Diffusion Tensor Imaging Abnormalities in the Uncinate Fasciculus and Inferior Longitudinal Fasciculus in Phelan-McDermid Syndrome

Abstract

Background:

This cohort study utilized diffusion tensor imaging tractography to compare the uncinate fasciculus and inferior longitudinal fasciculus in children with Phelan-McDermid syndrome with age-matched controls and investigated trends between autism spectrum diagnosis and the integrity of the uncinate fasciculus and inferior longitudinal fasciculus white matter tracts.

Methods:

This research was conducted under a longitudinal study that aims to map the genotype, phenotype, and natural history of Phelan-McDermid syndrome and identify biomarkers using neuroimaging (ClinicalTrial NCT02461420). Patients were aged three to 21 years and underwent longitudinal neuropsychologic assessment over 24 months. MRI processing and analyses were completed using previously validated image analysis software distributed as the Computational Radiology Kit (http://crl.med.harvard.edu/). Whole-brain connectivity was generated for each subject using a stochastic streamline tractography algorithm, and automatically defined regions of interest were used to map the uncinate fasciculus and inferior longitudinal fasciculus.

Results:

There were 10 participants (50% male; mean age 11.17 years) with Phelan-McDermid syndrome (n = 8 with autism). Age-matched controls, enrolled in a separate longitudinal study (NIH R01 NS079788), underwent the same neuroimaging protocol. There was a statistically significant decrease in the uncinate fasciculus fractional anisotropy measure and a statistically significant increase in uncinate fasciculus mean diffusivity measure, in the patient group versus controls in both right and left tracts (P ≤ 0.024).

Conclusion:

Because the uncinate fasciculus plays a critical role in social and emotional interaction, this tract may underlie some deficits seen in the Phelan-McDermid syndrome population. These findings need to be replicated in a larger cohort.

Introduction

Autism spectrum disorder (ASD) is a common and phenotypically heterogeneous neurodevelopmental disorder.^1,2 There are a growing number of different gene defects implicated in ASD. These different genetic causes may culminate in presentations of ASD through common pathways.¹ As a result, it is compelling to study ASD by examining genetic disorders associated with a high prevalence of ASD.

Phelan-McDermid syndrome (PMS), caused by either a 22q13.3 deletion encompassing the SHANK3 gene or SHANK3 mutation, is one compelling genetic model of ASD. PMS is characterized by intellectual disability, generalized hypotonia, delayed or absent speech, normal to accelerated growth, and mild dysmorphic features.^3,4 PMS is a rare, although underdiagnosed, disorder. The true prevalence of PMS is unknown; however, according to the PMS Foundation, over 2000 individuals have been diagnosed with PMS throughout the world. Up to 84% children with PMS have ASD.⁵ SHANK3 encodes a scaffolding protein located at the postsynaptic density of glutamatergic synapses and has important functions in synaptic scaffolding and regulating dendritic spine morphology.^6-8 SHANK3 mutations have also been identified in approximately 1% patients with ASD,⁹ and mouse models lacking Shank3 show autistic features.^8,10-16 Thus, PMS is a valuable genetic model of ASD to study the effects of SHANK3 on ASD clinical features and brain networks.¹⁷

Because ASD is characterized by deficits in social behavior, it is important to study relevant brain networks and phenotypic correlates in PMS. Social behavior, learning, and expression are all factors important in the ASD phenotype and are dependent on integrated activity of an extended, overlapping cortical socioemotional processing network.¹⁸ There has been a wealth of foundational data showing support for a long-distance cortical processing deficit in ASD as a contributing factor to core social deficits in ASD.^19,20 Two major tracts that together expand a circuit running between frontal, temporal, and occipital regions and are also involved in socioemotional functions are the uncinate fasciculus (UF) and inferior longitudinal fasciculus (ILF).²¹ The UF plays a role in episodic memory,²² emotional recognition and empathy,^23-25 and social reward,²² and the ILF plays a role in object and facial recognition^26,27 and more broadly in semantic learning²⁸ and reading comprehension.²⁹ Furthermore, the UF and ILF have been shown through diffusion tensor imaging (DTI) to have altered microstructure in children with ASD.^21,30-36 These two tracts were investigated in this study. To examine specificity of our findings, we also examined the corticospinal tract (CST) as a control tract.

