PHOSPHOLIPID COMPOSITION OF POSTMORTEM SCHIZOPHRENIC BRAIN BY ³¹P NMR SPECTROSCOPY

Abstract

Cell membrane abnormalities due to changes in phospholipid (PL) composition and metabolism have been implicated in schizophrenia pathogenesis. That work has generally assessed membrane phospholipids from non-neural tissues such as erythrocytes and platelets. High resolution ³¹P NMR spectroscopy was used to characterize PLs of gray matter in postmortem brain for 20 schizophrenics, 20 controls, and 7 patients with other mental illnesses (psychiatric controls). Tissues from frontal, temporal, and occipital cortices were extracted with hexane-isopropanol, and ³¹P NMR spectra were obtained in an organic-solvent system to resolve the major PL classes (based on headgroups) and subclasses (based on linkage at the sn-1 position). Surprisingly, repeated-measures multivariate analysis of variance revealed no overall differences among the groups. There were no significant differences (p<0.05) among the three groups for any individual PL subclass, including lysophospholipids. The sum of all phosphatidylethanolamine headgroups was significantly lower for schizophrenics than for controls or psychiatric controls in frontal cortex. The present results are minimally correlated with previous results for aqueous PL metabolites on these same samples. The metabolite changes measured by in vivo ³¹P MRS in schizophrenia do not appear to reflect PL concentration changes. The present results offer very little support for the phospholipid hypothesis of schizophrenia.

Introduction

Cell membrane abnormalities associated with changes in phospholipid (PL) composition and metabolism have been implicated in schizophrenia pathogenesis (1–3). Studies of erythrocytes and platelets from schizophrenic patients have suggested that alterations in membrane PL composition may play a role in the pathophysiology of schizophrenia (4, 5). The results of these previous findings have been mixed, but they have generally suggested altered concentrations of phosphatidylserine (PS), phosphatidylcholine (PC), phosphatidylethanolamine (PE), and/or phosphatidylinositol (PI) in schizophrenics compared to controls (5–8). Although erythrocytes and platelets are convenient models for some aspects of neurometabolism, changes measured in peripheral tissue may not reflect changes in neuronal membranes, and may be influenced more readily by extraneous factors. For example, Hibbeln et al. (9) found that smoking status and dietary intake predicted erythrocyte membrane fatty-acid composition for schizophrenic patients. Moreover, membrane PL composition varies greatly among different body tissues, consistent with widely different functions. Thus it is essential to measure PL composition and metabolism in the brain itself.

In vivo ³¹P MRS can be used to measure compounds involved in PL metabolism, primarily low-molecular-weight, phosphomonoester (PME) precursors and phosphodiester (PDE) degradation products (10). Although the detailed findings of various studies are not totally consistent, in general, in vivo ³¹P MRS studies do suggest alterations of PL metabolism in schizophrenia (11). Because PLs themselves in intact cell membranes are solid-like and have reduced molecular mobility, their MR signals are very broad and typically not visible in vivo.

An alternative to in vivo MRS is to perform high-resolution NMR spectroscopy on extracts of samples from postmortem brain. Tissue PLs can be characterized using ³¹P NMR in vitro after organic solvent extraction and suspension in a special reagent to prevent aggregation and yield narrow resonances (12, 13). Although high performance liquid chromatography (HPLC) is often used for PL analysis, it is labor intensive, and can require substantial sample modifications and multiple analyses. High resolution ³¹P NMR has the advantages of detection and quantitation of numerous PLs in a single assay, the detection of unsuspected and unidentified PLs, and its nondestructive nature. Because the number of P-atoms per molecule is known for generic PLs, molar concentrations can be determined for PLs of unknown side chain composition. Although there are obvious disadvantages to analysis of postmortem samples (14), unlike in vivo, precise quantitation of well resolved resonances of individual PLs or their metabolites can be performed.

In an earlier, preliminary study on nine subjects (15), we used ³¹P NMR of tissue extracts from frontal cortex of postmortem brain to detect PL differences between schizophrenics and controls. Here we used high resolution ³¹P NMR spectroscopy in an organic-solvent system to characterize PLs by class (headgroup) and subclass (linkage at the sn-1 position) in gray matter of frontal, temporal, and occipital cortices of postmortem brain for 20 schizophrenics, 20 controls, and 7 patients with other mental illnesses (psychiatric controls). We chose gray matter because schizophrenia is considered to be largely a gray-matter disease, although white-matter abnormalities have been detected (16).

