0346
Defining subnetworks by rich-club architecture in 16p11.2 deletion syndrome reveals differential structural white-matter alterations1Fetal Neonatal Neuroimaging and Developmental Science Center, Division of Newborn Medicine, Boston Children's Hospital, Harvard Medical School, Boston, MA, United States
Synopsis
16p22.1 deletion syndrome has been implicated in disorders such as autism and is associated with developmental deficits, including delay in language acquisition. Widespread DTI white-matter alterations have been identified in patients but the structural organisation using network theory has yet to be investigated. Using rich-club nodes to stratify the connectome into rich-club, feeder and seeder subnetworks, we compute graph topological measures in children with 16p11.2 deletion. While rich-club regions were similar to those in Controls, differential alterations in connectivity and topology suggest a reorganisation of subnetworks in patients with the feeder subnetwork possibly compensating for deficits in the rich-club.
Introduction
Copy number variants of the chromosomal locus 16p11.2 have been associated with several neurodevelopmental disorders including autism and epilepsy1,2 with ~71% of patients going on to experience delays and deficits in speech and language acquisition3. Diffusion MRI studies on children with 16p11.2 deletion syndrome have found widespread, abnormal brain white-matter structure4 with selective involvement of language pathways5. Here we extend the analysis to study a network topological principle, the rich-club (RC), a core 'backbone' integral for effective brain function6,7. We investigate the network topology between 16p11.2 deletion patients and controls in terms of global measures, and within RC, feeder and seeder subnetworks as defined by partitioning brain regions according to their connection to core RC nodes8.Methods
Data were from Simons VIP Consortium9. Subjects Twenty-one 16p11.2 deletion patients (aged 10.91±2.09 years, 11 males, 11 right-handed), and 18 controls (12.58±1.99 years, 10 males, 11 right-handed). MRI Data acquired on a 3T Tim Trio Siemens included a T1-Weighted MPRAGE (TE/TR/TI=1.64/2530/1200ms; flip angle=7°; voxel size=1mm3; FOV=256mm) and a DWI sequence (30 gradient directions at b=1000s/mm2; one b=0s/mm2; TR/TE=10s/80ms; voxel size=2mm3; FOV=256mm). Processing (Fig1) T1-Weighted cortical surface was parcellated into 68 regions10 and mapped to b0 space11. DWI were corrected for minor motion and co-registered to the b0 volume12. Following whole brain, white-matter seeded interpolated streamline tractography13 (angular threshold=45°, no FA threshold) a 68x68 connectome, W, was computed and weighted by number of tracts connecting pairwise nodes. W was then normalized by the total number of streamlines in the network. Global network analysis Network measures were calculated from W. These included: mean transitivity, global efficiency, and nodal density, degree and strength14. Subnetwork analysis Group-averaged connectomes, Wgroup, were computed separately for patients and controls where edges present in at least 90% of each group were retained and averaged. The RC coefficient was computed with increasing degree (k) from Wgroup (https://sites.google.com/site/bctnet) before being normalised, Фnorm, by 1000 random degree-, strength- and density-distribution preserved networks, Фrand. RC nodes were initially chosen if they had a degree greater than a kth-threshold defined as the first kth instance when Фnorm>115 and also significantly different to Фrand. Final RC nodes were chosen by ranking their degree and strength and choosing the top 13 nodes (20% of total connectome regions). W was then divided into an 1) RC subnetwork of 13 RC nodes and their inter-connections; 2) feeder subnetwork of non-RC nodes which 'feed' into the RC; and 3) seeder subnetwork of non-RC nodes connecting to other non-RC regions8. The same global network measures were calculated for each subnetwork. All network measures between groups were compared with unpaired t-tests.Results
Global network analysis No significant group differences in network measures were found (Fig2, Table 1). Subnetwork analysis An RC regime was observed from k=8 and 9 for patients and controls, respectively, yielding 19 and 22 regions. The 13 RC nodes were largely similar in both groups, with the exception of left precentral appearing for controls, and left supramarginal gyrus for patients (Fig1, Table 2). Significant group differences were found in all subnetworks and across measures (Fig3, Table 1).Discussion
An RC topology was identified in the 16p11.2 cohort with nodes similar in location to controls and to those typically reported6,16,17. The only difference was inclusion of the left supramarginal gyrus in patients, a region associated with language function. These results suggest a preserved RC topology in 16p11.2. However, subnetwork analysis revealed many significant group differences. RC subnetwork density and degree as well as seeder subnetwork density, degree and strength were significantly decreased in patients. These findings suggest lower structural connectivity in compartments that are central (RC) and peripheral (seeder) to brain organisation. These structural alterations align with the distributed changes found from tract-based diffusion analysis4. In contrast feeder subnetwork transitivity, efficiency and strength were significantly increased in patients, possibly indicative of a support or mediating network compensating for a deficit in the RC (although not significant, equivalent measures in the RC were lower in patients). No significant differences were found in whole connectome analysis.Conclusion
We demonstrate that subnetwork analysis by RC stratification has the potential to reveal differential alterations in brain organisation with greater group differences in 16p11.2 deletion syndrome that could not be detected by whole brain connectome analysis. Whilst RC regions are largely preserved in location, patients exhibit significant reduction in structural connectivity with differential changes in network topology within its subnetworks which may be indicative of underlying compensatory mechanisms. Future work will ascertain the implications of these network aberrations in relation to language deficits that are characteristic of 16p11.2 deletion.Acknowledgements
We thank the Simons VIP Consortium for access to their data.References
1. Berman JI, Chudnovskaya D, Blaskey L, Kuschner E, Mukherjee P, Buckner R, et al. Abnormal auditory and language pathways in children with 16p11.2 deletion. NeuroImage Clin. 2015;9:50–7.
