MYELIN AND AXON ABNORMALITIES IN SCHIZOPHRENIA MEASURED USING MRI TECHNIQUES

Abstract

Background

In schizophrenia (SZ), disturbances in integration of activity among brain regions appear to be as important as abnormal activity of any single region. Brain regions are connected through white matter (WM) tracts, and diffusion tensor imaging (DTI) has provided compelling evidence for WM abnormalities in SZ. However, DTI alone cannot currently pinpoint the biological basis of these abnormalities.

Methods

In this study, we combined a myelin-specific and an axon-specific MRI approach to examine potentially distinct abnormalities of WM components in SZ. Magnetization transfer ratio (MTR) provides information on myelin content while diffusion tensor spectroscopy (DTS) provides information on metabolite diffusion within axons. We collected data from a 1x3x3cm voxel within the right prefrontal cortex WM at 4 Tesla and studied 23 patients with SZ and 22 age and sex matched healthy control participants.

Results

MTR was significantly reduced in SZ, suggesting reduced myelin content. By contrast, the apparent diffusion coefficient of N-acetylaspartate (NAA ADC) was significantly elevated, suggesting intra-axonal abnormalities. Greater abnormality of both MTR and NAA ADC correlated with more adverse outcomes in the patient group.

Conclusions

The results suggest that WM abnormalities in SZ include both abnormal myelination and abnormal NAA diffusion within axons. These processes may be associated with abnormal signal transduction and abnormal information processing in SZ.

INTRODUCTION

Diffusion tensor imaging (DTI) provides information about water molecule diffusion and yields three diffusion eigenvalues labeled λ₁, λ₂, and λ₃ from largest to smallest. The brain’s white matter (WM) contains axon fibers, and water molecular diffusion takes place along the long axis of these fibers [axial diffusivity (AD)= λ₁] more than perpendicular to it [radial diffusivity (RD)=(λ₂+λ₃)/2]. Fractional anisotropy (FA) reflects directionality of diffusion (isotropic vs. anisotropic). Finally, mean or apparent diffusion coefficient [ADC=(λ₁+ λ₂+λ₃)/3] reflects the distance traveled by a molecule in unit time, partly reflecting geometry of the surrounding space. Past DTI studies have provided strong evidence for widespread disruptions in WM integrity in schizophrenia (SZ). FA reductions are associated with passivity phenomena (1), auditory hallucinations (2), impairments in working memory (3, 4) and executive function (5–8), and abnormal fMRI connectivity (9, 10). A related literature provides evidence of deficits in integration of large-scale neuronal networks (11–13) and in expression of myelin- and oligodendrocyte-related genes postmortem in SZ (14). Thus, abnormal integration of activity across brain regions appears critical to SZ pathophysiology.

Although WM abnormalities are central to SZ as an abnormal connection syndrome (15–17), the link between DTI and brain function remains abstract because of the non-specific nature of the DTI signal (18). FA, AD, and RD abnormalities are commonly interpreted as reflecting loss of “white matter integrity” but its exact nature cannot be determined using DTI alone. Water exists in intra- and extracellular compartments and there is exchange of water molecules between the two. Thus DTI abnormalities may reflect multiple processes (demyelination, fiber crossing, axonal swelling or atrophy) and even different abnormalities in different cases (19).

Separate in vivo measures of axon and myelin integrity would be valuable to address this issue. Notably, indices of axonal diameter and myelin sheath thickness would allow one to predict whether signal conduction speed is abnormal in SZ brains. Here, we utilize two MR-based approaches to probe specific WM abnormalities in SZ: magnetization transfer ratio (MTR) and diffusion tensor spectroscopy (DTS). MTR relies on magnetization exchange between water molecules in different physical environments. In biological tissue, “bound” water molecules around myelin lipids exchange protons with “free” water molecules. This exchange can be measured using a magnetization transfer paradigm where signal from “bound water” is saturated and the loss of “free water” signal (reflecting transfer to “bound water”) is measured. The larger the WM myelin component, the greater is the proton exchange, and the higher the MTR. MTR is reduced in SZ, suggesting reduced myelin complement in this condition (20, 21) although a recent study reported partially discrepant results (22).

