurons use VGLUT1 (Fremeau et al., 2001, 2004; Herzog et al., 2001; Varoqui et al., 2002; Fujiyama et al., 2004). To selectively study thalamostriatal synaptic terminals, we used VGLUT2 immunolabeling. We confirmed that VGLUT2 immunolabeling provides a means for selectively viewing thalamostriatal terminals, and then used VGLUT2 immunolabeling to characterize the thalamic input to striatum at the electron microscopy (EM) level. Our results indicate that about 40% of the excitatory input to striatum arises from thalamus, and that thalamostriatal terminals somewhat more commonly contact direct pathway neurons than indirect pathway neurons.

MATERIALS AND METHODS Animals and experimental plan Results from 16 adult male Sprague–Dawley rats (obtained from Harlan, Indianapolis, IN) are presented here, and all animal use was carried out in accordance

with the National Institutes of Health Guide for Care and Use of Laboratory Animals, Society for Neuroscience Guidelines, and University of Tennessee Health Science Center Guidelines. Nine rats were used for EM immunolabeling, three additional rats were used for light microscopy (LM) immunolabeling, two rats were used for Phaseolus vulgaris-leucoagglutinin (PHAL) anterograde labeling of corticostriatal terminals, and two rats were used for PHAL labeling of thalamostriatal terminals.

PHAL injection To label thalamostriatal terminals, PHAL was injected into the parafascicular nucleus (PFN) of the intralaminar thalamus, and to label corticostriatal terminals, PHAL was injected into layer 5 of primary motor cortex (M1). The rats were deeply anesthetized with ketamine (0.33 ml/ 500g) and xylazine (0.16 ml/500g), and 2.5% PHAL (Vector Laboratories, Burlingame, CA) in 0.01 M sodium phosphate buffer (pH 8.0) was iontophoresed into PFN or M1 using 5 lA positive current pulses (7 seconds on, 7 seconds off) for 30 minutes. Coordinates were from the Paxinos and Watson (2009) rat brain stereotaxic atlas. The PHAL-injected rats were allowed to survive for 7–10 days before being sacrificed, and the four rats injected with PHAL, as well as the three rats used for LM VGLUT localization, were anesthetized and transcardially perfused with 100 ml normal saline (0.9% NaCl), followed by 400 ml of 4% paraformaldehyde, 0.1 M lysine, 0.1 M sodium periodate in 0.1 M sodium phosphate buffer (PB) (pH 7.4). Brains were removed and postfixed in the same fixative for another 4–6 hours at 4 C. Brains were then cryoprotected in 20% sucrose, 10% glycerol in 0.1 M PB at 4 C, and transverse 40-lm sections cut frozen on a sliding microtome. Sections rostral to the anterior commissure were used for VGLUT immunolabeling.

LM visualization of VGLUT Single or multiple immunofluorescence was carried out to examine the relative localization of VGLUT1 and VGLUT2 in striatal axons and terminals, and to determine the extent to which they were in separate terminals. For these studies we first determined whether a guinea pig VGLUT2 antibody and a rabbit VGLUT2 antibody labeled the same set of striatal terminals (Table 1). Then as the next step (having shown complete coincidence between the two anti-VGLUT2 in their labeling patterns), we examined the colocalization of VGLUT2 and VGLUT1 in striatal terminals using the rabbit anti-VGLUT2 and a guinea pig VGLUT1 antibody (Table 1). For these studies sections were incubated for 72 hours at 4 C either in the guinea pig anti-VGLUT2 (1:1,000) and rabbit anti-VGLUT2 (1:2,000), or in guinea pig anti-VGLUT1 (1:1,000) and rabbit anti-VGLUT2 (1:2,000). After incubation in primary


1355

Lei et al.

