Synaptic and extrasynaptic location of the receptor tyrosine kinase Met during postnatal development in the mouse neocortex and hippocampus

Animals

For the EM studies, C57Bl/6 mice, originally purchased from Jackson Laboratories (Bar Harbor, Maine), were bred in house at Weill Cornell Medical College. For the biochemical and cell culture studies, timed pregnant C57Bl/6 mice were purchased from Charles River (Wilmington, MA). Animals were provided free access to food and water and were housed in a 12 hour light:dark cycle. All research procedures using mice were approved by the Institutional Animal Care and Use Committee at Weill Cornell Medical College and the University of Southern California and conform to the 2011 Eighth Edition of the NIH Guide for the Care and Use of Laboratory Animals. All efforts were made to minimize animal suffering and to reduce the number of animals used.

Antibody Characterization

Primary antibodies used in this study are described in Table 1. The primary mouse anti-Met antibody used for immuno-EM and Western blotting has been shown to recognize Met in a variety of species, including mice and monkeys (Judson et al., 2011a; Judson et al., 2009). The antibody specificity was confirmed by the absence of signal following immunoblotting (Fig. 1A) and immunohistochemistry (Judson et al., 2011a; Judson et al., 2009), using tissue prepared from the cortex of mice in which the Met gene was deleted from the dorsal pallium.

Three mouse monoclonal antibodies were used to assess the efficacy of the subsynaptic fractionation procedure. Specifically, an anti-synaptophysin antibody was used as a marker of biochemical fractions containing synaptic vesicles, an anti-postsynaptic density protein 95 (PSD95) antibody was used as a marker of the postsynaptic density fraction and an anti-syntaxin 1 antibody was used as a marker of the presynaptic active zone fraction. For each antibody, we observed the same unique molecular weight band in Western blots of homogenates of postnatal day (P) 14 mouse cortex (Fig. 1B) as has been reported by the manufacturer (manufacturers’ information sheets) and in multiple publications (Anastasio et al., 2010; Bouvier et al., 2008; Louneva et al., 2008; Needleman et al., 2010; Phillips et al., 2001; Rebola et al., 2005).

For the PLA assay, three antibodies were used to label pre- and post-synaptic compartments in primary cultures of neocortical and hippocampal neurons. The same monoclonal anti-PSD-95 antibody described above was used to delineate the postsynaptic compartment. For presynaptic compartment labeling, polyclonal anti-neurexin 1 and anti-synapsin 1 antibodies were used to label the presynaptic compartment. The anti-synapsin 1 antibody recognizes the appropriate band that represents both the 1a and 1b forms at 77 and 80kD (Fig. 1B). The anti-pan neurexin 1 antibody is reported by the manufacturer to recognize multiple forms of neurexin 1. In Western blots of P14 mouse cortex, the antibody recognizes the β-forms of neurexin 1, but does not appear to recognize the α-forms (Fig. 1B). Staining of mouse primary neocortical and hippocampal neurons produced patterns of PSD-95, neurexin 1 and synapsin 1 immunoreactivity are identical with previous descriptions (Dean et al., 2003; Li et al., 2010; Mondin et al.; Zurner et al., 2011). We could not use the same monoclonal anti-Met antibody described above for the PLA assay as this assay requires two antibodies generated in different species. Therefore, we used a goat polyclonal anti-Met antibody: the antibody specificity was confirmed by the absence of signal following immunoblotting (Fig. 1A) using tissue prepared from the cortex of mice in which the Met gene was deleted from the dorsal pallium.

Immuno-electron microscopy

The distribution of Met in the neocortical and hippocampal neuropil was examined at P7, P14 and P21 using pre-embedding immuno-EM (Milner et al., 2011). Unless otherwise noted, all steps were performed at room temperature. Male mice (N = 3 at each age) were deeply anesthetized with sodium pentobarbital (150mg/kg i.p.) prior to transcardial perfusion first with 2ml 2% heparin in saline, followed by 10–20 ml 3.75% acrolein (Polysciences, Warrington, PA) and 2% paraformaldehyde (Electron Microscopy Sciences [EMS], Fort Washington, PA) in 0.1M phosphate buffer, pH 7.4 (PB). Brains were postfixed for 30 min in 2% acrolein and 2% paraformaldehyde in PB, and then stored in PB at 4°C until sectioned (less than 2 days). Coronal sections were cut (50µm thick) on a Leica VT1000S vibratome (Leica Microsystems, Deerfield, IL) and stored as free-floating sections in 30% ethylene glycol (Sigma, St Louis, MO)/ 30% sucrose (JT Baker, Phillipsburg, NJ) in PB at −20°C.

