Abstract
Article Figures and data Abstract Editor’s evaluation Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract Understanding cortical microcircuits requires thorough measurement of physiological properties of synaptic connections formed within and between diverse subclasses of neurons. Towards this goal, we combined spatially precise optogenetic stimulation with multicellular recording to deeply characterize intralaminar and translaminar monosynaptic connections to supragranular (L2/3) neurons in the mouse visual cortex. The reliability and specificity of multiphoton optogenetic stimulation were measured across multiple Cre lines, and measurements of connectivity were verified by comparison to paired recordings and targeted patching of optically identified presynaptic cells. With a focus on translaminar pathways, excitatory and inhibitory synaptic connections from genetically defined presynaptic populations were characterized by their relative abundance, spatial profiles, strength, and short-term dynamics. Consistent with the canonical cortical microcircuit, layer 4 excitatory neurons and interneurons within L2/3 represented the most common sources of input to L2/3 pyramidal cells. More surprisingly, we also observed strong excitatory connections from layer 5 intratelencephalic neurons and potent translaminar inhibition from multiple interneuron subclasses. The hybrid approach revealed convergence to and divergence from excitatory and inhibitory neurons within and across cortical layers. Divergent excitatory connections often spanned hundreds of microns of horizontal space. In contrast, divergent inhibitory connections were more frequently measured from postsynaptic targets near each other. Editor’s evaluation Hage and colleagues used channelrhodopsin-mediated 2 photon stimulation combined with genetic labeling to survey the synaptic connections from layers 4, 5 and 6 onto layer 2/3 pyramidal neurons in the adult mouse primary visual cortex (V1). This work not only confirms prior knowledge regarding synaptic connectivity made through paired intracellular recordings, but also provides important new parameters and constraints about these connections. The results will contribute to our understanding of the cortical information processing. https://doi.org/10.7554/eLife.71103.sa0 Decision letter eLife’s review process Introduction The receptive field properties of visual cortical neurons are governed by long-range synaptic connections formed across brain areas and local inputs formed between cells within a cortical area. Lack of detailed information about both inter-areal and local connections limits mechanistic understanding of how sensory information is propagated and transformed. So, too, does the absence of detailed information about the patterns of convergence of synaptic inputs to neurons in the recipient layers and divergence of outputs from the projecting layers. Positioned between the thalamic input layer 4 and output layer 5, L2/3 pyramidal cells (L2/3 PCs) represent a critical internal node in the canonical pathway. Although L2/3 neurons in primary visual cortex (V1) are known to receive direct input from dorsal lateral geniculate nucleus (dLGN) (Ji et al., 2016; Morgenstern et al., 2016; D’Souza et al., 2019) and from higher visual areas (D’Souza et al., 2016; Young et al., 2021), monosynaptic rabies tracing suggests the majority of presynaptic inputs arise from neurons within V1 (Liu et al., 2013; Wertz et al., 2015). These local inputs include well-established recurrent excitatory connections between L2/3 PCs, horizontal inhibition from nearby interneurons, and ascending inputs from L4 excitatory neurons that directly receive sensory information from thalamus. Local ascending projections from neurons in layers 5 and 6 to L2/3 have been described in anatomical studies of sensory cortex and are thought to represent feedback from later stages of cortical processing (Burkhalter, 1989; Yoshioka et al., 1994; Harris et al., 2019). Experiments using laser scanning photostimulation combined with glutamate uncaging in visual and somatosensory cortex of juvenile rodents have identified connections from excitatory neurons in L5 to L2/3 PCs (Dantzker and Callaway, 2000; Bureau et al., 2004; Shepherd et al., 2005; Shepherd and Svoboda, 2005; Yoshimura et al., 2005; Hooks et al., 2011; Xu et al., 2016). However, connections from L5 to L2/3 have been scarce or absent when probed via simultaneous electrophysiological recordings in rodent cortex (Thomson and Bannister, 