Neuropsychiatric systemic lupus erythematosus (NPSLE) is associated with poor outcome¹, but the understanding thereof is limited. Considering NPSLE is a systemic autoimmune disease with widespread involvement², it is of significance to study it from a global perspective. Therefore the aim of this study is to investigate alteration in the structural brain network of NPSLE and systemic lupus erythematosus (SLE) patients. The hubs of the brain network were also analyzed to explore the influence of NPSLE and SLE in individual regions.

Participants 20 female NPSLE patients (46 ± 12 years old) and 20 female SLE patients (44 ± 13) and 20 female healthy subjects (42 ± 14) without any neurological or psychological conditions or any physical disabilities were recruited.

Acquisition Diffusion-weighted images (DWIs) were acquired using single-shot EPI sequence in 15 gradient directions with b = 1000 s/mm² using a 3T scanner (Achieva TX scanner, Philips Healthcare). 2 for DWIs.

Post-processing Anatomical images were segmented into 90 brain regions. Whole-brain tractography was obtained using Diffusion Toolkit (trackvis.org/dtk/). Structural connections/connectivity were established between two regions when fiber tracts traversed both regions.

Network Analysis The network topological properties, including clustering coefficient, characteristic path length, normalized clustering coefficient, normalized characteristic path length, global and local efficiency were computed using the Brain Connectivity Toolbox³. The hubs of the network were identified from the normal subjects by using the betweenness centrality and nodal degree.

Statistical Analysis For between-group difference in the small-world properties, network efficiency and nodal characteristics between SLE patients, NPSLE patients versus healthy control subjects, repeated measures ANOVA with age, gender, and education as covariates followed by post-hoc independent samples t-tests were performed. p < 0.05 was considered as statistical significance.

The structural connectivity matrices of the 3 cohorts are shown in Fig. 1.

Global Network Consistent with previous studies^4,5, the structural network of our normal controls demonstrated small-world topology with greater clustering coefficient (γ»1) and equivalent characteristic path length (λ≈1) in comparison with the matched random network. Interestingly, both NPSLE and SLE patients also showed an overall intact small-world network when compared to random network. The brain network of NPSLE patients has lower clustering coefficient (p = 0.031), longer characteristic path length (p < 0.001), lower global (p = 0.010) and local efficiencies (p = 0.035) as compared to healthy subjects. These results suggested that there was alteration in both the long-range and short-range brain connections. For SLE patients, longer characteristic path length (p = 0.012) as compared to healthy subjects was observed.

Nodal Characteristics 9 brain regions were identified from healthy subjects as hubs: precuneus, insular, putamen, hippocampus, caudate, superior frontal cortex, thalamus, middle cingulum and globus pallidum. When comparing NPSLE patients with healthy subjects, (a) the nodal degree of middle cingulate cortex (p = 0.005) was decreased, (b) the local clustering coefficient of superior frontal cortex (p = 0.003), hippocampus (p = 0.027), caudate (p = 0.018) and putamen (p = 0.010) were decreased, (c) and the nodal efficiency of superior frontal cortex (p = 0.048) and caudate (p = 0.004) were decreased. For SLE patients, decreased nodal degree was observed in the middle cingulate cortex (p = 0.004) and caudate (p = 0.011).

Interestingly, there is a distinct difference between the brain networks of NPSLE and SLE patients despite diffuse changes in fractional anisotropy across whole brain, further supporting the notion that connectivity analysis is a reliable method to study systemic disease.

Anti-ribosomal P protein was found in SLE patients⁶, and animal experiment showed that this protein targeted neurons in cingulate and hippocampus, causing depressive-like behavior⁷. Our results provided further evidence that middle cingulate cortex and hippocampus were particularly vulnerable in patients with NPSLE. Atrophy in the caudate and putamen were observed in prior studies^8,9, again consistent with our finding that these regions are more susceptible to damage.

It is noteworthy that significant changes in the nodal characteristics were observed amid the absence of atrophy (data not shown), likely indicating the improved sensitivity to NPSLE using connectivity analysis. Given normal cognition is controlled by the caudate and hippocampus^10,11, a diminished role as hub for these region in the brain might explain symptoms related to cognitive impairment in NPSLE.

In summary, we have successfully demonstrated altered network topological properties and efficiency of NPSLE patients, and found diminished role of several brain regions as hubs. These results suggested that disruption of topology of structural brain network might be the hallmark of NPSLE that is characterized by multiple neuropsychiatric manifestations, and that brain network analysis is a reliable method to study systemic disease.