DTS measures the diffusion of intracellular metabolites such as N-acetylaspartate (NAA). Because NAA is located exclusively in neurons and almost exclusively in the cytosol where diffusion is less restricted than within organelles (23), NAA diffusion provides specific information about intra-neuronal structure. DTS measures are based on molecular Brownian motion and are independent of metabolite concentration and transverse (T2) relaxation times; therefore the NAA reductions (24) and NAA T2 abnormalities (25) observed in SZ do not confound NAA diffusion. DTS approaches have been validated in a variety of contexts, including as probes of cellular diffusion (26) and in seminal studies of axon diameter (27). The only clinical studies using DTS are in acute cerebral ischemia, where NAA ADC is significantly reduced, (28–30) and MELAS (mitochondrial myopathy, encephalopathy, lactic acidosis, and stroke-like episodes), where it is elevated (31). The DTS parameter of greatest interest in the present study is NAA ADC which has no directionality. By contrast, FA as well as AD and RD scale with directionality of structures in the voxel, and macroscopic curvature artifacts render them uninterpretable in a large voxel (see Discussion). NAA ADC is informative about axon abnormalities: demyelination with preserved axon diameter would leave NAA ADC normal while changes in NAA diffusion within axons with preserved myelination would modify NAA ADC.

NAA ADC can be determined by axonal geometry or by NAA distribution within axonal organelles (e.g. mitochondria). Interactions between axon health and myelin sheath thickness are complex and bidirectional. Larger axons have thicker myelin sheaths, and vice versa. The ratio between axon diameter and fiber diameter (defined as axon diameter + myelin sheath thickness) is termed the g-ratio. The g-ratio evolves during brain development and reaches a level of 0.6 in adulthood (32–34). Divergence from this optimal g-ratio in either direction is associated with abnormalities in conduction speed (35).

The combination of MTR and DTS affords the ability to probe axon vs. myelin-related abnormalities separately in the human WM. Based on strong evidence for myelination abnormalities (36–39) as well as a mechanistic relationship between developmental myelination and SZ (40, 41), we hypothesized that myelin sheath thickness and MTR are reduced in SZ. Given the paucity of published information on axon structure in SZ, we could not predict NAA ADC changes.

METHODS

Participants

Following approval by the McLean Hospital IRB, we recruited 22 healthy controls from the community and 23 participants with SZ from the clinical services at McLean Hospital. Demographic and clinical characteristics of the study participants are provided in Table 1. See Supplementary Materials for details of inclusion/exclusion criteria, participant screening, and standard study procedures.

Table 1.

Demographic and clinical characteristics of study participants

	Healthy Control (N=22)	Schizophrenia (N=23)	Statistical Evaluation
Age (y)	31.4±6.7	34.0±9.2	F(43,1)=1.19; p=0.286
Gender	12M, 10F	14M, 9F	χ2=0.18; p=0.668
BMI	23.4±2.8	26.9±4.5	F(42,1)=9.13; p=0.004
Education*	6.8±1.0	4.4±1.4	F(42,1)=42.70; p<0.001
Parental SES†	5.9±1.0	5.5±1.6	F(26,1)=0.46; p=0.505
Age at onset (y)	--	21.8±6.8
Lifetime number of suicide attempts	--	1.5±2.5
Lifetime number of hospitalizations	--	6.7±6.1
MADRS	--	10.2±10.4
YMRS	--	8.1±7.8
PANSS	--	49.8±14.0
Lithium (No.)	--	3
Anticonvulsants (No.)	--	4
SGAs (No.)	--	19
FGAs (No.)	--	2
CPZ equivalents	--	501±459
Benzodiazepines (No.)	--	7

SGA: second generation antipsychotic; FGA: first generation antipsychotic. Other abbreviations as in the text.

^†Parental SES calculated according to the Hollingshead scale.

