Associations of White Matter Integrity and Cortical Thickness in Patients With Schizophrenia and Healthy Controls

Abstract

Typical brain development includes coordinated changes in both white matter (WM) integrity and cortical thickness (CT). These processes have been shown to be disrupted in schizophrenia, which is characterized by abnormalities in WM microstructure and by reduced CT. The aim of this study was to identify patterns of association between WM markers and cortex-wide CT in healthy controls (HCs) and patients with schizophrenia (SCZ). Using diffusion tensor imaging and structural magnetic resonance imaging data of the Mind Clinical Imaging Consortium study (130 HC and 111 SCZ), we tested for associations between (a) fractional anisotropy in selected manually labeled WM pathways (corpus callosum, anterior thalamic radiation, and superior longitudinal fasciculus) and CT, and (b) the number of lesion-like WM regions (“potholes”) and CT. In HC, but not SCZ, we found highly significant negative associations between WM integrity and CT in several pathways, including frontal, temporal, and occipital brain regions. Conversely, in SCZ the number of WM potholes correlated with reduced CT in the left lateral temporal gyrus, left fusiform, and left lateral occipital brain area. Taken together, we found differential patterns of association between WM integrity and CT in HC and SCZ. Although the pattern in HC can be explained from a developmental perspective, the reduced gray matter CT in SCZ patients might be the result of focal but spatially heterogeneous disruptions of WM integrity.

Introduction

High-resolution structural MRI (sMRI) and diffusion tensor imaging (DTI) are 2 commonly used MRI methods to explore the underlying neurobiology of schizophrenia (SCZ). These approaches capture different properties of the brain that often have been used in isolation, but studies combining these techniques are beginning to emerge.^1–3 Although sMRI provides reliable measures of brain structure, DTI provides measures of tissue microstructure.⁴ One important measure obtained from sMRI is cortical thickness (CT), which is defined as the distance between the gray matter (GM)/white matter (WM) interface and the outer rim of gray matter. DTI, on the other hand, taps into diffusion properties of water in the brain.^5,6 Because neuronal fibers generally course together in the same direction, DTI is well suited to measure WM microstructure. One common measure extracted from DTI is fractional anisotropy (FA), which is a rotationally invariant measure of water diffusion in the brain.⁷ Higher FA implies greater diffusion in one direction, such as occurs in WM tracts as water flows more freely along the axons compared with through the cell wall. Alternatively, lower FA implies water diffusion equally in all directions, such as diffusion in cerebral spinal fluid.

DTI has demonstrated WM abnormalities in indi viduals at high risk for psychosis and in first-episode, medication-naive, and patients with chronic SCZ. These studies provide strong evidence for a primary role of WM abnormalities in the pathogenesis of this burdensome disorder.^8–10 Although there is considerable heterogeneity in the spatial localization of WM abnormalities,¹¹ recent meta-analyses show that deep (left) frontal and temporal WM regions appear to be more commonly affected.^8,12,13

The underlying affected WM tracts include the anterior thalamic radiation (ATR), corpus callosum (CC), and the superior longitudinal fasciculus (SLF). Fibers of the ATR connect the anterior and medial thalamic nuclei and the prefrontal cortex and are thought to be integral to the pathophysiology of SCZ.^14,15 The CC is the largest WM fiber bundle and has been suggested to play a role in abnormal hemispheric specialization and lateralization in SCZ.¹⁶ The SLF connects the temporal, parietal, and frontal regions and plays an important role in working memory,^17,18 which is typically impaired in SCZ.^19,20

In addition to DTI, sMRI studies also show regional GM changes, most notably reductions of the volume in the medial temporal lobe and ventricular enlargement.^21–24 More recently, sMRI studies have shown widespread CT reductions in frontal, temporal, and parietal regions in patients with SCZ.^25–28 Furthermore, CT is highly heritable^25,29,30 and has been suggested as a promising endophenotype for SCZ.

According to mechanical models of brain development, variations in CT are also thought to be modulated by axonal tension exerted between linked areas.^31,32 Consequently, the development of WM circuitry and CT should be closely interrelated. Early sMRI studies have reported global increases in WM volume and specific increases in WM density from childhood, through adolescence, and into adulthood.^2,33,34 Subsequently, voxel-based DTI studies have demonstrated significant increases in FA from childhood to adulthood,³⁵ suggestive of increasing myelination, fiber packing density, fiber coherence, or decreased axon diameter.³⁶ Similarly, GM volumes first increase during early childhood and start to decrease thereafter.³³ This is due to a considerable thinning of the cortex that occurs in children and continues throughout adolescence^37,38 and even beyond, during adulthood.³⁹ Thus, both CT and FA appear to show parallel changes, but the relationship between these 2 measures and their association with SCZ is yet unknown.

