In the published paper of Nakamura et al. (2007), the double primes referred to in Figs 1, 2, 7 and 8 were corrupted throughout the text and legends at an early conversion stage of production. We apologise for any inconvenience caused and, for clarity, reproduce the whole paper again here on pages 1033–1046.
2007) Transiently increased colocalization of vesicular glutamate transporters 1 and 2 at single axon terminals during postnatal development of mouse neocortex: a quantitative analysis with correlation coefficient. Eur. J. Neurosci., 26, 3054–3067., , , , , & (
Vesicular glutamate transporter 1 (VGLUT1) and VGLUT2 show complementary distribution in neocortex; VGLUT1 is expressed mainly in axon terminals of neocortical neurons, whereas VGLUT2 is located chiefly in thalamocortical axon terminals. However, we recently reported a frequent colocalization of VGLUT1 and VGLUT2 at a subset of axon terminals in postnatal developing neocortex. We here quantified the frequency of colocalization between VGLUT1 and VGLUT2 immunoreactivities at single axon terminals by using the correlation coefficient (CC) as an indicator in order to determine the time course and spatial extent of the colocalization during postnatal development of mouse neocortex. The colocalization was more frequent in the primary somatosensory (S1) area than in both the primary visual (V1) and the motor areas; of area S1 cortical layers, colocalization was most evident in layer IV barrels at postnatal day (P) 7 and in adulthood. CC in layer IV showed a peak at P7 in area S1, and at P10 in area V1 though the latter peak was much smaller than the former. These results suggest that thalamocortical axon terminals contained not only VGLUT2 but also VGLUT1, especially at P7–10. Double fluorescence in situ hybridization confirmed coexpression of VGLUT1 and VGLUT2 mRNAs at P7 in the somatosensory thalamic nuclei and later in the thalamic dorsal lateral geniculate nucleus. As VGLUT1 is often used in axon terminals that show synaptic plasticity in adult brain, the present findings suggest that VGLUT1 is used in thalamocortical axons transiently during the postnatal period when plasticity is required.
The postnatal development of neural circuitry in the mammalian neocortex is composed of precisely scheduled events including axonal elongation/ramification (Larsen & Callaway, 2005), dendritic branching (Maravall et al., 2004), synapse formation (Micheva & Beaulieu, 1996; De Felipe et al., 1997) and activity-dependent modification of the circuitry (for review, see Fox & Wong, 2005). Thus, developing cortical circuitry shows particular forms of plasticity, such as ocular dominance plasticity in the visual cortex (for review, see Hensch, 2005) and receptive field plasticity in the somatosensory barrel cortex (Stern et al., 2001), which is called ‘critical period’ plasticity. The critical period indicates that some molecular events demarcate the onset and duration of those particular forms of plasticity (for review, see Hensch, 2005). It is hence crucial to examine the temporal course of such molecular events that may serve as key processes in cortical circuitry formation.
l-Glutamate is the major excitatory neurotransmitter playing a central role in neocortical function, and is considered to be indispensable for maturation of neocortical circuitry. Vesicular glutamate transporters (VGLUTs) catalyse glutamate uptake from cytoplasm to synaptic vesicles (for reviews, see Fremeau et al., 2004b; Takamori, 2006), and are essential proteins for glutamatergic transmission. Of three VGLUT isoforms in mammals, VGLUT1 and VGLUT2 are the two major isoforms in the brain (for review, see Takamori, 2006). Interestingly, VGLUT1 is predominantly produced by neurons in telencephalic structures including neocortex, whereas VGLUT2 is largely expressed by neurons in diencephalic and lower brainstem regions (Ni et al., 1995; Hisano et al., 2000; Fremeau et al., 2001; Fujiyama et al., 2001; Herzog et al., 2001). Therefore, glutamatergic axon terminals of cortical and thalamic origins, respectively, have been assumed preferentially to display VGLUT1 and VGLUT2 immunoreactivities in the neocortex (Fujiyama et al., 2001). Furthermore, VGLUT2 is frequently used in the axon terminals that are known to show synaptic depression, such as thalamocortical, retinogeniculate and cerebellar climbing fiber terminals (Gil et al., 1997; Turner & Salt, 1998; Dittman et al., 2000; Yanagisawa et al., 2004). In contrast, VGLUT1 is often found in the terminals displaying synaptic facilitation and long-term synaptic plasticity, such as hippocampal glutamatergic terminals and cerebellar parallel fiber terminals (Creager et al., 1980; Bliss & Lømo, 1973; Ito & Kano, 1982; Dittman et al., 2000). These facts suggest that VGLUT isoforms are associated with the important synaptic characteristics.
We recently found transiently augmented colocalization of VGLUT1 and VGLUT2 in the primary somatosensory cortex during the early postnatal period by virtue of immunolabeling with antigen retrieval (Nakamura et al., 2005). This finding leads to a refinement of the simplified view that VGLUT subtypes are segregated in neocortical axon terminals, and suggests that VGLUT1 is temporarily expressed by thalamocortical terminals, which mainly use VGLUT2 in adulthood. The colocalization might be involved in thalamocortical synaptic plasticity specific to this period. In the present study, we attempted to quantify the frequency of VGLUT1–VGLUT2 colocalization at single axon terminals by using the correlation coefficient (CC), and reveal the exact time course and extent of the colocalization in developing mouse neocortical areas.
Materials and methods
The experiments were conducted in accordance with the rules of animal care by the Institute of Laboratory Animals, Graduate School of Medicine, Kyoto University. C57BL/6 mice were purchased from Japan SLC (Shizuoka, Japan). The day of birth was defined as postnatal day 0 (P0). All efforts were made to minimize the number of animals used.
