4985

Atlas-based and voxel-based symptom mapping to explore anatomical substrates of functional recovery after stroke in the mouse
Stefan Paul Koch1,2, Felix Knab1, Ulrike Grittner1, René Bernard1, Susanne Mueller1,2, Marco Foddis1, Janet Lips1, Philipp Euskirchen1, Norman Zerbe3, Peter Hufnagl3, André Rex1, Ulrich Dirnagl1, Tracy Deanne Farr4, Christoph Harms1, and Philipp Boehm-Sturm1,2

1Department of Experimental Neurology and Center for Stroke Research Berlin, Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health, Berlin, Germany, 2Cluster of Excellence NeuroCure and Charité Core Facility 7T Experimental MRIs, Charité - Universitätsmedizin Berlin, Berlin, Germany, 3Institute of Pathology, Charité – Universitätsmedizin Berlin, Berlin, Germany, 4School of Life Sciences, University of Nottingham, Nottingham, United Kingdom

Synopsis

Post-stroke functional recovery remains a poorly understood process. We aimed to identify anatomical regions that statistically explain a deficit in the staircase test after stroke in the mouse using two animal models (MCAO, photothrombosis). MATLAB toolboxes are presented for co-registration of MRI and histology data to the Allen brain atlas and implementation of atlas region-wise or voxel-wise mapping of behavioral deficits. Automatic techniques are shown for validation of MRI results on stained tissue sections.

Introduction

Post-stroke functional recovery remains a poorly understood process. Here, we introduce techniques adapted from voxel-based lesion symptom mapping1 in order to identify anatomical regions whose damage statistically explain a functional deficit and thus would be of highest interest for a targeted therapeutic intervention. To this end, we present custom MATLAB toolboxes for Allen atlas registration of MRI and histology data allowing brain region-based and voxel-based correlation of behavioral deficits to tissue damage after stroke in the mouse.

Methods

C57/Bl6 mice (n=60) were trained in the staircase test2 and randomized to undergo either 45 min middle cerebral artery occlusion (MCAO, final n=17), sham surgery (n=17) or photothrombosis (PT, n=9). n=13 (MCAO) died before endpoint, n=1 (MCAO) did not learn the task, and n=1/1 (MCAO/PT) did not have a lesion. Experiments and analysis were performed blinded to the condition of animals. 24 h after surgery, animals underwent T2-weighted (T2w) MRI (RARE, 32 contiguous 0.5 mm thick slices, FOV=(25.6 mm)2, matrix=256x196 zerofilled to 2562, TR/TEeff=4.2 s/36 ms, RARE factor 4,4 averages, TA=6:43 min) at 7 T (Bruker BioSpin, Ettlingen, Germany) using a 20 mm diameter Tx/Rx volume resonator coil (RAPID, Rimpar, Germany). A lesion mask was drawn by an experienced researcher and images+masks were warped to the Allen-Brain-Atlas using our publicly available MATLAB toolbox ANTX3 (Fig. 1). We correlated the degree of initial deficit (mean performance 2-6 d post surgery) i) region-wise with percent lesion volume in atlas regions or ii) voxel-wise with image intensity in the warped T2w image. 21 d post surgery, animals were sacrificed for histological validation. We performed DAB-enhanced NeuN staining on 20 µm thick slices (400 µm slice distance), center of mass of NeuN nuclei were automatically identified using Cell Profiler (http://cellprofiler.org/), and corresponding slices of the Allen brain atlas were automatically determined and registered to each histological slice using a custom ELASTIX-based (http://elastix.isi.uu.nl/) MATLAB toolbox (Fig. 2).

Results

MCAO animals had the largest difference in performance in the staircase test compared to shams (mean difference: 41.2%, (30.7-51.7%), p<0.001) whereas the deficit was smaller for PT animals (mean difference: 19.9% (6.8-33.0%), p=0.0049). We found a trivial correlation of initial deficit and total stroke volume on T2w images at 24 h (R=0.73). However, region-based and voxel-based symptom mapping revealed similarly high correlation in very specific areas involved in somatosensory and fear processing. Damage in more caudal structures (e.g. thalamus and midbrain) also correlated well with the deficit (Fig. 4).