1998; Reyes and Sakmann, 1999; Thomson et al., 2002; Lefort et al., 2009; Jiang et al., 2015). The L5 to L2/3 connection is most often observed from L5a and therefore thought to originate from intratelencephalic (IT)-type L5 pyramidal neurons (Dantzker and Callaway, 2000; Bureau et al., 2004; Shepherd et al., 2005; Shepherd and Svoboda, 2005; Yoshimura et al., 2005; Hooks et al., 2011; Xu et al., 2016). It was recently demonstrated that feedback projections from anteromedial or lateromedial visual areas preferentially target IT-type L5 neurons in V1 to generate loops between higher and lower visual areas (Young et al., 2021), raising the possibility that local ascending projections from L5 could transmit behavioral or brain-state-related information to superficial layers. A direct comparison of the translaminar excitation from layers 4–6 to L2/3 in adult rodent V1 has not yet been made and will be necessary to understand the contributions of feedforward- and feedback inputs to the response properties of L2/3 neurons. Likewise, direct comparisons of horizontal and vertical inhibition are needed to understand the role of local inhibition in shaping neuronal responses. Intralaminar inhibition from interneurons to nearby PCs has been described as abundant and nonselective when measured using two-photon glutamate uncaging (Packer and Yuste, 2011; Fino et al., 2013; Karnani et al., 2014). Such horizontal inhibition from somatostatin-expressing (Sst) interneurons shapes the receptive field properties of nearby L2/3 PCs in mouse visual cortex (Adesnik et al., 2012). Several patterns of vertical inhibition (mediated by translaminar inhibitory connections) have been described for Sst interneurons (Kapfer et al., 2007; Silberberg and Markram, 2007; Jiang et al., 2015; Anastasiades et al., 2016; Naka et al., 2019) and parvalbumin-expressing (Pvalb) interneurons (Olsen et al., 2012; Bortone et al., 2014). Furthermore, subtypes of Sst interneurons may be differentially integrated into distinct layer-specific subnetworks (Muñoz et al., 2017; Naka et al., 2019). While intra- and inter-laminar inhibitory inputs to L2/3 of V1 have been compared using one-photon photostimulation of GABAergic interneurons via either caged glutamate (Xu et al., 2016) or channelrhodopsin-2 (ChR2) expression (Kätzel et al., 2011), potentially unique roles of subclasses of GABAergic interneurons may be revealed by photostimulation techniques with increased genetic and spatial specificity. To this end, two-photon excitation targeted to individual opsin-expressing neurons within genetically defined subclasses of cells can provide improved genetic and spatial precision. Two-photon optogenetic stimulation capable of inducing action potentials (APs) was first demonstrated using rapid galvo scanning of cultured dissociated neurons expressing ChR2 (Rickgauer and Tank, 2009). Newly discovered and engineered opsin variants were subsequently combined with rapid scanning and patterned illumination methods to demonstrate feasibility of two-photon optogenetic stimulation in acute brain slices and in vivo (Andrasfalvy et al., 2010; Papagiakoumou et al., 2010; Prakash et al., 2012; Packer et al., 2012; Packer et al., 2015; Chaigneau et al., 2016; Ronzitti et al., 2017; Pégard et al., 2017; Shemesh et al., 2017; Forli et al., 2018; Mardinly et al., 2018; Yang et al., 2018; Marshel et al., 2019). In recent years, these methods have been combined with electrophysiological recordings to measure synaptic connectivity, with individual studies focused on recurrent excitatory connectivity within cortical layers (Izquierdo-Serra et al., 2018; Seeman et al., 2018) or distinct connectivity of subtypes of Sst interneurons (Naka et al., 2019). In this study, we examined the applicability of two-photon optogenetics to a probe a diverse array of potential synaptic connections by measuring the reliability and specificity of two-photon optogenetic stimulation using seven different Cre lines with either transgenic or adeno-associated virus (AAV)-mediated expression of the opsin ChrimsonR (Klapoetke et al., 2014). We then combined multiphoton stimulation with multicellular patch-clamp recording (up to four neurons) to measure the spatial distribution and functional properties of intralaminar and translaminar synaptic connections to neurons in L2/3 of primary visual cortex in the young adult mouse (722 total connections identified from 10,720 probed). Measures of within-layer connectivity were validated by comparing two-photon optogenetics