TABLE 1. Antibody Information Antibody

Type and host

Source

Catalog number

Vesicular glutamate transporter 1 (VGluT1)

Guinea pig polyclonal

Millipore/ Chemicon

AB5905


Guinea pig polyclonal

Millipore/ Chemicon

AB5907


Rabbit polyclonal

Sigma-Aldrich

V2514

D1 dopamine receptor

Rat monoclonal

Sigma-Aldrich

D-187

Phaseolus vulgaris agglutinin (EþL) (PHAL)

Rabbit polyclonal

Vector Labs

AS-2300

Antigen

Dilution used

Synthetic peptide from rat VGLUT1 C-terminus (amino acids 542-560): GATHSTVQPPRPPPPVRDY Synthetic peptide from rat VGLUT2 C-terminus (amino acids 565-582): VQESAQDAYSYKDRDDYS Synthetic peptide located near the C-terminus of rat VGLUT2 (amino acids 520-538): HEDELDEETGDITQNYINY 97 amino acid C-terminal fragment of human D1 fused to glutathione: LCPATNNAIE-TVSINNNGAA-MFSSHHEPRGSISKECNLVY-LIPHAVHSSE-DIKKEEAAGIARPLEKLPSA-LSVILDYDTD-VSLEKIQPITQNGQHPT Purified 275aa Phaseolus vulgaris agglutinin (EþL)

1:1,000 (LM) 1:5,000 (EM) 1:1,000 (LM) 1:5,000 (EM) 1:2,000 (LM)

1:500 (EM)

1:250 (LM)

Detail on the commercial source, catalog number, animal host, target antigen, and working concentration for the antibodies used in the present study. Information on antibody specificity testing is provided in the text.

antibody at 4 C with gentle agitation, the tissue was rinsed three times, and the secondary antibody incubation carried out. The sections were incubated for 2 hours at room temperature (with gentle agitation) in a secondary antisera mixture that contained an Alexa 594-conjugated goat anti-guinea pig IgG (to detect the guinea pig anti-VGLUT1 or anti-VGLUT2) and an Alexa 488-conjugated goat antirabbit IgG (to detect the rabbit antiVGLUT2). Both secondaries were from Chemicon (Temecula, CA) and were diluted at 1:200. Sections were then rinsed three times in 0.1 M PB, mounted on gelatincoated slides, and coverslipped with ProLong antifade medium (Molecular Probes, Eugene, OR). Sections were viewed and images captured using a Zeiss 710 confocal laser scanning microscope (CLSM), using a 40 oil or 60 oil objective. Z-stack serial images were collected at 1 lm (40 oil), or 0.5 lm (60 oil) steps from dorsolateral striatum. Note that some single-label tissue was also prepared using the peroxidase-antiperoxidase method as detailed in prior studies (Deng et al., 2006, 2007).

LM visualization of PHAL and VGLUT Immunofluorescence for VGLUT combined with immunofluorescence PHAL visualization was used to confirm VGLUT2 localization to thalamostriatal terminals. Sections from the cases with intralaminar thalamic or M1 injection of PHAL were incubated for 72 hours at 4 C in a primary antisera cocktail containing either guinea pig VGLUT1 or guinea pig VGLUT2 (1:1,000), and rabbit antiPHAL (Table 1). After incubation in the primary antibody cocktail at 4 C with gentle agitation, the tissue was rinsed three times and the sections incubated for 2 hours

1356

at room temperature (with gentle agitation) in a secondary antisera mixture that contained an Alexa 488-conjugated goat anti-guinea pig IgG (to detect the VGLUT) and an Alexa 594-conjugated goat antirabbit IgG (to detect the PHAL). Both the Alexa 488-conjugated goat antiguinea pig IgG and the Alexa 594-conjugated goat antirabbit IgG were from Molecular Probes and used at a 1:200 dilution. All sections were then rinsed three times in 0.1 M PB, mounted on gelatin-coated slides, and coverslipped with ProLong antifade medium (Molecular Probes). Sections were viewed using a Zeiss 710 CLSM.