^*Education code: 3: graduated high school; 4: part college; 5: graduated 2 year college; 6: graduated 4 year college; 7: part graduate/professional school; 8: completed graduate/professional school

Magnetic Resonance Imaging and Spectroscopy

See Supplementary Materials for details of anatomic imaging and voxel placement (Figure 1).

Figure 1.

Representative axial images depicting the location of our 1x3x3cm white matter voxel in the right prefrontal cortex.

MTR

The MTR experiment relies on measuring water signal magnitude in the presence and absence of a saturation pulse, which causes saturation of signal coming from “bound” water molecules. Because there is exchange between “bound” and “free” water molecules, the saturation pulse measurably attenuates the signal coming from “free” water molecules (measured as the water resonance following the pulse) (Figure 2A). MTR is calculated based on water signal intensity acquired in the presence (M_s) and absence (M_c) of the saturation pulse [MTR = (M_c−M_s)/M_c].

Figure 2.

(A) MR spectra of the water resonance acquired from a healthy control participant in the presence (red) and absence (black) of the saturation pulse. The reduction in magnetization of the water resonance (ΔM) and the “restricted water” resonance (amplified scale) are also shown. ΔM forms the basis of calculations for the magnetization transfer ratio (MTR). (B) MTR presented from healthy controls (green), schizophrenia patients (red), and from a phantom (blue). The region highlighted by the box exhibits direct RF off-resonance effects and hence was excluded from the calculations presented in this paper.

We used a BISTRO saturation pulse train (42) constructed with multiple hyperbolic Sec pulses (width=50 ms) with varied RF pulse amplitudes and applied at the beginning of a standard point-resolved spectroscopy (PRESS) sequence (prior to the 90 degree pulse) to saturate “bound-water” signal with a specific frequency offset (42, 43). Data were obtained in 50 Hz steps at a range of frequencies offset 400 to 1000Hz in either direction from the water signal and a single MTR number was calculated by averaging across frequencies. Saturation time (t_sat) was 2.6 s with TR/TE=3000/30ms and repetitions=2.

DTS measurements

The standard PRESS sequence was modified by incorporating diffusion gradients for DTS measurements. Bipolar diffusion gradients with 6 directions: [1,1,0] [1,0,1] [0,1,1] [−1,1,0] [−1,0,1] [0,−1,1] and one control [0,0,0] (totaling 7 spectra) were applied to calculate diffusion tensors of signal from water and metabolites. The applied b value was 1412 s/mm², calibrated by a home-made phantom with water ADC assumed to be 2.1×10⁻³ mm²/s at room temperature (~20°C) (44). In these measurements, TR/TE=3000/135 ms, diffusion time (D_t)=60 ms, repetitions=96 and 4 for metabolites and water diffusion measurements, respectively. Metabolite spectra were acquired with water saturation using VAPOR (45). Free induction decays were stored separately prior to averaging for correction of frequency- and phase-drifts, and eddy currents resulting from diffusion gradients or instability of machine hardware. Total experiment time including MTR and DTS measurements of water and metabolites was around 70 minutes. The water data from the DTS experiment are analogous to DTI data from the literature with one major exception: they were collected from a large 1x3x3 cm³ voxel. In addition, before or after each human subject study, we carried out a phantom DTS scan to correct for measurement errors from potential machine instability.

MRI and MRS data processing/analysis

An MR physicist (FD) processed all MRI/MRS data blind to diagnosis. Post-processing of the free induction decays including apodization, Fourier transformation, frequency, phase, and eddy current correction of individual spectra in the DTS experiment, as well as calculation of MTR and DTS constants were carried out using software provided in the Varian Console and home-grown software running on MATLAB. Note that MTR and DTS measurements depend on relative signal change with saturation RF pulse or diffusion gradient, respectively. We digitized the water or NAA signal (resonance peak area) and normalized it to baseline, i.e. to the signal without RF saturation in MTR or to that without diffusion gradients in DTS. The units for ADC, RD, and AD are mm²/second x10⁻³.

