A prospective longitudinal volumetric MRI study of superior temporal gyrus gray matter and amygdala–hippocampal complex in chronic schizophrenia

,


Introduction
Abnormalities in superior temporal gyrus (STG) gray matter and amygdala-hippocampal complex (AHC) are among the most consistently documented findings of schizophrenia in structural magnetic resonance imaging (MRI) studies (Honea et al., 2005;Shenton et al., 2001;Wright et al., 2000). Functionally, the STG includes primary and secondary auditory cortices and a language-related area, while the amygdala and hippocampus are critical for memory functions (Squire and Zola, 1996) and affect perception (Morris et al., 1996). Volume reduction in STG has been associated with auditory hallucinations (Barta et al., 1990) and thought disorder (Shenton et al., 1992) while volume reduction in AHC has been associated with negative symptoms (Anderson et al., 2002). The significance of these structures has been also highlighted by functional MRI reports of STG and AHC abnormalities in auditory perception, language, and memory tasks (Blakemore and Frith, 2000).
Several prospective longitudinal studies suggest progression of gray matter volume reduction in schizophrenia in the period just following onset (Nakamura et al., 2007). Greater gray matter decline in left posterior STG was found in schizophrenia patients 1.5 years after first hospitalization but not in patients with affective psychosis and controls (Kasai et al., 2003a). Progressive volume reduction of temporal structures has been also reported for whole temporal lobe (DeLisi et al., 1995), as well as Heschl's gyrus/planum temporale gray matter (Kasai et al., 2003b). Conjoint progression of Heschl's gyrus volume loss and mismatch amplitude decline  in first episode schizophrenia has also been reported. Moreover, progression in temporal lobe has been reported for patients with childhood-onset schizophrenia (Gogtay et al., 2004;Rapoport et al., 1999). In addition, reduction in left and right hemispheres and right cerebellum have been reported (DeLisi et al.,1997). However, other studies failed to find progressive volume reduction in temporal lobe structures, such as in temporal lobes and hippocampus (DeLisi et al., 1997), temporal lobe and STG (DeLisi and Hoff, 2005), hippocampal and temporal lobe volumes (Wood et al., 2001), total brain volume, hippocampus and amygdala (Whitworth et al., 2005), and hippocampus and total cortex (J. Lieberman et al., 2001).
The presence of progressive volume loss in chronic schizophrenia is even more controversial with some studies finding progressive changes and others failing to find them (see Table 1 for a comprehensive review of studies and regions involved). Among them, only two studies showed progressive volume changes in STG in chronic schizophrenia (Mathalon et al., 2001;van Haren et al., 2007).
Thus, the primary purpose of this study was to evaluate prospectively the presence of a progressive volume loss in temporal structures in chronic schizophrenia. We investigated gray matter volume of STG and AHC in chronic schizophrenia patients using high spatial resolution contiguous MRI sections and highly reliable manual region of interest measurements of these brain regions. The second purpose was to elucidate the relationship between volume changes and clinical symptoms as measured with the Positive and Negative Symptoms Scale (PANSS).

Subjects
Sixteen male patients with chronic schizophrenia and 20 healthy male comparison subjects participated in this study. Patients were recruited from the VA Boston Healthcare System-Brockton Division. The comparison subjects were recruited through newspaper advertisements. After a complete description of the study, written informed consent was obtained from all participants.
Demographic data for subjects in each group are presented in Table 2. Exclusion criteria for both groups were: 1) neurological illness or major head trauma, 2) previous treatment with electric convulsive therapy (ECT), 3) any history of alcohol or drug dependence, or 4) alcohol or drug abuse within the past 5 years. The age range for inclusion was 24 to 54 years. Comparison subjects were screened with the Structured Clinical Interview for DSM-III-R (SCID), nonpatient edition, by trained interviewers (M.E.S, P.N). No comparison subjects had an Axis I psychiatric disorder in themselves or a first degree relative. All patients were diagnosed with schizophrenia based on a SCID interview and a chart review by the same interviewers. All patients were receiving neuroleptic medication (typical, N=3; atypical, N=11; both, N=2). The mean dose was 392.8mg/day (SD=330.9) in chlorpromazine equivalents.

