Stereotyped Position of Local Synaptic Targets in Neocortex

See allHide authors and affiliations

Science  03 Aug 2001:
Vol. 293, Issue 5531, pp. 868-872
DOI: 10.1126/science.293.5531.868


The microcircuitry of the mammalian neocortex remains largely unknown. Although the neocortex could be composed of scores of precise circuits, an alternative possibility is that local connectivity is probabilistic or even random. To examine the precision and degree of determinism in the neocortical microcircuitry, we used optical probing to reconstruct microcircuits in layer 5 from mouse primary visual cortex. We stimulated “trigger” cells, isolated from a homogenous population of corticotectal pyramidal neurons, while optically detecting “follower” neurons directly driven by the triggers. Followers belonged to a few selective anatomical classes with stereotyped physiological and synaptic responses. Moreover, even the position of the followers appeared determined across animals. Our data reveal precisely organized cortical microcircuits.

The neocortex is a tissue of apparently impenetrable complexity (1). The cortical microcircuit, i.e., the intra- and interlaminar connections within a local neocortical region, is still largely unknown, although its characterization is essential to any theory of cortical function (2–4). The search for rules governing the cortical microcircuit (1–7) has revealed wide diversity of neurons (7–9), columnar (2, 5,10–12) and horizontal (13) connectivity, and distinct interlaminar and long-range projections (3,6, 14–17). Connections from cortical interneurons can be precise, targeting specific postsynaptic locations (18–20). However, connectivity rules among excitatory cells, which constitute the vast majority of cortical neurons, remain unclear. Some studies indicate that excitatory neurons are weakly interconnected in probabilistic patterns, whereby specificity can only be found at the statistical level (21–23). At the same time, because the number of different classes of neocortical neurons is still unknown and could approach several hundreds (24), any apparent lack of target specificity might result from heterogeneous sampling. Also, physiological studies indicate remarkable circuit specificity (5, 25).

We used an optical probing technique to detect postsynaptic targets of neurons in brain slices and then chose them for dual recordings (26–28). By imaging hundreds of neurons simultaneously while electrically stimulating a trigger cell, we optically detected which neurons (“followers”) were connected to it (Fig. 1) (29). This approach relies on imaging action potential activity in neuronal populations loaded with calcium indicators (30) and enables parallel probing of connections. We chose as triggers a homogeneous population of layer 5 pyramidal cells that project to the superior colliculus (i.e., corticotectal, or CT, neurons), identified by retrograde labeling (Fig. 1, A and B) (14–17, 31, 32). CT trigger somata were located at stereotyped positions (subpial depth = 620 ± 40 μm, n = 17 cells; 17 slices; 15 mice) (33) and had pyramidal morphologies (34) and thick apical dendrites (diameter = 2.75 ± 0.75 μm at 65 μm from cell body, n = 16) with multiple collateral branches (9.5 ± 2.7 along the initial 150 μm, n = 16) and a tuft in layer 1 (Fig. 2A; Web fig. 1) (35). We stimulated triggers with trains of action potentials using whole-cell recording in an artificial cerebral spinal fluid (ACSF) solution that strengthened synaptic transmission (36) and detected simultaneous somatic calcium transients in follower cells (Fig. 1, D and E) (37). We then established a second whole-cell recording from followers and used dual recordings to confirm monosynaptic connections in 17 trigger-follower pairs (Fig. 1F).

Figure 1

Optical probing. (A) Retrograde labeling of CT cells. Coronal V1 slice from a mouse previously injected with rhodamine beads in the superior colliculus. Note the band of fluorescently bead-labeled cells, ∼500 μm below the surface (top left). Scale bar, 500 μm. (B) Bead-labeled CT pyramidal neuron. DIC image of neurons in layer 5 (left); fluorescence reveals a bead-labeled pyramidal cell (right). Scale bar, 50 μm. (C) Fura-2AM loaded slice. Fields of layer 5 cells are loaded. Orientation as in (A). Scale bar, 100 μm. (D) Optical probing. Consecutive frames from a ΔF/F movie during trigger stimulation (averages of 20 frames each). Stimulation of trigger (black arrow, panels 2 and 4; blue trace, scales = 20 mV, 100 ms) results in a fluorescence transient (pixel brightening) in the trigger and in three neighboring cells (followers; green, red, and yellow arrows, panel 4). Scale bar, 50 μm. (E) Time-locked fluorescence transients from trigger (black) and followers (colored) during trigger stimulation (blue bar). Gray bars and numbers at top indicate frames in (D). (F) Dual recording confirms monosynaptic connection of trigger and follower. Whole-cell recordings from follower [red, (D) and (E)] during trigger stimulation (blue, top). The follower receives EPSPs (middle, average of 24 trials), which can cause it to spike (bottom).

