Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2010 Jan 27;30(4):1322-36.
doi: 10.1523/JNEUROSCI.5894-08.2010.

A neural basis for motor primitives in the spinal cord

Corey B Hart  1 , Simon F Giszter
Affiliations

A neural basis for motor primitives in the spinal cord

Corey B Hart et al. J Neurosci. .

Abstract

Motor primitives and modularity may be important in biological movement control. However, their neural basis is not understood. To investigate this, we recorded 302 neurons, making multielectrode recordings in the spinal cord gray of spinalized frogs, at 400, 800, and 1200 mum depth, at the L2/L3 segment border. Simultaneous muscle activity recordings were used with independent components analysis to infer premotor drive patterns. Neurons were divided into groups based on motor pattern modulation and sensory responses, depth recorded, and behavior. The 187 motor pattern modulated neurons recorded comprised 14 cutaneous neurons and 28 proprioceptive neurons at 400 mum in the dorsal horn, 131 intermediate zone interneurons from approximately 800 microm depth without sensory responses, and 14 motoneuron-like neurons at approximately 1200 microm. We examined all such neurons during spinal behaviors. Mutual information measures showed that cutaneous neurons and intermediate zone neurons were related better to premotor drives than to individual muscle activity. In contrast, proprioceptive-related neurons and ventral horn neurons divided evenly. For 46 of the intermediate zone interneurons, we found significant postspike facilitation effects on muscle responses using spike-triggered averages representing short-latency postspike facilitations to multiple motor pools. Furthermore, these postspike facilitations matched significantly in both their patterns and strengths with the weighting parameters of individual primitives extracted statistically, although both were initially obtained without reference to one another. Our data show that sets of dedicated interneurons may organize individual spinal primitives. These may be a key to understanding motor development, motor learning, recovery after CNS injury, and evolution of motor behaviors.

