Fig 1.
Rationale and experimental design.
(A) The experiments described here are based on the assumption that synapses formed between the same axon (green) and the same neuron (blue) will have similar activity histories, particularly when such synapses are located on the same dendrite. Such synapses are referred to as CI and CISD, respectively. For each CI synapse, a Ref synapse is selected that is connected to a different upstream neuron (white). (B) Given their similar activity histories, changes in the sizes of synapses belonging to the same CI pair might be expected to co-vary more than the sizes of two synapses innervated by different axons, i.e., a CI synapse and a Ref synapse (C).
Fig 2.
Comparing the size remodeling of CI and non-CI synapses.
(A) Cortical neurons growing on an MEA dish expressing cerulean-tagged SV2a (Cer:SV2, a presynaptic vesicle protein) and the postsynaptic scaffolding molecule PSD-95 tagged with EGFP (PSD-95:EGFP). (B) Enlarged view of region enclosed in stippled rectangle in (A) showing two CI synapses (green filled arrowheads) and two Ref synapses (empty arrowheads) at three time points obtained one day apart as part of week-long experiments in which PSD-95:EGFP images were collected at 30-min intervals. (C) Changes in PSD-95:EGFP fluorescence compared for the two CI synapses. (D) Similar comparisons made for each pair of a CI synapse and its respective Ref synapse over the same period. Pearson’s correlation (r) and Spearman’s rank correlation (ρ) coefficients measured for two-day periods (pink background) are shown for each comparison. Bars, 20 μm (A) and 5 μm (B). Source data provided in S1 Data.
Fig 3.
Size remodeling covariance of CI and non-CI synapses in monolithic networks.
(A) Distributions of size remodeling covariance values (Pearson’s correlation) for CI and non-CI synapse pairs (92 CI pairs from 24 neurons in 6 experiments). Inset: Same data shown as cumulative histogram. (B,C) Average (±SEM) size remodeling covariance for all CI and non-CI synapse pairs in spontaneously active networks (B) and after suppressing spontaneous activity with TTX (C). (D, E) Same as (B, C)—data pooled by experiment. (F) Distributions of range over mean values for CI and Ref synapses for the same two-day imaging periods. (G, H) Average (±SEM) of range over mean values for all CI and Ref synapses in spontaneously active networks (E) and after suppressing spontaneous activity with TTX (F). Statistical significance values based on two-tailed Mann-Whitney U tests. Source data provided in S1 Data.
Fig 4.
Diversifying network activity by a means of a cholinergic agonist.
(A) A 120-s recording of spontaneous activity in a network of cortical neurons growing on a MEA. Each line reports activity (action potentials) recorded from one of the 59 extracellular electrodes of the MEA. Each dot represents one action potential. An enlarged portion of the recording is shown in the bottom panel. Note the tendency of spontaneous activity in these networks to occur as synchronized, network-wide bursts. (B) The same network as in (A) after exposure to 20 μM of Carbachol. (C) Average (±SEM) size remodeling covariance for all CI and non-CI synapse pairs before exposure to Carbachol (27 CI pairs from 8 neurons in 2 experiments). (D) Average (±SEM) size remodeling covariance for all CI and non-CI synapse pairs after exposure to Carbachol. Source data for (C) and (D) provided in S1 Data.
Fig 5.
Diversifying network activity by means of modular networks.
(A) Schematic diagram of a four-quadrant (4Q) MEA divided into two modules by means of a barrier containing a small number of channels. Neurons in the “presynaptic” module (left) were infected with a lentiviral vector encoding for EGFP:SynI (red). A number of axons crossed over into the “postsynaptic” module in which the expression of PSD-95:mTurq2 was induced in a small number of neurons (green). Note that axons of neurons in the “postsynaptic” module probably also crossed over to the presynaptic module, but these were for the most part invisible. (B) Axons within channels (top) and entering the postsynaptic module (bottom) were visualized by expressing EGFP in the presynaptic module. EGFP expression was done only for development purposes and not used in subsequent experiments. Bar, 20 μm. (C) A neuron in the postsynaptic module expressing PSD-95:mTurq2 (green) innervated by axons originating in the presynaptic module belonging to neurons expressing EGFP:SynI (red). A number of CI synapses within the region enclosed in the stippled rectangle are shown in the bottom panels (cyan arrowheads). Bars, 50 μm (top), 20 μm (bottom). (D) A 40-s recording of activity from the two modules (presynaptic, red; postsynaptic, green). An enlarged (6 s) portion of the recording is shown in the bottom panel. Note that some bursts spread from one module to the other, whereas others did not.