Figure 2

Trigger cell morphology, follower classification, and positions. (A) Examples of three CT triggers. Dendrites are black and axons are red. Scale bar, 200 μm. See also Web fig. 1 (35). (B) Cluster analysis of followers. Four clusters were found. The first one (green) groups “dangling” pyramidal cells, the third one (yellow) groups large triangular interneurons, and the fourth one (red) groups fusiform interneurons. The second cluster (green) groups incompletely reconstructed pyramidal cells and a special interneuron (magenta) follower. (C) Position of followers. Follower somata [colored circles, numbered as in (B)] plotted relative to the triggers (triangle). Background represents overlay of all optical probing fluorescence images (each set to 6% opacity), with regions made optically inaccessible by trigger recording electrode deleted. Darker areas are therefore the most frequently sampled regions. A, apical; M, medial; L, lateral; B, basal: orientation to triggers in all figures. Scale bar, 50 μm. (D) Distributions of distance and (E) direction of followers' somatic positions relative to those of the triggers. Colors as in (B).

We characterized the followers anatomically and physiologically. Followers belonged to three distinguishable morphological classes, confirmed with cluster analysis (Fig. 2, B and C) (38,39). Most followers were pyramidal (7/17; green in Figs. 2, B to E, and 3), followed by a second class of followers with fusiform somata (n = 6/17; red inFigs. 2, B to E, and 4, A to D) and a third class of followers (3/17; yellow in Figs. 2, B to E, and 4, E, F, and H) with large triangular somata. Finally, one follower had a small triangular soma with sparse dendrites (magenta in Figs. 2, B to E and 4, E, G, and I).

Pyramidal followers showed “dangling” basal dendrites that extended ∼300 μm into layer 6 (7) (Fig. 3A; 3/3 cells with completely reconstructed basal dendrites). Pyramidal followers had broad spikes (half peak width = 1.21 ± 0.29 ms, n = 5) (40) and adaptive firing (53% ± 18 decrease in rate during 800-ms current step, n = 4; Fig. 3B). They received depressing excitatory postsynaptic currents (EPSPs) from CT neurons (5/6; Fig. 3C) (41). Putative contacts from trigger axons occurred onto the followers' proximal basal and apical collateral dendrites (92 μm ± 55 from somata, n = 19; Fig. 3A, red circles).

Figure 3

Pyramidal followers. (A) Anatomical reconstructions. Each pyramidal follower is depicted twice: (i) top right, all followers from different slices are overlaid relative to the trigger cells (red triangle; scale bar, 100 μm), with follower axons shown in blue and trigger axons in white, and (ii) surrounding, follower cell bodies and dendrites, positioned so as not to overlap (scale bar, 70 μm). Cells are numbered as in Fig. 2. Note dangling basal dendrites and thin apical dendrites. Putative axonal contacts are indicated with red circles. (B) Regular-spiking firing of pyramidal follower. (C) Paired-pulse depression of CT to pyramidal follower EPSPs.

Fusiform followers had polarized dendrites, which were radially oriented, sparsely spiny, and beaded (Fig. 4A, 10.1 ± 3.4 spines/100 μm, 10.1 ± 4.2 beads/100 μm,n = 6). Axons ascended vertically (4/4) and, in one case, ramified extensively in a band 75 to 175 μm above the cell body before sending two long projections into layer 1. Electrophysiologically, these neurons showed broad spikes (1.4 ms ± 0.5, n = 4), weakly adaptive firing (18 ± 14% decrease, n = 3, Fig. 4B), and low-threshold spiking (LTS; 3/5; Fig. 4C). EPSPs from trigger cells were strongly facilitating (2/3; Fig. 4D). Putative contacts occurred primarily on secondary and tertiary dendritic branches at a characteristic distance from the somata (74 ± 36 μm, n = 40; Fig. 4A, yellow circles). These interneurons were classified physiologically as LTS (42) and resembled lower layers “ascending axon” cells (8) and some classes of bitufted and Martinotti cells (19, 23).

Figure 4

Fusiform followers: (A) Anatomical reconstructions. Fusiform followers and scale bar depicted as in Fig. 3. Note polarized dendrites and putative axonal contacts in yellow. (B) Adaptive firing of fusiform follower to depolarizing current step. (C) LTS evoked by hyperpolarizing pulse. (D) Facilitation of CT to fusiform EPSPs. Triangular followers: (E) Anatomical reconstructions. Triangular followers and scale bar as in Fig. 3. Putative axonal contacts are indicated by blue circles. (F) Intrinsic firing of triangular followers. Top: adaptive spiking. Bottom: LTS. (G) Intrinsic firing of small triangular follower. Top: pausing spike train. Bottom: lack of LTS. (H) Paired-pulse facilitation of CT to large triangular follower EPSPs. (I) Paired-pulse depression of CT to small triangular follower EPSPs.