To date no published studies have examined white matter tracts in PMS using DTI. We utilized DTI tractography to compare the UF and ILF in children with PMS with those in age-matched controls and investigated the relationship between ASD diagnosis and the integrity of the UF and ILF white matter tracts. We hypothesized that when comparing the PMS group versus controls, there would be greater integrity of the UF and ILF white matter tracts in the control group.

Materials and methods

Study design

This research was conducted under an ongoing, longitudinal study that aims to map the genotype, phenotype, and natural history of PMS and to identify biomarkers using neuroimaging (ClinicalTrial NCT02461420). Patients were recruited at one of four network sites: Icahn School of Medicine at Mount Sinai, Boston Children's Hospital (BCH), Rush University Medical Center (RUMC), and the National Institutes of Health (NIH). Inclusion criteria required patients to be English speaking, aged between three and 21 years, and have pathogenic 22q13 deletions including SHANK3 or pathogenic sequence variants of the SHANK3 gene. Of note, for loss-of-function point mutations in SHANK3 we used 58.571 kb as the deletion size.

Participants underwent longitudinal physical, medical, and neuropsychologic assessment over a period of 24 months. Brain magnetic resonance imaging (MRI) scans were only performed if clinically indicated and were not collected longitudinally. All MRI scans were transmitted to BCH for quality assurance and analysis.

Age-matched controls enrolled in a separate longitudinal study (through the NIH R01 NS079788) underwent the same standardized clinical imaging protocol and were imaged at BCH. Controls were recruited from the Boston area by targeted advertisements and letters in the community and at BCH, and through the research volunteer registry at BCH. Controls have no history of neurological or psychologic disease. Data were collected between 2013 and 2018.

Sample

A total of 18 MRI brain scans from 18 patients with PMS were available for analysis. Seven scans did not pass diffusion-weighted imaging quality assurance and were excluded. One scan was removed because of prior brain surgery. After excluding these scans, the present study included 10 subjects (50% male; for genotype see Table 1) and 10 age- and sex-matched controls with at least one high-quality diffusion-weighted MRI, for a total of 20 scans analyzed.

TABLE 1.

Genotype of Participants

Participant	Genotype
1	arr[hg19] 22q13.31q13.33(46967883-51211392)x1
2	arr[hg19] 22q13.33(51116107-51193680)x1
3	arr[hg19] 22q13.33(51127905-51234443)x1
4	arr[hg19] 22q13.33(51130264-51176099)x1
5	22q13 deletion (22q13.3 - q terminus)
6	SHANK3, NM_033517.1:c.5008A>T, p.K1670*, SHANK3, NM_033517.1: c.3872C>T, p.S1291L
7	ish 22q13.3(TelVysion22q-)
8	arr[hg19] 22q13.31q13.33(44831210-51178257)x1; der(22)t(2;22)
9	arr[hg19] 22q13.31-q13.33(44921022-51224252)x1
10	arr[hg19] 22q13.31-q13.33(46350579-51224252)x1

MRI acquisition

MRI brain scans were acquired on 3T GE DiscoveryMR 750, Siemens Skyra, and Siemens TrioTim scanners. Subjects were imaged under a consensus clinical imaging protocol that included a 1 mm³ sagittal T1-weighted MPRAGE (magnetization-prepared rapid acquisition with gradient-echo), 1 mm³ T2-weighted SPACE (sampling perfection with application-optimized contrasts by using different flip angle evolution), 30 high-angular-resolution diffusion-weighted images (DWI) (30 b = 1000 s/mm² images, 5 b = 0 s/mm² images, matrix = 128 × 128, contiguous slice thickness = 2 mm), and reversed phase-encoding directions for distortion compensation, covering the entire brain. Patients were sedated when clinically indicated. Controls were not sedated.