For immunofluorescence labeling, 40 mice (five mice including both sexes per each age) of P1, P3, P5, P7, P10, P14, P22 and 12 weeks old (adult) were used. For in situ hybridization histochemistry, three P7 male mice and three 12-week-old male mice were used. Mice were deeply anesthetized with cooling on ice (P1–P3) or sodium pentobarbital (5.0 mg/100 g body weight; P5–P22 and adult mice). The mice were perfused transcardially with 5 mm sodium phosphate-buffered 0.9% (w/v) saline (PBS, pH 7.2), followed by perfusion for 30 min with 3–30 mL of 0.1 m sodium phosphate buffer (PB, pH 7.2) containing 4.0% (w/v) formaldehyde (4FA-PB). The brains were removed and postfixed at 4 °C overnight in the same fixative. After cryoprotection with 30% (w/v) sucrose in PBS, the brains were cut into 50-µm-thick frontal sections on a freezing microtome. For immunoelectron microscopy, three P7 male mice and three 12-week-old male mice were anesthetized, and perfused with PBS and then with 4% (w/v) paraformaldehyde and 0.05% (v/v) glutaraldehyde in 0.1 m PB. The brains were removed, postfixed with 4% (w/v) paraformaldehyde in 0.1 m PB for 4 h (adult) or overnight (P7) at 4 °C, and cut into 50-µm-thick frontal sections by a vibratome (Microslicer DTK-1000; Dosaka, Kyoto, Japan).
Affinity-purified anti-VGLUT1 guinea-pig IgG (Fujiyama et al., 2001) and affinity-purified anti-VGLUT2 rabbit IgG (Hioki et al., 2003) were produced against C-terminal nonadecapeptides of rat VGLUT1 and C-terminal nonacosapeptides of rat VGLUT2, respectively. The specificity of these antibodies has been characterized in adult rat tissue by Western blotting and immunohistochemistry (Fujiyama et al., 2001; Hioki et al., 2003) and in mouse tissue by preabsorption of antibody with antigen peptides (Nakamura et al., 2005).
Double immunofluorescence labeling
The frontal sections were cut into the hemispheres, one for VGLUT1 and VGLUT2 double immunolabeling, and the other for negative control staining with a mixture of normal IgGs (see below). To enhance VGLUT immunoreactivity, we applied antigen retrieval (Nakamura et al., 2005) to sections for the immunolabeling study: the sections were heated at 80 °C for 60 min in 10 mm citrate-NaOH (pH 6.0). After rinses in PBS, sections were incubated overnight with a mixture of 5 µg/mL guinea-pig anti-VGLUT1 and 5 µg/mL rabbit anti-VGLUT2 antibodies in PBS containing 0.3% (v/v) Triton X-100, 0.25% (w/v) λ-carrageenan (Sigma; preventing adhesion between sections), and 1% (v/v) donkey serum (PBS-XCD). For negative control labeling, the alternate hemispheres were incubated with 5 µg/mL each of normal guinea-pig and rabbit IgGs (Santa Cruz Biotechnology, Santa Cruz, CA, USA) instead of the anti-VGLUT antibodies. The sections reacted with anti-VGLUT antibodies and normal IgGs were separately postfixed in 4FA-PB for 15 min to immobilize IgGs, and then collected in the same incubation chamber to perform the following secondary antibody reaction in a competitive manner. The sections were incubated for 2 h with 50 µg/mL each of AlexaFluor647-conjugated anti-(guinea-pig IgG) goat antibody and AlexaFluor488-conjugated anti-(rabbit IgG) goat antibody (Molecular Probes, Eugene, OR, USA) in PBS-XCD, followed by 15 min of postfixation in 4FA-PB. Sections were then incubated with 1 µg/mL DAPI (Molecular Probes) in PBS for 5 min. All the antibody incubations were carried out at room temperature, followed by two washes in PBS containing 0.3% (v/v) Triton X-100. Sections were mounted on aminopropyltriethoxysilane-coated slides (Matsunami, Kishiwada, Japan), and coverslipped with 90% (v/v) glycerol, 2.5% (w/v) triethylenediamine (an antifading reagent) and 20 mm Tris-HCl, pH 7.2.
Low-magnification images were obtained with a confocal laser-scanning microscope, LSM5 PASCAL (Carl Zeiss). AlexaFluor488 and AlexaFluor647 were excited with 488- and 633-nm laser beams and observed through 510–530-nm and ≥ 650-nm emission filters, respectively. The negative control sections showed only very weak homogeneous background. Neocortical areas were identified according to brain atlases (Paxinos et al., 1991; Paxinos & Franklin, 2001). Cortical layers were determined on the basis of anti-VGLUT2 immunoreactivity and DAPI fluorescence.
Confocal images at high magnification were taken from both the double immunofluorescence-labeled sections and the corresponding negative controls. These images were obtained with a 63× oil-immersion objective lens (NA = 1.40; HCX PL APO CS, Leica Microsystems, Wetzlar, Germany) attached to a confocal microscope (TCS SP2; Leica). To minimize deviations in image acquisition, we performed confocal microscopy under constant conditions throughout the same analysis. The focus plane was set to be 2–4 µm beneath the surface of sections, where the highest immunofluorescence was observed. Scanning was done at a pinhole size of 1.0 Airy unit, and with line averaging of 16 times. Dual color images were captured in a sequential manner: AlexaFluor647 was excited with a 633-nm He/Ne laser at 10% power (visible acousto-optical tunable filter), and observed through an emission prism window of 650–700 nm (gain, 485.6 V; offset, −0.30); AlexaFluor488 was then excited with a 488-nm Ar laser at 5% power (visible acousto-optical tunable filter), and observed through an emission prism window of 500–535 nm (gain, 491.2 V; offset, 1.30). The images were stored in 512 × 512 pixels of 12-bit TIFF file format. Each pixel corresponds to an area of 35.76 × 35.76 nm tissue. To circumvent complexity of description, we hereafter refer to fluorescences of AlexaFluor488 and AlexaFluor647 as green and red, respectively.
