MHC class I molecules are present both pre- and postsynaptically in the visual cortex during postnatal development and in adulthood

OX-18 Antibody Specifically Labels MHCI Protein.

To examine the localization of MHCI proteins in the cortex over development, a pan-specific monoclonal mouse antibody, OX-18, was used to label MHCI proteins in most of our experiments. OX-18 recognizes many MHCI variants and has long been used for immunoaffinity purification and immunodetection of rat MHCI molecules in the immune system (12, 13). OX-18 was generated against an epitope of the extracellularly conserved α3-domain of MHCI RT1A expressed by all rat strains (12). Significant evidence supports its specificity in non-neuronal tissue and also shows that OX-18 specifically recognizes MHCI in the brain. First, OX-18 identified MHCI proteins from rat brain lysate in Western blots at the expected molecular weight (3) (Fig. 1A). This band pattern was replicated by another pan-MHCI antibody, a rabbit polyclonal antibody, that recognizes a different epitope of MHCI RT1A (14) (Fig. S1A). Second, OX-18 immunofluorescence in rat cortical slices (Fig. 1) resembled the pattern of mRNA from in situ hybridization (3). Third, immunofluorescence patterns in rat cortical slices generated by two MHCI antibodies—OX-18 and F16-4-4 (F16), a second monoclonal pan-MHCI antibody raised against a distinct epitope—are similar (Fig. S1B). Fourth, immunofluorescence intensities of surface MHCI were markedly attenuated in dendrites of β2m^−/− mice, which exhibit minimal plasma membrane expression of MHCI (Fig. S2). Finally, the ultrastructural localization of OX-18–labeled MHCI was virtually identical to that labeled by F16 (see data below). Taken together, these results strongly suggest that OX-18 specifically recognizes MHCI proteins.

MHCI Proteins Are Expressed in Rat Visual Cortex Throughout Development.

To examine the expression of MHCI proteins in developing rat visual cortex, two techniques were used: biochemistry and immunohistochemistry. In standard Western blots of isolated visual cortical cellular membranes, MHCI proteins were measured using the OX-18 antibody at four postnatal ages (P2, P9, P23, and adult). Several bands of MHCI proteins are observed at all ages, likely representing co- and posttranslational modifications such as glycosylation (Fig. 1A and Fig. S1A). MHCI proteins were most abundant in early postnatal cortex and decreased with age to comparably low levels during plasticity and into adulthood.

Three ages were investigated with immunohistochemistry: P7, P23, and adult. Sections of visual cortex were immunostained with OX-18, and all cell nuclei were stained with DAPI. MHCI proteins were expressed at all ages examined (Fig. 1 B–D), and the percentage of total cell somata expressing MHCI proteins increased significantly from P7 to P23 and adult (Fig. 1B; P7, 8.6 ± 1.0%, n = 9 sections through cortex; P23, 39.4 ± 2.6%, n = 8 sections; adult, 33.3 ± 2.6%, n = 4 sections; P < 0.001). Qualitatively, MHCI protein expression in P7 animals was highest in layer V, lower in superficial layers, and almost nonexistent in layer VI (Fig. 1C). In older animals, MHCI expression was distributed throughout the superficial layers and remained low in layer VI. (Fig. 1C). OX-18 staining was densely punctate on a diffuse background in proximal dendrites at all ages examined (Fig. 1 C and D). There were also immunostained puncta in the neuropil at all ages, suggestive of dendritic and/or axon terminal staining. These results show that MHCI molecules are present in early postnatal cortex during the peak of synaptogenesis and suggest that MHCI proteins may be important in the early establishment of synaptic connections in addition to their established role in later postnatal synaptic refinement (3, 4).

MHCI Proteins Are Found in Synapses, Axon Terminals, and Dendrites of Layer V Visual Cortex Throughout Development.

To investigate the subcellular distribution of MHCI, immuno-EM was performed on layer V of P7, P23, and adult visual cortex. Initially, experiments were performed using immunoperoxidase to visualize OX-18 labeling. Results showed synaptic and dendritic MHCI localization; however, because of the diffusion of the peroxidase labeling in neuronal profiles, precise quantification of its ultrastructural localization was impossible (Fig. S3A). Therefore, subsequent experiments were performed using two complementary methods of immuno-EM to determine where MHCI was located in the neuropil (Fig. 2). Each method has a distinct advantage. Preembedding immuno-EM retains membrane integrity but sacrifices antibody access to protein-dense regions such as the PSD of the synapse. Conversely, postembedding immuno-EM greatly facilitates antibody access but sacrifices membrane integrity. The secondary antibodies specifically identified MHCI proteins with both methods of immuno-EM (Fig. S3 B and C and Dataset S1). Moreover, the pattern of staining using immuno-EM matched that using immunohistochemistry in that there were fewer gold particles in layer VI compared to layer V (Dataset S1). Results from these two techniques provide a clear and more accurate representation of the subcellular localization of the protein.

