The experimental procedures were approved in advance by the Malmö/Lund Animal Research Ethics Committee (permit number and approval-ID: M32-09 and M05-12). Initial surgery was performed under propofol anesthesia, and all efforts were made to minimize suffering. Our EEG recordings were characterized by a background of periodic 1–4 Hz oscillatory activity, periodically interrupted by large-amplitude 7–14 Hz spindle oscillations lasting for 0.5 s or more. These forms of EEG activities are normally associated with deep stages of sleep. The pattern of EEG activity and the blood pressure remained stable and did not change with noxious stimulation throughout experiments.
Adult cats (N = 14) were prepared as previously described. Briefly, following an initial anesthesia with propofol (Diprivan® Zeneca Ltd., Macclesfield Cheshire, UK), the animals were decerebrated at the intercollicular level and the anesthesia was discontinued. The animals were artificially ventilated and the end-expiratory CO2, blood pressure, and rectal temperature were continuously monitored and maintained within physiological limits. Mounting in a stereotaxic frame, drainage of cerebrospinal fluid, pneumothorax, and clamping the spinal processes of a few cervical and lumbar vertebral bodies served to increase the mechanical stability of the preparation. The dorsal part of the pars intermedia of the left cerebellum was exposed to allow microelectrode access to the AIN. An additional exposure was made of the brainstem/spinal cord junction between the base of the skull and the first cervical vertebra. All exposed areas were covered in paraffin oil to prevent tissue drying.
Recordings and Stimulation
Patch clamp pipettes or metal microelectrodes (tungsten-in-glass microelectrodes, exposed tip 10–20 μm) were advanced to target the AIN as previously described [3, 4]. All neurons included in this study were putative glutamatergic projections neurons, based on the preponderance of short (<25 ms) interspike intervals and intermediate spike-widths . We recorded neurons from both forelimb and hindlimb regions of this nucleus, as identified using the location of the cf receptive field of the afferent Purkinje cells (Fig. 1). This location can be mapped out using electrical stimulation of the skin (0.1 ms pulses of 1.0 mA applied through percutaneous needle electrodes )—if the cfs of the locally afferent Purkinje cells are activated by the stimulation, characteristic local field potentials [6, 8] and postinhibitory rebound responses of the DCN neurons can be recorded . In this way, the location of the cf receptive field can be identified.
In order to stimulate the spinocerebellar and spinoreticulocerebellar tracts, which provide direct mf synaptic inputs to the interpositus nuclei, we placed a tungsten-in-glass microelectrode (exposed tip 50–150 μm) for stimulation laterally at the border between the spinal cord and brainstem. Using this stimulation microelectrode, mf field potentials recorded inside the AIN were routinely evoked at threshold intensities of <20 μA (single stimulus pulse of 0.1 ms duration), suggesting an effective recruitment of directly and synaptically activated (via the lateral reticular nucleus) mf synapses. In addition, we used electrical skin stimulation (pair of percutaneous needle electrodes with 5–10 mm spacing, stimulated at 1 mA shocks with 0.1 ms duration) to recruit another putative pool of spinocerebellar mfs. Cutaneous input is known to activate parts of the spinocerebellar neuron population, and since the other pathway mediating cutaneously activated mf input, the main cuneate nucleus does not terminate in the AIN ; potent excitatory responses evoked from the skin  are likely due to spinocerebellar mfs which should be at least partly non-overlapping with the population of mfs activated from the brain stem. The skin stimulation used was verified to not activate the afferent cfs and evoked a monophasic excitatory response .
In order to activate cfs, a second stimulation electrode was placed in the inferior olive, where low-threshold cf responses (evoked at <10 μA) could be evoked in the pars intermedia of the cerebellar cortex and in the AIN [4, 15].