data to independently collected, gold standard multicellular recordings from the same cortical region and developmental stage. Additionally, we performed targeted patching and validation of optically identified connections in a subset of experiments. Consistent with the canonical microcircuit, L4 excitatory neurons (labeled by Rorb-Cre or Scnn1a-Cre) and inhibitory interneurons within L2/3 displayed the highest rates of connectivity onto L2/3 PCs, while recurrent excitatory connectivity was sparse. Furthermore, we observed translaminar connections from L5 IT-type excitatory neurons (labeled by Rorb-Cre or Tlx3-Cre), and vertical inhibition from Pvalb and Sst interneurons in L4 and L5. Multicellular recordings allowed us to identify individual presynaptic cells with divergent connections onto multiple postsynaptic targets in L2/3. Diverging outputs from inhibitory interneurons were most often confined to postsynaptic cells within 100 µm of each other. By contrast, translaminar connections from individual excitatory neurons in L4 and L5 often diverged to target distant postsynaptic cells separated by hundreds of microns. Many classes of synaptic connections displayed broad and skewed distributions in the amplitudes of excitatory and inhibitory postsynaptic potentials (EPSPs and IPSPs) – with a small minority of connections generating very strong responses. Differences in the strength, kinetics, and short-term dynamics of synaptic responses were found across presynaptic and postsynaptic subclasses of cells; however, the amplitudes of less common synaptic connections were often similar to the more canonical connection paths when probed at the single-cell level. Results We measured intra- and translaminar synaptic connectivity to L2/3 in primary visual cortex of young adult mouse using two-photon stimulation of ChrimsonR-expressing neurons. We used layer-specific Cre lines – Penk-Cre (L2/3 IT), Scnn1a-Cre (L4 IT), Rorb-Cre (L4 and L5 IT), Tlx3-Cre (L5 IT), and Ntsr1-Cre (L6 coritcothalamic [CT]) to stimulate excitatory neurons. Pvalb-Cre and Sst-Cre were used to stimulate subclasses of inhibitory interneurons in multiple layers. Photostimulation of possible presynaptic cells has the potential to greatly increase the throughput of synaptic physiology experiments, but the approach does not provide the verification and record of presynaptic APs for all tested connections that one obtains from multicellular patch-clamp recording. Additionally, while current injection into a patch-clamped cell reliably yields single-cell precision, photostimulation has the potential to activate nearby photosensitive neurons (Anastasiades et al., 2018). Therefore, we began our study by carefully measuring the sensitivity and specificity of two-photon stimulation across all Cre lines used in subsequent mapping experiments. Characterization of two-photon stimulation of ChrimsonR-expressing neurons A number of methods have been developed for two-photon stimulation of neurons, with varying trade-offs in terms of reliability, spatial specificity, ability to stimulate multiple neurons, and complexity (Emiliani et al., 2015). We pursued galvo-based rapid spiral scanning of ChrimsonR-expressing neurons (Rickgauer and Tank, 2009; Prakash et al., 2012; Packer et al., 2012; Klapoetke et al., 2014) as a method to stimulate presynaptic cells (Figure 1A). We tested the reliability and specificity of two-photon stimulation using two methods of ChrimsonR expression. We first utilized the Ai167 TIGRE2.0 line, which we previously found generated robust one-photon-evoked photocurrents (Daigle et al., 2018). The second method improved the spatial specificity of photostimulation by targeting trafficking of ChrimsonR to the soma and proximal dendrites of neurons (Baker et al., 2016). Specifically, we utilized a combinatorial approach in which we performed stereotaxic injections AAV providing Cre-dependent expression of soma-targeted-ChrimsonR (flex-ChrimsonR-EYFP-KV2.1) into primary visual cortex of Cre-positive mice. Figure 1 with 3 supplements see all Download asset Open asset Characterization of two-photon-evoked spiking. (A) Examples of action potentials (APs) evoked by two-photon stimulation of a Penk-Cre neuron with indicated axial offsets. Each offset includes 10 overlaid sweeps. Stimulus duration was 10 ms (red shaded bar). (B) Cumulative probability of light-evoked spiking versus photostimulus power for the indicated Cre lines crossed to Ai167 or injected with AAV carrying soma-targeted ChrimsonR. (C) The