PubMed Disclaimer

Figures

Figure 1.
Figure 1.
Data types and processing used in the study. A, Recordings were made using a custom tetrode system. Individual units were isolated using PCA and EM algorithms incorporating the tetrode co-occurrence and covariations. Sample waveforms of four units from the highest amplitude tetrode channels are shown in each trace 1–4 for units whose rasters are shown with all 17 channels isolated in a frog in B. Multielectrode single and multiunit activity was recorded at 400–500, 700–800, and 1000–1200 μm depth at the L2/L3 segment border in a spinalized frog. B, Typical rasters of spike activity in a frog. C–G show sample data recorded in two trials (Trial 1, C–E; Trial 2, F, G): C, raw EMG; D, integrated and filtered EMG recordings; E, rasters on the corresponding trial; F, rasters on a second trial in the frog. Note general crude resemblance of firing structures in E–G from a second frog. G, Estimated rate from a unit in F using a Gaussian filter. H, Sites in spinal cord from which data were obtained: dorsal horn (DH), 400 μm; intermediate zone (IZ) gray, 800 μm; ventral horn (VH), 1200 μm. Recording occurred along penetrations at the L2/L3 border 500 μm from the midline.
Figure 2.
Figure 2.
Data analysis approach. Spike data were converted from analog recordings (type A in each analysis) to isolated unit spike times (type B) used in spike-triggered averages, to continuous rate estimates (type A) used for linear correlation based methods and to symbol string representations (15 ms bins, 8 symbols for level, type C) used for information-based analyses. Electromyograms were rectified filtered and used in ICA to obtain activation patterns of independent components or sources (type A) and their projection weights to muscles (type D). The resulting (type A) EMG and component waveforms were used in linear correlation methods and converted to symbol strings (15 ms bins, 8 symbols for level) for use in information techniques. Raw and rectified unfiltered EMG was used in spike-triggered averages to obtain significant facilitations and their magnitudes (type D). Comparisons and statistical tests of hypotheses were then made using the symbol string representations (C), the analog time series (A), and the projection weights and STA facilitation weights (D).
Figure 3.
Figure 3.
Comparisons across depths. Units responses were evoked in response to muscle palpation, light touch stimulation, or manipulation of limb position around the workspace in quiescent frogs or by activating spinal behaviors with strong skin stimulation. A, Quiescent responses. We obtained a range of neuron responses at 400–500 μm depth with muscle palpation (“muscle palp”), light touch (“light touch”), and passive limb movements (“limb movement”) of quiescent spinal frogs, but no neurons responded clearly and repeatably in this way at 700–800 or 1000–1200 μm. B, Informational comparisons between neurons recorded at different depths. Mutual information was not higher with components or EMG in neurons at 1000–1200 μm or neurons at 400–500 μm with the exception of light touch responses (plotted in C–E). At 700–800 μm, there was a strong preference for components (as elaborated further in Fig. 4 and here in C–E). DH, Dorsal horn; IZ, intermediate zone; VH, ventral horn. C, Neurons with responses to muscle palpation showed no clearly stronger relationship in their fractional mutual informational with either EMGs or ICs, with four neurons possessing better associations to EMG (points above line) and four neurons possessing better associations to the ICs (points below the equality line). D, Neurons responding to light touch stimuli had mutual information fractions uniformly better associated with ICs. All points were below the equality line (binomial distribution probability <0.001). E, Cells responding to manipulation of the limb showed no clearly better fractional mutual information association to either EMGs or ICs (Z score of 0.44 for frequency counts, p = 0.67, n = 20).
Figure 4.
Figure 4.
Comparisons of relations of 131 interneurons recorded in intermediate zone at 700–800 μm (diagram at bottom) with EMG and with components. Units had no clear sensory fields in quiescent frogs as shown in Figure 3. The neurons were related to components and to the associated individual EMGs comprising the component projection, and the relationships were calculated and compared. We tested both linear regression and mutual information measures. In each plot, only those neurons in which jackknifing of mutual information measures showed could be compared and differed with 95% confidence are shown (131 in all C–F). A, Comparison of Shannon self-information or entropy in the three sets of muscle-derived time series with which neuron MI is compared in C–F. These are not significantly different from one another (SDs shown). The neural entropy is also shown and on average was significantly larger than the EMG derived series. B, Lag 0 linear regression (and lagged regressions not shown) of neuron firing to components and EMG using a 15 ms window showed no significant differences between variance captured by EMG or components. C, Fraction of possible MI of neurons with components were routinely higher for the best-matched component than with the highest information EMG found in the projection of the component (111 of 131, p < 0.0001). D, The absolute MI of the neurons with best-matched component was greater than with the highest information EMG found in the projection of the component (102 of 131, p < 0.0001). E, The fraction of possible MI of neurons with the best-matched component for the neuron was routinely higher with components than with the lowest information EMG found in the projection of the component (131 of 131, p < 0.0001). F, Absolute MI of neurons with the best-matched component was greater than with the lowest information EMG found in the projection of the component (124 of 131, p < 0.0001). Mutual information measures for intermediate zone thus support spinal representation of pattern and motor drive in intermediate zone neurons in terms of premotor drive components, not individual muscles EMGs. DH, Dorsal horn; IZ, intermediate zone; VH, ventral horn.