MRI processing

All MRI processing and analyses were completed using previously published and validated image analysis software distributed as the Computational Radiology Kit (http://crl.med.harvard.edu/). For each scan, the T2-weighted image was aligned to the T1-weighted image, and an intracranial cavity was segmented using a previously validated multispectral method.³⁷ DWI data were screened for intravolume motion and scanner artifact by expert raters, and artifactual volumes were removed. None of the PMS scans analyzed were affected by artifact, and no volumes were removed. Three b = 0 volumes were removed from one of 10 PMS scans due to patient motion. All control scans analyzed were not affected by artifact, and no volumes were removed. Magnetic susceptibility distortion was mitigated with FSLtopup,³⁸ and intervolume motion correction was performed via affine registration of each DWI volume to the mean b = 0 s/mm² image. DWI was up sampled to the T1-weighted image and skull stripped with the intracranial cavity segmentation. A single tensor diffusion model was estimated using robust least squares in each brain voxel from which fractional anisotropy $[\text{FA}=\frac{\sqrt{({\lambda }_1-{\lambda }_2{)}^2+({\lambda }_1-{\lambda }_3{)}^2+({\lambda }_2-{\lambda }_3{)}^2}}{\sqrt{2}\sqrt{{\lambda }_1^2+{\lambda }_2^2+{\lambda }_3^2}}]$ and mean diffusivity [MD = (λ₁ + λ₂ + λ₃)/3] were computed.

Tractography

For each subject, a whole-brain connectivity was generated using a stochastic streamline tractography algorithm that combines the speed and efficacy of deterministic decision making at each voxel with probabilistic sampling from the space of all streamlines.³⁹ A single streamline was stochastically initialized within each white matter voxel, and streamlines were propagated with multiple steps per voxel. Streamlines were terminated if the step angle exceeded 35 degrees, or if the FA fell below 0.20.

The left and right UF, ILF, and CST were selected from the whole-brain connectivity using automatically defined regions of interest (ROIs), for a total of six tracts selected per subject. The CST was used as a control tract because this tract would not be affected in ASD. The mean was measured over all tract voxels. The ROIs were automatically delineated on the subject scans using a fully automatic, multitemplate approach. Briefly, a template library was constructed from whole-brain DTI of 15 healthy individuals, with each scan in its native space. For each template, scalar FA and color maps of the principal diffusion directions were computed from the DTI, and ROIs were hand drawn by an expert rater on the color map. To delineate the same white matter ROIs in the native space of each subject scan, the following procedure was performed for every template: the template DTI was aligned to the subject DTI using a nonlinear, dense registration, and the dense deformation field was then used to resample the template white matter ROIs to the subject DTI using nearest neighbor interpolation.⁴⁰ Consensus ROIs were computed from the aligned template ROIs using the STAPLE algorithm.⁴¹

ROI delineation

The UF and ILF tracts were investigated. The ILF was selected from the whole brain connectivity using two selection ROIs located in (1) the anterior temporal lobe and (2) the occipital lobe with a ventral limit of the white matter of the lingual and fusiform gyrus and a dorsal limit of the white matter of the joining splenium fibers.⁴² To avoid nearby aberrant fibers, two regions of exclusion were manually placed on (1) a single midsagittal slice encompassing the corpus callosum and (2) a single axial slice at the brainstem. Similarly, the UF was selected from the whole brain connectivity using two selection ROIs located in (1) the anterior temporal lobe and (2) the external capsule⁴² (Fig 1). The UF anterior temporal lobe ROI was constrained to the anterior medial aspect of the anterior temporal lobe to minimize selection of non-UF fibers. To further optimize fiber selection, three regions of exclusion were manually placed on (1) a single coronal slice posterior to the anterior temporal lobe spanning the full hemisphere, (2) a single whole-brain midsagittal slice, and (3) sagittal slices on the extreme capsule.

FIGURE 1.

Example images of the ILF (A and B) and UF (C–E) overlaid on a fractional anisotropy map in two different patients with PMS. (A) Left midsagittal slice showing a representative ILF in a patient with PMS and the two ROIs (Catani et al.¹⁸) used to capture the ILF: (1) dark pink: anterior temporal lobe label and (2) purple: posterior occipital lobe label. (B) A left midsagittal slice showing the sagittal ROE. Note: The axial brainstem ROE slice was applied to generate the ILF indicated, but is not shown. (C) A left side midsagittal slice showing a representative UF in a patient with PMS. (D) Left side cutaway view showing all three planes (axial, sagittal, and coronal) with orange labels showing the whole-brain sagittal ROE slice and the hemisphere coronal ROE slice. (E) Left side cutaway view showing the two ROIs (Catani et al.¹⁸) used to capture the UF: (1) dark pink anterior temporal lobe label (2) light pink external capsule label. Note: The sagittal extreme capsule ROE slice was applied to generate the UF indicated, but is not shown. ILF, inferior longitudinal fasciculus; PMS, Phelan-McDermid syndrome; ROE, region of exclusion; ROI, region of inclusion; UF, uncinate fasciculus.

