Ultrastructure of light-activated axons following optogenetic stimulation to produce late-phase long-term potentiation

Abstract

Analysis of neuronal compartments has revealed many state-dependent changes in geometry but establishing synapse-specific mechanisms at the nanoscale has proven elusive. We co-expressed channelrhodopsin2-GFP and mAPEX2 in a subset of hippocampal CA3 neurons and used trains of light to induce late-phase long-term potentiation (L-LTP) in area CA1. L-LTP was shown to be specific to the labeled axons by severing CA3 inputs, which prevented back-propagating recruitment of unlabeled axons. Membrane-associated mAPEX2 tolerated microwave-enhanced chemical fixation and drove tyramide signal amplification to deposit Alexa Fluor dyes in the light-activated axons. Subsequent post-embedding immunogold labeling resulted in outstanding ultrastructure and clear distinctions between labeled (activated), and unlabeled axons without obscuring subcellular organelles. The gold-labeled axons in potentiated slices were reconstructed through serial section electron microscopy; presynaptic vesicles and other constituents could be quantified unambiguously. The genetic specification, reliable physiology, and compatibility with established methods for ultrastructural preservation make this an ideal approach to link synapse ultrastructure and function in intact circuits.

Fig 1. Viral expression of ChR2-GFP and mAPEX, and induction of optical L-LTP.

(A) The rAAV was designed to co-express ChR2-GFP and mAPEX under a single synapsin promoter in neurons. During translation, the P2A peptide self-cleaves to yield the two proteins. (B) Cultured hippocampal neurons infected with the rAAV co-expressed ChR2-GFP and mAPEX1. Scale bars = 25 μm. (C) ChR2-GFP was co-expressed with mAPEX2 in the same axons in the area CA1 (dotted white lines). Yellow and magenta lines indicate ChR2-GFP only and mAPEX2 only puncta, respectively. Scale bars = 2 μm. (D) GFP fluorescence images (left) and experimental configurations (right) in ipsilateral (top, ipsi) and contralateral (bottom, contra) hippocampal slices. Scale bars = 500 μm. (E) Electrical (top) and optical (bottom) EPSPs evoked by a train of 50 Hz stimulations, recorded from CA1 middle stratum radiatum (C57B/6J strain). Responses to the first 40 pulses are shown and the stimulation artifacts are clipped from the eEPSP data. The inset shows HFS protocol for LTP induction. Scale bars = 1 ms, 2 mV (electrical), 1 mV (optical). (F) Optical HFS induced L-LTP in the Schaffer collateral and commissural pathways. Left: Representative oEPSP traces from ipsilateral (top) and contralateral (bottom) slices before (light shaded line) and 3 hr after (solid line) HFS. Right: Time course of oEPSP slope (mean ± SEM) showing optical L-LTP in ipsilateral (pink) and contralateral (purple) slices. Addition of APV blocked L-LTP (grey). The number of slices is indicated in parentheses. Scale bars = 4 ms, 1 mV.

Fig 2. Optical stimulation of CA3→CA1 commissural fibers was sufficient for L-LTP.

(A) Representative image of a contralateral slice with CA3 cut off from CA1. Scale bar = 500 μm. (B-C). LTP induced by electrical (B) and optical (C) HFS in intact and cut slices. The example traces show EPSPs recorded before (light shaded lines) and 3 hr after HFS (solid lines). Optical LTP magnitude (C) at 30 and 60 min post-HFS was significantly different between intact and cut slices (oEPSP slope F_{(1, 17)} = 4.74, p < 0.05; Time F_{(3, 51)} = 0.95, p > 0.05; Interaction F_{(3, 51)} = 6.60, p < 0.0001; repeated measures two-way ANOVA with Bonferroni’s post-hoc tests). Scale bars = 2 ms, 2 mV (B); 1 ms, 0.5 mV (C). (D) Summary data for oEPSP slope change at different time points following optical HFS in intact and cut slices. The intact slices showed significant changes in LTP magnitude in the last 120 min of recordings (F_{(1.28, 12.8)} = 5.75, p < 0.05; repeated measures one-way ANOVA). The box plots show medians and interquartile ranges, with whiskers extending from minimum to maximum values. The number of slices used for each condition is indicated in parentheses.

Fig 3. mAPEX2-catalyzed labeling and 3DEM identification of rAAV-targeted axons.

(A) Workflow for processing of hippocampal slices following optical L-LTP. Vibratome sections from the fixed area CA1 underwent tyramide signal amplification (TSA) catalyzed by mAPEX2 to deposit Alexa Fluor dye in the targeted axons. The dye-labeled section was then stained with heavy metals and embedded into epoxy before being cut into serial thin sections. The dye-containing axons in a subset of the sections were immunolabeled with 5 nm gold particles, followed by gold enhancement. Scale bars = 100 μm. (B and B’) A low magnification tSEM image of the area CA1 stratum radiatum after immunolabeling (Step 8 in A). Axonal boutons indicated by red contours were positively labeled with 5 nm gold particles (red arrowheads). Area indicated by black rectangle is enlarged in B’. Scale bar = 1 μm in B, 250 nm in B’. (B”) A tSEM image of colloidal gold particles (5 nm and 15 nm). To visualize the 5 nm particles more clearly, this image was acquired at 1 nm/pixel and scaled to the same magnification as B’. (C and C’) Same as B and B’, imaged after gold enhancement (Step 9 in A). Scale bar = 1 μm in C, 250 nm in C’. (C”) A tSEM image of enhanced gold particles. The numbers indicate diameters in nm (also see S5 Fig).

Fig 4. Post-embedding immunogold labeling for Alexa Fluor dyes deposited by mAPEX2-driven TSA reaction allowed for 3DEM identification of rAAV-targeted axons, while maintaining excellent ultrastructure.

(A1-A3) Three adjacent serial tSEM images of an axon containing immunogold labeling (red arrowhead). Scale bar = 250 nm. (B1) 3D reconstruction of the labeled axon (green) shown in A, which contained immunogold labels (black spheres) and synaptic vesicles (magenta spheres). Two mitochondria (purple) were associated with one of the two boutons. Red rectangle represents a portion of this axon shown in A1-A3. (B2) Same axon as B1, with reconstructions of spines (yellow) forming synapses (red) with this axon. Note, s1 was a branched spine with one of the heads forming a synapse with another axon. The second spine (s2) and its PSD could be reconstructed only partially because they were located at the end of the tSEM image series. Both axonal boutons are multi-synaptic, with each bouton forming synapses with two spines originating from different dendrites. (B3) Same as B2, rotated along the horizontal axis 180° to provide a different view of synapses and spines. One of the synapses at mitochondria-containing bouton was perforated (also see S1 Video). Red rectangle represents a portion of this axonal bouton shown in C1-C4. Scale cube = 250 nm per side. (C1-C4) Four adjacent serial tSEM images of the gold-labeled (red arrowhead) axonal bouton, forming synapses with two dendritic spines (s1 and s2; PSDs indicated by black arrows). Scale bar = 250 nm.