Clinical evaluations
Handedness was assessed with the Edinburgh Inventory (Oldfield, 1971). Subjects' and parental socioeconomic status were measured by the Hollingshead two-factor index (1= highest, 5=lowest). All subjects were given PANSS at initial scan (time 1) but not at rescan (time 2) ( Table 2). In the present study, we focused on correlations between regions of interest and clinical positive and negative symptoms.

MRI image acquisition and processing
MR images were acquired with a 1.5-T General Electric scanner (GE Medical Systems, Milwaukee) at the Brigham and Women's Hospital in Boston. Imaging methods have been described in detail elsewhere (Wible et al., 1995). The same acquisition protocol was used at time 1 and time 2. The acquisition protocol included two MRI pulse sequences. The first sequence resulted in contiguous spoiled gradient-recalled images (repetition time=35 ms, echo time=5 ms, one repetition, 45° nutation angle, 24-cm field of view, number of excitations=1, matrix=256×256 [192 phase-encoding steps]×124). Voxels were 0.9375×0.9375×1.5 mm. Data were formatted in the coronal plane and analyzed as 124 coronal 1.5-mm-thick slices. The second acquisition sequence resulted in axial series of contiguous double-echo (proton density and T 2 -weighted) images (repetition time=3000 ms, echo time=30 and 80 ms, 24-cm field of view, and an interleaved acquisition with 3.0-mm slice thickness). The voxel dimensions were 0.9375×0.9375×3.0 mm. This latter pulse sequence was used to measure the volume of the total intracranial contents (brain, CSF, connective tissue, and blood vessels). An anisotropic diffusion filter (Gerig et al., 1992) was applied to both spoiled gradient-recalled and T 2 images to reduce noise prior to processing. The intensity information from both the spoiled gradient-recalled and T 2 images was used in a fully automated segmentation program to classify tissue into gray matter, white matter, and CSF. An iterative expectationmaximization algorithm estimated image intensity inhomogeneities, applied intensity corrections on the basis of these estimates, and then classified tissue on the basis of the same set of signal intensity parameters for all subjects (Wells Wet al., 1996). Images were realigned by using the line between the anterior and posterior commissures and the sagittal sulcus to correct head tilt and then were resampled into isotropic voxels (0.9375×0.9375×0.9375 mm 3 ). Manual drawings of regions of interest were performed on the realigned and resampled coronal slices.

Regions of interest
The STG and AHC were delineated manually on a work-station by a rater who was blinded to the diagnosis and the time of scan (initial scan or rescan) with the established procedure described in our previous report (Hirayasu et al., 1998) (Fig. 1). Three raters (TY, MN, KL) who were blind to group membership independently drew the regions of interest on five subjects, resulting in a high interrater reliability (left and right anterior/posterior STG, intraclass correlation coefficients (ICC)=0.981/0.995 (left anterior/posterior) and 0.985/0.990 (right anterior/posterior); left and right anterior/posterior AHC, ICC=0.982/0.990 (left anterior/ posterior) and 0.980/0.990 (right anterior/posterior)).

Statistical analyses
We used t-tests to assess group differences at initial scan (time 1) and rescan (time 2) in age, handedness, socioeconomic status, parental socioeconomic status, education, WAIS-III, interscan interval, and total intracranial contents. The interscan interval (time between time 1 and time 2) was significantly different between groups (Table 2).
2.5.1. Group differences in relative volume at time 1 and time 2 scans and volume change over time-Cross-sectional group differences at time 1 and time 2 in regions of interest were evaluated using relative volumes ([absolute volume of region of interest]/[intracranial contents]× 100 (%)) we used repeated measures analysis of variance (ANOVA) with group (schizophrenia or healthy comparison) as between-subject factor, and region (STG or AHC), subdivision (anterior or posterior), and side (left or right) as withinsubjects factors. We evaluated volume change in regions of interest over time by using the percent of change as the dependent variable. Percent volume change was calculated with the following formula: ([absolute volume at time 2]−[absolute volume at time 1])/[absolute volume at time 1]×100 (%). An analysis of covariance (ANCOVA) with interscan interval as a covariate, group (schizophrenia or healthy comparison) as between-subject factor, subdivision (anterior or posterior) and side (left or right) as within-subjects factors was performed for STG and AHC, respectively. Finally, a regression analysis was performed to examine the effect of time between two scans on the percent volume change of the anterior/posterior STG and anterior/posterior AHC for both schizophrenia patients and healthy comparisons.