Sections that included the dorsal hippocampus and overlying somatosensory cortex (Fig. 2A) were processed for the immunohistochemical localization of Met. Sections were washed in PB, incubated for 30 min in 0.5% sodium borohydride (Sigma-Aldrich) in PB, and rinsed again in PB. After transfer to a cryoprotectant solution (25% sucrose, 10% glycerol [JT Baker] in 50mM phosphate buffer), sections underwent a freeze-thaw cycle, after which they were rinsed first in PB followed by 0.1M Tris-saline, pH 7.6 (TS), and blocked in 0.5% bovine serum albumin (BSA, Sigma) in TS for 30 min. Following two washes in TS, sections were incubated in the anti-Met primary antibody (1:300 in 0.1% BSA in TS) for 1 day at room temperature followed by 4 days at 4°C. After rinsing in TS, sections were incubated in biotinylated horse anti-mouse IgG (1:400 in 0.1% BSA in TS, Vector Laboratories, Burlingame, CA) for 30 min. Sections were rinsed in TS and processed using the ABC Vectastain Elite kit at twice the recommended dilution (Vector Laboratories). Met-immunolabeling was visualized with 3,3’-diaminobenzidine (Sigma) and hydrogen peroxide (EMS) in TS. Sections were rinsed in TS, then rinsed and stored in PB prior to embedding.

Tissue sections were embedded and sectioned for EM using standard procedures (Milner et al., 2011). Briefly, sections were incubated in 2% osmium tetroxide (EMS) in PB for 1 hour, rinsed with PB, and dehydrated through a graded series of alcohol and propylene oxide (EMS). Sections were embedded in EMbed 812 (EMS) and flat mounted between Aclar sheets (Honeywell, Morristown, NH) at 60°C for 3 to 5 days. For each animal, the areas of interest (hippocampus and somatosensory cortex) were excised from one embedded tissue section, mounted on EMbed chucks, and sectioned on a Leica ultracut UCT ultratome (Leica Microsystems). The ultrathin sections (70nm thick) were collected on 400 mesh copper grids (EMS) and counterstained with uranyl acetate and Reynold’s lead citrate (EMS).

Sampling and analysis

Sections were viewed and photographed on a Tecnai Biotwin transmission electron microscope (FEI, Hillsboro, OR). For each mouse, one block from the hippocampus and one from the somatosensory cortex was analyzed. Met-immunolabeled profiles were counted and classified in the stratum radiatum, 55 µm distal from the hippocampal CA1 pyramidal cell layer, and in superficial layer V of the somatosensory cortex, approximately 350 (P7) to 400 (P14, P21) µm deep to the pial surface. To insure optimum labeling and to allow quantitative comparisons between groups, only grids near the plastic/tissue interface were selected (Milner et al., 2011). Within these areas, two to three grid squares, representing 6050 to 9075 µm² of neuropil, were sampled; our previous studies have determined that this amount of sampling is sufficient to represent the population (Barker-Gibb et al., 2001; Milner et al., 2001b). Images of all fields containing immunolabeled profiles were acquired at 16,500× using Advanced Microscopy Techniques software (v3.2, Advanced Microscopy Techniques, Woburn, MA).

Cellular profiles were classified according accepted morphological criteria (Peters et al., 1991). Briefly, dendritic shafts contained regular microtubular arrays and were usually postsynaptic to axon terminal profiles. Dendritic spines also were usually postsynaptic to axon terminal profiles and sometimes contained a spine apparatus. Unmyelinated axons were smaller than 0.2 µm, round, contained a few small synaptic vesicles and lacked a synaptic junction in the plane of section (Milner and Bacon, 1989; Milner et al., 2001a). Axon terminals had numerous small synaptic vesicles and had a cross-sectional diameter greater than 0.2 µm. Glial profiles were distinguished by the presence of glial filaments and/or gap junctions (astrocytes), by the absence of microtubules, and/or by their tendency to conform to the boundaries of surrounding profiles (Barker-Gibb et al., 2001; Glass et al., 2001). Ambiguous profiles, which were more prevalent at P7, were classified as unknown. Asymmetric synapses were characterized by the larger size of the post-synaptic density whereas symmetric synapses had pre- and post-synaptic densities of equal size. Appositions were defined as those contacts between profiles in which no interposing glial processes were seen but that lacked any recognizable synaptic specializations. The number of each class of Met-immunolabeled profiles was tabulated and the data presented as mean ± SEM. Significant differences in the distribution between neocortex and hippocampus, and within a region across the three developmental ages, were determined using Pearson’s χ² test. The cross-sectional area and minimum diameter of terminal profiles in which the entire perimeter could be distinguished were measured using Microcomputer Imaging Device software (MCID, Imaging Research Inc., Ontario, Canada).

Proximity Ligation Assay (PLA)

Primary neuronal cultures using tissue from embryonic day (E) 18 hippocampus or neocortex were prepared as described previously (Kaech and Banker, 2006). For each culturing session, tissue from a single litter (7–10 total, including male and female embryos) was pooled. Briefly, tissue was incubated in 0.25% trypsin-EDTA (Invitrogen Life Technologies, Grand Island, NY), then dissociated into a single cell suspension by mechanical trituration. Cells were plated at a density of 1 × 10⁴ cells/cm² onto coverslips (Carolina Biological Supply Company, Burlington, NC) coated with 100µg/ml poly-D-lysine (Sigma). Cells were cultured at 37°C in 5% CO₂ in Dulbecco’s Modified Eagle’s Medium (Invitrogen) supplemented with 5% fetal bovine serum. After 2 hr, by which time the cells had adhered to the substrate, the coverslips were transferred to a new dish and the cells co-cultured above a layer of astrocytes in defined Neurobasal medium (Invitrogen) supplemented with 2% B27 (Invitrogen) and 1% L-glutamine (Invitrogen).