EM immunolabeling for VGLUT1 or VGLUT2 In EM single-label studies we characterized the ultrastructure of thalamostriatal terminals in comparison to corticostriatal terminals using immunolabeling for VGLUT2 and VGLUT1, respectively. For the EM studies, rats were deeply anesthetized with 0.8 ml of 35% chloral hydrate in saline, then exsanguinated by transcardial perfusion with 100 ml of 6% dextran in PB, followed by 400 ml of 3.5% paraformaldehyde / 0.6% glutaraldehyde / 15% saturated picric acid in PB (pH 7.4). The brain of each rat was removed, postfixed overnight in 3.5% paraformaldehyde / 15% saturated picric acid in PB, and then sectioned at 50 lm on a vibratome. Tissue was subsequently processed with guinea pig anti-VGLUT1 or antiVGLUT2. Sections were first pretreated with 1% sodium borohydride in 0.1 M PB for 30 minutes followed by incubation in 0.3% H2O2 solution in 0.1 M PB for 30 minutes. To carry out conventional single-label immunohistochemistry, sections were incubated for 72 hours at 4 C in primary antiserum diluted 1:5,000 (VGLUT1) or 1:5,000



(VGLUT2) with 0.1 M Tris buffer containing 4% normal goat serum / 1.5% bovine serum albumin. Sections were then rinsed and incubated in donkey anti-guinea pig IgG diluted 1:80 in 0.1 M Tris buffer (ph7.4), followed by incubation in the appropriate guinea pig PAP complex diluted 1:200 in 0.1 M Tris buffer (pH 7.4), with each incubation at room temperature for 1 hour. The sections were rinsed between secondary and PAP incubations in three 5-minute washes of PB. Subsequent to the PAP incubation, the sections were rinsed with three to six 10-minute washes in 0.1 M PB, and a peroxidase reaction using diaminobenzidine (DAB) carried out. After the PB rinses the sections were immersed for 10–15 minutes in 0.05% DAB (Sigma, St. Louis, MO) in 0.1 M PB (pH7.2). Hydrogen peroxide was then added to a final concentration of 0.01% and the sections were incubated in this solution for an additional 15 minutes, then washed six times in PB. Some sections to be viewed by LM were mounted onto gelatin-coated slides, dried, and dehydrated, cleared with xylene, and coverslipped with Permount (Fisher Scientific, Pittsburgh, PA). Tissue to be examined by EM was rinsed, dehydrated, and flatembedded in plastic as described below.

VGLUT2 and D1 immunolabeling We also double-labeled tissue for simultaneous visualization of VGLUT2-immunolabeled thalamostriatal terminals and D1-immunolabeled neurons for EM viewing using methods similar to those described previously (Reiner et al., 2000, 2003; Lei et al., 2004; Deng et al., 2006). Numerous published studies show that D1 dopamine receptors are referentially localized to those striatal neurons that have their major projection to GPi/SNr and a collateral projection to the GPe (Gerfen et al., 1990; LeMoine and Bloch, 1995; Deng et al., 2006; Lobo et al., 2006; Doyle et al., 2008; Shuen et al., 2008). The D1enriched type of striatal projection neuron also preferentially contains substance P and is termed the direct pathway striatal neuron type. By contrast, the type of striatal projection neuron that projects only to the GPe is rich in enkephalin and the D2-type dopamine receptor, but poor in the D1-type dopamine receptor (LeMoine and Bloch, 1995; Deng et al., 2006; Wang et al., 2006; Doyle et al., 2008). This neuron type is termed the indirect pathway striatal neuron type. Tissue from three of the same animals was used as in our single-label EM studies of VGLUT localization. The sections were first pretreated with 1% sodium borohydride in 0.1 M PB for 30 minutes followed by incubation in 0.3% H2O2 solution in 0.1 M PB for 30 minutes. VGLUT2 was then visualized using immunolabeling as described above. These sections were subsequently washed six times in PB and immunohistochemical labeling using a rat monoclonal anti-D1 antibody (Table 1) was