Using preembedding immuno-EM, silver-enhanced gold particles in single sections of randomly selected areas of layer V from each age were classified for localization within four subcellular compartments: synaptic structures, non-synaptic dendrites, myelinated axons, and undefined (Fig. 2 A and B). Synaptic structures included presynaptic axon terminals and postsynaptic dendritic spines or dendritic shafts as well as axon terminals containing SVs (Fig. 2A). In addition to immunogold labeling with OX-18, antibody F16 was used to verify that OX-18 was indeed specifically labeling MHCI in immuno-EM. Silver-enhanced immunogold particles were observed throughout the neuropil at all ages examined. Most MHCI proteins were localized to synaptic structures [Fig. 2B (P7, 24.9 ± 4.5%; P23, 29.9 ± 3.0%; adult, 41.5 ± 4.7%; adult-F16, 41.3 ± 7.9%) and Dataset S1] and dendrites [Fig. 2B (P7, 40.8 ± 1.6%; P23, 64.3 ± 1.6%; adult, 42.5 ± 9.1%; adult-F16, 44.6 ± 5.9%) and Dataset S1]. No myelinated axons were observed at P7, and only one was seen at P23, but many were observed in the adult, of which very few contained immunogold particles [Fig. 2B (P23, 0.00%; adult, 0.5 ± 0.5%; adult-F16, 0.0 ± 0.0%) and Dataset S1]. Immunogold particles were also observed in undefined regions [Fig. 2B (P7, 34.3 ± 5.8%; P23, 5.9 ± 4.7%; adult, 15.5 ± 14.4%; adult-F16, 14.1 ± 2.0%) and Dataset S1]. The almost identical distribution of MHCI with OX-18 and F16 supports our conclusion that OX-18 specifically labels MHCI molecules.

To investigate the association of MHCI molecules with membranes, we classified all preembedding silver-enhanced gold particles observed within synapses as unassociated, associated with plasma membrane, or associated with intracellular membranes such as SVs, mitochondria, and endoplasmic reticulum. Immunogold particles were considered to be associated with membrane if the center of the silver-enhanced particle measured less than 45 nm from the nearest membrane (15). Some particles were associated with plasma membranes [Fig. 2C (P7, 8.3 ± 0.8%; P23, 7.4 ± 0.2%; adult, 17.2 ± 6.1%) and Dataset S1], but most were associated with intracellular membranes [Fig. 2C (P7, 71.8 ± 8.2%; P23, 66.5 ± 16.5%; adult, 52.8 ± 2.8%) and Dataset S1]. However, at all ages examined, a population was membrane-unassociated [Fig. 2C (P7, 19.9 ± 7.4%; P23, 26.2 ± 16.7%; adult, 30.0 ± 3.3%) and Dataset S1].

Compared with preembedding immuno-EM (Fig. 2C), more immunogold particles from postembedding immuno-EM experiments were associated with the plasma membrane [Fig. 2 D and E (P7, 16.3 ± 3.8%; P23, 32.4 ± 2.8%; adult, 38.1 ± 5.5%) and Dataset S1] and fewer were classified as intracellular membrane-associated [Fig. 2E (P7, 68.8 ± 3.8%; P23, 60.3 ± 1.0%; adult, 60.16 ± 7.3%) and Dataset S1] and unassociated [Fig. 2E (P7, 15.0 ± 0.0%; P23, 7.3 ± 1.8%; adult, 1.8 ± 1.8%) and Dataset S1]. Unfortunately, the overall distribution of gold particles could not be accurately measured in tissue from postembedding immuno-EM because of the decreased membrane integrity from postembedding immuno-EM processing.