Thus, given the wealth of evidence for abnormalities of WM integrity and GM thickness in SCZ, this study aimed to investigate whether WM characteristics in specific WM tracts and circumscribed disruptions of WM microstructure is related to alterations in GM thickness. Because it is almost impossible to accurately determine where exactly WM tracts connect to cortical neurons using MRI, we modeled associations across the entire cortical surface. On the level of DTI, we used 2 distinct approaches to test different patterns of WM abnormalities. Both approaches were based on the assumption that WM aberrations do not necessarily occur in spatially overlapping regions across individuals. Instead, diffuse WM abnormalities (with slight decreases in FA) might occur at different locations along the same WM tract or might be present as focal abnormalities with dramatic decreases (ie, a 2 SD drop in FA in at least 50 contiguous voxels) in FA in spatially distinct brain regions.⁴⁰ There is no evidence that these lesion-like WM abnormalities, referred to as “potholes,” arise during normal development.⁴¹

Methods

Participants

The Mind Clinical Imaging Consortium (MCIC) study of SCZ^42,43 obtained sMRI and functional MRI (fMRI) scans on a total of 328 subjects from 4 participating sites: Massachusetts General Hospital in Boston (MGH), University of Iowa (UI), University of Minnesota (UMN), and University of New Mexico (UNM). The institutional review boards (IRBs) at each of the 4 sites approved the study protocol. After complete description of the study to the participants, written informed consent was obtained from each participant. The patient group (SCZ) included subjects with a Diagnostic and Statistical Manual of Mental Disorders IV (DSM-IV) diagnosis of SCZ, established through administration of structured clinical interviews (see below), and reviewed case files by trained clinicians (who completed 2 days of in-person training and testing using videotaped training materials with “gold standard ratings” to establish a cross-site, interrater reliability >85%). Healthy controls (HCs) were included if they had no history of a medical or Axis I psychiatric diagnosis. All participants were required to be at least 18 years of age and no older than 60, and to be fluent in English. Participants were excluded if they had a history of neurologic disease, psychiatric disease other than SCZ, history of a head injury, history of substance abuse or dependence within the past month, severe or disabling medical conditions, contraindication to MR scanning or intelligence quotient <70 (based on the reading subtest from the Wide Range Achievement Test [WRAT-III]).

The final sample with complete and high-quality sMRI and DTI data comprised 136 HC participants and 118 patients with SCZ.

Clinical Measures

All study participants underwent an extensive clinical diagnostic assessment that included either the Structured Clinical Interview for DSM-IV (SCID), Patient Edition (I/P)/Non-patient Edition (I/NP)⁴⁴ or the Comprehensive Assessment of Symptoms and History (CASH,⁴⁵ see also table 1 and ref. ⁴²⁾. Premorbid cognitive achievement was estimated by the WRAT-III⁴⁶; parental socioeconomic status (SES) was determined using the Hollingshead index⁴⁷ and handedness was determined using the Annett Scale of Hand Preference.⁴⁸ Severity of positive and negative symptoms was rated using the Scale for the Assessment of Positive Symptoms (SAPS) and the Scale for the Assessment of Negative Symptoms (SANS).^49,50

Table 1.