carried out, using a brown DAB reaction to visualize the D1 immunolabeling, as described above. Further details about the specificity of the anti-D1 are provided below. For each case, some sections were mounted onto gelatincoated glass slides, dried, dehydrated, cleared with xylene, and coverslipped with Permount (Fisher Scientific) for LM viewing. Tissue to be examined at the EM level was rinsed, dehydrated, and flat-embedded in plastic, as described in the following section. In the tissue prepared by double-DAB labeling, VGLUT2-immunolabeled terminals can readily be distinguished from D1-immunolabeled dendritic spines and dendrites of striatal neurons because they are morphologically distinct structures. Moreover, VGLUT2 is not found in striatal neurons, and thus VGLUT2-immunolabeling does not label the intrastriatal terminals, dendrites, or spines of striatal neurons (Fremeau et al., 2001, 2004). Finally, D1 immunolabeling of excitatory intrastriatal synaptic terminals is rare (only 3.1% of asymmetric axospinous synaptic terminals immunolabel for D1) and very light, and can generally be distinguished from the intense labeling of excitatory intrastriatal synaptic terminals obtained with VGLUT2 immunolabeling (Hersch et al., 1995; Lei et al., 2004). Thus, the use of double-DAB labeling did not significantly confound our EM interpretations or analysis.

Preparation of tissue for EM Following immunolabeling as described above, sections processed for EM viewing were rinsed in 0.1 M sodium cacodylate buffer (pH 7.2), postfixed for 1 hour in 2% osmium tetroxide (OsO4) in 0.1 M sodium cacodylate buffer, dehydrated in a graded series of ethyl alcohols, impregnated with 1% uranyl acetate in 100% alcohol, and flat-embedded in Spurr’s resin (Electron Microscopy Sciences, Fort Washington, PA). For the flat-embedding, the sections were mounted on microslides pretreated with liquid releasing factor (Electron Microscopy Sciences). The Spurr’s resin-embedded sections were examined light microscopically for the presence of VGLUT-immunolabeled axons and terminals in striatum, and in some cases D1þ structures as well. Pieces of embedded tissue were cut from the dorsolateral (motor) striatum and glued to carrier blocks, and ultrathin sections were cut from these specimens with a Reichert ultramicrotome. The sections were mounted on mesh grids, stained with 0.4% lead citrate and 4.0% uranyl acetate using an LKB Ultrastainer, and finally viewed and images captured with a JEOL 2000EX electron microscope.

Antibodies used Both guinea pig VGLUT antisera used here (Table 1) are highly selective for their target antigens (Fremeau et al., 2001; Montana et al., 2004). VGLUT1 antibody specificity


1357

Lei et al.

has been demonstrated by western blot analysis of rat cerebral cortex (Melone et al., 2005), and by immunogen block of retinal immunolabeling (W€assle et al., 1998). Melone et al. (2005) also showed that immunofluorescence with Chemicon anti-VGLUT1 nearly completely overlapped that for a previously well-characterized antibody against VGLUT1, although its target was called the brain-specific Na-dependent inorganic phosphate cotransporter (BNPI) at that time (Bellocchio et al., 1998). Montana et al. (2004) showed the specificity of the VGLUT2 antiserum in western blots of rat cerebral cortex, and W€assle et al. (2006) reported that preadsorption of the VGLUT2 antiserum with its immunogen peptide blocked immunostaining in mouse retina. VGLUT2 is also known as the differentiation-associated Na-dependent inorganic phosphate cotransporter (DNPI). The amino acid sequence for the immunogen for the rabbit VGLUT2 antibody used here (Table 1) is identical to that in mouse and human VGLUT2 and has no homology to VGLUT1. Western blotting by the manufacturer confirms antibody specificity. The anti-PHAL antibody (Vector) was generated against Phaseolus vulgaris agglutinin (EþL), and its selectivity is shown by the absence of labeling in tissue that has not been injected with PHAL. Western blots have shown that the anti-D1 rat monoclonal antibody used here selectively recognizes the D1 C-terminus protein as a single protein band at the predicted size of 65–75 kDa, but not the closely related D2, D3, D4, or D5 (Hersch et al., 1995). The distribution of D1þ perikarya in rat brain using this antibody is identical to that obtained by in situ hybridization (Gerfen et al., 1990; LeMoine and Bloch, 1995), as well as with a wellcharacterized and selective rabbit polyclonal anti-D1 antibody (Levey et al., 1993; Hersch et al., 1995). Notably, the mouse monoclonal anti-D1 antibody labels about half of the perikarya in rat striatum, which mainly represent the neurons of the direct pathway (Hersch et al., 1995; Deng et al., 2006).