We collected Creatine (Cr) and Choline (Cho) data along with NAA in our DTS studies. The SNR is lower for these metabolites than for NAA but it was possible to carry out analyses of Cho in our DTS data. This is valuable because Cho is compartmentalized differently from NAA, i.e. more of it is found in astrocytes than neurons (46). If Cho and NAA ADC show differential patterns in SZ, this would support the neuron—selective significance of NAA ADC.

Statistical approach

All analyses were carried out using SPSS (V.18). The statistical plan had three stages: tests of data quality, tests of our main hypotheses, and exploratory tests of associations between multiple variables in the dataset. First, two-sample t-tests and chi-square tests compared sample characteristics and signal-to-noise ratio (SNR) for the 135 ms DTS-PRESS spectrum (MRS data quality measure) across groups.

Second, we entertained our two main hypotheses: (1) that there would be a reduction in MTR and (2) NAA ADC may be abnormal in schizophrenia. These hypotheses were tested using two general linear models with MTR and NAA ADC as outcomes and diagnosis as predictor. Given the richness of the data we collected, we carried out parallel secondary analyses on NAA RD, AD and FA, and water RD, AD, and FA. Since age and smoking can impact white matter health, we reran the main analyses with these variables as covariates.

Third, we carried out a series of correlation analyses using Pearson’s coefficients (or Spearman where specifically mentioned for variables with skewed distribution). We examined correlations between the various diffusion variables in order to detect any possible structured covariance. We also examined the relationship between MTR and NAA ADC with age, education level, and BMI for the full dataset, and MTR and NAA ADC with duration of illness, lifetime number of suicide attempts, lifetime number of hospitalizations, NAART score, MCAS score, CPZ equivalents, PANNS, YMRS, and MADRS scores for the SZ group. In addition, we carried out ANOVAs with sex and race as independent variables and MTR or NAA ADC as the dependent variable. We did not control for multiple comparisons in any of these exploratory analyses because our goal was to allow detection of even modest relationships so they could be pursued in future studies. We were willing to accept the risk of Type-I error inherent in this approach because these didn’t concern our primary hypotheses and we did not have adequate power in this small clinical sample to correct for multiple comparisons.

RESULTS

See Table 1 for demographic and clinical variables; the two groups were well-matched with the exceptions usually noted in samples of patients with schizophrenia: BMI and participant educational attainment. In order to assess the reliability of our measures, we first carried out a test-retest study (see Supplementary Materials).

Magnetization Transfer Ratio (MTR) Spectroscopy

MTR measurements are described in Figure 2. Data from a phantom aqueous NaCl-solution showed a very low MTR of <1%, a face-valid finding because there is no “bound” water in an aqueous solution. In the human brain, “bound” water molecules (i.e. those interacting with lipids and proteins) cause a loss of water signal intensity, leading to a non-zero MTR. There was a significant reduction in MTR in SZ as compared with healthy controls [F(36,1)=5.339, p=0.027] and this remained when age and smoking were added as covariates [F(36,1)=5.682, p=0.023] (Table 2).

Table 2.

MTR and DTS data summary.

	Normal Control	Schizophrenia	Statistical Evaluation	Effect Size (Cohen’s d)
MTR	0.17±0.02	0.15±0.03	F(36,1)=5.339, p=0.027	0.78
NAA RD	0.15±0.04	0.17±0.05	F(41,1)=1.674, p=0.203	0.44
NAA AD	0.33±0.09	0.40±0.10	F(41,1)=4.189, p=0.047*	0.74
NAA ADC	0.21±0.05	0.25±0.05	F(41,1)=6.348, p=0.016*	0.80
NAA FA	0.48±0.16	0.52±0.16	F(41,1)=0.696, p=0.409	0.25
Water RD	0.53±0.06	0.59±0.10	F(41,1)=4.865, p=0.033*	0.73
Water AD	0.76±0.09	0.87±0.14	F(41,1)=8.417, p=0.006*	0.93
Water ADC	0.61±0.05	0.68±0.11	F(41,1)=7.687, p=0.008*	0.82
Water FA	0.24±0.11	0.25±0.07	F(41,1)=0.422, p=0.519	0.11

Statistical analyses are described in the text.

