Operant conditioning of cortical cell and muscle response patterns
Eaton, Ryan W.
MetadataShow full item record
<bold>Part I:</bold> In primates, corticomotoneuronal (CM) cells have sufficiently strong synaptic linkages to motoneurons to mediate post-spike facilitation in spike-triggered averages of muscle activity. We investigated the degree to which activity of CM cells and their target muscles could be independently controlled by operantly conditioning their relative activation levels. In two Macaca nemestrina monkeys, single cortical neurons were recorded with moveable microwires chronically implanted in the caudal bank of the pre-central gyrus. Rectified EMG of 12 distal forelimb muscles was recorded with sub-cutaneously implanted pairs of wires in each muscle in one monkey from which 35 unique cell-muscle pairs were operantly conditioned. In the other animal, surface electrodes were placed over wrist flexor and extensor muscles (9 unique cell-muscle pairs). Spike-triggered averages of rectified EMG were compiled while the monkeys performed a force target-tracking task about the wrist. Twenty-four CM cells, with post-spike effects at latencies between 6 and 16 ms in one or more forelimb muscles, were selected for activity dissociation conditioning. Monkeys performed an operant conditioning task through which relative activation of the CM cell and a target muscle could be explored (44 unique cell-muscle pairs). Cell and muscle activity controlled the position of a cursor on a screen, with cursor position determined by concurrent cell spike rates (C) and EMG activity of a target muscle (M); one activity assigned the horizontal direction the other, vertical. Monkeys received fruit sauce rewards for holding the cursor in target positions requiring at least four combinations of increased (+) and suppressed (-) activation relative to levels observed during force generation, namely C+M+; C+M-; C-M+ and C-M-. The monkeys learned to reciprocally activate cells and target muscles exhibiting post-spike facilitation (42 out of 44 cell-muscle pairs), in both directions. For muscles with post-spike suppression (4 out of 4 pairs), monkeys learned to co-activate these cell-muscle pairs. These results indicate that cortical cells with direct synaptic linkages to motoneurons can be flexibly activated relative to their target muscles. Further, CM cell-muscle activity dissociations can be rapid, robust, reversible and are subject to volitional control. Calculation of mutual information between CM cell and muscle activation patterns during reciprocal dissociation events, compared to force target-tracking, implicates involvement of other upstream sources. Activity independence between correlationally-linked components within a neural circuit favors strategies for brain computer interface (BCI) control in which individual neurons are each assigned an individual degree-of-freedom of device output. <bold>Part II:</bold> Operant conditioning of neural activity has typically been achieved under controlled behavioral conditions using food reinforcement. To reward cell activity during unconstrained behavior, we sought midbrain sites whose stimulation would support operant responding. Three nemestrina monkeys learned to perform a manual step-tracking task rewarded by fruit sauce. We found sites in nucleus accumbens and surrounding striatum whose stimulation could maintain task performance and verified that response rates increased monotonically with increasing pulse frequency and amplitude. We recorded activity of single neurons with moveable microwires chronically implanted in the precentral gyrus and documented neural modulation with a force-guided target-tracking task. We attempted to condition increased firing rates first with the monkey in the training booth and then during free behavior in the cage using the Neurochip—a head-fixed, autonomous recording and stimulating system. Spikes occurring above baseline rates triggered single or multiple electrical pulses to the reinforcement site (1 mA, 0.2 ms biphasic current pulses). This rate-contingent, unit-triggered stimulation was made available for periods of 1 to 3 minutes separated by 3 to 10 minute time-out periods without stimulation, regardless of cell activity. During in-booth sessions feedback was presented as vertical cursor movement and auditory clicks. During in-cage conditioning, barely audible clicks occurred during each spike-triggered stimulation event. In-booth conditioning produced increases in single neuron firing probability after transition to intracranial reinforcement in 48 of 58 cells. Reinforced cell activity could rise > 5 times that of non-reinforced activity, doubling in most sessions. Activity peaks typically occurred during the first 10 seconds of each time-in period and, for many cells, activity remained elevated above baseline for the full period. In-cage conditioning produced significant increases in post-transition activity in 21 out of 33 sessions. In-cage effects peaked later and lasted longer than in-booth effects but were often comparatively smaller, between 13 and 18 percent above non-reinforced activity. The difference in responding in the two conditioning environments could be due to the dynamic range of candidate cell firing rate, robustness of the reinforcement site and differing levels of attention and competing behaviors in the booth and cage. Controls indicate that stimulation of the reinforcement site did not directly evoke increased cell activity. In several sessions, neighboring, synaptically-linked motor cortex neurons were recorded simultaneously with the trigger cell, revealing network involvement in eliciting conditioned rate changes in the stimulation-triggering neuron.