Quantitative analysis of colocalization of VGLUT1 and VGLUT2
We employed the CC (Manders et al., 1992), intensity correlation quotient (ICQ; Li et al., 2004) and overlap coefficient (OC; Manders et al., 1993) as indices of the frequency of colocalization between VGLUT1 and VGLUT2 at putative single axon terminals (see Supplementary material, Appendix S1). These indices are all based on pixel-based quantitative analysis of confocal images. We used non-deconvolved single confocal images, but not deconvolved image stacks, for the analysis, because we tried deconvolution of image stacks and noticed it had little effect on colocalization analysis with CC (see supplementary Fig. S1). CC between VGLUT1 and VGLUT2 immunofluorescence signals in the confocal images can range from −1 to 1, where 1 means perfect overlap, and 0 means random distribution of the two antigens. ICQ for the fluorescence signals can show a range from −0.5 to 0.5, where 0.5 means perfect overlap of two antigens, and 0 means random overlap. For better comparison with CC and OC, we used 2× ICQ. OC for the fluorescence signals can range from 0 to 1, where 0 means no overlap, and 1 indicates perfect overlap.
Depending on NA of the lens, the limits of spatial resolution for the AlexaFluor488 and AlexaFluor647 dyes, respectively, were approximately 0.61 × 519/1.40 = 226 nm and 0.61 × 668/1.40 = 291 nm, which were sufficient for the size of axon terminals (0.5–3.0 µm in diameter) but not for small round vesicles (30–50 nm in diameter). In addition, VGLUT1 and VGLUT2 immunoreactivities in the rodent neocortex were assumed to be localized in glutamatergic axon terminals in adults (Fujiyama et al., 2001; Herzog et al., 2001) and in postnatal development (Minelli et al., 2003; Boulland et al., 2004; Nakamura et al., 2005). We thus consider that the present quantitative results indicate the frequency of VGLUT1–VGLUT2 colocalization at single glutamatergic axon terminals.
Exclusion of pixels saturated or not containing excitatory axon terminals
In 12-bit confocal images, pixels with an intensity of 0 or 4095 are deemed to be lacking linearity of fluorescence signals. We therefore carefully determined gain and offset values not to contain pixels with 0 or 4095 intensity when capturing confocal images. Only a few pixels with 0 or 4095 intensity in the red or green colour channels were excluded from calculation of CC, ICQ and OC.
In addition, pixels that are located in neuronal somata, dendritic processes, axonal fibers or glial processes should be excluded from the quantitative analysis, because the present analysis is intended for glutamatergic axon terminals. It has been reported that exclusion of such pixels improves fidelity of quantitative colocalization analysis for axon terminals (Silver & Stryker, 2000). For objective selection of the pixels containing glutamatergic axon terminals, we defined thresholds according to background immunofluorescence measured from the negative control sections. When both VGLUT1 and VGLUT2 immunofluorescence signals of a pixel were less than the thresholds, the pixel was considered to be devoid of glutamatergic axon terminals, and excluded from the calculation.
Because the maximum intensity in an image of normal IgG-treated control sections fluctuated between images presumably due to thermal noise arising in the photon detector, we instead used the 999/1000 quantile of intensity (i.e. the highest intensity after removal of the highest 0.1% of data) in each colour channel as the thresholds. For each region of an animal, we defined the threshold of each colour channel to be the mean value from three images of the control sections. Thus, the following quantitative colocalization analysis was carried out after exclusion of pixels that were not considered to be located at glutamatergic axon terminals.
Shuffling pixels in confocal images
Randomized confocal images were prepared by shuffling of pixels within an image and used for calculation of the coefficients and quotients, as described elsewhere (Costes et al., 2004). After the exclusion of pixels unsuitable for the present analysis, shuffling was done only for pixels in the green channel of an image, and the shuffled pixels were paired with pixels in the unchanged red channel. For each confocal image, shuffling was repeated three times, and the mean of coefficients from the three shuffled data was calculated as a representative value for the image unless otherwise stated.
All the calculations of coefficients and quotients from each confocal image, as well as pixel shuffling and calculation of the 999/1000 quantile of intensity, were done with custom-made plug-in programs (available at http://www.mbs.med.kyoto-u.ac.jp/imagej/index.html) combined with ImageJ software (National Institute for Health, Bethesda, MD, USA). The numerical data were imported to Excel software (Microsoft, Redmond, WA, USA) for further calculation. Finally, results were plotted into graphs with Igor Pro 5 software (Wave Metrics, Portland, OR, USA).
Double immunoelectron microscopy
Double labeling for VGLUT1 and VGLUT2 was made by the silver-intensified immunogold method combined with immunoperoxidase staining, as described previously (Fujiyama et al., 2006) with some modifications. Briefly, the vibratome sections were placed overnight at 4 °C in a mixture of 1 µg/mL guinea-pig anti-VGLUT1 antibody and 0.2 µg/mL rabbit anti-VGLUT2 antibody. Sections were incubated with a mixture of 1.4-nm gold particle-conjugated anti-(guinea-pig IgG) goat antibody (Nanoprobes, Yaphank, NY, USA) and 1 : 100-diluted biotinylated anti-(rabbit IgG) donkey antibody (Jackson Immunoresearch, West Grove, PA, USA) overnight at 4 °C. After silver-development with the HQ Silver enhancement kit (Nanoprobes), the sections were further incubated for 4 h with avidin-biotinylated peroxidase complex (elite ABC; Vector Laboratories, Burlingame, CA, USA) in PBS, and reacted with 0.02% (w/v) diaminobenzidine (DAB)-4HCl (Dojindo, Tokyo, Japan) and 0.001% (v/v) H2O2 in 50 mm Tris-HCl (pH 7.6) for 20–40 min. The sections were placed for 60 min in 1% (w/v) OsO4 in 0.1 m PB, counterstained for 1 h with 1% (w/v) uranyl acetate, dehydrated and flat-embedded in epoxy resin (Luveak 812; nacalai tesque, Kyoto, Japan). Once the resin was polymerized, the tissue specimens were cut into ultrathin sections on an Reichert-Nissei Ultracut S (Leica) ultramicrotome. The ultrathin sections were mounted on mesh grids and examined with an electron microscope (H-7100; Hitachi, Tokyo, Japan).