^*p<0.05

DTS

DTS measurements in a phantom preparation ([NAA] = 12 mM, [Cr] = 8 mM and [Cho] = 3 mM, pH=7.0) demonstrated isotropic diffusion, as expected in a structure-free medium (FA ≤ 0.08 for all three chemicals), and ADCs were in good agreement with published values (0.64±0.08, 0.74±0.06 and 0.92±0.06 for NAA, Cr and Cho, respectively) (44). The spectra obtained during a typical in vivo DTS experiment are shown in Figure 3. Note that we show water-suppressed spectra for simplicity, although we also collected water-unsuppressed spectra for calculation of water diffusion. See Supplementary Materials for discussion of SNR in our DTS data.

Figure 3.

Sample water-suppressed MRS spectra showing the modulation of metabolite signal with diffusion gradients. The top row shows data from a typical healthy control, the bottom row from a schizophrenia patient. In each row, the leftmost spectrum is with no gradients applied, and the next 6 spectra show a variety of x/y/z gradients as shown. Note the variable decrement in metabolite signal with differing gradients, giving rise to the calculation of the diffusion tensor.

NAA ADC was significantly elevated in SZ when compared with controls [F(41,1)=6.348, p=0.016] and this was true when age and smoking are added to the model [F(41,1)=5.500, p=0.023]. This measure was not correlated with MTR (R= −0.242, p=0.156). Water ADC was also elevated in SZ [F(41,1)=7.687, p=0.008] but not correlated with MTR (R= −0.129, p=0.447). For reasons discussed below, we did not consider FA a primary outcome (NAA FA and water FA in SZ were not significantly different from control in this study; see Table 2 for details). In addition, Cho ADC was not significantly different between healthy control and SZ groups (0.20±0.06 and 0.22±0.05, respectively; [F(37,1)=0.227, p=0.637]).

As expected, there were numerous statistically significant correlations (not shown) among the NAA diffusion parameters as well as among the water diffusion parameters (exceptions were NAA RD-AD, NAA FA-ADC, and water FA-ADC). By contrast, there were no correlations between any water and any NAA diffusion parameters, suggesting these measures were at least partially independent although they share some common mechanisms (e.g. intra-axonal water and NAA diffusion).

MTR, DTS, and Demographic/Clinical Variables

MTR correlated negatively with BMI (R= −0.449, p=0.005) and positively with education level (R=0.440, p=0.006) in the full dataset. It also correlated with number of lifetime suicide attempts (Spearman R= −0.553, p=0.014) and number of lifetime hospitalizations (R= −0.620, p=0.005) among patients (Supplementary Figure 1). NAA ADC correlated negatively with NAART score (R= −0.462, p=0.040) and education level (R= −0.393, p=0.010). By contrast, water ADC was not correlated with NAART score (R=0.309, p=0.173) or education level (R= −0.141, p=0.367). Because the two groups were not matched for BMI and education we also ran the MTR/BMI (R= −0.591, p=0.008), MTR/education (R= −0.118, p=0.592), and NAA ADC/education (R= −0.544, p=0.011) correlations within the SZ group only. No other correlation analysis was statistically significant, despite the liberal approach of not correcting for multiple comparisons.

DISCUSSION

We applied a combined MTR-DTS approach to probe microstructural WM abnormalities in chronically ill SZ patients. We implemented several data quality measures: a test-retest study in healthy controls, phantom calibration of each human acquisition, frequency/phase/eddy current correction of individual diffusion spectra, and SNR calculation for individual spectra. We found that MTR was reduced and NAA ADC was elevated in SZ, suggesting that WM pathology in SZ is driven by both myelination deficits and axon abnormalities.