In another previous study (17) we measured the concentrations of the metabolic precursors of PL metabolism, the PMEs phosphoethanolamine (pe) and phosphocholine (pc), as well as the PDE degradation products glycerophosphoethanolamine (gpe) and glycerophosphocholine (gpc) in the same samples that are the subject of the present study. Increased gpc was found in all three brain regions for schizophrenics relative to controls. Here we also examine possible correlations between the present PL results and those previous results on aqueous PL metabolites in the same samples (17).

Methods

The protocol was approved by the Human Research Advisory Committee of the University of Arkansas for Medical Sciences. Autopsies were obtained with the informed consent of next of kin. Written informed consent was obtained from each subject assessed premortem. Diagnoses were made postmortem using the Diagnostic Evaluation after Death (DEAD) protocol (18). DEAD uses information from multiple sources, including a structured chart review and interviews with kin. The diagnosis of schizophrenia in this instrument relies on the criteria of Feighner et al. (19). Information obtained from family members and treating physicians was incorporated when available. For the majority of cases the medical record was the major or sole source of information. Diagnoses were made independently by two research-trained, board-certified psychiatrists. A previous study (20) confirmed that schizophrenia diagnoses made with DEAD in a large group of elderly subjects agreed with those made according to DSM-III-R (21). Many patients also had been administered the Brief Psychiatric Rating Scale (BPRS), the Abnormal Involuntary Movement Scale, and/or the Neurobehavioral Cognitive Status Examination.

The patient groups consisted of 20 schizophrenics (S) (age 71±12 years, 15 M, 5 F), 20 controls (C) (age 71±8 years, 20 M, 0 F), and 7 psychiatric controls (PC) (age 63±12 years, 6 M, 1 F) (4 bipolar disorder; 1 each posttraumatic stress disorder, personality disorder, substance abuse). The groups were matched for age, race, and postmortem interval (PMI), except the PMI of the psychiatric control group (14.6±10.0 hours) was not matched to that of either the schizophrenic (4.1±1.7 hours) or control (3.9±0.6 hours) groups. The 11 schizophrenics whose medication status was known definitively had all been on first-generation antipsychotics. Four of 5 psychiatric controls whose medication status was known had been on first-generation antipsychotics, and one on second-generation antipsychotics. Controls had no history of neurologic or psychiatric disorders.

All brains were examined by a neuropathologist (REM) for pathological changes that would preclude inclusion in this study. At the time of gross autopsy, the brainstem and the left cerebellar and cerebral hemispheres were processed for special studies (see below). The right cerebrum and cerebellum were fixed for 10–14 days in 20% phosphate-buffered formalin, and then sectioned for gross examination and extensively sampled for paraffin embedding and histological examination. Sections were stained with hematoxylin-and-eosin stain, and with the Sevier-Munger modification of Bielschowski's stain for evaluation of senile changes (neuritic plaques and neurofibrillary tangles). Subjects with neuropathological changes, including Alzheimer's disease as based on Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) criteria (22), were excluded.

Tissues used in this study were obtained fresh at autopsy. These were dissected fresh, frozen in liquid nitrogen, and stored at −85 °C until used. Samples were taken from the left hemisphere in frontal, temporal, and occipital cortices. Frontal and temporal cortices were chosen because they have been strongly implicated in the pathophysiology of schizophrenia. Occipital cortex was chosen as a control region, since it has not been strongly implicated in schizophrenia. All samples were taken by a single individual with extensive experience in brain anatomy to insure that identical regions were sampled for all brains.

For frontal cortex, samples were taken from Brodmann area 10. Approximately, the section was anterior to a vertical cut 2-cm posterior to the apex of the frontal pole and superior to the orbitofrontal gyrus. For temporal cortex, samples were taken from the central region of left superior temporal gyrus in Brodmann area 22. For occipital cortex, samples were taken from Brodmann area 18, superior to primary visual cortex. As much white matter as possible was removed before the solvent extraction procedure. It is impossible to measure white matter contamination exactly, but for the cortical regions studied here, we estimated it to be on the order of 5–15%, consistent with a previous study (23).

Based on a visual inspection of the partially thawed surface of each frozen sample, 0.2–0.5 grams of primarily gray matter was isolated and deposited into a centrifuge tube on ice. Cold hexane-isopropanol (3:2) was then added at a volume ratio of 18:1, based on an assumed tissue density of 1.0 gram/mL. The mixture was homogenized in the tube by manual grinding with a teflon rod for 5 min, followed by sonication in a water-cooled (~15 °C) cup-horn flow cell assembly (Fisher Model 50 Sonic Dismembrator) for 5 min. Homogenates were then centrifuged at 2500×g for 10 min at 8 °C. The supernatants were then removed, subdivided into two separate glass vials, and the headspaces filled with nitrogen. The samples were stored at −85 °C. One of each set of extract subsamples was used for immediate (within a few days) ³¹P NMR analysis (see below), and the other was reserved for possible future analyses, with 0.1% butylated hydroxytoluene added to retard oxidation.