EM analysis Analysis and quantification was carried out on random fields using digital EM images in nine rats (R1, R2, R4, R7, R8, R9, CR1, CR2, CR5). We focused on dorsolateral somatomotor striatum at the level of the anterior commissure, which is poor in striosomes (although not entirely devoid) and the major target of intralaminar thalamus (Gerfen, 1992; Desban et al., 1993; Berendse and Groenewegen, 1994; Wang et al., 2007). We used a reference series of sections immunolabeled for mu opiate receptor prepared previously (Deng et al., 2007) to aid in selection of the striosome-poor part of dorsolateral striatum. Thus, our findings mainly reflect matrisomal synaptology. We performed the analysis in the upper 5 lm of the sections,

1358

in which labeling was optimal, and avoided the very surface, where histology was poor. The size of terminals was determined by measuring them at their widest diameter parallel to and 0.1 lm before the postsynaptic density, and spines were identifiable by their small size, continuity with dendrites, prominent postsynaptic density, and/or the presence of spine apparatus (Wilson et al., 1983). Dendrites were identifiable by their size, oval or elongate shape, and the presence of microtubules and mitochondria. For VGLUT1 and VGLUT2, counts of labeled and unlabeled synaptic terminals on spines and dendrites were made to ascertain the percent of axospinous and axodendritic terminals in rat striatum that possess VGLUT1 or VGLUT2. Note that as projection neurons are the predominant neuron type in the striatum and the only type to possess dendritic spines, all VGLUT axospinous endings and the vast majority of VGLUT axodendritic endings are on projection neurons. Some small fraction of axodendritic VGLUT synaptic contacts, however, are on striatal interneurons. The data are presented as group means (6SEM) for the various traits analyzed for seven rats for VGLUT1 (R1, R2, R4, R8, CR1, CR2, CR5) and six rats for VGLUT2 (R1, R2, R4, R7, R8, R9), unless otherwise stated. In any event, the means with terminals pooled across animals closely resembled the group means when calculated from the mean data (e.g., terminal size) of the individual animals analyzed. In general, we used pooled data when animal numbers or terminal counts were low, or to derive smoother size frequency distribution curves. Three rats were analyzed to determine the relative frequencies of VGLUT2þ synaptic terminals on D1 spines and dendrites (R7, R8, R9). Note that in tissue in which D1 immunolabeling is optimized (i.e., about half of spines and dendrites are D1-positive), D1negative spines and dendrites are likely to largely belong to D2-type striatal projection neurons, as recently also noted by Day et al. (2006). Thus, we used the D1 immunolabeling to reach conclusions about the relative distributions of VGLUT2 terminals on direct and indirect pathway striatal projection neurons. We did not use D2 immunolabeling directly to identify D2-positive spines and dendrites, since D2 is found on a high percentage of D1 striatal neurons (Deng et al., 2006). Typically about 75 VGLUT1-immunolabeled terminals and 50 VGLUT1negative terminals were characterized for the seven VGLUT1 cases analyzed for single-labeling, and about 125 VGLUT2-immunolabeled terminals and 115 VGLUT2negative terminals were characterized for the six VGLUT2 cases analyzed for single-labeling. In the VGLUT1VGLUT2 double-labeling studies, about 150 labeled terminals and seven unlabeled terminals were analyzed per case. Finally, for the VGLUT2-D1 double-label studies, about 150 VGLUT2-immunolabeled terminals and 180



Figure 1. Low power (A,B) and 40 light microscopic views (C,D) of VGLUT1 (A,C) and VGLUT2 (B,D) immunolabeling of rat brain. Images A and B present views at a coronal level anterior to the anterior commissure; note the presence of VGLUT2þ thalamostriatal terminals in cortical layer 4 in image B and the conspicuous absence of VGLUT1þ cortical terminals in cortical layer 4 in image A. The 40 views reveal the presence of numerous varicosities in striatum immunolabeled for VGLUT1 (C) and VGLUT2 (D), with the varicosities more than twice as abundant for VGLUT1 as VGLUT2.