Demographics and Number of Potholes per Acquisition Site and Diagnostic Group

		n	Sex		Age		WRAT-IIIRT		Parental SES		Handedness		Number of Potholes
			F	F
			n	%	Mean	SD	Mean	SD	Mean	SD	Mean	SD	Mean	SD
MGH	HC	23	9	39	40a	10	51.91	3.95	3.04a	0.98	1.04	2.93	0.76a	1.30
	SCZ	34	9	26	37a	10	46.25	6.61	3.13a	0.94	1.06	3.09	13.72a	13.28
UI	HC	51	28	55	31a	10	50.24	4.27	2.90	0.46	0.47	2.03	3.75a	3.55
	SCZ	19	4	21	32a	10	49.11	4.84	2.47	0.70	1.26	3.78	8.63a	5.60
UMN	HC	22	10	45	29a	9	50.86	3.68	2.23a	0.75	0.77	2.16	3.14a	3.40
	SCZ	27	8	30	31a	10	46.04	6.57	2.67a	0.92	1.54	3.01	13.63a	10.93
UNM	HC	40	7	18	29a	12	49.90	4.77	2.40	0.84	0.85	2.06	12.35a	12.22
	SCZ	38	8	21	33a	13	47.58	5.88	3.03	1.23	0.86	2.29	26.03a	12.90
Total	HC	136	54	40	31	11	50.52b	4.30	2.67	0.79	0.73	2.22	5.90b	8.46
	SCZ	118	29	25	34	11	47.11b	6.13	2.88	1.02	1.14	2.94	17.10b	13.29

Note: WRAT-IIIRT, reading subtest of the Wide Range Achievement Test-III; SES, socioeconomic status; handedness, Annett Handedness Scale; MGH, Massachusetts General Hospital; UI, University of Iowa; UMN, University of Minnesota; UNM, University of New Mexico; SCZ, patients with schizophrenia; HC, healthy controls. A series of ANOVA models were performed to detect significant differences of age, WRAT-IIIRT Score, parental SES, handedness, and gender between acquisition sites and diagnostic groups.

^aSignificantly different between acquisition sites on the basis of a linear regression with subsequent Scheffé post hoc tests (P < .05).

^bSignificantly different between SCZ and HC (P < .05).

Antipsychotic history was collected as part of the psychiatric assessment using the PSYCH instrument,⁵¹ and cumulative and current antipsychotic exposure was calculated using the chlorpromazine (CPZ) conversion factors (see online supplementary material 1.1).

Structural Image Acquisition

Structural MRI data were acquired with either a Siemen’s 1.5-Tesla (MGH, UI, and UNM) or a Siemen’s 3-Tesla (UMN) MR scanners. The T₁-weighted structural brain scans at each of the 4 sites were acquired with an in-plane resolution of 0.625 × 0.625 mm, a slice thickness of 1.5mm, and a flip angle of 7 degrees. MGH and UNM used a Siemen’s 1.5-Tesla scanner with repetition time (TR) = 12, echo time (TE) = 4.76, and number of excitations (NEX) = 1. UI used a GE 1.5-Tesla Genesis Signa scanner with TR = 20, TE = 6, and NEX = 3. UMN used a Siemen’s 3-Tesla scanner with TR = 2530, inverse time (TI) = 1100, TE = 3.79, and NEX = 1.

All DTI images were acquired at each site with a 2mm isotropic resolution. MGH used a Siemens Sonata 1.5-Tesla scanner with TR = 8900, TE = 80, B values of 0 and 700, NEX = 1, and 60 directions. UI used a Siemens TRIO 3-Tesla scanner with TR = 9500, TE = 90, B values of 0 and 1000, NEX = 4 and 6 directions. UNM used a Siemens Sonata 1.5-Tesla scanner with TR = 9800, TE = 86, B values of 0 and 1000, NEX = 4 and 12 directions. UMN used a Siemens TRIO 3-Tesla scanner with TR = 10 500, TE = 86, B values of 0 and 1000, NEX = 2 and 12 directions.

Cross-site MRI acquisition calibration and reliability were established in a preceding study using human phantoms, following guidelines developed by the biomedical informatics research network (BIRN) test bed for morphometry.^52,53

Structural Image Data Processing

Structural MRI data from 3 consecutive volumes were registered, motion corrected, averaged, and ana lyzed in an automated manner with atlas-based FreeSurfer software suite (http://surfer.nmr.mgh.harvard.edu, Version 4.0.1). This process included volumetric segmentation, cortical surface reconstruction (see also online supplementary material 1.2 ^54–58), and the estimation of total intracranial volume (ICV).⁵⁹ The applied method has been thoroughly evaluated and previously applied in a wide variety of settings, including multicenter studies.