At 14 days in vitro (DIV), proximity ligation assays were performed using the Duolink in situ PLA kit (Olink Bioscience, Uppsala, Sweden (Soderberg et al., 2006) to assess the co-localization of Met with proteins found exclusively in the presynaptic or postsynaptic compartment. Coverslips were fixed with 4% PFA for 15 min at room temperature, washed in PBS, and permeabilized with 0.1% Triton X-100 in PBS for 15 min. The coverslips were incubated for 1 hr at room temperature in the blocking solution provided in the kit, followed by an overnight incubation with anti-Met (1:25) antibody in conjunction with anti-PSD-95 (1:100), anti-synapsin 1 (1:100) or anti-neurexin 1 (1:100) antibodies. For each culturing session, one primary antibody was omitted from some coverslips to control for potential nonspecific PLA signals generated by a single antibody. After two washes with PBS, coverslips were incubated with oligonucleotide-linked secondary antibodies (PLA probes) provided in the kit. Subsequent hybridization, ligation and amplification steps were performed according to the manufacturer’s protocol, after which coverslips were incubated with Cy²-labeled secondary antibodies (Jackson ImmunoReseach) for 20 min at room temperature. After two washes with PBS, coverslips were mounted with ProLong® Gold Antifade reagent (Invitrogen). A fluorescent signal is generated only when the two PLA probes are in close proximity (40nm).

For each labeling combination, five different fields per coverslip, derived from at least 3 independent culturing sessions, were examined using an Olympus Fluoview FV1000 confocal microscope (60× oil object). Digital images were collected and downloaded into Adobe Photoshop CS5 (v12.1, Adobe Systems Incorporated, San Jose, CA) to assess the presence or absence of the PLA signal. For quantification of PLA signal, the total number of PLA clusters in five random fields (900 µm²) was counted for each labeling combination. Typically, PLA signal has been expressed per cell. In our culture system, however, it is not possible to define the number of cells contributing to the processes in any one field. Therefore, we determined the area fraction of each box occupied by Met immunostaining using Image J. PLA signal was then expressed as the number of PLA clusters/Met signal in each sampled field. For each labeling combination, the PLA value for each field was averaged to give a single value for each culturing session. For each brain region, a one-way ANOVA was used to determine overall significance. If significant differences were observed, this analysis was complemented by a Tukey multiple comparison post-test.

Generation and characterization of subcellular fractions

Synaptomsome and subsynaptic fractions were generated according to published protocols (Louneva et al., 2008; Phillips et al., 2001). All procedures were performed on ice or at 4°C. All chemicals are from Sigma unless otherwise noted. Mice at P7, P14 and P21 were anesthetized with saturated Piramal Healthcare isoflurane (Clipper Distributing Company, St Joseph, MO), decapitated and the brains removed. The neocortex or hippocampus was dissected in freshly oxygenated artificial cerebrospinal fluid and processed immediately to generate synaptosomal and subsynaptic fractions (Fig. 2B). Tissue from multiple male and female mice (P7 cortex = 14; P7 hippocampus = 24; P14 and P21 cortex = 5; P14 and P21 hippocampus = 10) was pooled to yield 450–600mg (P7) or 250 – 350 mg (P14, P21) wet weight starting material. Pooled tissue was homogenized with a glass homogenizer (Wheaton, Millville, NJ) in 320mM sucrose, 0.1mM CaCl₂, 1mM MgCl₂ containing a protease inhibitor cocktail. The sucrose concentration of the homogenate was adjusted to 1.25M, overlaid with 1M sucrose, and centrifuged at 100,000 g. The synaptosome fraction, located at the sucrose interface, was recovered and washed twice in 0.1mM CaCl₂, and a sample taken for analysis. The remaining synaptosomes were pelleted by centrifugation at 40,000 g. The pellet was solubilized by incubation in 20mM Tris-HCl, pH 6.0, 1% Triton-X 100, 0.1mM CaCl₂, and then centrifuged at 40,000 g. The resulting supernatant is considered the synaptic vesicle fraction, whereas the pellet consists of pre- and post-synaptic membranes. The pellet was solubilized in 20mM Tris-HCl, pH 8.0, 1% Triton-X 100, 0.1mM CaCl₂, and then centrifuged at 40,000 g. The supernatant contains presynaptic membranes, and the pellet contains the postsynaptic density fraction. The synaptic vesicle and presynaptic membrane fractions were acetone precipitated. The precipitates and the pellet containing the postsynaptic density fraction were resuspended in 5% sodium dodecyl sulfate (SDS). The protein concentration of each fraction (synaptosome, synaptic vesicle, presynaptic and postsynaptic density) was determined using the Dc protein assay (Bio-Rad, Hercules, CA).