Large triangular followers showed broader, more stellate dendrites than fusiform cells (Fig. 4E, major/minor axis ratio,r, of binding polygon; r fusiform = 1.6 ± 0.28, r triangular = 1.2 ± 0.08; t test P < 0.05), with spines and beads (13.3 ± 3.9 spines/100 μm, 12 ± 5.7 beads/μm). Axons also ascended vertically (3/3) but, unlike those of fusiform neurons, gave rise to a descending branch at a stereotyped distance from the somata (52.5 ± 2.5 μm, n = 3). Electrophysiologically, these neurons showed broad spikes (1.18 ± 0.23 ms, n = 2), adaptive firing (57 ± 31% decrease, n = 2, Fig. 4F, top), and, in one case, rebound spiking (Fig. 4F, bottom). These neurons received facilitating EPSPs from the triggers (2/2; Fig. 4H). Putative contacts from the trigger axons were on secondary and tertiary dendrites (125 ± 60 μm,n = 18). These interneurons were classified physiologically as LTS and resembled triangular cells with ascending axons described in mouse lower layers (8).

Finally, a small triangular follower showed polarized aspiny and moderately beaded dendrites oriented obliquely and an axon that projected medially before bifurcating into ascending and descending branches (Fig. 4E). Spikes were narrow (mean = 0.93 ms) and adapted (40% decrease) and showed a pause in all trials (Fig. 4G). This neuron was classified physiologically as fast spiking (FS) and received large, strongly depressing EPSPs (Fig. 4I).

In numerous ways, trigger-follower connections appeared specific and stereotyped from animal to animal. Pyramidal followers were different from CT cells: Their apical dendrites were thinner (1.53 ± 0.31 μm, n = 6; t testP < 0.001) and had fewer collateral branches (2.2 ± 2.1 SD, n = 6; t test P≪ 0.001). Indeed, the absence of CT neurons among the seven pyramidal followers indicates that CT neurons preferentially drove non-CT neurons to spike in our experiments. Connections from CT neurons to follower interneurons also appeared stereotyped. To characterize the background population of interneurons in mouse layer 5, we reconstructed 48 randomly sampled nonpyramidal layer 5 cells (43). These interneurons showed great heterogeneity in their morphologies, most of them very different from follower interneurons (see Web fig. 2A, Table 1) (35). Indeed, multivariate statistics ruled out that interneuron followers were randomly drawn from layer 5 interneurons [multivariate analysis of variance (MANOVA); Wilks' lambda = 0.8; P < 0.05] (44). Furthermore, cluster analysis independently showed that interneuron followers represent only a few of the different types of layer 5 interneurons (see Web fig. 2B) (35).

In addition, we noticed that even the somatic positions of the followers were predictable from animal to animal (Fig. 2C). Although triggers were located at a restricted subpial depth, the locations of the three major classes of followers relative to their triggers appeared precisely determined (Fig. 2, D and E), with variances much smaller than that of the depths of the triggers. Fusiform followers were always located below and at a distance of ∼50 μm (51.2 ± 12.2 μm, n = 6) from triggers, forming a semicircle (Fig. 2, C to E). Meanwhile, triangular followers were located ∼65 μm above triggers (66.2 ± 9.5 μm,n = 3). At the same time, pyramidal followers were located in a narrow horizontal band that included triggers (6/7 followers). We tested the positional stereotypy of followers using four independent statistical approaches: (i) The positions of follower interneurons fell along preferred directions from the triggers (Rayleigh's vector average, R fusiform = 0.74, P < 0.05;R triangular = 0.96, P < 0.05), (ii) these directions were not skewed by the positions of all optically probed cells (test of mean directions; P < 0.05), (iii) the observed fusiform cells' position below their triggers was significantly different from the expected one, based on the positions of all probed somata (χ2 = 7.4,P < 0.01), and (iv) the stereotyped regions occupied by follower interneurons' dendrites were poorly correlated to the regions covered by trigger axons (pairwise linear regression of axonal and dendritic probability density contour plots;R fusiform = 0.29;R triangular = 0.03) (45,46).

Using a method to optically reveal connections, we encountered stereotyped local synaptic circuits in neocortex. Although our results could be influenced by many experimental factors, they can nevertheless reveal stereotyped circuits only if they indeed exist. Neurons activated by CT cells fell into a few anatomically and physiologically homogeneous types, which represented a small subset of those present in layer 5 (see Web fig. 2) (4, 8, 19,35, 47, 48). Moreover, even the position of these targeted neurons appeared determined in different animals and was remarkably precise, indicating robust developmental control of circuit formation.

  • * To whom correspondence should be addressed. E-mail: jkjames{at}


Stay Connected to Science

Navigate This Article