Fig 6.
Size remodeling covariance of CI and non-CI synapses in modular networks.
(A) Distributions of size remodeling covariance values (Pearson’s correlation) for all CI and non-CI synapse pairs (271 CI pairs from 29 neurons from 8 experiments). Inset: Same data shown as cumulative histogram. (B) Average (±SEM) size remodeling covariance for all CI and non-CI synapse pairs. (C) Same as (B)—data pooled by experiment. (D) Distributions of size remodeling covariance values for all CISD (that is, same axon, same dendrite) and non-CI synapse pairs (91 CISD pairs from 29 neurons from 8 experiments). Inset: Same data shown as cumulative histogram. (E) Average (±SEM) size remodeling covariance for all CISD and non-CI synapse pairs. (F) Same as (E)—data pooled by experiment. Statistical significance values based on two-tailed Mann-Whitney U tests. Source data provided in S1 Data.
Fig 7.
Measuring maximal detectable covariance values.
(A) A synapse selected to serve as a template for excitation light intensity modulation, based on a range over mean value similar to the average range over mean values measured over 48-h periods for PSD-95:EGFP and PSD-95:mTurq2 puncta belonging to CI synapses; (~37%; Fig 3G and ~35%, not shown, respectively). (B) The intensity of the excitation laser was modulated to match changes in the fluorescence of selected synapses such as that shown in (A). Intensity was modulated around nominal laser intensities (expressed here as 100%). (C) Pseudocolor images of a segment of a dendrite expressing PSD-95:mTurq2 at five time points (gray arrows in B). The image on the right is an inverted grayscale of the first image (t = 0). Bar, 10 μm. (D) Comparisons of fluorescence covariance for pairs of PSD-95:mTurq2 puncta connected by short arcs in (C) (right-hand side). Note the marked but imperfect fluorescence covariance for such pairs. (E) Distribution of covariance values (Pearson’s and Spearman’s correlations) for 1,223 comparisons made for 100 synapses from 4 different neurons. Source data for (E) provided in S1 Data.
Fig 8.
Size similarity of synapses belonging to the same CISD pairs.
(A) PSD-95:mTurq2 fluorescence values for synapses belonging to the same CISD pairs. Each point represents the PSD-95:mTurq2 fluorescence at the first time-point of a time lapse series. Data shown are for the most stringently selected CISD pairs (40 pairs; see also S5 and S6 Figs). (B) Same as (A) except that here data points represent fluorescence values averaged over 24-h time periods (24 measurements). (C) A similar analysis performed for PSD sizes of 124 pairs of CISD synapses taken from the data set published by [21]. Each point represents a comparison of the PSD sizes (in originally published metrics) of one such pair. To locate such pairs, the data set containing detailed information on 1,700 synapses was filtered to locate synapses connecting the same axons and the same dendrites. Shaft synapses or synapses for which no spine volume data were provided were excluded. Eighty-five CISD synapse groups composed of pairs (72), triplets (10), quadruplets (2) and quintuplets (1) were found. Groups containing more than two synapses were broken into all possible pairwise comparisons. (D) Same as (C) but for spine volumes (in μm3). (E) Contributions of specific activity histories, neuron/dendrite-wide processes, and spontaneous processes to glutamatergic synapse remodeling. The fractional contributions were calculated as explained in main text. Source data for (A–D) provided in S1 Data.