Chronic pain is a highly debilitating condition, which afflicts almost 30% of the population worldwide. The burden for the healthcare systems as well as for the suffering patients is high. Therefore, new insights into the processes of pain chronification and the formation of a pain memory are urgently needed.

By means of fMRI, we established a first mouse model to study the initial processes of pain chronification, and hereby assessed the changes of brain functional connectivity during this early induction phase.

All measurements on 12-week old male C57Bl/6N mice (controls: n=8; chronification group: n = 22) were conducted on a 4.7T Bruker Biospec (200mT gradient system, quadrature mouse head coil) using BOLD fMRI (GE EPI sequence: TEef 25.035ms; TR 2000ms NEX 2; FOV 15x15mm; matrix 64x64; slice thickness 0.5mm; 22 slices axial, 750 repetitions).

To initiate potential chronification processes, the right hind paw of the animals was stimulated during the fMRI-session with three sets of consecutive - two non-noxious (40°C and 45°C) and two noxious (50°C and 55°C) - thermal stimuli, (controls only 45°C, repeated 12 times) lasting 20 seconds.

The experiment lasted 14 days, consisting of fMRI measurements every other day, resulting in seven measurements per animal. To exclude peripheral sensitization effects, the Hargreaves Plantar Test was performed the day after each fMRI measurement. For analysis, we first used the general linear model approach to detect voxels that coupled significantly (FDR corrected) to the stimulation protocol. Next, an in-house 3D brain atlas was used to identify the significantly activated voxels of 196 brain structures. BOLD activation volume and peak amplitude were determined and the mean time course of the BOLD-Signal was extracted for each structure.

Graph-theoretical analyses based on cross-correlations of these time courses (after removal of global signal) were used to construct stimulus specific networks. To ensure ideal topological comparisons, the resulting networks were thresholded to contain the same number of connections. Significant changes in connectivity^1-3 between all brain structures were identified using Student's t-test.

Computation of hub properties⁴ and Blondel-Communities⁵ were used to identify changes of interaction and the network parameters path length, cluster-coefficient and small-world-index were used to characterize changes in flow of information over the time course of the experiment.

First, no significant changes in paw withdrawal latency and therefore no peripheral sensitization could be shown for either the controls or the chronification group.

Activated volume and peak amplitude showed a significant transient decrease around the fourth measurement in most brain structures. Hence we assumed that the repetitive noxious stimulation does not just affect a subset of brain structures.

The networks of both groups were shown to be of small-world-topology. Differences in the small-world-index resulted from changes in the average path length, where the controls and temperatures 40, 50 and 55°C of the chronification group were relatively static, in contrast to the 45°C of the chronification group, which showed great undulations over time. This may reflect changes in temperature discrimination between non-noxious and noxious stimuli. The overall higher cluster-coefficient of the chronification group indicated an enhanced and more efficient information processing.

The evaluation of changes in interactions between the structures (Blondel Communities) revealed highly relevant alterations again of many structures like thalamus and hippocampus as well as sensory and motor cortices.

Most strikingly, the emotional-affective component of pain processing was changed: a profound and lasting dissociation of cingulate, limbic and frontal association cortex as well as the septum from the sensory and motor cortices, and an association towards structures of the reward system (Nucleus accumbens and the olfactory tubercle) could be shown for the chronification group, but not for the controls (Fig. 1).

Our results therefore demonstrate that pain chronification might indeed start in the brain, even before peripheral or behavioral symptoms can be detected.