Mechanisms of spinal learning related to a simple motor task in a mammalian system
Even in the absence of brain input, the spinal cord has the ability to learn dynamically complex motor tasks such as stepping and standing with repetitive training. (Hodgson, Roy et al. 1994, Harkema, Hurley et al. 1997, Edgerton, Tillakaratne et al. 2004, Fong, Cai et al. 2005). Although there is substantial behavioral evidence of the spinal plasticity following injury, molecular mechanisms of plasticity are largely unknown. To examine the mechanisms of plasticity or learning in spinal cord of locomotor trained animals proved difficult due to the complexity the circuitry involved.
In an attempt to understand the synaptic mechanisms involved in the performance of motor tasks in mammals, we adopted a simple model of learning in the rat paw withdrawal learning (PaWL) to the hindlimbs of mice (Jindrich, Joseph et al. 2009). Learning to avoid a shock by limb withdrawal was originally reported by Horridge (Horridge 1962, Chopin and Buerger 1976) in insects, later adapted for spinal rats (Chopin and Buerger 1976), (Grau, Crown et al. 2006). In PaWL paradigm, spinal mice dorsiflex the hindpaw in response to position-dependent mild electric shocks to the tibialis anterior muscle and learn to hold the paw to minimize the net exposure to shocks. Although important molecular substrates of instrumental learning in the spinal rats have been validated by pharmacological manipulations (Joynes and Grau 2004, Gomez-Pinilla, Huie et al. 2007, Ferguson, Bolding et al. 2008), neither physiological characterization nor cellular mapping of learning-associated spinal neurons has been examined. This simple in vivo learning model provides an opportunity to identify molecules and neural pathways that mediate spinal learning in mammals, utilizing genetic and molecular tools of key learning-associated signaling molecules in the brain (Gomez-Pinilla, Huie et al. 2007, Joseph, Tillakaratne et al. 2012).
Through EMG analysis in tibialis anterior hindlimb muscle, the primary ankle flexor involved in the task, we showed that the proprioceptive learning during PaWL is mediated through specific sensory-motor spinal network (Chapter 3). Specifically, 1) We measured muscle activity using chronically and acutely implanted EMG electrodes during PaWL testing in several hind limb muscles of Master and Yoked groups, and showed that the sustained hindpaw dorsiflexion during learning is mediated through the activation of the TA motor pool in learned (Master) but not in the failed (Yoked) group (Chapter 3); 2) We compared EMG activity in the TA muscle under passive paw withdrawal learning condition, where electrical shock is removed and only the proprioceptive foot position is applied to complete spinal transected mice, we showed under these conditions, mice cannot learn to maintain the dorsiflexed hindpaw position (Chapter 3); 3) By selectively blocking the TA muscle or the cutaneous afferents, only blocking of muscle afferents led to failure of PaWL (Chapter 3).
Using the activity dependent marker c-fos in combination with the transynaptic pseudorabies virus (PRV) or the learning marker, CaMKII, we identified activated neurons in specific laminae in mice that successfully learned the PaWL task (Chapter 4). We found that the number of activated neurons (Fos+) were higher in the ipsilateral spinal cord of both groups of mice that received contingent (Master group) and non-contingent shocks (Yoked group) to the TA, but the number of activated neurons was inversely correlated with the degree of learning, only in the mice receiving the contingent shocks. CaMKII immunoreactivity in the ipsilateral spinal cord was highest in L4 segment in the lumbosacral spinal cord in Master mice. These mice also had more Fos+ CaMKII+ neurons in laminae IV-VI compared to the yoked mice. Pseudorabies virus (PRV) that labels functionally connected circuitry, when injected into tibialis anterior, the primary muscle involved in PaWL, showed that majority of activated neurons in L4 and activated CaMKII+ neurons in laminae IV-VI were also PRV+.
We then disrupted PaWL by interfering with cAMP response element binding protein (CREB) function in CaMKII expressing neurons using the conditional CREBIR transgenic mice, that express CREB repressor protein (CREBIR) with tamoxifen (Chapter 4). We found that disrupting CREB function in these transgenic mice while not affected the early learning, interfered PaWL when retested 24 hrs. later. Furthermore, the failure to learn the PaWL task was correlated with decreased levels of pCaMKII, and pCREB proteins in only in the CREBIR mice when induced by tamoxifen. Our data showed that compared to wildtype mice, the mutant mice showed reduced of number of activated spinal neurons in laminae IV-VI of the spinal cord. For example, wild type spinal mice contained higher numbers of both Fos+ and CaMKII+ in lateral lamina II and laminae IV-VI at L4-L5 compared to CREBIR mice. The location of active spinal neurons in mice with normal CREB function, but not observed in transgenic mice with targeted CREB disruption, suggests that that these neurons are important in Paw withdrawal learning. These results indicate that specific neurons in the L4 segment are part of the circuitry associated with hindlimb flexion during paw withdrawal learning. We showed that fewer neurons are activated in paw withdrawal learning than stepping, allowing the mapping of circuitry in this simpler and feasible task. Identification of the spinal networks for this simple motor task will be a step toward understanding cellular and synaptic mechanisms of more complex learning such as standing and stepping after a complete spinal cord transection.