average number of APs evoked per photostimulus, latency, jitter, and spatial resolution of photostimulation (measured using the same power used for mapping experiments) for all Cre line and opsin expression strategies used in this study. Error bars represent standard deviation across cells. Minimum power, latency, and jitter data were collected from a total of 123 neurons (8–19 neurons per Cre line-expression method combination). Lateral resolution was measured from a total of 66 neurons (5–12 per Cre line-expression method combination. Axial resolution was measured from a total of 78 neurons (6–13 per Cre line-expression method combination)). Figure 1—source data 1 Raw values for plots in Figure 1. https://cdn.elifesciences.org/articles/71103/elife-71103-fig1-data1-v2.xlsx Download elife-71103-fig1-data1-v2.xlsx To measure the sensitivity and specificity of two-photon stimulation, we used loose-seal cell-attached recordings to minimally disturb intrinsic cell properties (Figure 1A; 10 overlaid sweeps at each stimulus position). Across all ChrimsonR-expressing cells, the minimum power necessary to reliably generate one or more AP per stimulus ranged from 0.5 to 94 mW (10 ms photostimulus duration, measured from 8 to 19 neurons for each Cre line/expression method combination; Figure 1B). Although the distributions vary between Cre lines, we observed spiking in 85–100% of targeted cells in all Cre lines when using an 85 mW photostimulus. AAV-transfected neurons in the Pvalb and Sst lines were very sensitive to photostimulation – half of the sampled cells reliably generated APs in response to 8 mW photostimuli (Figure 1B). Based on these results, we used 85 mW as our primary photostimulation power for probing synaptic connectivity in experiments using Ai167 and excitatory Cre lines transfected with AAV. To avoid off-target activation of the more photosensitive AAV-transfected interneurons, we used a lower stimulus intensity of 35 mW, which reliably activated 86% (Pvalb) and 94% (Sst) of ChrimsonR-expressing interneurons tested. We measured the number of light-evoked APs, the latency to the first AP, and the associated temporal jitter (standard deviation of the latency) as a function of stimulus intensity (Figure 1—figure supplement 1). The mean and standard deviation of these values at the powers used for mapping experiments are presented in Figure 1C. Latency and jitter decreased with higher stimulus intensities, although the effect on jitter was less dramatic as the majority of cells displayed sub-millisecond jitter even with low-power stimuli. Some cells generated multiple APs per photostimulus (Figure 1A and C, Figure 1—figure supplement 1). Similarly, we observed instances of multiple light-evoked postsynaptic potentials (PSPs) in subsequent mapping experiments and utilized exponential deconvolution to aid in identifying such cases (see ‘Materials and methods’). Finally, we characterized the spatial resolution of photostimulation. The probability of light-evoked spiking decreased rapidly with lateral offset of the photostimulus target (full-width at half-maximum [FWHM] of Gaussian curve fit: 12–23 µm; Figure 1C, Figure 1—figure supplement 2), whereas greater axial offset was necessary to observe the same decrease in light-evoked spiking (FWHM: 38–75 µm; Figure 1C, Figure 1—figure supplement 3). This asymmetry in effective excitation volume is consistent with previous reports (Rickgauer and Tank, 2009; Prakash et al., 2012; Packer et al., 2012; Izquierdo-Serra et al., 2018) and suggests that unwanted off-target activation of cells will be most likely when two or more Cre-labeled cells are in near axial alignment. Both the latency and jitter of light-evoked spiking increased with either lateral or axial offset of the photostimulus location (Figure 1—figure supplement 2, Figure 1—figure supplement 3). As a result, while nearly all responsive neurons displayed AP firing with short latency (<15 ms) and low jitter (<2 ms) following on-target stimulation (143/144 cells, 99%, aggregated across Cre lines), only 30% of photoresponses measured with 10 µm of lateral offset met the same criteria (127/423 photostimulus locations). Such precise firing was nearly absent with 20 µm of lateral offset (9/396 photostimlus locations; 2.3%). The data further suggest that off-target photostimulation is unlikely to produce reliable, short-latency synaptic responses. Measurements of cell densities and estimation of off-target activation In addition to the spatial resolution of individual