Mouse models of immune dysfunction: their neuroanatomical differences reflect their anxiety-behavioural phenotype

Abstract

Extensive evidence supports the role of the immune system in modulating brain function and behaviour. However, past studies have revealed striking heterogeneity in behavioural phenotypes produced from immune system dysfunction. Using magnetic resonance imaging, we studied the neuroanatomical differences among 11 distinct genetically modified mouse lines (n = 371), each deficient in a different element of the immune system. We found a significant and heterogeneous effect of immune dysfunction on the brains of both male and female mice. However, by imaging the whole brain and using Bayesian hierarchical modelling, we were able to identify patterns within the heterogeneous phenotype. Certain structures-such as the corpus callosum, midbrain, and thalamus-were more likely to be affected by immune dysfunction. A notable brain-behaviour relationship was identified with neuroanatomy endophenotypes across mouse models clustering according to anxiety-like behaviour phenotypes reported in literature, such as altered volume in brains regions associated with promoting fear response (e.g., the lateral septum and cerebellum). Interestingly, genes with preferential spatial expression in the most commonly affected regions are also associated with multiple sclerosis and other immune-mediated diseases. In total, our data suggest that the immune system modulates anxiety behaviour through well-established brain networks.

Fig. 1. Immune system mutations have a highly heterogeneous effect on mouse brain anatomy.

A Nearly all brain structures showed a significant effect of strain evaluated using F-statistics from ANOVA. B The directional effect in females of the various mutant strains relative to the wild-type strains is visualised using t-statistics and shows a heterogeneous neuroanatomical phenotype. Regions larger or smaller in mutants relative to wild-type are given maroon-pink and blue-turquoise colours, respectively, if effects are <5% FDR. Saturated colours represent effects <0.01% FDR.

Fig. 2. Brain regions showed variations in susceptibility to volume changes due to immune system mutations.

A The probability of a brain region having a large effect size (d) magnitude in mutant strains. Probability of effect-sizes (x-axis) for various structures—B midbrain, C corpus callosum, D dorsal striatum, E thalamus—across the different mutant strains (y-axis), for both female (F) and male (M) mice. Vertical dashed lines represent effect sizes of ±1, and probability within this interval is shaded grey. Integrating the area of the probability density outside this interval (pink) provides the probability that immune system mutations result in a large effect-size magnitude. These four structures had the highest probability of large effect size magnitudes.

Fig. 3. Mutant strains with similar anxiety-behavioural phenotypes have similar neuroanatomy endophenotypes.

For each pair of strains, the dissimilarity of endophenotypes was assessed using Hellinger distance and visualised using a network (thicker edges imply greater similarity). A Strains with increased anxiety behaviours (red nodes) had similar endophenotypes and clustered together in the network. A similar pattern was seen for unchanged anxiety behaviours (dark purple), but not decreased anxiety behaviour (blue). Cxcr2 anxiety phenotype is not known and assumed unchanged (light purple). The inset plot shows that pairs of strains (represented as dots) with similar anxiety phenotype had similar neuroanatomy endophenotypes (p < 0.01 from permutation testing). B The effect size (η²) for the anxiety grouping was computed for each structure. The green colour bar represents at least medium effect sizes and saturates for large effect sizes. C The culmen and lateral septum were chosen as representative examples to illustrate large effect-size for anxiety grouping. The 95% credible interval of predicted volume for each strain (grey bars) and anxiety phenotype (coloured bars) are shown.

Fig. 4. Brain regions susceptible to immune system mutations have a preferential spatial expression of genes involved with multiple sclerosis (MS).

The top 25 brain structures with the highest effect-size magnitudes are shown in C (first row) and constitute the region-of-interest (ROI). For all genes in the mouse genome, preferential spatial expression was assessed using a fold-change measure (i.e., gene expression signal in ROI relative to the whole brain). A Genes associated with MS (solid line) had significantly higher expression in the ROI compared to the genome (two-sided KS test D = 0.05, p < 10⁻³). B MS genes showed four different clusters of temporal expression within the ROI. The shaded region represents the 95% confidence interval. C A representative example from each cluster was chosen to visualise gene expression signals within the ROI over the course of neurodevelopment. Each example was closest to its respective cluster’s centroid.