Laterality and manuality dependent and invariant activity

(A) Top row: schematic of cortical recording and food handling conditions. Bottom row: schematic of hand-nose distance for the ipsilateral (green) and contralateral (purple) hands bringing food to the mouth. (B-E) Simplified, non-exhaustive schematic of hypothetical cortical activity patterns under different types of dependence and invariance.

Head-fixed mice can handle food bimanually or unimanually

(A) The hand blocker apparatus consists of a modified arm rest that slides from side to side with flaps that can be rotated and locked in place to prevent one hand or the other from reaching the area immediately in front of and below the mouse’s mouth. (B) Example video frames showing a mouse handling food unimanually with the left hand (left), unimanually with the right hand (middle), and bimanually (right). (C) Example hand-nose distance traces of left unimanual (left, green), right unimanual (middle, purple), and bimanual (right, blue) food handling. (D) Example rasters of all recorded units from a bilateral fl-M1 recording aligned to the kinematic traces in (C). (E) Probe-average firing rate histograms (100 ms bins) of the data in (D).

Units in fl-M1 increase in activity around unimanual and bimanual transport-to-mouth events

(A) Hand-nose distance aligned to onsets of ipsilateral (left), contralateral (middle), and bimanual (right) transports-to-mouth, for an example right fl-M1 recording, plotted as a heat map with one row per event. White regions indicate oromanual/ingestion epochs belonging to previous or following holding/chewing-oromanual/ingestion cycles, which were excluded from the aligned traces and PETHs for the purposes of analyzing event-aligned data. (B) Mean (thick line) and S.D. (shaded area) hand-nose distance for the same data in (A). (C) Raster plot (grey lines) and trial-average PETHs (colored lines) for an example fl-M1 unit aligned to the same data in (A-B). (D) Baseline subtracted, peak-normalized, trial-average PETHs for each unit recorded from fl-M1 during the recording associated with (A-C), sorted by time to peak firing around bimanual transports-to-mouth. (E) Grand average (across units and events) baseline-subtracted PETHs for the units in (D).

Cortical activity in multiple areas increases around unimanual and bimanual transport-to-mouth events

(A) Mean (lines) and S.D. (shaded area) hand-nose distance traces for 13 recording sessions from 7 mice aligned to ipsilateral (left), contralateral (middle), and bimanual transports-to-mouth (right). (B) Z-scored baseline-subtracted firing rate aligned to transports-to-mouth averaged over events and units for 8 fl-M1 recordings from 6 mice. (C) As (B) for 4 fl-M2 recordings from 3 mice. (D) As (B) for 11 LOM recordings from 6 mice. (E) Z-scored baseline-subtracted firing rate aligned to transports-to-mouth averaged over probes, days, then mice in fl-M1 (violet), fl-M2 (gold), and LOM (teal). (F) As (E), but arranged by area and colored by condition. Left: fl-M1, middle: fl-M2, right: LOM. Green: ipsilateral, purple: contralateral, blue: bimanual.

Mice appearing in this paper

LOM shows a higher proportion of non-selective significant responses compared to forelimb areas

(A) Trial-average, baseline-subtracted, Z-scored PETHs for all recorded fl-M1 units (pooled across recordings and mice) aligned to ipsilateral (left), contralateral (middle), and bimanual (right) transports-to-mouth. Units are organized by condition(s) for which they show significant firing rate increases. (B) As (A), but for fl-M2 (C) As (A), but for LOM (D) Exploded Venn diagrams showing predicted proportions of significant responses under different types of dependence (qualitatively, and assuming symmetric laterality preferences). (E) Venn diagrams showing proportions of units (calculated per recording, then averaged over probes, days, and mice) that are significantly responsive during only one type of transport-to-mouth, to two types, or to all three in fl-M1 (left, 8 recordings from 6 mice), fl-M2 (middle, 4 recordings from 3 mice), and LOM (right, 11 recordings from 6 mice). (F) The same data as in (E) re-plotted as a Sankey diagram. Note the relative size of the dark red (contralateral only) band in fl-M1/fl-M2 and grey (responsive during all) band in LOM.

The fl-M2 shows the most lateralized transition-aligned activity, LOM the least lateralized

(A) example raster plots and PETHs (green: ipsilateral, purple: contralateral, blue: bimanual) for three example units - one with a strong contralateral preference but a weak bimanual preference (left), one with a weak ipsilateral preference but a strong bimanual preference (middle), and one with no strong preferences (right). (B) Cumulative distributions of preference indices for contralateral versus ipsilateral transports-to-mouth for fl-M1 (violet), fl-M2 (gold), and LOM (teal). (C) As (B), for absolute value of contralateral versus ipsilateral preference indices. (D) As (B), for bilateral transports-to-mouth versus each unit’s preferred unimanual transports-to-mouth. (E) Scatter plot of laterality preference indices vs manuality preference indices for all recorded units in fl-M1 (left, 8 recordings from 6 mice), fl-M2 (middle, 4 recordings from 3 mice), and LOM (right, 11 recordings from 6 mice). Small grey dots indicate no significant preference. Large colored dots indicate significant ipsilateral preference (green), ipsilateral and unimanual preference (lime), unimanual preference (yellow), contralateral and unimanual preference (peach), contralateral preference (purple), contralateral and bimanual preference (lavender), bimanual preference (light blue), or ipsilateral and bimanual preference (turquoise). (F) 2D histogram of the data in (E).