In situ hybridization histochemistry
Complementary DNA fragment of VGLUT1 (nucleotides 855–1788; GenBank accession number XM_133432.2) or VGLUT2 (nucleotides 848–2044; GenBank accession number NM_080853.2) was cloned into a vector pBluescript II KS (+) (Stratagene, La Jolla, CA, USA). Using the linearized plasmids as template, we synthesized sense and antisense single-strand RNA probes with a digoxigenin labeling kit or fluorescein labeling kit (Roche Diagnostics, Basel, Switzerland).
The procedures for nonradioactive in situ hybridization were as described elsewhere (Liang et al., 2000; Tochitani et al., 2001). In brief, for bright-field staining, free-floating sections were hybridized for 20 h at 70 °C with 1 µg/mL digoxigenin-labeled riboprobe for VGLUT1 or VGLUT2 in hybridization buffer containing 50% (v/v) formamide, 5× SSC (75 mm NaCl and 75 mm sodium citrate, pH 7), 2% (w/v) blocking reagent (Roche Diagnostics), 0.1% (w/v) N-lauroylsarcosine (NLS), and 0.1% (w/v) SDS. After washes and RNase treatment, hybridized sections were incubated with 1 : 1000-diluted sheep anti-digoxigenin antibody conjugated with alkaline phosphatase (Roche diagnostics) overnight at room temperature. After rinsing, the sections were reacted for several hours with 0.375 mg/mL nitroblue tetrazolium and 0.188 mg/mL 5-bromo-4-chloro-3-indolylphosphate (Roche Diagnostics) in 0.1 m Tris-HCl (pH 9.5), 0.1 m NaCl and 50 mm MgCl2. Sections were mounted on aminopropyltriethoxysilane-coated slides, cleared in xylene and coverslipped. Sense probes detected no signals higher than the background.
Double fluorescence in situ hybridization was performed as previously described in detail (Komatsu et al., 2005). In short, hybridization was done as described above, except that 1 µg/mL each of fluorescein-labeled VGLUT1 riboprobe and digoxigenin-labeled VGLUT2 riboprobe were used as a mixture. After washes and RNase treatment, hybridized sections were incubated with 1 : 2000-diluted peroxidase-conjugated anti-fluorescein sheep antibody (Roche Diagnostics) for 3 h, with TSA-Plus dinitrophenol kit (PerkinElmer, Wellesley, MA, USA) for 30 min, and then with a mixture of 1 : 250-diluted AlexaFluor488-conjugated anti-dinitrophenol rabbit antibody (Molecular Probes) and 1 : 1000-diluted alkaline phosphatase-conjugated anti-digoxigenin sheep antibody overnight. The sections were reacted with HNPP Fluorescence Detection kit (HNPP/FastRed TR; Roche Diagnostics) for 1 h and with 1 µg/mL DAPI (Molecular Probes) for 10 min. After mounting on aminopropyltriethoxysilane-coated slides, sections were coverslipped with an aqueous mounting medium, Permafluor (Beckman Coulter, Fullerton, CA, USA). Photographs were taken with LSM5 PASCAL. AlexaFluor488 and HNPP/FastRed TR reaction products, respectively, were excited with 488- and 543-nm laser beams and observed through 510–530- and ≥ 560-nm emission filers. The distributions of fluorescent mRNA signals for VGLUT1 and VGLUT2 were entirely different from each other, indicating lack of crossover between the two mRNA signals.
For the statistical analysis, data were evaluated with two-way anova with post-hoc Tukey's HSD test. The calculation was performed with Igor Pro 5. Statistical significance was determined at P < 0.01 or P < 0.05.
VGLUT1 and VGLUT2 immunoreactivities in developing mouse neocortex
VGLUT1 immunoreactivity in area S1 of the neocortex was almost absent in the superficial layers and very weak in layer VI at P3 (Fig. 1a and a′), clearly increased in all layers (Fig. 1b, b′ and c), and became almost homogeneously intense in adulthood except in layer IV (Fig. 1d). The primary motor (area M1) and visual (area V1) areas showed a similar increase in VGLUT1 immunoreactivity during postnatal development (Fig. 1e and f).
VGLUT2 immunoreactivity in the neocortex also increased during postnatal development, but the increase was less distinct than that of VGLUT1 immunoreactivity. In areas S1 and V1, layer IV showed the most intense VGLUT2 immunoreactivity from P3 to adulthood (Fig. 1a′′, b′′, c′, d′,f′). In area M1, where layer IV is absent, the deep part of layer III showed the strongest VGLUT2 immunoreactivity (Fig. 1e′). Layer I and the boundary between layers V and VI showed the second strongest VGLUT2 immunoreactivity, whereas layers II/III, V and VI showed weak VGLUT2 immunoreactivity throughout postnatal development.
Confocal images of the doubly immunofluorescence-labeled sections at higher magnification showed punctate immunoreactivities for VGLUT1 and VGLUT2 in the developing mouse neocortex (Fig. 2). These VGLUT1 and VGLUT2 immunoreactivities were often colocalized with each other, but the frequency of colocalization varied among cortical areas, layers and postnatal ages even under semiquantitative analysis by visual inspection. For instance, confocal images of layer IV barrels of P7 area S1 showed many punctate profiles immunopositive for both VGLUT1 and VGLUT2 (arrowheads; Fig. 2a′). In adult area S1 barrels, the overlap of VGLUT1 and VGLUT2 immunoreactivities was less frequently observed (Fig. 2b′). Confocal images of P7 area V1 layer IV and area M1 layer III showed a few punctate profiles positive for both VGLUT immunoreactivities (Fig. 2c′ and d′).