Samples were prepared for analysis in the CDCl₃-CH₃OH-H₂O (10:4:2) solvent system. After removing hexane-isopropanol under a stream of nitrogen and weighing the dried extract, 0.94 mL of CDCl₃:CH₃OH (10:4) was added. The mixture was vortexed briefly, then 134 µL of deionized H₂O containing 0.2 M Cs-EDTA at pH 6.0 was added. The total volume was determined so as to obtain a minimum 5-cm height (700 µL) of organic phase in the NMR tube and to exclude the aqueous phase from detection and facilitate shimming. The mixture was vortexed briefly again and the two phases were allowed to settle for at least 15 min before transferring them, lower phase first, to a 5-mm NMR tube, which was allowed to settle for at least an hour and subsequently equilibrated to a temperature of 25.0 °C in the NMR probe.

Phosphorus-31 NMR was performed at 161.9 MHz on a Varian Mercury 400 spectrometer using methods similar to those described previously (24). Spectra were acquired with WALTZ-16 ¹H decoupling (no nuclear Overhauser enhancement) with minimum post-acquisition delay. Typical conditions were: tip angle 84°; rf-pulse width, 5.6 µs; TR, 4.1 s; spectral width, 1 kHz in 8K points; typical number of transients, 496.

The resonances were quantified relative to the summed areas of all PL resonances using the program NUTS 2D (Acorn NMR, Palo Alto, CA). The FIDs were zero-filled once, multiplied by an exponential function (0.2 Hz line narrowing for resolution enhancement), Fourier transformed, and the resulting spectra phased and fit with Lorentzian-Gaussian lineshapes using a Simplex optimization algorithm. Because the resonances from PL classes often display complex shapes or even splitting due to the presence of molecular species with slightly different chemical shifts (24, 25), each resonance band was fit as the sum of 1 to 4 individual lines (Figure 1). However, even though these individual lines represent potentially identifiable molecular species of the various PL classes (24, 25), we report only the relative concentrations of PL classes without regard to molecular species. Chemical shifts were referenced to 85% H₃PO₄ at 0.00 ppm by setting the native PC peak to −0.51 ppm (25, 26). Intensity corrections for partial saturation were not required because this is a comparison of relative resonance intensities between subject groups, and absolute concentrations were not determined. The ³¹P spin-lattice relaxation times of PLs in this solvent system generally ranged from 1.2 to 2 s. Thus the TR and tip angle used here provide good overall signal-to-noise ratio (SNR). As long as SNR is good, the measured resonance intensities will be usable for relative quantitation (27). Absolute quantitation is usually not performed in this two-phase system due to variable partitioning effects for PI, PS, lyso-PLs and many quantification reference compounds (24). In our laboratory absolute quantitation is usually performed on PLs dispersed at low concentration in a single phase of aqueous detergent, although that system does not resolve PL subclasses as well as the present one (24). The aqueous-detergent analysis could not be performed here because, unlike the current analysis at higher PL concentration, it requires a ³¹P-NMR probe for 10-mm sample tubes, which is currently not available to us.

Figure 1.

161.9-MHz ³¹P NMR spectrum of a hexane-isopropanol extract of occipital cortex from a schizophrenic patient. The PC peak is off-scale. Chemical shifts are referenced to 85% H₃PO₄ at 0.00 ppm by setting the chemical shift of PC to −0.51 ppm (see text). U1–U3 are unidentified PLs. LPC usually occurred at the same chemical shift as PI under the conditions used here. Also shown are the individual lines used to fit the various PL class and subclass resonances. The numbers of lines used were: 1 each for U1, U2, SM, LPC, PI, PC_aa, PC; 2 each for LPE, PE_aa; 3 for PS; 4 each for PE_p, PE. U3, which appeared in only a few spectra, was not fit.

Statistical analyses [t-tests for independent samples, Pearson and Spearman correlations, repeated-measures multivariate analysis of variance (MANOVA), and post-hoc tests (Bonferroni and Scheffe)] were performed using Statistica (Statsoft, Tulsa, OK). The statistical significance of univariate results was confirmed by post-hoc testing.