Tract-Based Diffusion Data Processing

For the tract-specific FA values, we limited our analysis to participants of the MGH site due to their superior DTI acquisition protocol. We used a standard method, available in FMRIB Software Library (FSL) (http://www.fmrib.ox.ac.uk/fsl), for mitigating distortions induced by eddy currents and motion by registering the diffusion-weighted images to the b = 0 images after spatial normalization of the diffusion-weighted images (using the reconstructed high-resolution T₁-weighted images and an intrasubject affine registration method that maximizes the intensity contrast gradient of the image across the gray/white boundary).⁶⁰ The eddy current-corrected diffusion images were then processed in Trackvis (http://trackvis.org). Deterministic tensor streamline tractography was applied to the whole brain using the FACT method.⁶¹ Manual labeling of the whole-brain streamline solutions was then performed in Trackvis to isolate the tracts belonging to the WM pathway of interest (for more details, please refer to ref.⁶²). Specifically, we followed the protocol described by Wakana et al⁶³ to label the forceps major (CC-fmj) and minor (CC-fmi) of the corpus callosum, the anterior thalamic radiation (ATR) in both hemispheres (rh = right hemisphere and lh = left hemisphere) and the parietofrontal part of the rh/lh SLF (see also online supplementary figures 1 and 4). FA values were averaged over all the voxels that intersected with the streamlines of each tract.

No-Overlap Requirement Spatial Analyses of Diffusion Data

An overview of the no-overlap requirement spatial analyses (NORSA) or the “pothole” approach is available elsewhere.⁴⁰ Briefly, using spatially normalized images, a mean and SD images were created using all the individual FA images from the control group within each of the 4 sites. These mean and SD images were then used to transform all images into z-images for both the patients and controls. The z-transformation is performed within each site, using the mean and SDs for the controls at that site. For each generated z-image, an algorithm searches for contiguous clusters of voxels that drop below a set z-score, which was set at z = −2. To reduce the likelihood that the potholes are merely a chance finding the minimum volume of each pothole was set to 50 voxels (see also ref. ⁴¹⁾ and the number of potholes in each participant was used for further analyses.

Statistical Analyses

Group differences of basic demographics were examined with ANOVA models, Student t tests, and chi-square tests using SPSS 17.0.

Entire cortex vertex-wise analyses of CT were performed with FreeSurfer (for details please refer to online supplementary material 1.2). Statistical maps were generated by computing general linear models (GLM) of the effects of each predictor variable on CT at each vertex. Because of known confounding effects and in line with similar studies, we included age, gender, and acquisition site (for the pothole analysis) into the models as control variables.^{25,27,64–67} All CT results were corrected for multiple comparisons using a Monte-Carlo simulation (uncorrected results not reported). This procedure includes the following steps: (1) The initial vertex-wise threshold was set to P = .05 to form spatially contiguous areas of association (referred to as “cluster”). (2) The likelihood that a finding (cluster) of this size and magnitude (difference in thickness as specified by the VWT) would appear by chance, ie, when using repeated random sampling, was tested using Monte-Carlo simulation with 10 000 repeats. This results in a cluster-wise probability (CWP), which is reported as P-values throughout the results section. In addition, for our tract-specific analysis, CWP values were corrected for the number of tested WM tracts (CC-fmj, CC-fmi, rh/lh ATR, and rh/lh SLFP = 6 tracts) in our main analysis using the Bonferroni method.

Furthermore, we performed additional analyses to control for possibly confounding effects caused by handedness or cumulative antipsychotic exposure (see online supplementary material 2.2).

Results

Sample Characteristics

Patients with SCZ and HCs did not differ in age, parental SES, and handedness score (table 1). Patients had a significantly lower estimated premorbid cognitive function than controls (table 1). The clinical characteristics of the SCZ cohort are listed by acquisition site in online supplementary table 1.

Reduced White Matter Integrity and Cortical Thickness in Schizophrenia

As expected, patients with SCZ had reduced FA in the CC-fmj, lh-ATR, and rh-ATR when compared with HCs (see online supplementary table 2). They also had a significantly greater number of potholes (for the full sample: T = −7.9, df = 237, P < .001; see also table 1).

In accordance with previous studies^25,27,66 and using the same technique, entire cortex vertex-wise statistics revealed widespread bilateral thickness reductions in SCZ patients, most pronouncedly in the frontal lobe, temporal cortex, inferior parietal lobe, and occipital cortex (see online supplementary figure 2). There were no regions where CT was increased when compared with HCs.