Using the SCT or skin stimulation as test stimulation, we applied three different stimulation protocols to investigate whether plasticity in the input to the AIN neurons could be recorded. In most cases, more than one protocol was applied in the same experiment. When this was the case, the recording electrode was moved to a different location in the AIN, where neurons had substantially different location of their cf receptive fields (i.e., hindlimb versus forelimb, or proximal versus distal parts of the limb). We also moved the SCT stimulation electrode to recruit a different set of mfs, and also the skin stimulation used to evoke mf inputs was moved to a distinctly different location. The three different protocols that we used were as follows:
The combined SCT and skin burst stimulation protocol. The SCT electrode was stimulated with 15 pulses at 200 Hz, and the skin was stimulated 10 times at 333 Hz. With this configuration, the two inputs evoked largely overlapping time windows of excitation. The SCT stimulation intensity was typically 30–70 μA, in a couple of cases 100 μA.
The skin burst and simultaneous, single inferior olive (IO) stimulation protocol. The IO was stimulated once, and a skin burst stimulation of 50 pulses at 333 Hz was started 10 ms in advance in order for the first mf input to arrive at the same time as the cf input (the mf input evoked from the periphery needs at in the order of 10 ms to reach the cerebellar nuclei ).
The skin burst and delayed single IO stimulation protocol. A skin burst stimulation of 50 pulses at 333 Hz and at the time point of the last stimulation pulse, a single-pulse IO stimulation was applied.
For all three protocols, the bursts were repeated at 0.33 Hz for 10 min, i.e., for a total of 200 repetitions.
We quantified the responses obtained from a single-pulse stimulation, either to the SCT or to the skin, before and after a burst stimulation protocol. For the protocols involving skin bursts and the simultaneous or delayed IO stimulation, respectively, the responses were quantified using peristimulus histograms of raw spike time data (5 ms bin width). For the display and analysis of the combined skin burst and SCT stimulation protocol, we used a kernel density estimation (KDE) plot, i.e., each spike was replaced by a Gaussian distribution with standard deviation of 0.5 ms. The averaged sum of all Gaussian distributions transforms a discrete spiking pattern into a continuous function describing the spiking probability on a continuous time scale. The standard deviation of the kernels was set so that the total spiking probability function was smooth across neurons. This was done as the responses to the SCT stimulation were brief, which reduced the total number of spikes and made the responses more sensitive to chance distributions of single spikes. The KDE helped in reducing this problem. See Hoebeek et al.  for a more comprehensive discussion on KDE.
In all cases, the response was quantified as the mean firing frequency during the time window of the response, with the firing frequencies being obtained either from the KDE plots or the peristimulus histograms. To smooth the signal used in the analysis, the peristimulus histograms were filtered with a moving average of width 15 ms. The response onset was counted from the first occurrence of at least two consecutive bins with an activity that exceeded the baseline activity by at least two standard deviations. The end of the response was defined as the bin where the activity decreased to the threshold. For each cell, the time window for the response was initially calculated individually for every peristimulus histogram (i.e., control and all the post-protocol time points). Then the median start and end points of the responses were used to define the response time window for the cell, in which the response was quantified. The response was quantified as the mean net activity during the defined time window. For responses evoked by SCT, the responses obtained were typically evoked between 1.5 to 4.0 ms after the onset of the stimulation. The KDE in itself did not allow a rigorous setting of the time limits, but histograms of the raw data provided a support for the chosen time limits in a similar fashion as above. For responses evoked by the skin, the quantified data was typically evoked within a response latency time window of 10–30 ms after the onset of the skin stimulation.
Subsequently, the relative response for each set of single-pulse stimulations was compared to the relative response before onset of the burst protocol and the change in response from each cell was analyzed in separate consecutive time spans of 10 min (time points). The null hypothesis that there was no net change in the response was tested using Wilcoxon signed-rank test. The signed-rank test was computed for each time point by comparing the total number of cells with the number of cells with a positive response. The probability for the outcome is calculated, assuming there is a 50% probability for each cell having a positive change in response.