Western blot analysis, using previously published protocols (Eagleson et al., 2011; Judson et al., 2009), was used to assess the distribution of Met in select subcellular compartments and the efficacy of the fractionation procedure. A sample of each fraction from the same region at the same age, as well as Precision Plus Protein Standards (Bio-Rad), was run on a single gel. Equal amounts of protein (30µg) were separated by SDS-polyacrylamide gel electrophoresis (7.5% gel) under reducing conditions and transferred overnight to a supported nitrocellulose membrane (Whatman Optitran BA-S85, Sigma). The membrane was stained for total protein with the Pierce Reversible Protein Stain Kit for Nitrocellulose Membranes (Pierce) to verify equal lane loading. Each membrane was probed sequentially with antibodies directed against Met (1:500), syntaxin-1 (1:2500), synaptophysin (1:1,000,000), and PSD-95 (1:2000), all diluted in blocking buffer - 5% blotto (nonfat dry milk, Cell Signaling Technology, Beverly, MA) in Tris buffered saline, 0.05% Tween-20 (TBST). Specifically, membranes were blocked in blocking buffer for 2 hours at room temperature, followed by incubation with primary antibody overnight at 4°C. Membranes were rinsed in TBST and incubated in a horseradish peroxidase-conjugated anti-mouse secondary antibody (1:5000 in blocking buffer, Jackson Immunoresearch, West Grove, PA) for one hour at room temperature. After rinsing in TBST, membranes were reacted with either Pierce SuperSignal West Dura Extended Duration Substrate (PSD-95, synpatophysin and syntaxin 1, Thermo Scientific, Rockford, IL) or Pierce SuperSignal West Femto Maximum Sensitivity Substrate (Met, Thermo Scientific) to visualize immunoreactive bands. Membranes were stripped between each primary antibody using OneMinute Western Blot Stripping Buffer (GM Biosciences, Rockville, MD).

Digital images of the Western blots were acquired and analyzed using a CCD camera coupled to a UVP BioImaging System using VisionWorksLS Image Acquisition and Analysis Software (v7.0.1, UVP, Upland, CA). The density of each band was measured, a global background subtraction applied, and the resulting values normalized for total protein loading (Brown et al., 2005; Gustin et al., 2010) based on the Pierce Reversible Protein Stain.

Digital illustrations

Images of tissue processed for immuno-EM were acquired using Advanced Microscopy Techniques software (v3.2, Advanced Microscopy Techniques). Images of primary cultures processed using the PLA method were acquired using an Olympus Fluoview FV1000 confocal microscope (60× oil object). Images of the Western blots were acquired using a CCD camera coupled to a UVP BioImaging System using UVP Imager software (v7.0.1, UVP, Upland, CA). Graphs were prepared in SigmaPlot (v7.0, SPSS Incorporated, Chicago, IL). Images and graphs were converted to TIFF format and imported into Adobe Photoshop CS5 (v12.1, Adobe Systems Incorporated, San Jose, CA), where they were cropped, adjusted for brightness and contrast, and sharpened for preparation of final figures.

The distribution of Met-ir in hippocampal CA1 changes across development

Regardless of age, Met-ir is present in dendrites, dendritic spines, axons, terminals and glia in the CA1 stratum radiatum in the dorsal hippocampus. In dendritic shafts, Met-ir is patchy and often associated with the plasma membrane or endomembrane structures in the cytoplasm (Fig. 3A). Occasionally, Met-ir is on the plasma membrane of dendritic profiles in the region in which they contact each other (Fig. 3B). In dendritic spines, Met-ir is distributed consistently in the head near the post-synaptic junction (Fig. 3C). Intermittent, narrow patches of Met-immunoreactive membranes also are detected in small unmyelinated axons (Fig. 3D and F); however, no Met-labeling was seen in unmyelinated axons. Axon terminals immunoreactive for Met contain numerous small synaptic vesicles; at all ages, the majority of Met-labeled terminals are less than 0.5 µm in diameter (Table 2; Fig. 3D). However, some terminals, especially at P7, have minimum diameters as large as 0.85 µm (Table 2; Fig. 3E). Within the large terminals, peroxidase reaction product is associated with clusters of vesicles near the synaptic specialization (Fig. 3E). There is limited glial cell immunolabeling, with these profiles usually conforming to the boundaries of the neuropil and lacked astrocytic fibrils (Fig. 3G).

Quantitative analysis revealed that the distribution of Met-ir across different subcellular elements was significantly different across postnatal development (Table 3; Fig. 4A). At P7, postsynaptic profiles (i.e., dendrites and dendritic spines) account for approximately 1 in 3 labeled profiles, but their number decline such that they represent 1 in 5 at P14 and only 1 in 20 at P21 (Fig. 4A). In contrast, the percent of presynaptic profiles is relatively similar between the three ages. Specifically, axons are the most frequently labeled profile at all ages, ranging from 45% at P7 and P14 to 63% at P21 (Fig. 4A). Moreover, the proportion of labeled terminals remains constant over the three ages (Fig. 4A). As a consequence of the changes in the type of Met-ir profiles over the first three weeks postnatal, the ratio of pre- to post-synaptic profiles changes from near equivalence at P7 and P14 to predominantly presynaptic at P21 (Fig. 4B).