Difference in the frequency of colocalization between VGLUT1 and VGLUT2 immunoreactivities across neocortical layers
Of the three indices of frequency of colocalization, we mainly used CC for the following quantitative analyses, because CC was most reliable when the frequency of colocalization is reduced to chance level by pixel shuffling (see supplementary Table S1). We first examined the frequency of colocalization of VGLUT1 and VGLUT2 immunoreactivities at putative single axon terminals across different layers within neocortical areas. Using the doubly immunofluorescence-labeled sections of P7 and adult mice, we took consecutive confocal images (three images per row, 35–66 rows) along the axis perpendicular to the plane of the pial surface in area S1 barrel subfield and area V1. The mean CCs of each row were plotted against relative depth from the pia mater (Fig. 3).
In area S1 of P7 mice, the CC was largest (0.5–0.6) at layer IV barrels and smaller (0.0–0.2) in the other layers (Fig. 3a). This could not simply reflect the strongest VGLUT2 immunofluorescence in layer IV, because layer I showed CCs of approximately 0 despite its second strongest VGLUT2 immunofluorescence in areas S1 and V1 (Fig. 3, gray traces). In adult area S1, layers IV and II/III, respectively, showed smaller (0.2–0.4) and larger (0.2–0.4) CCs than those of P7 mice (Fig. 3b). In contrast, the CCs at layers I, V and VI showed almost no change from P7 to adulthood (0.0–0.2). P7 area V1 displayed a distinct result from P7 area S1 (Fig. 3c): the CCs ranged from −0.1 to 0.1 in all cortical layers, indicating less frequent VGLUT1–VGLUT2 colocalization in area V1 than in area S1 at P7.
We performed a similar analysis in a direction tangential to the pial surface in layer IV of area S1 barrel subfield at P7 and in adulthood to compare CCs of barrels to those of septa (three images per column, 26 columns). As shown in Fig. 4a, the barrels at P7 showed higher CCs (0.5–0.7) than the septa (0.4–0.5). However, in adulthood, the difference in CCs between the barrels and septa was unclear (Fig. 4b).
Postnatal changes in frequency of colocalization between VGLUT1 and VGLUT2 immunoreactivities
We then compared the CCs from the confocal images of various neocortical areas at different postnatal ages. We focussed on layer IV, because layer IV barrels showed the largest CCs among all layers of P7 area S1 (Fig. 3a). For area M1, in which layer IV is not defined, we instead chose the deep part of layer III. We also examined the striatal patch regions, where little VGLUT1–VGLUT2 colocalization at single axon terminals was reported (Fujiyama et al., 2004; Nakamura et al., 2005). The patch was determined by its stronger and weaker VGLUT2 immunoreactivity before P10 and after P14, respectively, than that of the matrix (Nakamura et al., 2005; Fujiyama et al., 2006). Data for areas V1 and M1 at P1, where VGLUT1 immunoreactivity was almost negative, were removed from the present results. Figure 5a shows changes in the CCs during postnatal development (n = 5; mean ± SD). Highly significant main effects on the CCs were found for both ages (F6,140 = 255.1, P = 3.2 × 10−63) and areas (F4,140 = 16.6, P = 2.1 × 10−14) by repeated-measures two-way anova with a significant interaction between ages and areas (F24,140 = 6.1, P = 1.9 × 10−12; data of P1 were not included for this anova).
In area S1 barrels, the CCs were increased from P1 to P7 and decreased from P7 to P10; a prominent peak of the CC (0.65 ± 0.11, mean ± SD) was shown at P7 (Fig. 5a). In the barrels, the CC at P7 was significantly different from those of any other ages (P = 1.19 × 10−5 to 0.00164; Tukey's HSD test) except at P5 (0.50 ± 0.12, P = 0.0994). The CCs in the barrels showed a slight, but nonsignificant, increase from P22 to adulthood (P = 0.996).
Neuronal lesions in the ventrobasal thalamic nuclei drastically reduced VGLUT2 immunoreactivity in layer IV of adult rat area S1 (Fujiyama et al., 2001), indicating that the vast majority of VGLUT2 immunoreactivity in layer IV may be located at thalamocortical axon terminals. To examine the contribution of thalamocortical terminals to the colocalization, we calculated CCs from the pixels positive for VGLUT2 immunoreactivity, and thereby estimated CCs at thalamocortical terminals. The CCs were 0.66 ± 0.11 at P7 and 0.26 ± 0.04 in adulthood, suggesting the presence of colocalization at thalamocortical terminals. Furthermore, CCs in VGLUT1-positive pixels were slightly lower (0.64 ± 0.13) at P7 and higher (0.29 ± 0.05) in adulthood than those in VGLUT2-positive pixels, although the differences were not significant.
Within area V1, the CCs significantly increased from P3 to P10 (P = 0.00341; Fig. 5a). After P10, the CCs decreased slightly, but not significantly (P = 0.840–0.999). Area V1 hence displayed a small peak in CCs (0.15 ± 0.05) at P10. In area M1 and the striatal patch, developmental changes in the CCs were not statistically significant (P = ∼1.00). The results for the striatal matrix were very similar to those of patch regions (data not shown).
