Redox Dysregulations in Schizophrenia Revealed by in vivo NAD+/NADH Measurement
Sang-Young Kim1,2, Bruce Cohen3, Scott Lukas4, Cagri Yuksel2, Dost Ongur2, and Fei Du1,2

1McLean Imaging Center, McLean Hospital, Harvard Medical School, Belmont, MA, United States, 2Psychotic Disorders Division, McLean Hospital, Harvard Medical School, Belmont, MA, United States, 3Program for Neuropsychiatric Research, McLean Hospital, Harvard Medical School, Belmont, MA, United States, 4Behavioral Psychopharmacology Research Laboratory, McLean Hospital, Harvard Medical School, Belmont, MA, United States

Synopsis

In this work, we demonstrated the feasibility of 31P MRS-based in vivo intracellular redox state quantification at 4 T. We applied this novel method to investigate oxidative stress in the frontal lobe of chronic and first-episode SZ as well as first-episode BD patients. We found evidence for striking Rx reductions in SZ, with every chronic SZ patient showing an Rx of at least one standard deviation below the control mean. Rx reduction was of even greater magnitude among first-episode SZ patients. This study illustrates the power of examining in vivo brain redox dysregulation (measured as Rx) in psychiatric disorders.

Purpose

Balance between the redox pair of nicotinamide adenine dinucleotides (oxidized NAD+ and reduced NADH), reflects the oxidative state of cells and the ability of biological systems to carry out energy production. A growing body of evidence suggests that an “immuno-oxidative” pathway including oxidative stress and mitochondrial dysfunction may contribute to disruptions in brain activity in schizophrenia (SZ). The aim of this study is to assess possible redox imbalance in SZ patients by using novel in vivo 31P MRS technique.

Method

The participants included 35 healthy controls, 21 chronic SZ, 12 first-episode SZ, and 16 bipolar (BD) patients. All participants underwent diagnostic imaging at a 3 T and 31P MRS measurements were performed on a 4 T MR scanner. The spectral model for NAD is previously described1. In-house MATLAB software (version 7.9, R2009b) was developed for simulating the 31P spectra containing NAD+, NADH and a-ATP resonance. In addition, Monte Carlo simulations were performed to assess the accuracy of our NAD quantification in vivo. All 31P spectra were from a 6*6*4 cm voxel in the frontal lobe, acquired using a 200-µs hard RF pulse with outer volume saturation2, and its flip angle (nominal 90°) was adjusted to achieve an optimal NMR signal. All NAD analyses were extracted from a study of creatine kinase and ATP synthetase reaction rates2 with a magnetization saturation transfer strategy, where 31P spectra were acquired in the absence (control) and presence (6 spectra with varying saturation times) of a saturating pulse train centered on the r-ATP frequency. The saturation pulse train (bandwidth 140 Hz) has no effect on the spectral region containing NAD+ and NADH. For NAD+ and NADH quantification, two unknown parameters (i.e., signal intensity and line-width of a-ATP, NAD+, and NADH) were determined using nonlinear least-square fitting of the model output to the resonance signals of NAD+, NADH and a-ATP obtained from in vivo 31P MRS experiments. The fitting errors were calculated by dividing standard deviation of residual signal by mean value of fitted signal. The concentrations were determined using a-ATP signal as an internal reference, in which its brain concentration was set to 2.8 mM1,3. Thus, intracellular redox state (i.e., Rx=NAD+/NADH) could be calculated from the concentrations of NAD+ and NADH.

Results and Conclusions

We demonstrated excellent agreement between simulated and experimentally measured 31P spectra and Monte Carlo simulations with different a-ATP SNR values suggest excellent fitting accuracy for SNR (>40) used for processing our in vivo data. Figure 1a presents in vivo 31P spectra from a healthy control participant (top) and a chronic SZ patient (bottom), showing excellent spectral quality and high SNR. Figure 1b shows a magnified version of the relevant spectral region (stippled green box in Figure 1a) with NAD+ and NADH fitting results and residual signal from the same participants [healthy (left) and chronic SZ (right)]. Overall fitting error was less than 5 % and comparable between groups. Figure 2a displays box-and-whisker plots showing in vivo Rx data in chronic SZ patients and matched healthy participants. Figure 2b shows in vivo Rx data in first-episode SZ patients (N=12) and age-matched (i.e., young) controls (N=15) as well as patients with a first-episode of bipolar disorder (BD, N=16). Rx was significantly reduced by 35% in chronic SZ. This finding was driven by a 47% NADH elevation in chronic SZ with no significant change in NAD+. In addition, first-episode SZ patients had a highly significant 46% reduction in Rx compared with healthy controls as well as a more modest 23% reduction compared with first-episode BD patients. Figure 3a shows summed patient and control spectra and the difference between them. To better visualize the NAD region, we show magnified spectra for each group in range of -9 ppm to -11.5 ppm in Figure 3b, clearly demonstrating an elevated NADH signal at -10.63 ppm in chronic SZ compared to controls, without an apparent difference in NAD+. NAD measures and Rx show strong age-dependence in healthy individuals3. NADH and Rx for SZ and healthy participants are plotted against age in Figures 4a and 4b, demonstrating replication of Rx decline in healthy participants and an extension of this finding to SZ patients. In addition, the youngest participants (age<25) have lower Rx than older ones, apparently reversing the age-dependent trend. These findings provide an evidence for redox imbalance in the brain in all phases of SZ, potentially reflecting oxidative stress, and suggest that Rx may become a biomarker for treatment response and for screening of new therapeutic approaches targeting oxidative stress in SZ.

Acknowledgements

Supported by MH094594(DO), NARSAD(DO), NARSAD(FD), MH092704(FD), and Shervert Frazier Research Institute(BMC).

References

1. Lu M, et al. Magn Reson Med 2014; 71: 1959-1972.

2. Du F, et al. JAMA Psychiatry 2014; 71: 19-27.

3. Zhu XH, et al. Proc Natl Acad Sci USA 2015; 112: 2876-2881.

Figures

Figure 1. (a) In vivo representative 31P spectra from healthy participant (upper) and a chronic SZ patient (bottom), respectively, at 4T. Each major resonance is labeled. (b) By fitting the spectral region containing NAD signal, signal components for NAD+ and NADH could be quantified. Fitting results from the healthy participant are on the left and chronic SZ patient are on the right. Spectral pattern difference due to higher level of NADH in the chronic SZ patient is clearly visible.

Figure 2. (a) Box and whisker plot showing Rx in the control and chronic SZ groups. Individual data are also shown on the right. Bottom and top boundaries of each box indicate 25th and 75th percentiles, with lower and upper whiskers indicating 1st and 99th percentiles, respectively. Inside each box is a horizontal line indicating median values, and a filled square indicating mean values. (b) Box and whisker plot showing Rx in the control, first-episode SZ and first-episode BD groups.

Figure 3. (a) The summed spectra for healthy (black line) and chronic SZ group (red line) are shown after frequency alignment and normalization of peak intensity to a-ATP (an internal reference). The difference spectrum (blue line) is obtained by subtracting SZ spectra from that of controls. (b) The summed spectra magnified in the NAD region ranging from -9.0 ppm to -11.5 ppm, with the stippled vertical line marking the NADH resonance frequency at -10.63 ppm.

Figure 4. Age dependence of intracellular NADH level (a) and Rx (b) in the healthy (black filled circle) and chronic SZ groups (red filled circle). Trend lines are shown for each group in the same colors. In addition, data from first-episode SZ (red open circles) and age-matched healthy control groups (black open circles) are presented.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
4125