VGLUT2-negative terminals were characterized for each case analyzed. Chi-square and t-tests were used for statistical analysis of the results. Images presented here were prepared using Adobe Photoshop CS (San Jose, CA). Contrast enhancement and/or sharpening were performed on some images.

RESULTS LM localization of VGLUT2 versus VGLUT1 in intrastriatal terminals In single-labeled tissue we observed that the striatum was enriched in terminals that immunolabeled for VGLUT1 as well as in terminals that immunolabeled for VGLUT2 (Fig. 1). Consistent with prior evidence that exci-

tatory thalamic projection neurons use the vesicular glutamate transporter VGLUT2 for packaging glutamate in synaptic vesicles, we observed that layer 4 of cortex was enriched in VGLUT2þ terminals (Fig. 1) (Herzog et al., 2001; Fremeau et al., 2001, 2004; Varoqui et al., 2002). By contrast, VGLUT1þ terminals were prominently localized to cortical layers 2–3 and 5–6 (Fig. 1), consistent with prior evidence that excitatory cortical neurons use VGLUT1 (Fremeau et al., 2001, 2004; Herzog et al., 2001; Varoqui et al., 2002). In striatum, numerous varicosities were observed in tissue immunolabeled for VGLUT1 or VGLUT2, with the VGLUT1þ varicosities more abundant than the VGLUT2þ varicosities (Fig. 1). To confirm that our VGLUT1 and VGLUT2 antisera detected separate populations of terminals in striatum,


1359

Lei et al.

we carried out immunofluorescence double-labeling, viewed by CLSM. We first compared terminals labeled with guinea pig anti-VGLUT2 to those labeled by rabbit anti-VGLUT2 in the same tissue. We examined Z-stacks at high magnification of multiple fields in dorsolateral striatum. This revealed that the immunofluorescent labeling only penetrated 5–7 lm from the surface, and labeling was only optimal within a 4 lm zone from the surface. In this zone in which labeling was optimized, we found that all intrastriatal puncta (i.e., >0.5 lm wide structures representing presumptive terminals) labeled with guinea pig anti-VGLUT2 were also immunolabeled with rabbit antiVGLUT2, and vice versa (Figs. 2A,C,E, 3A,C,E). This then allowed us to use rabbit anti-VGLUT2 and guinea pig antiVGLUT1 in double-label studies to determine if VGLUT1 and VGLUT2 are in separate populations of terminals in the striatum. We again found that immunofluorescent labeling for both antibodies only penetrated 5–7 lm from the surface. We quantitatively analyzed Z-stacks of 6–16 fields at high magnification in each of three high-resolution CLSM images of dorsolateral striatum from each of three rats, within the 4 lm zone from the surface. In the separate VGLUT1 and VGLUT2 images we used thresholding with ImageJ to measure the areas occupied by VGLUT1 and VGLUT2 terminals and preterminal axons. Overall, we found that VGLUT1 puncta occupied 2.73 times more territory than VGLUT2 puncta in dorsolateral striatum, reflecting either greater size and/or greater abundance. In merged VGLUT1–VGLUT2 red-green images, we then measured the very small area occupied by double-labeled terminals. Our results showed that only 1.4% of intrastriatal puncta area labeled with rabbit antiVGLUT2 was also immunolabeled with guinea pig antiVGLUT1 (Figs. 2B,D,E, 3B,D,E), and only 0.55% of intrastriatal puncta area labeled for VGLUT1 also immunolabeled for VGLUT2 (Fig. 2B,D,E). Thus, our evidence suggests that VGLUT1 and VGLUT2 are in nearly separate populations of terminals in the striatum, and that VGLUT1 terminals occupy about 2.5 times more territory than VGLUT2 terminals.