opsin-expressing neurons, the likelihood of generating off-target spiking will be influenced by the density of photosensitive cells. Across Cre lines and expression methods used in this study, the average densities of labeled cells ranged from ~2000 to 22,000 cells/mm3 (Figure 2A–C). As expected, labeling with interneuron Cre lines was relatively sparse in comparison to excitatory Cre lines (Figure 2A–C). For each Cre line, the density of cells labeled using AAV was slightly lower or approximately equal to the corresponding density measured in Ai14 (tdTomato reporter) mice (Figure 2B; on average, density with AAV was 84% of densities measured in Ai14). In contrast, cell densities measured from Ai167 mice suggest a minority of cells within a targeted subclass were labeled (Figure 2C; on average, density with Ai167 was 25% of densities measured in Ai9 or Ai14). Incomplete or sparse labeling has been observed with other combinations of Cre lines and some TIGRE1.0 or TIGRE2.0 reporter lines ( Madisen et al., 2015; Daigle et al., 2018; Bounds et al., 2021). Although the sparse labeling by Ai167 may be serendipitous to the extent that it limits off-target photostimulation, the more thorough labeling and inclusion of a soma-targeting motif with AAV-mediated expression motivated us to use it as our primary strategy in this study. Data collected from individual Cre lines using either method of expression (Rorb and Sst) are directly compared in subsequent sections (Figure 4B and C, Figure 6B and C, Figure 7—figure supplement 3). When targeting excitatory subclasses, labeled cells were observed in the expected cortical layers (Figures 3-5), and the intrinsic electrophysiological features of labeled Pvalb and Sst interneurons resembled previous reports (Figure 3—figure supplement 1). In total, the data suggest that the targeted Cre line-defined subclasses were appropriately labeled. Notably, in all experiments, the photosensitive cells constitute a minority of all neurons as estimates of total neuron densities in visual cortex of the mouse range from 92,000 to 214,000 cells/mm3 (Figure 2B and C; Cragg, 1967; Heumann et al., 1977; Rockel et al., 1980; Keller et al., 2018). Figure 2 with 1 supplement see all Download asset Open asset Labeled cell densities and estimates of off-target activation. (A) Example images for all Cre lines and expression method combinations used in this study. Two-photon z-stacks were collected at the conclusion of mapping experiments. Each image shown is a maximum intensity projection of a 40 µm subset of the axial/z dimension. Scale bars = 60 µm. The positions of labeled cells were manually annotated in three dimensions (colored dots, z location not represented here). (B) Cumulative distributions of labeled cell densities measured across multiple experiments using adeno-associated virus (AAV) (dashed lines). Distributions are color-coded by Cre line as in panel (A). Open triangles represent the mean density of labeled cells measured from Ai14 adult mice crossed to the same Cre lines. Filled gray triangle represents the average reported total neuron density in mouse visual areas (Keller et al., 2018). (C) Same as panel (B) for experiments using Ai167-mediated ChrimsonR expression. (D) For 100 randomly selected neurons in the Tlx3:AAV dataset (overlaid green circles), the relative location of all identified neighboring cells s plotted and color-coded by estimated off-target activation probability (note warmer colors near the origin). Neighbors with an off-target activation probability < 0.01 are represented by smaller gray circles. (E) Same as panel (D) for 100 starter neurons from the Sst:AAV dataset (green circles). (F) Estimation of off-target activation for Tlx3:AAV experiments. Left panel: heatmap of the number of neighbors per cell at combined lateral and axial offsets. With increasing lateral offsets, each bin corresponds to a concentric disc of increasing volume (Figure 2—figure supplement 1), and therefore, an increasing number of neighbors are observed. panel: heatmap of the photostimulus panel: heatmap of the number of activated off-target neighbors per targeted cell estimated as the of the number of neighbors and the probability of off-target activation. The across all estimates the average number of off-target cells activated by photostimulation in Tlx3:AAV experiments and is plotted as an in panel represents the location of the targeted Same as panel (F) for Sst:AAV experiments. The same were used in each The estimated number of activated off-target