Finally, using a biochemical approach, synaptosome, PSD, and SV fractions were enriched to examine the synaptic localization of MHCI proteins (Fig. 2F). Western blots were also probed with PSD-95 and synaptophysin (SYP) antibodies and show enrichment of the PSD and SV fractions, respectively. MHCI proteins are present and represented equally in both fractions, supporting the conclusion that MHCI is found both pre- and postsynaptically at all ages examined. These data also show that MHCI proteins are associated with SVs, in addition to being found in the synaptic cleft and PSD.

MHCI Proteins Are Expressed in Many Layer V Synapses Both Pre- and Postsynaptically Throughout Development.

To characterize the localization of MHCI proteins within the synapse, preembedding immuno-EM was performed (Fig. 3A and Fig. S4A). MHCI proteins were clearly present at many synapses in layer V at all ages examined. To determine if the synaptic localization of MHCI changes over development, the percentage of total synapses (with a presynaptic terminal, synaptic cleft, and PSD) containing at least one silver-enhanced immunogold particle was quantified at P7, P23, and adult stages of development. In addition to labeling with OX-18, antibody F16 was used for comparison in the adult. Although a majority of synapses contained MHCI at P7, a much smaller proportion of synapses contained MHCI at P23 and in the adult (Fig. 3B). The decrease in the number of MHCI-positive synapses from P7 to P23 and in the adult was highly significant [Fig. 3B (P7, 66.6 ± 1.0%; P23, 26.3 ± 0.8%; adult, 27.3 ± 1.9%; F16, 35.0 ± 5.3; P7 vs. P23, adult, and F16; P < 0.01) and Dataset S1].

Our preembedding immuno-EM data show unequivocally that MHCI proteins are present in both presynaptic axon terminals and postsynaptic dendritic shafts and spines of layer V neurons in visual cortex (Fig. 3 A–D and Fig. S4A). Some synapses contained silver-enhanced immunogold particles both pre- and postsynaptically, whereas others were labeled exclusively in the presynaptic terminal or postsynaptic dendritic shafts and spines (both pre- and postsynaptic: P7, 41.8 ± 10.2%; P23, 22.5 ± 2.5%; adult, 9.5 ± 1.2%; F16, 10.6 ± 3.7%; P7 vs. adult, P < 0.05; presynaptic only: P7, 39.1 ± 3.1%; P23, 61.3 ± 1.3%; adult, 73.9 ± 0.6%; F16, 36.6 ± 11.6%; adult vs. F16, P < 0.05; postsynaptic only: P7, 19.2 ± 7.2%; P23, 16.3 ± 3.8%; adult, 16.6 ± 1.7%; F16, 52.8 ± 7.9; F16 vs. P7, P23, and adult, P < 0.05; Dataset S1). These observations were made on randomly selected single sections and thus may inaccurately assess the frequency of MHCI molecules. Nevertheless, the numbers clearly illustrate that synapses of the visual cortex contain both pre- and postsynaptic MHCI proteins.

To determine if MHCI proteins were associated directly with the active zone or PSD, each silver-enhanced immunogold particle within 0.5 μm of the synaptic membrane of the synapse was measured from the center of the particle to the closest synaptic membrane. Immunogold particles were distributed throughout pre- and postsynaptic structures (Fig. 3 C and D and Dataset S1), but few particles were associated directly with the presynaptic membrane [Fig. 3C Inset (P7, 1.7 ± 1.7%; P23, 1.7 ± 1.7%; adult, 0.8 ± 0.8%) and Dataset S1] or postsynaptic membrane [Fig. 3D Inset (P7, 0.0 ± 0.0%; P23, 1.4 ± 1.4%; adult, 11.6 ± 3.9%) and Dataset S1]. In the adult synapse, more immunogold particles were associated with the postsynaptic membrane than at either P7 or P23, but the numbers remained extremely small. This result likely reflects a lack of penetration of the antibody into the synaptic cleft, a common drawback to preembedding immuno-EM (16).