Figure 5.
Figure 5.
Interneurons segregated into groups based on mutual information matching to components in intermediate zone (IZ; 800 μm depth). Components were related to primitives identified previously by both statistical and physiological means (Hart and Giszter, 2004; Kargo and Giszter, 2008; Kargo et al., 2009). Interneuronal firing was phasically related to primitive/component group activity. A, Rasters of two neurons in two frogs with strong relationships to the hip extensor component/primitive are shown. The rates of the neurons clearly relate to component amplitudes and the burst or pulse-like behavior. These neurons were also used in STA analysis. B, Two neurons best related to the period preceding (above) or the onset (below) of the hip extensor component burst. C, Neurons best related to one of the most frequent components (the hip flexor component) were aligned with component burst/pulse onsets (defined as 3 SDs above baseline noise). The neural activity was then binned to form a population peri-onset time histogram (PSTH), and the component burst/pulses that were aligned in this way were also averaged. It can be seen that the aggregate activity of the informationally related neuron population can be related especially well to the motor burst onsets and duration. However, strong preceding and following population activity can also be seen as might be expected based on the mutual information criteria for selection used (see Results). DH, Dorsal horn; IZ, intermediate zone; VH, ventral horn.
Figure 6.
Figure 6.
Spike-triggered averaging of EMG postspike facilitations of interneurons. We used both STA of raw EMG that minimizes likelihood of false positives and STA of rectified unfiltered EMG to examine facilitation or inhibition. All effects were facilitatory. A, Spike-triggered average of raw EMG for a putative motor unit from the ventral horn (1200 μm depth) to the gluteus muscle of a frog. Shown for comparison with interneuron spike-triggered averages. B, Spike-triggered averages of raw recorded EMG with an interneuron in which we recorded 505 spikes. The best tetrode channel waveform is displayed at the top. The average waveform, the significant peaks (asterisks), and the Chronux calculated confidence limits are shown. The unit had significant STA peaks for RA, SM, GL, and VE. Each postspike facilitation effect was excitatory and also significant in STA of rectified EMG. The most closely related component had significant projections to RA, SM, GL, and RI. The binary inner product correlation measure of the two would thus be 0.75. C, Comparison of the distribution of binary inner product correlations of 46 neurons with significant STA postspike facilitations matched to best components (rectified EMG distributions shown) are compared with those predicted by distributions of (1) random connections or (2) random connections drawn from the distribution of observed components from mutual information matches in our data (calculated with Monte Carlo simulations with 1000 iterations). Cumulative distributions are shown. These were used in Kolmogorov–Smirnov tests. The high number and distributions of good matches of components to STA postspike facilitations were unlikely to have arisen by chance sampling of the random distributions (p < 0.01). D, Log–log correlation of the STA postspike facilitation peak strengths with the component weight strengths. Open circles represent the corresponding STA peaks measured at the same latency. Correlation of STA peak strengths with component weights were 0.65. The significance of the relation can be calculated assuming that each connection and its strength is an independent observation [r = 0.65, t statistic of 12.4, n = 214 (taken from STA peaks), p < 1e−10]. Alternatively, only individual neurons are considered independent observations (significant STA peaks: t statistic of 5.5, DOF of 45, p = 1e−6; all matched STA peaks: t statistic of 2.04, n = 34, p < 0.05). The regression presented here provides a good visual representation by collapsing all parameters onto a plane. More correct multiple input/output regressions were also performed, and outcomes are given in Results. Together, these results support a specific interneuronal distribution system providing drive for primitives.
See this image and copyright information in PMC

Comment in

  • On the origins of modularity in motor control.
    Delis I, Chiovetto E, Berret B. Delis I, et al. J Neurosci. 2010 Jun 2;30(22):7451-2. doi: 10.1523/JNEUROSCI.1562-10.2010. J Neurosci. 2010. PMID: 20519519 Free PMC article. No abstract available.

References

    1. Abbott LF. Where are the switches on this thing? In: van Hemmen JL, Sejnowski TJ, editors. Problems in Systems Neuroscience. Ed 23. Oxford: Oxford UP; 2006. pp. 423–431.
    1. Abeles M. Quantification, smoothing and confidence limits for single-units histograms. J Neurosci Methods. 1982;5:317–325. - PubMed
    1. Alaburda A, Russo R, MacAulay N, Hounsgaard J. Periodic high-conductance states in spinal neurons during scratch-like network activity in adult turtles. J Neurosci. 2005;25:6316–6321. - PMC - PubMed
    1. Bateson PJ. The active role of behaviour in evolution. Biol Philos. 2004;19:283–298.
    1. Bell AJ, Sejnowski TJ. An information-maximization approach to blind separation and blind deconvolution. Neural Comput. 1995;7:1129–1159. - PubMed

Publication types

MeSH terms

LinkOut - more resources