Areal specializations in the morpho-electric and transcriptomic properties of primate layer 5 extratelencephalic projection neurons

Abstract

Large-scale analysis of single-cell gene expression has revealed transcriptomically defined cell subclasses present throughout the primate neocortex with gene expression profiles that differ depending upon neocortical region. Here, we test whether the interareal differences in gene expression translate to regional specializations in the physiology and morphology of infragranular glutamatergic neurons by performing Patch-seq experiments in brain slices from the temporal cortex (TCx) and motor cortex (MCx) of the macaque. We confirm that transcriptomically defined extratelencephalically projecting neurons of layer 5 (L5 ET neurons) include retrogradely labeled corticospinal neurons in the MCx and find multiple physiological properties and ion channel genes that distinguish L5 ET from non-ET neurons in both areas. Additionally, while infragranular ET and non-ET neurons retain distinct neuronal properties across multiple regions, there are regional morpho-electric and gene expression specializations in the L5 ET subclass, providing mechanistic insights into the specialized functional architecture of the primate neocortex.

Keywords: CP: Neuroscience; Non-human primate; Patch-seq; RNA sequencing; cortico-spinal; dendrite; motor cortex; neocortex; physiology; pyramidal neuron; temporal cortex.

Figure 1.. Infragranular glutamatergic neuron subclass and areal distinctions in the expression of ion channel- and neurofilament-related genes in the primate neocortex

(A) The expression of morpho-electric related genes in infragranular glutamatergic cell types was compared between the MCx and TCx of the macaque (n = 6 animals) and human (n = 5 donors). Datasets are from previously published snRNA-seq taxonomies of the human and macaque M1 and middle temporal gyrus (MTG). Infragranular glutamatergic subclasses: L5 extratelencephalic (L5 ET), L5 intratelencephalic (L5 IT), L6b, L6 IT Car3, L5/6 near-projecting (L5/6 NP), L6 corticothalamic (L6 CT) and L6 IT. (B) UMAP visualization of the expression of morpho-electric related genes in infragranular glutamatergic neurons in the human (n = 14,581 nuclei) and macaque neocortex (n = 17,618 nuclei). Individual cells are colored according to their subclass annotations from the original publications. Boxed regions correspond to regions shown in (C). (C) The same UMAPs as in (B) but focused on the L5 ET subclass and color coded by region. (D) Expression of example genes that differentiate L5 ET neurons from non-L5 ET neurons in both human and macaque regardless of neocortical area. All comparisons between L5 ET and non-ET neurons are significant at FDR-corrected p < 2.08e–19, subclass effect, three-way ANOVA. The gray region highlights the L5 ET subclass in each plot (n = 50 nuclei per violin). (E) Examples of differentially expressed genes in L5 ET neurons in the M1 vs. MTG in both human and macaque (n = 50 nuclei per violin). All comparisons between areas are significant at FDR-corrected p < 0.0009, area effect, two-way ANOVA. Violin plots show normalized probability density (width) and median expression (white dots) in log2(counts per million + 1) normalized by gene for each species. See also Figure S1 and Tables S1 and S2.

Figure 2.. Applying Patch-seq to study subclass and cross-areal differences in the electrophysiological properties of infragranular glutamatergic neurons

(A) Schematic of macaque tissue regions collected for recordings and of the corticospinal (CS) neuron retrograde labeling procedure. Tissue was collected from the motor cortex (MCx; red) or the temporal cortex (TCx; black). (B and C) Epifluorescence image of the tdTomato expression in retrogradely labeled CS neurons (B) in the hand-wrist area of M1 (scale bar, 250 μm) (C) and in the cervical spinal cord near the virus injection sites (scale bar, 1 mm). (D) Patch-seq sample alignment within a dendrogram of cell type clusters (defined by reference taxonomy from Jorstad et al.). Shown is the number of Patch-seq samples (bars) mapped to each t type that passed quality control (correlation strength is shown with filled circles; quality control criteria are shown in Figure S2). (E) Top: UMAP of the MTG 10× snRNA-seq dataset (n = 10,971 nuclei), with each of the 7 infragranular glutamatergic subclasses coded by color. Bottom: Patch-seq samples (n = 177) projected on to the same space, colored by subclass labels from correlation-based mapping in (D). (F) Distribution of correlation scores (circles) and number of CS samples (n = 16) mapped with high confidence (bars) to MTG (left) and M1 (right) macaque reference taxonomies. See also Figure S2.

Figure 3.. Subclass and areal variation in the intrinsic membrane properties of infragranular excitatory neurons

(A–D) UMAP visualization of electrophysiology features color-coded by (A) transcriptomic subclass, (B) acute versus culture preparation, (C) cortical area, and D) subclass predicted by logistic regression classification (n = 333 recordings). (E–G) Performance of the classifier trained to predict transcriptomic/long-range projection subclass based solely on physiology. This classifier was used to assign cells without transcriptomics/retrograde labeling to an L5 excitatory subclass (UMAP in D). The scale corresponds to the average fraction of the row in the confusion matrix. (H) Median effect size of each factor in three-way ANOVA used to test for differences in each measured physiological feature. Whiskers represent 1.5× the interquartile range (n = 40 physiological features). See also Figure S3.