We examined the CST as a control tract. The left and right CSTs were selected using three ROIs. An ROI was drawn in the posterior limb of the internal capsule in the axial plane, with its inferior limit defined on the first axial slice superior to the anterior commissure and the superior limit defined on the axial slice where the lenticular nucleus separates the internal and external capsules.⁴² A second ROI was drawn in the cerebral peduncle, covering the inferior-superior-traversing white matter of the anterior pons.⁴² A final ROI covered the precentral gyrus. A region of exclusion was placed over the midline corpus callosum and cerebellar white matter.

Clinical measures

Patient phenotype was evaluated with neuropsychologic assessment at baseline, 12-month visit, and 24-month visit. ASD diagnosis was made based on clinical consensus using the Diagnostic and Statistical Manual for Mental Disorders, Fifth Edition,⁴³ and informed by evaluation with the Autism Diagnostic Observation Schedule, second edition, (ADOS2)⁴⁴ and Autism Diagnostic Interview-Revised.⁴⁵ The ADOS2 was given at baseline, 12 months into the study, and 24 months.

Best estimate intelligence quotient (IQ) was derived from standard scores on IQ tests and ratio IQ estimates in participants who received out-of-range testing with developmental cognitive tests. Specifically, as is standard in such studies⁴⁶ a hierarchy of tests was used, including the Mullen Scales of Early Learning⁴⁷ in 10 participants and the Stanford-Binet.⁴⁸

Statistical analysis

Wilcoxon sign rank testing was used to compare PMS with age-matched controls regarding UF FA and MD (left and right) and ILF FA and MD (left and right) DTI metrics. For patients, we used Spearman's rank correlation coefficients test to see if there was an effect of DTI metric on IQ. Owing to the small sample size, we used exact logistic regression to understand the relationship between ASD diagnosis and DTI metric while adjusting for age.

In posthoc analysis (given that control data was from a different study), we employed Benjamini-Hochberg (BH) false discovery rate procedure, to account for multiple comparisons with q = 0.05.⁴⁹ Because of small sample, we used less conservation BH correction approach instead of Bonferonni, but we used stricter cutpoint because we did not have numerous hypotheses to test. There were 12 comparisons when comparing DTI metric for patients with controls: three tracts [ILF, UF, CST] × two sides [left, right] × two metrics [FA, MD].

Results

Of the initial 18 subjects enrolled in the PMS clinical trial (NCT02461420) with MRI available for analysis, eight were excluded either due to failure to meet quality assurance or because of previous brain surgery. Ten PMS participant scan data were included in this cohort study. In both the PMS and control groups, 50% were male (χ²(1) = 0.000, P = 1.000). The mean age of the PMS group was 11.17 (6.17) years, and the mean age of the control group was 11.24 (5.96) years. The mean IQ of the PMS group was 27.1 (S.D. = 20.7, range = 3.4 to 61). A total of 80% had a diagnosis of ASD. The mean deletion size was 2457.8 kb (S.D. = 2694.1, range = 46 to 6320). In the sample, seven (70%) were administered ADOS Module 1, two (20%) were administered ADOS Module 2, and one (10%) was administered ADOS Module 3.

There was a statistically significant decrease in the UF FA measure in the PMS versus age-matched controls in both the right and left tracts (adjusted P ≤ 0.024), as well as a statistically significant increase in UF MD measure in the PMS group versus controls in both right and left tracts (adjusted P ≤ 0.024) (Fig 2). There was a decrease in the ILF FA measure in the PMS versus age-matched controls in both the right and left tracts, as well as an increase in ILF MD measure in the PMS group versus controls in the right and left tracts, but these changes were not statistically significant with adjusted P values (Fig 3). Furthermore, an investigation of DTI metrics in the CST was used as a control tract between the PMS and control groups. There was no statistically significant difference in DTI metrics in the CST of PMS versus controls.

FIGURE 2.