Several additional features of the DTS data lend face validity to our findings. For example, the water FA and ADC and the NAA FA and ADC values for healthy controls are similar to those observed in other DTI and DTS studies (e.g. (27, 47)). NAA is a larger molecule, diffuses more slowly than water, and is predicted to have a lower AD; this is exactly what we observe. In addition, since NAA is intracellular, it is predicted to have more anisotropic diffusion than water, and we find that NAA FA is higher than water FA in healthy controls. However, note that FA is sensitive to noise and the NAA resonance has lower signal-to-noise than that of water. Finally, our finding of reduced MTR is consistent with the elevated water RD in SZ since the latter is proposed as an indicator of myelin reduction.

The clinical significance of these findings was highlighted by the fact that low MTR was associated with worse educational attainment, markers of a more severe phenotype (lifetime hospitalizations and suicide attempts), and higher BMI; while higher NAA ADC was associated with lower NAART scores and educational attainment. Thus, for each measure, the direction of change in patients compared to controls (reduced MTR, elevated NAA ADC) was also associated with more adverse outcomes within the patient group. In addition, we did not see similar correlations with water ADC, suggesting that there is additional value in measuring NAA ADC. This suggests the WM abnormalities we observed in schizophrenia may be related to factors that determine functional outcomes in this condition. Although we cannot currently propose a mechanism for these effects, future studies can examine the relationship between specific cognitive functions or clinical symptoms and MTR/DTS in specific WM tracts.

Our findings suggest that both myelination (measured by MTR) and axonal abnormalities (measured by DTS) play a role in WM abnormalities in SZ. One interpretation consistent with our hypotheses is that myelin is reduced in SZ accompanied by an increase in intra-axonal space available for diffusion (increase in axonal diameter or reduced hindrance within axons). The axon-selectivity of the NAA ADC findings is further supported by the absence of similar changes in Cho ADC. The myelin and axon changes would act in concert to lead to abnormal signal transduction between brain regions in SZ. In addition, they would each reduce the anisotropy of water molecule diffusion, leading to well-documented DTI abnormalities. Consistent with this assumption, we also observe changes in water diffusion along with those in MTR and NAA ADC. The simultaneous elevation of NAA and water ADC in SZ is distinct from stroke where both measures are reduced (30). On the other hand, it is noteworthy that the MTR and DTS measures did not correlate with each other. This suggests the possibility of independent mechanisms leading to myelin and axon abnormalities. Future longitudinal studies in early stages of schizophrenia may be instructive in how these mechanisms evolve.

This interpretation suggests abnormalities in the g-ratio in schizophrenia. We cannot calculate a g-ratio from the current dataset since MTR and DTS data have different units. Therefore, we calculated NAA ADC/MTR as a simpler index. This ratio is 1.35±0.27 for controls and 1.66±0.39 for patients (mean±SD), a 24% between-group difference (p = 0.007). Although not a primary outcome of this study, we note that this difference is substantial and highly significant.

Although abnormalities in axonal geometry are one possible explanation for our DTS data, others are possible. For example, molecular diffusion properties can be affected by membrane permeability during anesthesia (48) and there may be membrane permeability abnormalities in SZ. In addition, abnormalities in the proportion of NAA localized in mitochondria, or in NAA cleavage in extracellular space as part of myelin synthesis (49) may also impact NAA ADC. Since mitochondrial abnormalities are reported in schizophrenia (50), abnormal NAA distribution in mitochondria is an attractive alternative hypothesis to be pursued.

The relationship between NAA diffusion and transverse T2 relaxation is intriguing because each measure reflects a related but distinct aspect of NAA’s microenvironment. While T2 relaxation is determined by spin-spin interactions between the index molecule and its immediate neighbors, diffusion reflects the distance traveled by a molecule in unit time. We recently reported, in a dataset partially overlapping with the current one, that NAA T2 relaxation time is shortened in SZ while that of water is prolonged (51). This contrasts with the current findings of elevated NAA and water ADC in SZ. This pattern suggests that WM microenvironment changes in SZ may be more complex than only axonal geometry changes and involve both greater NAA diffusion and more frequent interaction with other molecules. In addition, there is no significant correlation between NAA ADC and NAA T2 in our data (R=−0.214; p=0.218) suggesting these measures reflect independent processes. Deeper insight into this issue would come from analysis of NAA T2 relaxation for multi-exponential decay. Multiple-component T2 relaxation may offer a clue that NAA signal arises from molecules in different environments. Here, we calculated T2 times based on 4 echo times, not enough to explore multi-exponential decay. Future studies with more detailed T2 relaxation data are needed.