cells per stimulus versus labeled cell density for all Cre line and expression method combinations used in this study. lines represent estimates of off-target activation across a range of cell densities using two different photostimulus to the range of observed across Cre lateral at half-maximum [FWHM] = axial = 35 µm; gray lateral = axial = 66 µm; see also Figure 2—figure supplement 1). represents the reported of the total number of cells activated per photostimulus using two-photon of caged glutamate in visual cortex et al., of photostimuli generating a synaptic response following stimulation of multiple cells with a common connection probability plotted connection lines represent different values of line represents Figure data 1 Raw values for plots in Figure Download To our measurements of cell densities to estimates of off-target activation for each Cre line and expression we first all in the lateral and and axial these and the measurements of spatial resolution described (Figure 1), we estimated the probability of off-target activation for each and corresponding activation are for 100 randomly selected starter cells and all identified neighbors from a labeled line Figure and a relatively sparse line Figure – both transfected with AAV. For across all measuring the number of neighbors at combined lateral and axial were generated and an was to for possible off-target cells of the (see ‘Materials and methods’). The (Figure and represent the average number of labeled neighbors per targeted measurements of the spatial resolution of photostimulation (Figure 1—figure supplement 2, Figure 1—figure supplement were used to the probability of off-target activation at the same lateral and axial (Figure and The of these two the number of off-target activated neighbors at each (Figure and Finally, the across all estimates the total number of activated off-target cells per targeted cell off-target for Tlx3:AAV and off-target for These estimates were generated for the used in this study (Figure and for photosensitive neurons (Figure Figure 2—figure supplement at densities to cells/mm3 average reported density of all neurons in mouse visual Keller et al., 2018). The estimates that unwanted off-target activation is with sparse interneuron Cre lines off-target < Figure but higher when using more labeled excitatory Cre lines (up to off-target for Ntsr1-Cre combined with Figure photostimulation with similar spatial resolution to experiments in which a of neurons are photosensitive to either opsin expression or use of caged will likely activate multiple cells per target (Figure line In with these two-photon of caged glutamate was previously estimated to activate excitatory neurons at each photostimulus location on an estimated density of cells/mm3 in visual et al., Figure gray Therefore, of using Cre lines can the extent of unwanted off-target activation. Additionally, synaptic responses from off-target activation are expected to originate from the subclass of cells labeled by the utilized Cre Finally, we effect off-target activation may have on estimates of synaptic connectivity from photostimulus responses. The probability of generating an off-target synaptic response will on both the total number of cells activated and the connection probability of cells. we that off-target presynaptic cells similar connection to the targeted cells a common Cre line and in a similar anatomical we can the probability of at one cell a synaptic response as represents the per cell connection probability and represents the total number of cells activated per photostimulus targeted and For values of for the within this study (<2 1 that will be to nearly all from to 1 (Figure lines). is increased to activated neurons a two-photon stimulus in which the majority of neurons are as 0.5 (Figure gray In this it may be to abundant connectivity from however, the measured likelihood of responses can potentially patterns within It is also important to that as the number of off-target cells the potential for responses responses from a of cells will likely increase as In total, these that our two-photon optogenetic stimulation strategy will to synaptic connectivity, even as off-target activation and some of connectivity to presynaptic cells may We to these in our mapping experiments and connections identified by photostimulation using targeted Intralaminar connectivity measured by two-photon stimulation connectivity measured by paired recordings the associated with photostimulation, we began our study of synaptic