When compared within the same age, area S1 barrels showed significantly larger CCs than areas V1 and M1 and the striatal patch from P3 to adulthood (P = 1.19 × 10−5 to 0.0346; Fig. 5a). Areas V1 and M1 showed similar CCs to those of each other at the same age (P = 0.746–1.00). The CCs of both the areas were higher than those of the striatal patch at the same age, with significant differences at P7 (P = 0.00167) and P10 (P = 0.0467) between area V1 and the striatal patch. There was good agreement between the frequency of colocalization measured by CCs and that judged by visual inspection (Fig. 2a–d′′).
To further examine the developmental change in area S1 barrels, we compared three quantitative indices for colocalization analysis, CC, OC and 2× ICQ, for the same images used above (Fig. 5b). The OCs were 0.794–0.890 with little change during postnatal development. If the OC indicates frequent colocalization, it is very inconsistent with the results obtained by the CCs or visual inspection. Even after shuffling pixels, OCs remained 0.694–0.880, making interpretation of this coefficient difficult. Nevertheless, the difference between OCs of the original and shuffled images was largest at P7. By contrast, the 2× ICQs showed very similar results to the CCs (P = 0.062–0.644; two-tailed paired t-test) except at P7 (P = 0.0449). When we shuffled pixels, 2× ICQs were 0.070–0.136 with a small peak at P7, showing a moderate positive correlation (R2 = 0.621) between the 2× ICQs of original and shuffled images. On the other hand, the CCs after pixel shuffling were around 0 (−0.011 to 0.005; R2 = 0.068 vs. the CCs of original images). Thus, although both 2× ICQ and CC readily detected developmental changes in VGLUT1–VGLUT2 colocalization in the barrels, the latter is superior in that CC = 0 always represents random distributions of VGLUT immunoreactivities (CC shuffled in Fig. 5b).
Electron-microscopic colocalization of VGLUT1 and VGLUT2 immunoreactivities on single axon terminals in P7 barrels
To directly demonstrate colocalization of VGLUT1 and VGLUT2 at single axon terminals, we examined area S1 of P7 and adult mice electron microscopically via the double labeling method. In layer IV barrels at P7, VGLUT1 immunoreactivity as silver grains was frequently colocalized with VGLUT2 immunoreactivity visualized as DAB reaction products on single axon terminals forming asymmetric synapses (Fig. 6a and b). In the same specimens, we also observed axon terminals immunopositive for VGLUT1 alone (Fig. 6d) or VGLUT2 alone (Fig. 6c). In the barrels of adult mice, however, most VGLUT2-immunoreactive large axon terminals were negative for VGLUT1 (Fig. 6e), although a few smaller axon terminals showing intense VGLUT1 immunoreactivity displayed weak VGLUT2 immunoreactivity (not shown).
Expression of VGLUT1 and VGLUT2 mRNAs in neurons of the developing neocortex and thalamus
Because a clear peak of VGLUT1–VGLUT2 colocalization at putative single axon terminals was detected in area S1 barrels at P7, we explored possible origins of those axon terminals containing both VGLUT subtypes by in situ hybridization histochemistry for VGLUT1 and VGLUT2 mRNAs on P7 mouse brain with a reference to adult brain. VGLUT1 and VGLUT2 mRNA signals were exclusively detected in neurons (see supplementary Fig. S2). VGLUT1 mRNA signals in the neocortex including areas S1 and V1 were moderate at P7 (Figs 7a and 8a) and intense in adulthood (Figs 7c and 8c). The specific nuclei of the thalamus including the ventral posteromedial nucleus (VPM), posterior nuclear group (Po) and dorsal lateral geniculate nucleus (LGd) showed weak VGLUT1 mRNA signals at both ages, whereas the nonspecific thalamic nuclei containing the intralaminar and midline nuclear groups showed almost no VGLUT1 mRNA signals (Figs 7a and c, and 8a and c). The LGd and Po displayed weaker VGLUT1 mRNA signals than the VPM at both ages.
VGLUT2 mRNA signals were, in contrast, intense in almost all thalamic nuclei at both ages with some variations in intensity (Figs 7b and d, and 8b and d). Within the neocortex, VGLUT2 mRNA signals were very weak at both ages, but detected in a layer-specific manner in signal-enhanced images (Figs 7e′ and f′, and 8e′ and f′), suggesting specificity of these signals. VGLUT2 mRNA-expressing neurons were densely distributed in layers II-IV, but sparsely scattered in layers V and VI at both ages (Figs 7e′ and f′, and 8e′ and f′). Interestingly, VGLUT2 mRNA signals were absent from the upper part of layer V (layer Va) in both areas S1 and V1 at both ages.
Double fluorescence in situ hybridization histochemistry clearly demonstrated coexpression of VGLUT1 and VGLUT2 mRNAs in neurons of the thalamus and neocortex. Almost all VGLUT2-expressing neurons showed weak VGLUT1 mRNA signals in the VPM (Fig. 7g and h) and Po (not shown) both at P7 and in adulthood, as well as in the LGd of adult mice (Fig. 8h). In P7 LGd, however, VGLUT1 mRNA signals were very weak, and coexpression with VGLUT2 mRNA was less evident (Fig. 8g). In layers II–IV of the neocortex, almost all VGLUT1 mRNA-expressing neurons showed very weak but significant VGLUT2 mRNA signals at both ages (Figs 7i and j, and 8i and j). In layers V and VI, however, only a small population of VGLUT1 mRNA-expressing neurons showed VGLUT2 mRNA signal. Thus, both thalamocortical projection neurons and layers II–IV glutamatergic neurons may account for the origins of VGLUT1- and VGLUT2-coexpressing axon terminals in the neocortex.