Significant Associations Between Tract-Specific Fractional Anisotropy and Cortical Thickness

Associations between CT at each vertex and tract-specific FA were tested separately in the groups of MGH schizophrenia patients (n = 34) and MGH healthy controls (n = 23). After correction for multiple comparisons at the cluster level using Monte-Carlo simulation, there were negative associations between tract-specific FA in 5 WM tracts and 16 cortical brain regions in HCs (see online supplementary material 2.1; figures 1 and 2; table 2 for cluster-wise P-values). There were no positive associations between FA and CT in controls and there were no significant associations between FA and CT in the group of SCZ patients. When controlling for multiple comparisons at the level of the 6 different tracts using Bonferroni correction (ie, P < .05/6 = 0.0083), the following negative associations between CT and tract-specific FA in HCs remained significant: FA in the CC-fmj and CT in the rh temporal pole and inferior temporal gyrus; FA in the CC-fmi and CT in the cuneus; FA in the lh ATR and CT in the lh rostral and caudal middle frontal gyrus as well as lateral orbitofrontal cortex; FA in the lh SLFP and CT in the lh middle and inferior temporal gyrus as well as FA in the rh SLFP and CT in the rh middle and inferior temporal gyrus and superior frontal gyrus (labeled with an asterisk in table 2; see also online supplementary figure 3). Some of these associations may be direct, ie, areas of reduced CT are locally adjacent to the endpoints of the WM tracts (see online supplementary figure 4).

Fig. 1.

Cortical statistical maps displaying negative associations between tract-specific fractional anisotropy and cortical thickness in healthy controls (mapped onto the surface of the FreeSurfer average brain). Cluster-wise probability values (corrected for multiple comparisons) are represented according to the color code and are all <0.05. Abbreviations: lh, left hemisphere; rh, right hemisphere; SLFP, superior longitudinal fasciculus–parietofrontal part; ATR, anterior thalamic radiation; CC-fmj, corpus callosum forceps major, and CC-fmi, minor. For color version, please see figure online.

Fig. 2.

Cortical maps displaying an association of the number of potholes with cortical thickness in schizophrenia patients (CWP = 0.00575; TalX, Y, Z = −39.8, −64.1, 3.4, respectively). For color version, please see figure online.

Table 2.

Associations Between Fractional Anisotropy in Manually Labeled White Matter Pathways and Cortical Thickness in Healthy Controls

Tract	Number of Cluster	Max	NVtxs	Size (mm2)	TalX	TalY	TalZ	CWP	Annot
CC-fmj	1	−3.977	1880	1178.86	41.1	−5.2	−35.6	0.0055*	Inferior temporal
CC-fmi	1	−5.189	1682	1075.82	40.9	9.4	−32.3	0.012	Temporal pole
	2	−4.175	1852	882.06	7.4	32.5	11.4	0.0367	Rostral anterior cingulate
	3	−4.078	1562	865.81	64.1	−35	−5.9	0.0412	Middle temporal
	4	−3.598	1824	1342.34	16.5	−67.3	19.4	0.0025*	Cuneus
	5	−2.556	1621	964.1	13	40.3	38.1	0.0232	Superior frontal
lh ATR	1	−3.87	4931	2916.5	−28.6	29.5	32.3	0.0002*	Rostral middle frontal
	2	−3.337	1846	1373.21	−18.2	51.4	−14.7	0.0015*	Lateral orbitofrontal
lh SLFP	1	−5.662	1948	1236.76	−56.1	-28.7	−12.9	0.0037*	Middle temporal
	2	−2.295	1787	926.06	−33.5	13.3	−31	0.03	Temporal pole
rh SLFP	1	−3.471	1906	1160.88	15.6	42.6	34.8	0.003*	Superior frontal
	2	−3.465	2533	1446.04	64.1	−35.1	−5.5	0.0002*	Middle temporal

Note: NVtxs, number of vertices; TalX, Y, Z, Talairach coordinates; CWP, cluster-wise P-value; Annot, FreeSurfer annotation label.

*Association remains significant after controlling for multiple comparisons at the level of the 6 different tracts using Bonferroni correction (P < .05/6 = 0.0083).

When running the same GLM for the whole sample with diagnosis as a between-subject factor and the interaction between tract-specific FA and diagnosis, the negative association between FA in the lh ATR and CT in the lh superior frontal gyrus was significantly larger in HC compared with the same in SCZ. Similarly, the negative association between FA in the rh-ATR and CT in the rh superior frontal and rh superior temporal gyrus was significantly larger in HC compared with SCZ (see online supplementary table 2).