Population dynamics are more conserved across conditions in LOM than forelimb motor cortices

(A) Mean bimanual transport-to-mouth aligned activity for an example fl-M1 recording projected onto the top three principal components recovered from the data (left), and the activity along the top two principal components plotted against each other over time (right) for transports-to-mouth (blue). (B) As (A) for an example LOM recording. (C) Boxplot of fraction of explained variance accounted for by the top principal component of transport-to-mouth aligned average activity for all mice (see Table 1 for mice included for each condition), organized by area then condition. (D) As (C), but organized by condition then area. (E) Left: Event aligned neural activity associated with bimanual transports-to-mouth in an example fl-M1 recording projected onto the top principal component identified from ipsilateral (green), contralateral (purple), and bimanual (blue) transport-to-mouth responses. Right: Event aligned neural activity associated with ipsilateral (green), contralateral (purple), and bimanual (blue) transports-to-mouth projected onto the top two bimanual principal components. Blue traces are the same as in (A). (F) As (E) for an example LOM recording. (G) Left: Boxplot of subspace angles between the top principal components for each area and condition pair. Right: Expected distributions of subspace angles under laterality/manuality invariance (light gray) and maximum laterality/manuality dependence (i.e. orthogonality, dark gray) compared to the real data (red bars: condition pair/area medians from left panel). Areas fl-M1 and fl-M2 have been pooled to ensure at least two mice were included for each condition pair. (H) As (G) for alignment index between the top ten principal components.

Statistical analyses (linear mixed-effects models)

Population activity throughout food handling reorganizes between conditions in forelimb motor cortices but not LOM

(A) Pairwise correlations for an example fl-M1 recording during continuous ipsilateral (left), contralateral (middle), and bimanual (right) food handling. Ordering of units is consistent between panels. (B) As (A) but for simultaneously recorded LOM activity. (C) Correlation between the fl-M1 (left) and LOM (right) population correlation matrices for each condition, for the recording in (A-B)= (D) Boxplot of correlation between population correlation matrices for each pair of conditions for all recordings (fl-M1: 7 recordings from 6 mice, fl-M2: 6 recordings from 4 mice; LOM: 11 recordings from 6 mice), organized by area.

General linear model decoding of continuous food handling kinematics

(A) Schematic of GLM decoding. Sliding windows of neural activity (left) are multiplied by a matrix (time lags × units) of fitted model coefficients (middle) to reconstruct food handling kinematics on a moment-to-moment basis (right: kinematics are shown in black, model prediction in light blue). (B) Example kinematic trace (black) and decoded kinematics (light blue) from fl-M1 spiking activity. (C) Boxplots of average decoding accuracy (cross-validated R2) across areas (fl-M1: 7 recordings from 6 mice, fl-M2: 6 recordings from 4 mice; LOM: 11 recordings from 6 mice), dimensions, and conditions. Green: ipsilateral unimanual kinematics, purple: contralateral unimanual kinematics, yellow: ipsilateral bimanual kinematics, light blue: contralateral bimanual kinematics.

Kinematic decoding is preserved between conditions in LOM but not forelimb motor cortices

(A) Schematic of decoder generalization. Coefficients fit to neural and kinematic data from one condition (e.g. unimanual contralateral) are used to decode neural data from another condition, i.e. unimanual handling with the other hand or bimanual handling with the same hand. The cross-condition R2 is then compared to the R2 for the within-condition decoder for a held-out dataset from the same condition as the within-condition training data. (B) Example of decoder generalization. The same bimanual contralateral kinematic trace from Figure 9A is shown in black, along with its decoding from the decoder trained on bimanual contralateral data in light blue. The purple trace shows the kinematics decoded from the same activity with the decoder trained in the contralateral unimanual condition. (C) Left: average percentage change in decoding accuracy (ΔR²/R²₀) when generalizing across limbs and conditions for each area (fl-M1: 7 recordings from 6 mice, fl-M2: 6 recordings from 4 mice; LOM: 11 recordings from 6 mice). Right: expected distributions of ΔR²/R²₀ under perfect generalization (light gray) and no generalization (chance cross-condition decoding accuracy, dark gray), compared to the real data (red bars: condition pair/area medians from left panel)

Recording site locations

(A) Schematic dorsal view of a mouse brain annotated with locations of all fl-M1 (violet), fl-M2 (gold), and LOM (teal) recordings analyzed in this paper. (B) Merged brightfield-epifluorescence images of brain slices with pseudocolored fluorescent probe track locations for example fl-M1 (top, magenta), fl-M2 (middle, yellow), and LOM (bottom, teal) recordings.

Principal components analysis and related analyses without pooling fl-M1 and fl-M2

(C) Boxplots of fraction of explained variance accounted for by the top principal component of transport-to-mouth aligned average activity, organized by area then condition. (D) As (C), for participation ratio. (G) Left: Boxplots of subspace angles between the top principal components for each area and condition pair. Right: Expected distributions of subspace angles under laterality/manuality invariance (light gray) and maximum laterality/manuality dependence (i.e. orthogonality, dark gray) compared to the real data (red bars: condition pair/area medians from left panel). (H) As (G) for alignment index between top ten principal components.