In the present study, we quantitatively examined postnatal changes in colocalization of VGLUT1 and VGLUT2 immunoreactivities at axon terminals in neocortical areas, mainly using CC as an index. CCs varied in an area-, layer- and age-dependent manner, indicating heterogeneity of VGLUT subtype expression at neocortical glutamatergic axon terminals. Of the areas and layers analysed, layer IV barrels of area S1 showed a prominent peak of the frequency of VGLUT1–VGLUT2 colocalization at P7, and area V1 layer IV displayed a smaller peak at P10. Double in situ hybridization histochemistry confirmed coexpression of VGLUT1 and VGLUT2 mRNAs at thalamic relay neurons and some cortical neurons, supporting the colocalization of the two VGLUTs at single axon terminals in the neocortex.
Origins of axon terminals expressing both VGLUT1 and VGLUT2 in area S1
We investigated the coexistence of VGLUT1 and VGLUT2 mRNAs in neurons of the neocortex and thalamus, both of which are assumed to be two major origins of intracortical glutamatergic axon terminals. The present in situ hybridization histochemical findings in adult mice were in good agreement with previous reports using adult rats (De Gois et al., 2005; Barroso-Chinea et al., 2007); a small amount of VGLUT2 mRNA was coexpressed in VGLUT1-producing layer IV neurons of adult area S1, and VGLUT1 and VGLUT2 mRNAs were coexpressed in neurons of adult thalamic relay nuclei, including the VPM and LGd. The present study further demonstrated the coexpression of VGLUT mRNAs in neocortical and thalamic neurons at P7. Together, both neocortical and thalamic neurons might contain both VGLUT1 and VGLUT2 at their axon terminals distributed in neocortex.
In the present quantitative colocalization analysis, layer IV barrels of P7 area S1 showed the most frequent colocalization of VGLUT1 and VGLUT2 at single axon terminals. One would presume that the frequent colocalization in layer IV barrels may be attributable mostly to axon terminals of thalamic origin rather than those of intracortical origin, as barrels receive massive thalamocortical inputs from the VPM (Jensen & Killackey, 1987). In fact, almost all VGLUT2 mRNA-expressing neurons in the VPM showed VGLUT1 mRNA signals at P7 as well as in adulthood. Furthermore, frequent colocalization of VGLUT1 and VGLUT2 was evident at axon terminals in area S1 layer IV as early as P3, which is consistent with the establishment of a thalamocortical glutamatergic response by P4 in mouse barrels (Yanagisawa et al., 2004). Given that VGLUT1 mRNA signal increased from P0 to P10 in the rat VPM (De Gois et al., 2005), a rise in VGLUT1–VGLUT2 colocalization at the barrels from P1 to P7 could be explained by increasing arborization of thalamocortical axons and by up-regulation of VGLUT1 expression in VPM neurons.
After P7 the colocalization in the barrels decreased as the animal maturated. VGLUT1 mRNA up-regulation in the neocortex as well as ramification of axon collaterals of neocortical neurons may lead to an increase in the number of axon terminals containing VGLUT1 alone, which may result in the decrease seen in the CCs. However, when the CC in layer IV barrels was calculated for those pixels positive for VGLUT2, the CC at P7 was much larger than that in adulthood, indicating that the decrease in the colocalization from P7 to adulthood may also be attributable to the net decrease of colocalization in VGLUT2-positive axon terminals. The developmental decrease in CC might be caused by the decrease of VGLUT1 production in thalamocortical neurons or by the increase of cortical neurons that produce a large amount of VGLUT1 and a trace amount of VGLUT2 as discussed below.
VGLUT1–VGLUT2 colocalization was frequent in barrels but not in septa at P7 area S1 layer IV. Whereas the barrels receive massive inputs from the VPM, the septa exclusively receive inputs from the Po (Lu & Lin, 1993). In P7 mice, VGLUT2 mRNA signal was intense in both the Po and the VPM, whereas VGLUT1 mRNA signal was clearly weaker in the Po than in the VPM. VGLUT2-expressing thalamocortical axons of Po neurons to the septa may hence contain a smaller amount of VGLUT1 than those of VPM neurons to the barrels at P7. In adulthood, the frequency of VGLUT1–VGLUT2 colocalization in the septa was moderate and similar to that in the barrels. Local axon collaterals of layer IV glutamatergic neurons, which express both VGLUT subtypes, might account for the colocalization in adult septa.
Throughout the entire thickness of layer II/III in adult area S1, the frequency of colocalization between VGLUT1 and VGLUT2 was moderate and almost the same as that in layer IV. This colocalization could not be attributable to VPM-derived thalamocortical axon terminals containing both VGLUT1 and VGLUT2, because thalamocortical inputs from the VPM to area S1 are abundant only in the deep part, but not in the superficial part, of layer II/III (Jensen & Killackey, 1987). Almost all VGLUT1-expressing cells in layer IV showed weak VGLUT2 mRNA signals in adult mice (Fig. 7) and rats (De Gois et al., 2005), although the signal was much less than that observed in the thalamus. Given that layer IV glutamatergic neurons of rat area S1 send their axons massively to layers II/III (Petersen & Sakmann, 2000; Feldmeyer et al., 2002; Cho et al., 2004), those axon terminals containing both VGLUT subtypes are more likely to be of layer IV origin. VGLUT1–VGLUT2 colocalization in layer II/III of area S1 was more frequent in adulthood than at P7, indicating late onset of the colocalization in this layer. The late onset may reflect maturation of intracortical axon collaterals of layer IV neurons expressing both VGLUT1 and VGLUT2. Thus, layer II/III axon terminals containing both VGLUT subtypes may be derived primarily from layer IV glutamatergic neurons.