Mean UF DTI metric versus age of patients and controls. Plot of mean DTI metric versus age of patients and controls for (A) FA UF left (B) FA UF right (C) MD UF left, and (D) MD UF right. DTI, diffusion tensor imaging; FA, fractional anisotropy; MD, mean diffusivity; UF, uncinate fasciculus.

FIGURE 3.

Mean ILF DTI metric versus age of patients and controls. Plot of mean DTI metric versus age of patients and controls for (A) FA ILF left, (B) FA ILF right, (C) MD ILF left, and (D) MD ILF right. DTI, diffusion tensor imaging; FA, fractional anisotropy; ILF, inferior longitudinal fasciculus; MD, mean diffusivity.

There was no significant difference found within the PMS group between any DTI metric and diagnosis of ASD. Of note, only two of the 10 PMS participants did not have ASD. There was not a statistically significant difference found within the PMS group between any DTI metric and best estimate IQ. There was not a statistically signification correlation between any DTI metric and deletion size.

Discussion

This is the first study to date to use DTI in a PMS population to investigate potential microstructural changes in tracts critical for social and emotional interaction. This cohort study utilized DTI tractography to compare the UF and ILF in 10 children with PMS with age-matched controls and investigated the integrity of the UF and ILF white matter tracts.

In this study, there was a statistically significant decrease in the UF (and increase in the MD) in both the right and left tracts of PMS versus age-matched controls. The CST was used as a control tract, and no difference was found in DTI metrics in the PMS versus control groups. This control tract investigation further supports the specificity of the DTI changes in the UF. Based on previous studies in conjunction with our PMS investigation, UF microstructural abnormalities have been highlighted as an important tract underlying aspects of emotional and social behavior. The UF is a medial frontotemporal long-range white matter tract that links anterior portions of superior, middle, and inferior temporal lobes, including the hippocampus and amygdala, with the insular and orbitofrontal cortex.^50-52 The UF has been shown to play a critical role in emotional recognition,^23,24 self-awareness,⁵³ social reward processing,²² and proper name retrieval.⁵⁴ One study found that participants with UF lesions have deficits in an emotional empathy task, highlighting the role of the UF in social interaction and the ability to relate to others.²⁵ Past ASD DTI studies have found significant, yet conflicting, results when examining the UF in patients with ASD versus controls.^{21,30-34,55-57}

In this PMS study, the ILF FA DTI metric values were found to be decreased, and the ILF MD DTI increased, in the PMS group when compared with age-matched controls, but these findings were not statistically significant with adjusted P values. The ILF is the main pathway connecting brain structures that play a critical role in not only facial recognition but also modulation of visual stream socially salient information.⁵⁸ Visual processing and emotional recognition are crucial for effective social communication.^59-64 The ILF extends from the occipital cortex into the temporal lobes, which mediates connections to the superior temporal sulcus, fusiform face area, and the amygdala.^65-67 Previous ASD DTI studies have found changes in the ILF microstructure in ASD when compared with controls, reinforcing that the ILF is a relevant and important white matter tract to study in the context of ASD. This study also investigated a potential relationship between ASD diagnosis and DTI metric within the PMS group. No significant differences were found between any DTI metric and ASD diagnosis or IQ within the PMS group.

Both the UF and ILF have temporal and amygdala connectivity, which provides further support to study these two relevant tracts.^50-52,65-67 A previous positron emission tomographic study detected localized dysfunction of the left temporal polar lobe and amygdala hypoperfusion in children with PMS⁶⁸ suggesting that there may be abnormalities in amygdalo-cortical connectivity that may contribute to the social interaction deficits found in PMS. By studying white matter tracts in the context of a known genetic mutation, we can now also lay the foundation to study the role that SHANK3 may play in specific axonal tracts.

Limitations

One main limitation of this study was the small sample size. It is possible that the small sample size of this study is one factor that potentially prevented ILF differences between patients and controls from maintaining statistical significance. In addition, there was a wide age range in this study, and age significantly affects the development of both the UF and ILF.