We repeated these experiments with postembedding immuno-EM, a method that enables greater antibody access (Fig. 3E and Fig. S4B). Results using this method also clearly show MHCI protein expression at many synapses in layer V at all ages examined. Although a majority of synapses contained MHCI at P7, almost half of the total number of synapses contained MHCI at P23 and in the adult (Fig. 3F). The decrease in the number of MHCI-positive synapses from P7 to P23 and adult was significant [Fig. 3F (P7, 89.0 ± 2.7%; P23, 49.4 ± 4.3%; adult, 46.6 ± 1.8%; P7 vs. P23, and adult, P < 0.01) and Dataset S1]. Using this method, it is clear that MHCI proteins are present in both presynaptic axon terminals and postsynaptic dendritic shafts and spines of layer V. Some synapses contained immunogold particles both pre- and postsynaptically, whereas others were labeled exclusively in the presynaptic terminal or postsynaptic dendritic shafts and spines (both pre- and postsynaptic: P7, 22.6 ± 9.0%; P23, 23.4 ± 1.6%; adult, 26.7 ± 13.3%; presynaptic only: P7, 48.7 ± 1.3%; P23, 34.7 ± 4.2%; adult, 41.7 ± 1.7%; postsynaptic only: P7, 28.7 ± 7.7%; P23, 42.0 ± 5.9%; adult, 31.7 ± 11.7%; Dataset S1). Together, results from two immuno-EM methods confirm the localization of MHCI in synapses, both pre- and postsynaptically.

To measure MHCI association with the active zone or PSD, each postembedding immunogold particle within 0.5 μm of the synaptic membrane of the synapse was measured from the center of the particle to the closest synaptic membrane. Immunogold particles were distributed throughout pre- and postsynaptic structures (Fig. 3 G and H and Dataset S1), and some particles were associated directly with the presynaptic membrane [Fig. 3G Inset (P7, 13.0 ± 7.0%; P23, 12.5 ± 12.5%; adult, 30.8 ± 3.5%) and Dataset S1]. More immunogold particles were associated with the postsynaptic membrane [Fig. 3H Inset (P7, 29.2 ± 4.2%; P23, 25.0 ± 1.9%; adult, 48.8 ± 13.8%) and Dataset S1]. Synaptic membrane association increased from P7 to adulthood (Fig. 3 G Inset and H Inset). This result may reflect a more accurate estimate of the distribution of MHCI proteins at the synapse than the preembedding results. Together, our EM experiments revealed that MHCI is localized at the synapse on some SVs and directly at the membrane of the synaptic cleft. The distribution of MHCI both pre- and postsynaptically suggests that MHCI proteins may function on both sides of the synapse to influence connectivity in the visual cortex.

Animals.

All protocols were approved by the Institutional Animal Care and Use Committee. Long–Evans rats aged P7, P23, and adult (sexually mature female rats) were deeply anesthetized and treated as described in SI Materials and Methods.

Subcellular Fractionation.

Visual cortex from P7, P23, and adult Long–Evans rats was homogenized and fractionated using adaptations of established protocols (22–25).

Immunoblot Analysis.

Total cellular membranes were isolated from primary visual cortex and processed as previously reported (3). Both anti-MHCI antibodies (ab52922 and OX-18) show similar molecular weight band patterns on immunoblots (Fig. S1A).

Immunohistochemistry.

Cortical sections were immunostained and imaged. Control sections processed without primary antibody to test the specificity of the secondary antibody showed negligible background compared with those processed with OX-18.

Preembedding Immuno-EM.

Sections were processed as described previously (26). Because detergent was not used, antibody penetration was limited to superficial portions of the sections. Therefore, most of the ultrathin sections were collected from ≈2 μm below the surface.

Postembedding Immuno-EM.

Sections of primary visual cortex were processed and cryoembedded as described previously (27).

EM Imaging.

Ultrathin sections were examined in a Philips BioTwin 120 electron microscope at 80 kV and 11,000× magnification. Electron micrographs were taken with a 2,000 × 2,000-resolution CCD camera (Gatan) and processed with Digital Micrograph software (Gatan).

Data Analysis.

Microsoft Excel and GraphPad Prism were used to analyze data, compute statistics, and generate graphs. For all quantifications, statistics were calculated from two animals per age. All error bars show the SE of the mean. All statistics were quantified by one-way ANOVA with a Tukey post hoc test.

Morphological Subgroup Definitions.

A synapse was defined as a presynaptic axon terminal filled with SVs, which was apposed to a postsynaptic dendritic spine or dendritic shaft containing a PSD, and separated by a synaptic cleft (28). Synaptic structures included classic synapses as well as axon terminals with SVs but no visible postsynaptic apposition. Dendrites were non-synaptic dendrites without a PSD and not apposed to a presynaptic axon terminal. Myelinated axons included axons with clear myelination and no axon terminal. Undefined includes structures that cannot be classified in the previously defined categories. Cell somata were not included.