At all ages, the majority of Met-immunoreactive terminals do not form a synaptic specialization in a single plane of section (Table 4). Of the terminals that do form synaptic specializations, the targets and types of synapse formed by Met-immunoreactive terminals varies depending on age (Table 2). At P7, most Met-labeled terminals form symmetric synapses on dendritic shafts. However, by P14 as well as P21, few Met-immunoreactive terminals synapse on dendrites. Instead, most of the Met-labeled terminals form asymmetric synapses on dendritic spines. At all three ages the number of terminals containing Met-ir is relatively constant.

In addition to neuronal elements, an increasing number of Met-immunoreactive glial profiles are found in the neuropil as development proceeds. At P7, glial cells represent only 3% Met-labeled profiles (Fig. 4A). At later ages, however, as many as 1 in 5 Met-immunoreactive profiles are glial (Fig. 4A).

At all developmental ages, Met-ir is predominantly located in presynaptic profiles in the somatosensory cortex

Like CA1 of the hippocampus, Met-immunoreactive profiles in superficial layer V of somatosensory cortex are found in pre- and postsynaptic profiles as well as glia (Fig. 5). Within dendritic shafts, Met-ir is found near the plasma membrane sometimes affiliated with the perisynaptic zone (Fig. 5A) or adjacent to the region contacted by another dendrite (Fig. 5B). Met-ir is located in small terminals that synapse on unlabeled dendritic spines (Fig. 5C) or in clusters of vesicles in terminals that appose terminal-spine synaptic complexes (Fig. 5D). As seen in the hippocampus, the majority of Met-labeled terminals in the cortex were less than 0.5 µm in diameter (Table 2). As noted for the hippocampus, Met-ir also labels limited membrane domains of small unmyelinated axons (Fig. 5E and F).

The distribution of Met-ir across different types of profiles is overlapping, but not identical, between the two regions at all ages examined (Table 3; Fig. 6A; P7 hippocampus versus neocortex, p < 0.001, d.f. = 4, χ²; P14 hippocampus versus neocortex, p < 0.001, d.f. = 4, χ²; P21 hippocampus versus neocortex, p < 0.001, d.f. = 4, χ²). Indeed, Met-labeled presynaptic profiles already predominate in superficial layer V of neocortex at P7 (Fig. 6B), primarily due to an increased number of Met-labeled axons and the very limited staining of dendritic spines. It should be noted that even unlabeled spines are rare in this layer at P7, consistent with the onset of robust spinogenesis in the rodent neocortex during the second postnatal week (Meller et al., 1968; Miller and Peters, 1981). Nonetheless, even at P14 and P21, when spine number increases dramatically, Met-labeled spines are observed rarely, while the number of Met-labeled axons remains high. Thus, at all ages, axons represent the most frequently labeled profile, ranging from 70% at P7 to 85% at P21 (Fig. 6A), whereas spines represent the least (Fig. 6A, less than 1% at all ages). The proportion of Met-labeled dendrites and terminals are higher at P7 and P14 than at P21 (Fig. 6A). In contrast, Met-labeled glial profiles are uncommon during the first two postnatal weeks (Fig. 6A), but increase in number during the third postnatal week (Fig. 6A), although not to the same extent as seen in the hippocampus.

Like the hippocampus, the majority of Met-labeled terminals do not form a synaptic specialization in a single plane of section (Table 4). Of the terminals that do form synaptic specializations, the targets and types of synapses formed by Met-labeled terminals vary depending on age (Table 4). At P7, most Met-labeled terminals form symmetric synapses on dendritic shafts. However, few to no Met-labeled terminals synapse on dendritic shafts at P14 and P21, respectively. As seen in the hippocampus, most of the Met-immunoreactive terminals form asymmetric synapses on dendritic spines at P14 and P21. Unlike the hippocampus, the number of terminals containing Met-ir greatly diminishes from P7 to P21.

PLA localizes Met presynaptically in growing neurons in vitro

PLA provides a refined approach at the light microscope level to determine co-localization of Met with other proteins that are located in subsynaptic compartments. A fluorescent signal is only observed when the PLA probes used to detect the two proteins are within 40nm (Soderberg et al., 2006). In primary cultures of hippocampal (Fig. 7) and neocortical (Fig. 8) neurons after 14 days in vitro, there was extensive Met immunostaining of neuronal processes (anti-Met panels in Figs. 7 and 8). Co-localization of pre- and postsynaptic markers with Met was evident from strong PLA fluorescent signals in the presence of antibodies directed against Met and synapsin 1 (Figs. 7A–F, 8A–F) or Met and neurexin 1 (Figs. 7G–L, 8G–L), indicating the localization of Met in the presynaptic compartment. PLA using antibodies directed against Met and the postsynaptic marker PSD-95 revealed positive signals in hippocampal cultures (Fig. 7M–R). In contrast, no PLA signals using antibodies against Met and PSD-95 could be detected in neocortical cultures (Fig. 8M–R). Quantification of the PLA signal confirmed these observations (Fig. 9). Co-localization of Met differed significantly across the three markers in hippocampal [F(2,6) = 6.09, P = 0.04] and neocortical [F(2,6) = 9.58, P = 0.01] cultures. In the hippocampus, Tukey’s post-hoc comparisons of the three markers demonstrate that there is significantly more PLA signal for synapsin 1/Met than for PSD-95/Met at P < 0.05. A similar analysis for neocortex demonstrates significantly more PLA signal for synapsin 1/Met and for neurexin 1/Met than for PSD-95/Met at P < 0.05. Together, these data demonstrate a significant presynaptic localization of Met in vitro that has not been previously appreciated.