Figure 4.. Key physiological differences between L5 ET neurons and other infragranular glutamatergic neuron subclasses across the MCx and TCx

(A) Physiological measures differentiate L5 ET neurons from other infragranular glutamatergic neuron subclasses, ranked by effect size. (B) Left: representative voltage responses (Vm) to chirp current stimuli (I inj). Right: impedance (Z) amplitude profiles of chirp responses. (C) Resonant frequency (fR) distinguishes L5 ET neurons (n = 147) from non-ET neurons (n = 177) in both the MCx and TCx. Whiskers represent 1.5× the interquartile range. (D) Representative Vm responses to hyperpolarizing step I inj. (E) Maximum input resistance is lower in L5 ET neurons (n = 145) than non-ET neurons (n = 173) in both the MCx and TCx. (F) Representative spike waveforms of each cell type for the first action potential in response to a series of 1-s-long DC injections. (G) The derivative of the downstroke (Min dV/dt) is faster in L5 ET neurons (n = 148) than non-L5 ET neurons (n = 176). (H) Violin plots of differentially expressed example genes from Patch-seq data from L5 ET (n = 72) versus non-ET infragranular neurons (n = 92); see Table S2 for subgroups. All comparisons between L5 ET and non-ET neurons are significant at FDR-corrected p < 0.01, subclass effect, three-way ANOVA. Throughout (B)–(H), colors are coded to cell subclass regardless of brain region. See also Figures S3 and S4 and Tables S3 and S4.

Figure 5.. The dendritic morphology of L5 ET neurons is distinct between the MCx and TCx

(A) Individual L5 ET neuron morphologies from the TCx and MCX aligned to an average neocortical template, with dendrites shown in dark blue and axons in light blue. (B) Comparison of L5 ET neurons for each region. Overlaid dendritic morphologies (blue) with soma locations (yellow triangles) are shown on the left. Shown on the right are histograms displaying average dendrite branch length for apical (facing right) and basal (facing left) dendrites by neocortical depth for all reconstructed neurons. Shading shows ± SD about mean, and soma locations are shown by open black circles. (C) Effect size (Cohen’s d) for morphological features that distinguish L5 ET neurons in the MCx versus the TCx. All features are different between areas at FDR-corrected p < 0.01, two-sided t test. (D) Box-and-whisker (1.5× interquartile range) plots of example morphological features that are different between L5 ET neurons in the MCx (n = 12 reconstructions) versus the TCx (n = 11 reconstructions). (E) Performance of the classifier trained to predict neocortical area based solely on dendritic morphology of L5 ET neurons. See also Figure S5.

Figure 6.. Physiological features distinguish L5 ET neurons in the MCx versus the TCx

(A) Physiological measures that differentiate L5 ET neurons between neocortical areas, ranked by effect size. (B) Left: representative voltage responses (Vm) to chirp current stimuli (I inj). Center: impedance (Z) amplitude profiles of the chirp responses. Right: resonant frequency (fR) is higher in MCx L5 ET neurons (n = 77) compared with TCx neurons (n = 70). Whiskers represent 1.5× the interquartile range. (C) Representative first-spike waveforms for L5 ET neurons in the TCx and MCx during 1-s-long direct current (DC) injections. (D and E) L5 ET neurons in the (D) TCx (n = 70) have a deeper fast AHP (fAHP) than L5 ET neurons in the MCx (n = 78) and are more symmetrical (E), as indicated by a lower up/down ratio of the maximum/minimum spike dVdt. (F) Bottom left: example voltage response of an MCx L5 ET neuron in response to the first current injection that drove spiking and the voltage response to the preceding current step (gray). Top left: the same traces expanded in time to show firing onset. Right: the thick line is average maximum instantaneous rate of all MCx L5 ET neurons (red), and the shaded region is the 95% confidence interval with 10 randomly sampled curves from individual neurons. (G) Same as in (F) but for an L5 ET neuron in the TCx (black). Arrows denote individual examples of nonlinear increases in the maximum instantaneous firing frequency. (H) The initial slope of the maximum instantaneous firing rate is greater in L5 ET neurons in the TCx (n = 69) compared to the MCx (n = 78). (I) Representative responses (center) to the largest current injections (bottom) probed during extended FI curve protocols in the MCx (left) and TCx (right), with time-expanded traces of the first and last spikes shown at the top. (J and K) The ratio of the (J) width and (K) maximum dVdt of the last to first spike as a function of the number of spikes in a train. For each plot, the thick line is the average of all neurons (MCx, n = 42; TCx, n = 26), and the shaded region is the 95% confidence interval with 15 randomly sampled curves from individual neurons included. Arrows denote individual examples in which depolarization-induced block of action potential generation occurred. See also Figure S6 and Table S5.

Figure 7.. Subthreshold oscillations are associated with alpha- and beta-band firing frequencies in MCx L5 ET neurons

(A and B) Representative voltage response of (A) MCx L5 ET and (B) TCx L5 ET neurons to a 10-s, near-rheobase current injection. (C and D) Corresponding isolated interspike segments that meet the minimum window required for analysis. (E and F) Autocorrelograms from the example isolated interspike segments in the MCx (E) and (F) TCx. (G and H) Fast Fourier transforms (FFTs) from the example isolated interspike segments in the (G) MCx and (H) TCx. A black line represents the spline fit of individual FFT segments (colored lines). (I) Oscillatory frequency in the autocorrelogram as a function of isolated sweep segment membrane potential (MCx, n = 79 observations from 52 neurons; TCx, n = 66 observations from 49 neurons). (J and K) FFT peak frequency (J) and peak strength (K) of the spline fitted curve as a function of the average sweep segment membrane potential (MCx, n = 91 observations from 59 neurons; TCx, n = 82 observations from 53 neurons). (L) Firing rate (modal instantaneous spike frequency; STAR Methods) during spiking epochs as a function of FFT peak frequency (MCx, n = 122 observations from 42 neurons; TCx, n = 42 observations from 29 neurons). See also Figure S7.