2. Steinman KJ, Spence SJ, Ramocki MB, Proud MB, Kessler SK, Marco EJ, et al. 16p11.2 deletion and duplication: Characterizing neurologic phenotypes in a large clinically ascertained cohort. Am J Med Genet A. 2016;170(11):2943–55.
3. Hanson E, Nasir RH, Fong A, Lian A, Hundley R, Shen Y, et al. Cognitive and behavioral characterization of 16p11.2 deletion syndrome. J Dev Behav Pediatr JDBP. 2010 Oct;31(8):649–57.
4. Chang YS, Owen JP, Pojman NJ, Thieu T, Bukshpun P, Wakahiro MLJ, et al. Reciprocal white matter alterations due to 16p11.2 chromosomal deletions versus duplications. Hum Brain Mapp. 2016;37(8):2833–48.
5. Ahtam B, Link N, Hoff E, Ellen Grant P, Im K. Altered structural brain connectivity involving the dorsal and ventral language pathways in 16p11.2 deletion syndrome. Brain Imaging Behav. 2018 Apr 9;
6. Heuvel MP van den, Sporns O. Rich-Club Organization of the Human Connectome. J Neurosci. 2011 Nov 2;31(44):15775–86.
7. Heuvel MP van den, Kahn RS, Goñi J, Sporns O. High-cost, high-capacity backbone for global brain communication. Proc Natl Acad Sci. 2012 Jul 10;109(28):11372–7.
8. Schirmer MD, Chung AW. Structural subnetwork evolution across the life-span: rich-club, feeder, seeder. In: Connectomics in Neuroimaging Workshop. LNCS, Springer; 2018. p. 134–43.
9. Simons Vip Consortium. Simons Variation in Individuals Project (Simons VIP): a genetics-first approach to studying autism spectrum and related neurodevelopmental disorders. Neuron. 2012 Mar 22;73(6):1063–7.
10. Fischl B. FreeSurfer. NeuroImage. 2012 Aug 15;62(2):774–81.
11. Jenkinson M, Smith S. A global optimisation method for robust affine registration of brain images. Med Image Anal. 2001 Jun;5(2):143–56.
12. Jenkinson M, Bannister P, Brady M, Smith S. Improved optimization for the robust and accurate linear registration and motion correction of brain images. NeuroImage. 2002 Oct;17(2):825–41.
13. Wang R, Benner T, Sorensen A, Wedeen V. Diffusion Toolkit: A Software Package for Diffusion Imaging Data Processing and Tractography. In: International Society for Magnetic Resonance in Medicine. 2007. p. 3720.
14. Rubinov M, Sporns O. Complex network measures of brain connectivity: Uses and interpretations. NeuroImage. 2010 Sep;52(3):1059–69.
15. Opsahl T, Colizza V, Panzarasa P, Ramasco JJ. Prominence and control: the weighted rich-club effect. Phys Rev Lett. 2008 Oct 17;101(16):168702.
16. Crossley NA, Mechelli A, Scott J, Carletti F, Fox PT, McGuire P, et al. The hubs of the human connectome are generally implicated in the anatomy of brain disorders. Brain J Neurol. 2014 Aug;137(Pt 8):2382–95.
17. Grayson DS, Ray S, Carpenter S, Iyer S, Dias TGC, Stevens C, et al. Structural and functional rich club organization of the brain in children and adults. PloS One. 2014;9(2):e88297.
Figures
Figure 2 - Boxplots of global mean network measures between controls and 16p11.2 deletion patients. No significant group differences were found.