As a negative control, no fluorescent signals could be detected in the presence of anti-Met (Fig. 7S–X, 8S–X), anti-synapsin 1, anti-neurexin 1 or anti-PSD-95 antibodies when used individually. As would be expected, strong fluorescent PLA signals could be detected in the presence of synapsin 1 and neurexin 1 antibodies in hippocampal and neocortical cultures, with no PLA signal detected in the presence of neurexin 1 and PSD-95 or synapsin 1 and PSD-95 antibodies in neocortical cultures (data not shown). In hippocampal cultures, however, occasional faint PLA signals could be observed in the presence of neurexin 1 and PSD-95 or synapsin 1 and PSD-95 antibodies (data not shown).

Biochemical distribution of Met in the pre- and post-synaptic compartments of the synapse

The immuno-EM and PLA analyses indicate that, in addition to its presence at extra-synaptic sites, such as axons, a significant proportion of Met-ir is associated with the developing synapse. Throughout the period of synaptogenesis in vivo, Met-ir is present specifically in axon terminals, spines and proximate to postsynaptic densities. A biochemical fractionation approach was used to expand the analysis of the localization of Met at the synapse and to determine, with a method that is not dependent upon tissue fixation methods, the relative distribution of the receptor across 1) presynaptic membranes associated with the active zones, 2) the postsynaptic density, and 3) synaptic vesicles and presynaptic membranes within the terminal but outside the active zones.

We first evaluated the efficacy of the fractionation method, using specific markers for different components of the synapse and found minor cross-contamination between the fractions (Fig. 10), consistent with previous reports (Anastasio et al., 2010; Bouvier et al., 2008; Lee et al., 2008; Louneva et al., 2008; Needleman et al., 2010; Phillips et al., 2001; Rebola et al., 2005; Rebola et al., 2003). In general, the signal for the synaptic vesicle membrane protein, synaptophysin, is enriched in the synaptic vesicle fraction. It should be noted, however, that this fraction likely also includes intracellular membranes from other intracellular organelles as well as presynaptic plasma membrane located outside the active zone. Less signal is found in the presynaptic active zone fraction (Fig. 10), and likely represents vesicles that had been docked at or fused with the active zone at the time of tissue harvest (Tao-Cheng, 2006). Little to no synaptophysin signal is detected in the postsynaptic fraction. As expected, syntaxin-1 is present in both presynaptic fractions (Fig. 10). Syntaxin-1, however, also is detected in the postsynaptic fraction, although at much lower levels. This indicates the presence of some presynaptic membranes in this fraction. The postsynaptic density protein PSD-95 is almost exclusively located in the postsynaptic fraction (Fig. 10), indicating little cross-contamination of postsynaptic membranes in either of the presynaptic fractions.

Met-ir is present in synaptic fractions generated from hippocampus and cortex at all ages examined (Fig. 11), consistent with an association of Met-ir with the synapse. Previous studies have emphasized the postsynaptic localization of Met in hippocampal synapses (Kawas et al., 2012; Tyndall and Walikonis, 2006). In contrast, we observed Met-ir associated with the presynaptic nerve terminal (synaptic vesicle and presynaptic membranes, including the active zone) fractions as well as the postsynaptic density fraction in the hippocampus at P7 and P14 (Fig. 11). By P21, although Met-ir is not visible in the synaptosomal fraction itself, Met-ir still is associated with both pre- and post-synaptic fractions. The localization of Met to cortical synapses has not been examined previously. Similar to the hippocampus, Met-ir can be observed in pre- and post-synaptic fractions across all stages of development (Fig. 11). These data highlight a significant distribution of Met protein to presynaptic nerve terminals during early postnatal development.

Regional and temporal variation in subcellular distribution of Met

Presynaptic (axons and terminals) and postsynaptic (dendrites and spines) neuronal elements are Met-immunoreactive, although the relative abundance varies between regions and postnatal ages that were examined. We note these differences with caution, as such variations may be dependent upon the specific neocortical layers and hippocampal subfields analyzed. In general, the analyses revealed that axons are the most abundant Met-labeled profile, consistent with previous light microscopic observations indicating that, during early periods of active axonal growth and synapse development, Met-ir is preferentially, although not exclusively, trafficked to the axonal compartment in restricted fiber tracts in the mouse and monkey forebrain (Judson et al., 2011a; Judson et al., 2009). This represents in part the transport of Met receptors to presynaptic terminals, but may also include positioning of Met in axonal membranes to influence other developmental processes, such as modulating the extent of axonal branching in terminal fields. It is unlikely that Met is involved in axon pathfinding, because in the absence of Met signaling, major forebrain fiber tracts still develop normally (Judson et al., 2010). We did not, however, examine terminal fields for possible arborization differences.