Significant Associations Between White Matter Potholes and Cortical Thickness

Our cortex-wide analysis did not show any associations between the number of potholes and CT in HCs (n = 130). However, in SCZ patients (n = 111), the number of potholes correlated with reduced CT in the left inferior and medial temporal gyrus, left fusiform and left lateral occipital brain area (figure 2). For subanalyses regarding the effects of handedness and antipsychotic medication please refer to online supplementary material 2.1.

Discussion

Using 2 very different forms of assessing WM characteristics and a well-validated analytic approach for measuring CT across the brain, we identified distinct associations between WM characteristics and regional thinning of the cortical ribbon. In HCs, FA in the corpus callosum, anterior thalamic radiation, and superior longitudinal fasciculus were negatively correlated with CT in associated brain regions, whereas this presumably normal relationship was disrupted in patients with SCZ. In contrast, WM potholes, ie, relatively large lesion-like regions with markedly reduced FA, were more prevalent in SCZ patients and associated with reduced thickness in temporal-occipital brain regions, specifically in patients.

Furthermore, we confirmed findings of previous studies, demonstrating reduced FA in a number of WM tracts,^15,62,68,69 an increased number of WM potholes^40,70 and substantially reduced CT in frontal, temporal, and occipito-parietal areas in patients with SCZ compared with HCs.^{25–28,64,66}

Increased WM integrity of specific WM tracts may be associated with improved global and local brain network connectivity and permit a more efficient organization of neuronal resources during normal development. In particular, superior connectivity might allow for appropriate synaptic plasticity, which could influence regional CT in healthy adults. In line with that, fMRI studies in healthy children suggest that developmental changes in patterns of brain activity involve a shift from diffuse to more focal activation, likely representing a fine-tuning of relevant neural systems with experience.⁷¹ Interestingly, GM thinning occurs at the same time and in the same locations as brain growth (particularly in the frontal lobe).⁷² However, the exact mechanism behind cortical thinning during normal development is not known. Cortical thinning during puberty has been associated with the loss of unneeded connections and synaptic pruning and may parallel a reduction in neuropil and glial cells.⁷³ Decreased GM and increased WM volume and integrity may also reflect progressive age-related myelin coating of fibers,⁷⁴ which is also assumed to occur in the lower cortical layers.³⁷ It is interesting to note that both greater cortical thinning and increasingly dense and ordered WM fiber tracts have also been associated with improved cognitive performance.^72,75

Negative relationships between WM integrity and GM characteristics have been observed previously.^2,3 Giorgio et al (2008)² were one of the first to report that the age-related decrease in GM density in the right lateral prefrontal cortex correlates with FA increases in the superior region of the corona radiata in healthy adolescents and young adults. Based on this negative association, they suggested that age-related changes may be occurring simultaneously in cortical GM and WM microstructure and that these changes may be a consequence of experience- and use-dependent regional development.^73,76 More recent results by Tamnes et al (2010)³ support this conclusion. They also found significant, although somewhat weaker, negative correlations between CT in a large number of brain regions and FA in underlying tracts in healthy children, adolescents, and adults. These adaptive mechanisms, shaping neural circuits and thus providing a biological basis for ongoing development of cognitive abilities and behavior,⁷⁷ might be impaired in SCZ. The lack of the seemingly typical relationship between cortical GM and WM microstructure could be the morphological correlate of this disturbance.

In addition to the disturbance of general developmental mechanisms, there is evidence for more dramatic local WM disruptions such as potholes. “Potholes” as defined in this study represent relatively drastic disruptions in WM integrity that may occur at different “points of weakness” in different individuals and are unlikely to be associated with normative development. Early onset as well as typical adult SCZ patients have been shown to carry a higher load of potholes compared with HCs.^40,41 Given the lesion-like nature of potholes, it is not surprising that the amount of potholes correlates with reduced CT in SCZ patients. These results are in line with the existing evidence for a dysconnectivity syndrome in SCZ. To date, the underlying biology of potholes remains to be clarified but possible reasons are related to inflammation,^78,79 microvascular changes,⁷⁹ or oligodendrocyte pathology.