Layer I of areas S1 and V1 showed little colocalization of the VGLUT subtypes at single axon terminals both at P7 and in adulthood. Although lesions in the rat ventrobasal thalamic nuclei resulted in a marked decrease of VGLUT2 immunoreactivity in area S1, VGLUT2 immunoreactivity partially remained in layer I, suggesting other origins of VGLUT2 immunoreactivity in layer I than the ventrobasal thalamic nuclei (Fujiyama et al., 2001). The present in situ hybridization histochemical analysis revealed that neurons in the nonspecific thalamic nuclei including the intralaminar and midline nuclei expressed VGLUT2 mRNA, but not VGLUT1 mRNA, both at P7 and in adulthood. These neurons send their axons massively to the striatum (Groenewegen & Berendse, 1994) and provide thalamostriatal axon terminals possessing VGLUT2 but not VGLUT1 (Fujiyama et al., 2004), which may lead to little colocalization of VGLUT1 and VGLUT2 at single axon terminals in the striatum. In addition to the thalamostriatal axons, these neurons have been reported to provide thalamocortical axons, which often reach layer I (Groenewegen & Berendse, 1994). Thalamocortical axon terminals from these nonspecific nuclei should thus contain VGLUT2 alone, possibly contributing to infrequent colocalization in layer I.
Functional significance of colocalization of VGLUT1 and VGLUT2
VGLUT1-containing synapses are known often to show synaptic facilitation and long-term plasticity such as long-term potentiation and depression, whereas VGLUT2-loaded synapses tend to display synaptic depression (for reviews, see Kaneko & Fujiyama, 2002; Fremeau et al., 2004b; Takamori, 2006). In fact, Schaffer collateral synapses in the hippocampus of VGLUT1 knock-out mice, which expressed VGLUT2 alone, were reported to show synaptic depression more rapidly in response to repetitive stimulation, and recover more slowly than those from wild-type animals, whose Schaffer collaterals predominantly use VGLUT1 (Fremeau et al., 2004a). Thus, it has been pointed out that synapses containing VGLUT1 and those with VGLUT2 may have differential synaptic properties.
Although VGLUT1 and VGLUT2 have shown very similar properties in their transport activity (for reviews, see Kaneko & Fujiyama, 2002; Fremeau et al., 2004b; Takamori, 2006), it has been revealed recently that VGLUT1, but not VGLUT2 or VGLUT3, interacts with endophilins A1 and A3, which are associated with clathrin-mediated endocytosis of synaptic vesicles (Verstreken et al., 2002; Schuske et al., 2003). VGLUT1 and endophilins have been reported to be colocalized at single axon terminals in the rat hippocampus and cerebellar cortex (De Gois et al., 2006; Vinatier et al., 2006; Voglmaier et al., 2006). The interaction between VGLUT1 and endophilins might hence confer unique characteristics of VGLUT1-loaded vesicles different from those of VGLUT2-laden ones, possibly via further recruitment of cytosolic protein complexes involved in vesicle recycling and trafficking. Indeed, this interaction was shown to accelerate VGLUT1 recycling after transmitter release (Voglmaier et al., 2006). Moreover, expression of VGLUT1 and VGLUT2 has been recently reported to be regulated in a bidirectional and opposite manner in mature neocortical neuronal cultures, depending on neuronal activity (De Gois et al., 2005): both in protein and mRNA levels, hyperactivity down-regulated VGLUT1 and up-regulated VGLUT2, whereas blockade of neuronal activity up-regulated VGLUT1 and down-regulated VGLUT2. This implies that, at axon terminals coexpressing both VGLUT subtypes, the ratio between the amounts of the two subtypes can be modulated by neuronal activity. Thus, glutamatergic axon terminals showing colocalization of VGLUTs might change their synaptic properties, including the speed of vesicle recycling, in an activity-dependent manner.
It has recently been reported that repetitive whisker deflections of P7–12 mice produce augmenting EPSP responses in barrel cortex neurons, whereas the same stimuli evoke decrementing responses to repetitive stimuli after P13 (Borgdorff et al., 2007). The authors have discussed that the change of augmenting to decrementing responses are caused by the decrease of recruitment of neurons in the somatosensory system and the growth of GABAergic inhibitory system during development. However, thalamocortical synapses of 2- to 3-week-old mice are known to show synaptic depression to repetitive electrical stimuli (Gil et al., 1997; Yanagisawa et al., 2004), and the synaptic depression is also considered to underlie the decrementing responses after P13 reported by Borgdorff and colleagues. In contrast to the 2- to 3-week-old mouse brain, no synaptic depression was observed in thalamocortical synapses of P4 mouse pups (Yanagisawa et al., 2004), suggesting that the synaptic characteristics of developing thalamocortical synapses were different from those of mature synapses. Because synaptic depression or facilitation depends mainly on presynaptic mechanisms (for review, see Thomson, 2000), the transient colocalization of VGLUT1, which appears to be rapidly recycled after transmitter release (Voglmaier et al., 2006) and often related to facilitative synapses, at VGLUT2-containing thalamocortical terminals could neutralize their synaptic depression in the barrels during the early postnatal period. Later, thalamocortical neuronal activity might increase, which could in turn down-regulate VGLUT1 expression at those synapses. Thus, the transiently suppressed synaptic depression at VGLUT2-dominant thalamocortical synapses during the early postnatal period may lead to an increase in glutamate release during excitatory burst discharges. This mechanism to enhance glutamatergic transmission during the build-up period of thalamocortical transmission may serve to establish mature thalamocortical synapses in area S1 barrels.
We appreciate Mr Akira Uesugi and Mrs Keiko Okamoto-Furuta for assistance with electron microscopic and photographic techniques and Dr Roberto Ireneo Gavinio for providing helpful comments on the manuscript. This study was supported by Grants-Aid for Scientific research from the Ministry of Education, Culture, Sports, Science and Technology of Japan. Grant numbers: 16200025, 17022020, 17022024, 17650100, 18020013, 18500262, 18700341 and 19700317.
- Area M1
primary motor area
- area S1
primary somatosensory area
- area V1
primary visual area
intensity correlation quotient
dorsal lateral geniculate thalamic nucleus
posterior thalamic nuclear group
vesicular glutamate transporter
ventral posteromedial thalamic nucleus.