LM localization of VGLUT2 versus VGLUT1 in corticostriatal and thalamostriatal terminals To confirm that our labeling of VGLUT2 was specific for thalamostriatal terminals, we performed immunolabeling for VGLUT2 or VGLUT1 on sections in which thalamic terminals in striatum had been anterogradely labeled with PHAL from the PFN, or cortical terminals had been anterogradely labeled with PHAL from M1 (Figs. 4–6). We used PHAL rather than BDA10k for these studies because of the proclivity of BDA10k to track retrogradely and yield collateral labeling (Reiner et al., 2000). Thus, injections of cortex with BDA10k could yield some retrograde trans-

1360

port to thalamic neurons projecting to both cortex and striatum, potentially yielding collateral BDA10k labeling of thalamic terminals in striatum. Similarly, injections of PFN with BDA10k could yield some retrograde transport to cortical neurons projecting to both thalamus and striatum, potentially yielding collateral BDA10k labeling of cortical terminals in striatum. We thus used PHAL for anterograde labeling, which shows little such retrograde collateral labeling (Chen and Aston-Jones, 1998). For cortical injections, we confirmed there was no thalamic retrograde labeling, and for thalamic injections we confirmed there was no cortical retrograde labeling. We examined multiple fields at high magnification in high-resolution CLSM images within the 4-lm zone from the surface in which VGLUT labeling is optimal, in 10–30 images from each of our PHAL cases. Because of the narrow focal plane, PHALþ fibers were relatively sparse in any individual field. Both individual isolated PHALþ puncta (sometimes with associated short preterminal axons) and longer PHALþ fibers with regular varicosities were observed. Cortical and thalamic PHALþ axons were about 0.2–0.4 lm in diameter, and the varicosities were >0.5 lm in diameter. Because isolated varicosities and those associated with short axons (8 lm, VGLUT1þ varicosities were observed on average every 5.02 lm of corticostriatal axon length. For PHALþ thalamostriatal fibers longer than >8 lm, VGLUT2þ varicosities were observed on average every 4.07 lm of thalamostriatal axon length. Thus, VGLUT1 in striatum is highly specific for corticostriatal terminals, and VGLUT2 in striatum is specific for thalamostriatal terminals. Moreover, our results suggest that at least 90% of corticostriatal terminals contain VGLUT1 and at least 90% of thalamostriatal terminals contain VGLUT2. Note that because some puncta may, in fact, have been the tortuous portions of axons in crosssection, it may be that all corticostriatal terminals contain VGLUT1 and all thalamostriatal terminals contain VGLUT2. Note that we tested immunolabeling for synaptophysin using a mouse monoclonal antibody in an effort to better define terminals in the PHAL and VGLUT tissue, but the resolution of the synaptophysin immunolabeling at the high magnification required in our studies was not adequate to substantially aid in our unambiguous discernment of synaptic terminals.



Figure 2. CLSM images in (A,C,E) present a single field from a double-labeled section, showing the complete colocalization of immunolabeling in rat striatum for guinea pig (GP) anti-VGLUT2 (A) and rabbit (Rb) anti-VGLUT2 (C), as shown by the complete labeling overlap in the merged image (E) for (A,C). Conversely, CLSM images in (B,D,F) present a single field from a double-labeled section, showing the near absence of colocalization of immunolabeling in rat striatum for guinea pig (GP) anti-VGLUT1 (B) and rabbit (Rb) anti-VGLUT2 (D), as shown by the meager overlap in the merged image (F) for (B,D).


1361

Lei et al.

Figure 3. Detail of each CLSM image shown in Figure 2. Images in (A,C,E) present magnified views of the upper left from Fig. 2A,C,E, respectively. Similarly, images (B,D,F) present magnified views of the upper left from Fig. 2B,D,F, respectively. These magnified views make it more possible to resolve individual terminals, and thereby confirm: 1) the complete colocalization seen in rat striatum for guinea pig (GP) anti-VGLUT2 (A) and rabbit (Rb) anti-VGLUT2 (C), as further evidenced by the complete labeling overlap in the merged image (E) for (A,C); and 2) the near absence of colocalization in rat striatum for guinea pig anti-VGLUT1 (B) and rabbit anti-VGLUT2 (D), as shown by the absence of evident overlap in the merged image (F) for (B,D).