A recent study in SCZ patients and HCs corroborates our pothole findings.¹ In this study, the authors followed up on an earlier finding in the same cohort demonstrating drastically increased radial diffusivity in an area of the left temporal cortex most probably containing fibers from the inferior and superior longitudinal fasciculus.⁸⁰ They found an association with decreased CT in the posterior cingulate cortex in patients, but not in controls. The markedly increased radial diffusivity in the WM region of interest (ROI) in patients in this study resembles our pothole definition and can be interpreted as resulting from demyelination or changes to the axonal cytoskeleton.⁸¹ Previous studies in patients suffering from multiple sclerosis came to complementary conclusions: Lesions in certain WM tracts seem to cause GM atrophy in connected regions via impaired axonal transport, anterograde, and retrograde degeneration of cell bodies or secondary neuronal loss.^82–84 Similarly, the number of WM hyperintensities—which have a similar definition as potholes—have been researched extensively^85–87 and have also been found to be associated with hippocampal atrophy in patients with Alzheimer disease.⁸⁸

Limitations

The present findings must be interpreted within the context of the study limitations. First, abnormal patterns of associations between WM characteristics and CT in SCZ patients may be influenced by the effects of antipsychotic medications. However, abnormal WM integrity and reduced cortical GM thickness have been shown to occur in persons with a high risk of developing SCZ and among neuroleptic-naive and very young patients with a first episode of SCZ,^89,90 implying the involvement of disturbed neurodevelopmental rather than pharmacological mechanisms. Furthermore, in our own (see online supplementary Results) and in the vast majority of previous studies^8,41,91 cumulative antipsychotic exposure was not related to FA in any of the tracts studied or the number of potholes. When we covaried for cumulative antipsychotic exposure, the effect of WM potholes on CT in SCZ patients remained highly significant.

Second, due to the multisite study approach, our results could have been influenced by differences in MR scanner field strength and diffusion-weighted sequence acquisition parameters. However, it also provided increased statistical power, which enabled us to isolate relationships between WM potholes and CT in an unbiased manner, ie, without averaging across predefined a priori regions of interest. Furthermore, MRI findings that are platform independent (eg, independent of field strength or diffusion directions) would be more likely to be illness related. Several groups have studied the robustness of CT measurements using the same technology as in our study. Measurements across field strengths were found to be highly reliable⁹² and only slightly biased.⁹³ Similarly, the NORSA approach includes a standardization step within site and has been shown to produce relatively consistent results across site.⁴¹ We covaried for the effects of acquisition site in all statistical models.

Third, a subanalysis comparing MCIC participants who were included in this analysis and participants who were excluded due to missing or low-quality brain imaging or other phenotype-related data revealed that excluded patients had higher positive, negative, and disorganized symptoms. Therefore, our findings may not apply to the most severely affected patients.

Finally, the interpretation of diffusion MRI studies is difficult due to our limited knowledge about the exact biological causes of diffusion anisotropy changes. Dividing WM tracts into subcomponents or using additional measures extracted from diffusion-weighted images, such as radial and axial diffusivity may help to identify underlying WM processes.⁹⁴ Changes in radial diffusivity have been found to accompany WM maturation in HCs,³ whereas axial diffusivity, the average magnitude of water molecule displacements parallel to WM pathways, is indicative of axonal health.^95,96 In line with the developmental hypothesis, the associations between WM integrity and CT in our study (which we observed only in HCs) were driven by radial diffusivity (see online supplementary table 4).

Conclusions

In conclusion, this study revealed differential patterns of association between WM integrity and CT for HCs and SCZ patients. Although the pattern in HCs is characterized by slight decreases of CT with increasing FA and can be explained from a (normal) developmental perspective, reduced GM thickness in SCZ patients may be interpreted as the result of focal but spatially heterogeneous disruptions of WM integrity.

Supplementary Material

Supplementary material is available at http:// schizophreniabulletin.oxfordjournals.org.

Funding

National Institutes of Health/National Center for Research Resources (P41RR14075); Department of Energy (DE-FG02-99ER62764); The Mind Research Network, Morphometry Biomedical Informatics Research Network (1U24, RR021382A); Function Biomedical Informatics Research Network (U24RR021992-01); National Institutes of Health/National Center for Research Resources (MO1 RR025758-01); Shared Instrumentation Grants (1S10RR023401, 1S10RR019307, and 1S10RR023043); National Institutes of Health (K99/R00 EB008129 to A.Y.); National Institute of Mental Health (K08-MH068540 to T.W.); the NARSAD Young Investigator Award (to S.E.); Deutsche Forschungsgemeinschaft (to S.E.).