Correlations between percent volume change or absolute volume and
clinical measures-Spearman's correlations were used to examine correlations between percent change and clinical symptom scales. Spearman's correlations were also used to examine correlations between absolute volume of time 1 and clinical symptom scales. In these exploratory analyses, we used p<0.05 as the cutoff value for statistical significance for each region.

Demographic and neuropsychological group differences
There were no significant group differences in age, handedness, and parental socioeconomic status (Table 2). Patients had lower socioeconomic status, education, WAIS-III total score relative to comparison subjects, which is consistent with a functional deterioration brought by the disorder.

Cross-sectional differences in relative volume at time 1 and time 2
In the ANOVA with group, region (total STG or total AHC), subdivision (anterior or posterior), and side (left or right) for relative volume at time 1 and time 2, the groups were significantly different at both time points (time 1: F=15.51, df=1, 34, p<0.0001, time 2: F=18.30, df=1, 34, p<0.0001). Interaction of group by region by subdivision was significant both at time 1 and time 2 (time 1: F=7. 15, df=1, 34, p=0.01, time 2: F=5.78, df=1, 34, p=0.02). To follow-up on this interaction, we performed ANOVA with group, subdivision, and side separately for STG and AHC at time 1 and time 2. Interaction of group by subdivision was significant only in STG at both time points (time 1: F=4.69, df=1, 34, p=0.04, time 2: F=6.28, df=1, 34, p=0.017). In ANOVA with group for anterior and posterior STG, the main effect of group was significant only in the posterior subdivision at both time points (time 1: F=7.07, df=1, 34, p=0.01, time 2: F=8.08, df=1, 34, p=0.008) but not significant in the anterior subdivision. These results indicate the reduced volume of bilateral posterior STG in schizophrenic patients at both time 1 and time 2 (Table 3, Fig. 2a). In AHC, we performed ANOVA with group, subdivision, and side both at time 1 and time 2. The groups were significantly different at both time points (time 1: F=22.81, df=1, 34, p<0.0001, time 1: F=22.95, df=1, 34, p<0.0001). There were no significant interactions of group by subdivision at time 1 and time 2 (Table 3, Fig. 2b).

Volume change
In STG, ANCOVA conducted on percent volume change over time with interscan interval as a covariate and group, subdivision, and side as between-and within-subjects factors did not show significant group differences (F=0.01, df=1, 34, p=0.91) or significant interactions (Table  3, Fig. 3a). In AHC, ANCOVA conducted on percent change over time with interscan interval as a covariate and group, subdivision, and side showed no significant group differences (F=0.18, df=1, 34, p=0.67) or significant interactions (Table 3, Fig. 3b). We also performed one factor ANCOVAs on both STG and AHC (with volumes collapsed for left and right sides) with interscan interval as covariate and group as between-subject factor. Neither region showed significant main effect of group. Finally, we performed one factor ANCOVAs on both STG and AHC (with volumes for left and right sides separately) with interscan interval as covariate and group as between-subject factor. Only right posterior AHC reached a trend level (F=3.07, df=1, 34, p=0.09) ( Table 3, Fig. 3b).
In the regression analyses for schizophrenia and normal controls, there were no regions which showed a significant relationship between volume change and interscan interval in either the schizophrenia or normal control group. Moreover, ANOVA with group and interscan interval to compare regression slopes between the two groups did not show significant interactions of group by interscan interval in any of the regions.