Based on the EM analyses, only a subpopulation of axons and terminals are Met-immunoreactive and we currently cannot identify the specific neuronal populations that give rise to the immunolabeled profiles. Axons and terminals present in the stratum radiatum include Schaffer collaterals of CA3 pyramidal cells, local GABAergic interneurons, as well as projections from the medial septum, ventral tegmental area, raphe and locus ceruleus (Amaral and Witter, 1989; Colom et al., 2005; Jones and Moore, 1977; Klausberger, 2009; Moore and Halaris, 1975; Swanson, 1982). Within superficial layer V of somatosensory cortex, axons and terminals arising from pyramidal neurons in the same region as well other cortical areas and local GABAergic interneurons are included. It is unlikely, however, that GABAergic neurons contribute to the Met-ir axons and terminals in either region, as cells arising from the ganglionic eminences do not express Met during normal development (Eagleson et al., 2011; Judson et al., 2009). Moreover, double label in situ hybridization in neocortex revealed excitatory neurons express Met, but GABAergic neurons do not (Eagleson et al., 2011). Similarly, pyramidal neurons located in the principal cell layer of CA1 or in infragranular layers of somatosensory cortex (Judson et al., 2009), rather than local interneurons, are the likely source of Met-immunoreactive dendrites and spines we observe in stratum radiatum and superficial layer V, respectively.

The development of Met-labeled terminals differs between the neocortex and hippocampus. Most notably, while the number of labeled terminals remains constant in the hippocampus between P7 and P21, there is a rapid and pronounced decline in Met-ir terminals in the neocortex over the same time period. Intriguingly, in both hippocampus and cortex, the majority of Met-labeled terminals do not form a synaptic specialization. Indeed, in both regions across all ages examined, about half the Met-labeled terminals are not found in close apposition to any other neuronal element. At P14 and P21, asymmetric synapses on dendritic spines represent the majority of synapses formed by Met-ir terminals in both hippocampus and neocortex. This is consistent with excitatory, rather than inhibitory, neurons contributing to Met-ir axons and terminals, as discussed above. In contrast, at P7, symmetric synapses on dendritic shafts predominate among Met-labeled terminals forming synaptic contacts. Although, in adults, such a morphology is typically associated with inhibitory synapses, it has been suggested that one of the first steps in spine formation in cortical pyramidal cells involves the formation of a symmetric contact between an axon and dendritic shaft that subsequently matures into an asymmetric synapse between a terminal and dendritic spine (Miller and Peters, 1981). Thus, the symmetric synapses formed by Met-ir terminals at P7 may represent the first stages in the maturation process of excitatory synapses. This would be consistent with the lack of co-localization of Met transcript with markers of inhibitory neurons in the developing neocortex (Eagleson et al., 2011).

The relative contribution of glial cells to Met-immunoreactive profiles within the neuropil varies across development and between regions. Few Met-labeled glial processes are present at the end of the first postnatal week. After this time, there is a marked increase in the proportion of Met-immunoreactive glial profiles in the hippocampus, but only a modest increase in the neocortex. Currently we cannot distinguish the specific type of glial cell associated with Met-ir in either region; however, the absence of fibrils suggests that they are not fibrous astrocytes. Previous analyses, using double-label in situ hybridization, indicated an absence of Met expression in astroglia and oligodendroglia in the P14 mouse somatosensory cortex, although specific microglial markers were not examined (Eagleson et al., 2011). Thus, the Met-ir observed in the current study may be associated microglial cells, an intriguing possibility as these cells have been implicated in synapse elimination and refinement during normal development (Hoshiko et al., 2012; Paolicelli et al., 2011; Schafer et al., 2012; Tremblay et al., 2010). Met expression also has been reported in astrocytes and oligodendrocytes, although usually either in vitro or in the context of an injury response (Kitamura et al., 2007; Lalive et al., 2005; Machide et al., 2000; Nagayama et al., 2004; Shimamura et al., 2007; Shimazaki et al., 2003; Yan and Rivkees, 2002). Future studies will aim to clarify the identity of Met-labeled glial processes.

We focused our EM analyses on the developing male neocortex and hippocampus, because in general, in vivo and electrophysiological analyses of Met expression and function only include males (Akimoto et al., 2004; Judson et al., 2009; Judson et al., 2010; Qiu et al., 2011). In vitro studies typically utilized primary cultures that include a mixed population of neurons from male and female embryos (Akimoto et al., 2004; Akita et al., 2008; Finsterwald and Martin, 2011; Lim and Walikonis, 2008; Nakano et al., 2007; Tyndall et al., 2007; Tyndall and Walikonis, 2006; 2007). However, sex differences in neuronal architecture, plasticity and circuitry, primarily arising from the influence of gonadal hormones, have been reported in the hippocampus (for example (Bian et al., 2012; Bowers et al., 2010; Cooke and Woolley, 2005; Fester et al., 2012; Gould et al., 1991; Isgor and Sengelaub, 2003; Lebron-Milad and Milad, 2012; Mitsushima et al., 2009; Romeo et al., 2004) and neocortex, including the somatosensory region (Dawson et al., 2009; Kritzer, 1998; Tobet et al., 1993; Urban-Ciecko and Mozrzymas, 2011). Thus, it is reasonable to hypothesize that there are sex differences in the expression and function of proteins involved in synapse and circuit formation, such as Met. Extensive age and sex-based analyses will address this hypothesis in future studies.