1362



Figure 4. CLSM views of immunofluorescence for VGLUT1 (A) or VGLUT2 (B) in fields with fluorescent PHAL labeling of corticostriatal axons and terminals (C,D). Note that corticostriatal terminals in (C) immunolabel for VGLUT1 but those corticostriatal terminals in (D) do not immunolabel for VGLUT2. This can be seen more clearly in the merged images (E,F).


1363

Lei et al.

Figure 5. CLSM views of immunofluorescence for VGLUT1 (A) or VGLUT2 (B) in fields with fluorescent PHAL labeling of thalamostriatal axons and terminals (C,D). Note that thalamostriatal terminals in (C) do not immunolabel for VGLUT1 but those thalamostriatal terminals in (D) do immunolabel for VGLUT2. This can be seen more clearly in the merged images (E,F).

1364



Figure 6. Detail of CLSM images shown in Figures 4 and 5. Images in (A,C,E) present magnified views of the lower left from images Fig. 4A,C,E, respectively. Similarly, images (B,D,F) present magnified views of the upper left from images Fig. 5B,D,F, respectively. These magnified views make it more possible to resolve individual terminals, and thereby confirm: 1) PHAL-labeled corticostriatal varicosities that are evident as such by their thickness (arrows) are characteristically immunolabeled for VGLUT1 (A,C,E); and 2) PHAL-labeled thalamostriatal varicosities that are evident as such by their thickness (arrows) are characteristically immunolabeled for VGLUT2 (B,D,F).


1365

Lei et al.

EM localization of VGLUT1 and VGLUT2 At the EM level, we found that VGLUT2þ terminals tended to be rounded, and formed asymmetric synaptic contacts with spine heads and dendrites of striatal neurons (Fig. 7). VGLUT1þ terminals also formed asymmetric synaptic contacts with spine heads and dendrites of striatal neurons, although VGLUT1þ terminals tended to be more varied in size and shape (Fig. 8). Counts of random striatal fields indicated that 85.5% of VGLUT1þ terminals synapse on spines and the remainder on dendrites (Table 2). By contrast, 66.8% of VGLUT2þ terminals synapsed on spines, and the remainder on dendrites. The relative spine versus dendrite targeting for VGLUT1 was significantly different from that for VGLUT2 by chi-square. Taking all VGLUT1þ and VGLUTþ synaptic terminals into consideration, our results indicate that thalamic terminals constitute about 40% of all striatal VGLUTþ terminals. We also found that 33.4% of axospinous asymmetric synaptic terminals immunolabeled for VGLUT2, while 65.9% of axospinous asymmetric synaptic terminals immunolabeled for VGLUT1 (Table 2), a significant difference by ttest. Since the sum of these two frequencies (99.3%) approximates 100%, and since the cortex and thalamus are the only known sources of excitatory input to striatal projection neuron spine heads (Gerfen, 1992), our EM results suggest that VGLUT2 immunolabeling detects all (or nearly all) axospinous thalamostriatal terminals on striatal projection neuron spines and VGLUT1 immunolabeling detects all (or nearly all) corticostriatal axospinous terminals on striatal projection neurons. Moreover, these results suggest that about 35% of striatal projection neuron spines receive thalamic input and about 65% receive cortical input. Note, however, that when we combined VGLUT1 and VGLUT2 immunolabeling for tissue from two of the rats used in the VGLUT1 and VGLUT2 single-label studies, we found that only 96.4% of axospinous synaptic terminals labeled for both. Thus, given that our LM data suggest that no more than about 1% of all axospinous terminals contain both VGLUT1 and VGLUT2, we cannot rule out the possibility that a small percent (