Correlations between percent change or absolute volume and clinical measures
There were no significant correlations between percent change in the anterior, posterior, and total regions in STG and total clinical symptom scales on PANSS. The scores of total negative PANSS negatively correlated at trend level (r s =−0.48, p=0.06) with percent change of left anterior AHC. Total positive PANSS scores negatively correlated with absolute volume of time 1 in the left total STG (r s =−0.59, p=0.02). In addition, hallucinatory behavior in PANSS (r s =−0.74, p=0.001) significantly negatively correlated with absolute volume of time 1 in the left total STG (Fig. 4).

Discussion
No statistically significant progressive volume changes over a three-year scan time interval in STG and AHC in chronic schizophrenia were found in this study despite using the same region of interest definitions and methodology as in Kasai et al. (2003a) that reported significant progressive volume loss in the left STG over 1.5 year in first episode patients. Thus, given that significant volumetric reductions were not detectable in the period twice as long as that used in the first episode study, this result strongly suggests a greater rate of volume loss in first episode relative to chronic schizophrenia patients. The small volume change (all ROI reductions were less than 3.0%) and small to medium effect sizes (0.006-0.56) across all the regions imply that, at least for the three-year interval used in this study, putative volume reductions, if present, happen at a very slow rate. It may be the case that reductions of this magnitude are very difficult to detect in small ROI such as STG and AHC although they may be easier to detect with larger brain and CSF regions (DeLisi, 2008;Hulshoff Pol and Kahn, 2008). The long duration of illness in the subjects in the present study may have been an additional reason for the small volume changes reported here.
These different rates of progression have important implications for our understanding of brain changes in schizophrenia. Thus, for the STG, volume changes seem to be large initially after onset and then become relatively static in a chronic phase. If this scenario is correct, it would suggest the importance of early interventions insofar as pharmacotherapy and psychosocial treatment can mitigate progression.
The presence of significant cross-sectional relative volume reduction in the bilateral posterior STG at both time 1 and time 2 reported in this study is in line with previous results (Anderson et al., 2002;Shenton et al., 2001;Wright et al., 2000). We also found significant volume reduction in amygdala and hippocampus in schizophrenia in concordance with several previous studies. On the other hand, some post mortem studies did not show volume reduction in schizophrenia. However, post mortem findings could be complicated by several aging factors.
Volume change over time reached trend level in the right posterior AHC, which may imply a slow progressive volume loss in chronic schizophrenia. A recent volumetric MRI study reported significant association between right hippocampal volume reduction and total PANSS change in chronic schizophrenia, suggesting the possibility of progressive reduction of right hippocampus (Panenka et al., 2007). Only a longitudinal study with multiple measurement points can address this possibility and such a study is currently in progress in our laboratory.
We note that limitations of this study include a small sample size, only male participants, only two time scan points, and different interscan intervals for the patient and control groups. While we used interscan interval as a covariate to compensate statistically for the group difference in the interscan intervals, having the same interscan interval for both controls and patients would have been more desirable. In addition, atypical antipsychotic administration has been reported to be associated with changes in gray matter volume in humans, both with a possible neuroprotective effect (Konradi and Heckers, 2001;J. A. Lieberman et al., 2005;Nakamura et al., 2007) and also with a possible increase in gray matter loss reported in humans (Cahn et al., 2002) and in animals (Dorph-Petersen et al., 2005). In our study, 13 out of 16 schizophrenia patients used atypical neuroleptic drugs. Finally, because we did not measure whole brain volume or whole brain gray matter, nor any ROI other than STG and AHC, we cannot state whether cross-sectional volume reductions are specific to these ROIs, nor can we be sure whether our findings are specific to the amygdala or hippocampal component of the AHC.