Met at the synapse

The demonstration of Met-ir in axon terminals and dendritic profiles at the ultrastructural level is consistent with the biochemical localization of Met in synaptosomes and across both pre- and post-synaptic compartments. We note there are technical considerations when comparing quantitative data across different methodological platforms that each has their own strengths and limitations. First, the ultrastructural analyses focused on discrete subregions of the hippocampus and neocortex, providing precise subcellular information on Met localization in the context of regional circuitry. In contrast, the entire neocortex and hippocampus from multiple mice were required to have sufficient starting tissue to generate biochemical subsynaptic fractions. This method is not impacted by factors such as antibody penetration and fixation, but does not provide anatomical information. It is likely that, using the biochemical fractionation method, at least some of the Met detected in the postsynaptic fractions may arise from contamination of this fraction with presynaptic proteins, as indicated by the presence of syntaxin 1 in this fraction. For the primary cell culture studies, embryonic neurons intrinsic to the neocortex or hippocampus were grown, but this excludes presynaptic contributions from subcortical sources. Finally, using the PLA method, there was no indication of postsynaptic localization in vitro in the neocortex; postsynaptic labeling in the hippocampus, however, was observed, which is consistent with previous reports (Kawas et al., 2012; Tyndall and Walikonis, 2006). We interpret the absence of postsynaptic labeling in cultures of neocortical neurons cautiously. The data clearly demonstrate that Met is located presynaptically in vitro, but the absence of postsynaptic localization may reflect a culture-specific phenomenon of a difference in synaptic maturation in vitro, or the possibility that Met distribution in neurons growing in vitro is not spatially associated with PSD-95, which is required for PLA to generate a signal. It is important to emphasize, however, that despite these caveats, the goal of the present study was to determine, with multiple methodologies, the presence of Met at synapses during peak periods of synaptogenesis in the neocortex and hippocampus. Each of the three different techniques provides evidence of a previously unappreciated presence of Met in presynaptic axon terminals.

TrkB, another member of the receptor tyrosine kinase family, also is present at pre- and postsynaptic sites in the hippocampus, although the majority of phosphorylated TrkB is localized to presynaptic profiles (Spencer-Segal et al., 2011). Thus, it will be important to determine if there is a similar differential profile of activated phospho- and total Met in the neocortex and hippocampus. Perhaps more intriguingly, the relative distribution of Met-ir is dependent on the stage of development and the brain region examined. Similar ultrastructural and biochemical analyses have been used to elucidate the subsynaptic localization of a variety of receptors and MHC class I proteins (Bouvier et al., 2008; Hettinger et al., 2001; Muly et al., 2007; Needleman et al., 2010; Rebola et al., 2005; Rebola et al., 2003). Several comparative examples are worth highlighting. Similar to our findings with Met, MHC I proteins are distributed pre- and postsynaptically, but appear to exhibit a more balanced distribution across synaptic compartments and do not show the dramatic reduction in expression during maturation (Needleman et al., 2010) that is seen with Met (Judson et al., 2009). In the adult, the subsynaptic localization of group II metabotropic glutamate receptors differs between the basolateral amygdala (BLA) and the bed nucleus of the stria terminalis (BNST), with the proportion of postsynaptic localization higher in the BLA; this difference corresponds to the ability to elicit more pronounced postsynaptic responses in the BLA compared to the BNST (Muly et al., 2007). Differences in the subsynaptic localization of the adenosine A_2A receptor in the hippocampus and striatum also have been reported, with a highly enriched postsynaptic localization in striatum and presynaptic localization in hippocampus (Rebola et al., 2005). This likely reflects the predominant postsynaptic mechanism through which activation of the adenosine A_2A receptor exerts its modulatory effect in the striatum, compared to a presynaptic mechanism in the hippocampus. Together with our data, and as noted in (Rebola et al., 2005), these studies indicate that it is not accurate to extrapolate data concerning receptor localization from one brain region to another or, indeed, from one stage of development to another. Our data also highlight potential differences in receptor localization in vitro and in vivo.

Multiple laboratories have now postulated a role for the Met receptor in excitatory synaptogenesis and plasticity. The specific molecular mechanisms through which Met exerts its modulatory effects, however, remain unknown. In studies focusing on the hippocampus, the localization of Met in the postsynaptic compartment has been highlighted (Kawas et al., 2012; Sharma, 2010; Tyndall and Walikonis, 2006), without reporting potential presynaptic contributions. The Met protein labeling in the neostriatum by corticofugal axons (Judson et al., 2009) raised the clear possibility that Met also is prominently localized presynaptically. The current report presents direct evidence that, even in neocortical and hippocampal regions, the presynaptic contribution of Met is substantial. Thus, to understand the role of Met in synapse development and function, and particularly related to pathogenesis in ASD, it will be critical to define the molecular interacting partners of Met and intracellular signaling cascades initiated in response to HGF pre- and post-synaptically.