Of note, two of the 10 PMS participants did not have ASD, so the analysis was unbalanced in the setting of a small cohort, which likely limited our ability to make further conclusions about the DTI findings in relation to ASD diagnosis. In other words, the main DTI findings of this investigation may be specific to PMS rather than ASD more broadly, although conclusions are limited with this small sample size. Although PMS is rare, through the ongoing longitudinal network study encompassing multiple institutions, we plan to grow our sample size for further studies (ClinicalTrial NCT02461420). In addition, in future studies with greater samples sizes, ADOS severity score should be evaluated as a continuous measure. Furthermore, in addition to including an estimate of Full Scale IQ, which is important, future directions with a greater sample size include investigating aspects of phenotypic measurements that more specifically relate to the function of the UF or ILF in PMS or ASD populations. For example, the Vineland Adaptive Behavior Scales, specifically the socialization domain, can be used to more effectively evaluate whether changes in UF white matter microstructure correlate with changes in specific adaptive behaviors in areas previously found to be functionally relevant in the UF and ILF. In addition, future work can utilize a low IQ control group to specify findings further to PMS-related differences and not merely to gross cognitive differences between PMS and control groups.

Conclusions

There are no published studies that have examined white matter tracts in PMS using DTI. We utilized DTI tractography to compare the UF and ILF in children with PMS with those in age-matched controls and investigated the relationship between ASD diagnosis and the integrity of the UF and ILF white matter tracts. Overall, this study found significant decrease in the FA (and increase in the MD) of the right and left UF in the PMS group when compared with age-matched controls. The UF plays a critical role in social and emotional interaction and may underlie some deficits seen in the PMS population. Further work with greater sample sizes needs to be done to study these two important tracts, ILF and UF, and their relationship to PMS and ASD. In addition, more work needs to be done to investigate on a neurological systems level the role of SHANK3 in different neural networks. Altogether, PMS provides a unique context to study contributions of SHANK3 to the development of ASD.

Appendix

Members of the Developmental Synaptopathies Consortium (DSC) – Phelan-McDermid Syndrome Group include: Mustafa Sahin, MD, PhD^a,b, Alexander Kolevzon, MD^c,d, Joseph Buxbaum, PhD^c,d,e,f, Elizabeth Berry Kravis, MD, PhD^g,h,i, Latha Soorya, PhD^j, Audrey Thurm, PhD^k, Craig Powell, MD, PhD^l,m, Jonathan A. Bernstein, MD, PhDⁿ, Simon Warfield, PhD^o, Benoit Scherrer, PhD^o, Rajna Filip-Dhima, MS^b, Kira Dies, ScM, CGC^b, Paige Siper, PhD^c, Ellen Hanson, PhD^p, Jennifer M. Phillips, PhD^q, Stormi P. White, Psy D^r.

Affiliations for above: ^aDepartment of Neurology, Boston Children's Hospital, Harvard Medical School, Boston, MA; ^bF.M. Kirby Neurobiology Center, Boston Children's Hospital, Harvard Medical School, Boston, MA; ^cSeaver Autism Center for Research and Treatment, Mount Sinai School of Medicine, New York, NY; ^dDepartment of Psychiatry, Mount Sinai School of Medicine, New York, NY; ^eDepartment of Genetics and Genomic Sciences, Mount Sinai School of Medicine, New York, NY; ^fDepartment of Neuroscience, Mount Sinai School of Medicine, New York, NY; ^gDepartment of Pediatrics, Rush University Medical Center, Chicago, IL; ^hDepartment of Neurological Sciences, Rush University Medical Center, Chicago, IL; ⁱDepartment of Biochemistry, Rush University Medical Center, Chicago, IL; ^jDepartment of Psychiatry, Rush University Medical Center, Chicago, IL; ^kPediatrics and Developmental Neuroscience Branch, National Institute of Mental Health, National Institutes of Health, Bethesda, MD; ^lDepartment of Neurobiology, University of Alabama at Birmingham School of Medicine, Birmingham, AL; ^mCivitan International Research Center, University of Alabama at Birmingham School of Medicine, Birmingham, AL; ⁿDepartment of Pediatrics, Stanford University School of Medicine, Stanford, CA; ^oComputational Radiology Laboratory, Department of Radiology, Boston Children's Hospital & Harvard Medical School, Boston, MA; ^pDepartment of Developmental Medicine, Boston Children's Hospital, Harvard Medical School, Boston, MA; ^qDepartment of Psychiatry and Behavioral Sciences, Stanford University School of Medicine, Stanford, CA; ^rCenter for Autism and Developmental Disabilities, University of Texas Southwestern Medical Center, Dallas, TX.