Discussion

We developed a set of fully automated tools for data-driven analysis of anatomical substrates of functional recovery after stroke in the mouse. From the fact that lesions in MCAO animals are mainly subcortical and these animals performed worse than PT animals (primarily cortical damage), we conclude that the subcortex is important to perform well in a behavioral test of somatosensory and motor function. Region-based and voxel-based symptom mapping was used to explore and quantify the importance of anatomical lesion location, a comprehensive validation of these findings through correlation with neuronal cell densities is work in progress but the necessary tools have been established. One limitation of our study is the exclusion of animals that underwent MRI but died before the final endpoint, a systematic strategy to statistically deal with this cohort is work in progress. Another limitation is that correlation of a behavioral deficit with damage, e.g. in more caudal brain regions, may simply be due to effects related to overall lesion volume and not to loss of the specific function of the region. Therefore, in order to gain a more mechanistic insight, we aim to target most prominent candidates with pharmacogenetic excitation/inhibition in future studies.

Conclusion

Atlas-based and voxel-based symptom mapping in the mouse is a powerful tool to explore mechanisms of functional recovery after stroke.

Acknowledgements

Work was supported by the Federal Ministry of Education and Research (BMBF) (grant number 01EO0801, Center for Stroke Research Berlin) and Deutsche Forschungsgemeinschaft (DFG) (Excellence Cluster NeuroCure).

References

1 Bates E, Wilson SM, Saygin AP, Dick F, Sereno MI, Knight RT et al. Voxel-based lesion-symptom mapping. Nat Neurosci 2003; 6: 448–50.

2 Emmrich J V, Neher JJ, Boehm-Sturm P, Endres M, Dirnagl U, Harms C. Stage 1 Registered Report: Effect of deficient phagocytosis on neuronal survival and neurological outcome after temporary middle cerebral artery occlusion (tMCAo) [version 1; referees: 2 approved]. F1000Research 2017; 6. doi:10.12688/f1000research.12537.1.

3 Koch S, Mueller S, Foddis M, Bienert T, von Elverfeldt D, Knab F et al. Atlas registration for edema-corrected MRI lesion volume in mouse stroke models. J Cereb Blood Flow Metab 2017; : 0271678X1772663. https://github.com/philippboehmsturm/antx

Figures

Fig.1: Allen atlas registration of T2w images using the MATLAB toolbox ANTX. T2w images acquired 24 h after stroke in the mouse are segmented into tissue compartments using SPMMouse and a synthetic image is generated. This image is warped to the atlas in ELASTIX. Nonlinear terms allow to compensate for potential swelling due to edema.

Fig. 2: Pipeline to register the Allen atlas to histological tissue slices.

Fig. 3: Strategy for automatic neuronal density measurement on histological slices. Representative results from rostral and caudal DAB-enhanced NeuN stained tissue slices from a mouse with MCAO on the right side. The best matching slice of the Allen template (2nd column) is nonlinearly transformed to the histological slice (1st column). The associated transformation is then applied to yield the corresponding atlas slice in the histological space (3rd column, regions in pseudo gray shades). Center of mass of NeuN positive nuclei (overlaid in pseudocolor red) can be detected automatically (columns 4 and 5), which allows cell counting within each atlas region.

Fig. 4: Atlas-based and voxel-based symptom mapping in MCAO animals. [A] Correlation of poststroke performance with lesion volume within each atlas region (ordered according to correlation coefficent, yellow: significant correlations after Bonferoni correction). [B] Voxels with significant correlation of T2w intensity and performance (permutation test, FDR corrected). For this approach, T2w were z-scored based on mean and std of the intact hemisphere and z-scores were correlated voxel-wise with poststroke performance. Numbers above slices show distance to Bregma in mm. [C] Correlation of T2w z-score and performance from two surviving clusters (each dot represents a mouse). [D] Anatomical labels of peaks of the ten statistically strongest surviving clusters.

Proc. Intl. Soc. Mag. Reson. Med. 26 (2018)
4985