Although DTS capitalizes on NAA diffusion in the same manner DTI does on water diffusion, brain NAA concentration is 5000-fold lower than brain water concentration (approximately 10mM and 50M, respectively). Thus, while DTI can achieve whole-brain coverage with milimetric voxels in minutes, we collect data from a single large voxel over many repetitions to obtain reliable DTS data. The large DTS voxel can be associated with macroscopic curvature effects: as axons course through the voxel they may curve and the DTS signal averaged over a large volume partly reflects curvature as opposed to diffusion. To address this confound, the results of a DTS study can be examined for macroscopic curvature effects. If macroscopic curvature effects are operative, RD and AD should covary (e.g., curved fibers would yield high RD and low AD, while straight fibers would yield low RD and high AD). If RD and AD vary independently of one another (as is the case with NAA), this suggests the findings are not secondary to macroscopic curvature. In addition, macroscopic curvature affects water and NAA parameters equally while biologically-specific changes cause independent variation in these parameters. The absence of covariation between NAA and water parameters in our study is reassuring in this regard. Nonetheless, we focused on ADC as the DTS parameter least affected by macroscopic curvature. This is because ADC has no directionality while all other DTS measures do.

In addition to these conceptual caveats, our study has several limitations. First is the potential for variable voxel placement, which could result in the inclusion of axons of different diameters in different brains, impacting the ADC measures. Our voxel is anchored by anatomical landmarks which improve reliability as demonstrated by our test-retest study. Collecting data from the entire brain is currently possible for MTR but not for DTS. Chemical shift imaging can collect high quality MRS data from the entire brain (52, 53) but these are challenging to implement with diffusion gradients. Second, the NAA signal we measure contains contributions from NAA and N-acetylaspartylglutamate (NAAG). NAAG is located both intra- and extracellularly (54) and our DTS measures may be confounded by this contamination. NAAG is similar to NAA in chemical structure so the two MRS signals are challenging to resolve. NAAG concentration in the PFC white matter is 1.5mM in healthy individuals (of which an unknown fraction is extracellular) (55) whereas NAA concentrations are usually calculated at 10mM (56). Therefore, we do not expect NAAG contribution to be a major factor in our results. Third, MTR is not a specific measure of myelin content. MTR abnormalities can arise from acute inflammation, edema, and other processes that impact brain water content (57). This limits the utility of MTR in pathologies where gross brain water abnormalities are seen. There is no evidence for such abnormalities in SZ and past applications of MTR in SZ have revealed subtle changes (20, 22). Others have used a “myelin water fraction” approach which takes advantage of the differential T2 relaxation properties of water molecules trapped within myelin (58) but this approach is technically challenging. Fourth, the SZ patients in this study were chronically ill and taking medications. DTI abnormalities are widely reported in chronically ill patients, in fact more consistently than in first-episode patients with SZ (59) or those with other psychiatric diagnoses (60, 61). In addition, we are not aware of any literature on psychotropic medications causing alterations in WM microstructure; there was no relationship between CPZ equivalents and MTR or NAA ADC in our study. Nonetheless, we cannot rule out medication effects; this needs to be addressed in future studies. Finally, we did not correct for cardiac gating effects which can impact the MRS signal. Since our MRS sequence involved interleaved acquisition, such effects were unlikely to impact the findings.

In conclusion, we used a novel MRI/MRS approach to probe WM abnormalities in SZ and provided in vivo evidence for both abnormal axon geometry and myelination. Our findings suggest that signal transduction speeds are abnormal in SZ, possibly leading to information processing abnormalities and cognitive deficits. Future studies will focus on early stages of illness and on abnormalities in specific WM pathways with specific cognitive or clinical presentations.