The precise neurobiological mechanism of structural abnormalities and their progression in schizophrenia is unclear. However, a growing body of evidence implicates abnormal interaction of neurons using excitatory amino acid neurotransmission and those using GABAergic neurotransmission (Lewis and Gonzalez-Burgos, 2006). Although still controversial, this reduction of GABAergic neurotransmission and resulting increased excitation is regarded as a possible cause of neuropil (dendrites and synapses) volume reduction. Interestingly, increased levels of glutamine in medial prefrontal cortex (Bartha et al., 1997;Olbrich et al., 2008) and anterior cingulate cortex and thalamus (Theberge et al., 2002(Theberge et al., , 2007, glutamate in dorsolateral prefrontal cortex and hippocampus (van Elst et al., 2005), and cerebrospinal fluid (Hashimoto et al., 2005)were found in first episode and early schizophrenia, while lower levels of glutamine and glutamate in anterior cingulate and thalamus (Theberge et al., 2003) and glutamate/glutamine in dorsolateral prefrontal cortex (Ohrmann et al., 2007(Ohrmann et al., , 2005 were found in chronic schizophrenia. These different levels of glutamatergic dysfunction in several regions at different stages in schizophrenia may explain the difference between progressive volume change in the early stage of schizophrenia and a lack of such changes in the chronic stage. Thus, in temporal lobe structures, the pathophysiological processes may differ between the early and late stages after the onset of psychosis in schizophrenia. Consistent with this hypothesis, there were reports of stable longitudinal mismatch negativity (MMN) deficits in test-retest measurements over 1-2 years duration in patients with chronic schizophrenia (Light and Braff, 2005). This contrasts dramatically with progression of deficits over 1.5 years in a longitudinal study of patients with initial measurements at first hospitalization and time 2 measurements 1.5 years later . Moreover, the Salisbury et al. (2007) study found the change in MMN to be correlated with volume reduction in Heschl's gyrus, consistent with evidence of a MMN generator in STG (Alho, 1995;Naatanen, 1992;Sabri et al., 2004), and with NMDA as a primary neurotransmitter contributing to MMN generation (Javitt et al., 1996;Kreitschmann-Andermahr et al., 2001;Umbricht et al., 2000). The absence of progressive volumetric changes in the STG and stable MMN deficits in chronic schizophrenia may thus reflect a common pathophysiological basis of increased GABAergicglutamatergic dysfunction in STG during the early course of schizophrenia and less in chronic schizophrenia.
In clinical correlations, the left anterior AHC volume reduction, mainly including left amygdala, correlated with total negative PANSS scores at trend level. This area has been reported to be related to negative symptoms of schizophrenia (Rajarethinam et al., 2001). It is unclear why, in our study, negative symptoms inversely correlated only with volume change over time, but not with relative or absolute volume. However, the present findings suggest that the volume change of the left amygdala over time may contribute to negative symptoms during a chronic course of schizophrenia. Smaller left total STG was associated with severity of hallucinatory behavior. Left STG has been reported to be associated with auditory hallucinations in schizophrenia in studies using several functional modalities (Copolov et al., 2003;Hoffman et al., 2003;Ropohl et al., 2004;van de Ven et al., 2005). Absolute volume of the left total STG also negatively correlated with total positive PANSS scores indicating that volume reductions of left STG may contribute to global positive symptoms in chronic schizophrenia.
In summary, progressive volume change of the STG and AHC was not found in chronic schizophrenia in our study suggesting that the absence of further volume reduction over time may characterize the chronic course in schizophrenia, in contrast with peri-onset period. In addition, percent change of left anterior AHC was correlated at trend level with negative symptoms, which suggests an important role of this region in pathophysiology of negative symptoms in chronic schizophrenia. The correlations between positive symptoms and the volume of left STG highlighted its possible involvement in hallucinations. Further studies with a follow-up over a longer period of time are needed to clarify the nature of course in schizophrenia.

Abbreviations
STG, superior temporal gyrus; AHC, amygdala-hippocampal complex.  Relative volume of the STG and AHC by hemisphere in patients with schizophrenia and healthy comparison subjects at time 1 and time 2 (horizontal black lines indicate means).  Correlation between percent change (left panel) or absolute volume at time 1 (middle and right panels) and clinical measure.   Relative volume of regions of interest at time 1 and time 2 and percent change in schizophrenia patients and healthy comparison subjects.