Results

Phospholipid Classes and Subclasses

Figure 1 shows the 161.9-MHz ³¹P NMR spectrum of a hexane-isopropanol extract of occipital cortex from a schizophrenic patient. The identities of the resonances of the PL classes and subclasses were confirmed by spiking an extract sample with standard compounds (24). Resonances were seen for lysophosphatidylethanolamine (LPE), alkyl, acyl-phosphatidylethanolamine (PE_aa), phosphatidylethanolamine plasmalogen (PE_p), PE, sphingomyelin (SM), PS, PI, lysophosphatidylcholine (LPC), alkyl, acyl-phosphatidylcholine (PC_aa), and PC. LPC usually occurred at the same or similar chemical shift as PI under the conditions used here. Because the amount of LPC is very small when unambiguously resolved from PI, which was seldom under our circumstances, it was not quantified. Unidentified resonances U1–U3 were also observed. The resonance U2 probably arises from cardiolipin (24). The resonance U3, which appeared in only a few spectra, was not fit in the quantitation. Also shown are the individual lines used to fit the various PL-class and -subclass resonances. Table 1 gives the average, relative PL concentrations for the subject groups in the three brain regions as the percent of total PL measured in the ³¹P NMR spectrum.

Table 1.

Phospholipid Concentrations (± standard deviations)^a in Three Regions of Postmortem Brain for Schizophrenics, Controls, and Psychiatric Controls

Brain	Frontal			Temporal			Occipital
Region→
Metabolite↓	Control	Schizophrenia	Psychiatric Control	Control	Schizophrenia	Psychiatric Control	Control	Schizophrenia	Psychiatric Control
LPE	0.036±0.11	0.077±0.16	0.16±0.30	0.036±0.11	0.056±0.14	0.094±0.17	0.67±0.77	0.65±0.63	0.62±0.43
PEaa	2.78±0.74	2.52±0.67	3.72±1.37	2.27±0.51	2.34±0.62	2.53±0.97	3.41±2.16	2.91±1.11	2.42±1.04
PEp	19.30±1.60	18.76±1.82	18.17±3.47	19.41±1.68	19.00±1.71	18.94±1.48	18.67±3.15	18.65±1.95	18.18±1.87
PE	16.22±1.52	16.38±1.40	16.91±2.50	16.11±1.69	16.44±1.40	17.13±1.41	15.71±1.94	15.90±1.63	16.75±1.70
total PE	38.30±0.79	37.67±1.12b,c	38.80±0.53	37.80±1.07	37.78±0.76d	38.60±0.80	37.78±1.19	37.47±0.93	37.35±0.75
SM	9.27±0.79	9.51±0.67	9.24±0.97	9.27±0.74	9.60±0.54	9.31±1.00	8.79±0.99	9.29±0.74e	9.10±0.59
PS	13.70±0.90	13.67±0.84	13.17±1.30	13.37±0.98	13.30±0.76	13.04±0.79	13.01±1.22	12.70±0.96	12.74±0.84
PI	3.35±0.34	3.31±0.37	3.26±0.23	3.24±0.34	3.22±0.37	3.31±0.19	3.36±0.38	3.48±0.48	3.46±0.37
PCaa	0.59±0.16	0.58±0.15	0.51±0.11	0.57±0.15	0.55±0.16	0.55±0.07	0.74±0.19	0.73±0.17	0.66±0.25
PC	33.13±1.64	33.51±2.00	32.96±1.99	33.99±2.22	33.75±1.59	33.75±1.20	33.36±2.09	33.54±1.39	33.86±1.02
total PC	33.72±1.57	34.08±1.93	33.47±1.98	34.56±2.17	34.31±1.48	34.30±1.13	34.10±1.95	34.26±1.30	34.51±0.85
SM/total PE	0.242±0.021	0.253±0.018e	0.238±0.026	0.245±0.018	0.254±0.014e	0.242±0.030	0.233±0.027	0.248±0.021e	0.244±0.020

^aAs percent of total PL measured in the ³¹P NMR spectrum.

^bSignificantly different from Control at p=0.047.

^cSignificantly different from Psychiatric Control at p=0.017.

^dSignificantly different from Psychiatric Control at p=0.022.

^eDifferent from Control at 0.05<p<0.10.

The results for the individual PLs for the combined schizophrenic and control groups did not correlate with PMI, age, sex, or race, except for temporal LPE, which correlated with sex (r=0.39, p=0.012), and occipital PC, which correlated with age (r=0.34, p=0.030). Metabolite concentration did not correlate with PMI in the unmatched PC group alone. When considering all three subject groups (S, C, PC), repeated-measures MANOVA did not show a significant effect of subject group, but did show a highly significant effect of brain region (p<10⁻⁵, F=8.4). Univariate tests for individual PLs showed no statistically significant group effect after post-hoc correction.

When considering only schizophrenics and controls, repeated-measures MANOVA showed a trend (p=0.092, F=1.9) for the effect of subject group, and a highly significant effect of brain region (p<10⁻⁵, F=10.9). Univariate tests showed no statistically significant difference between schizophrenics and controls for any individual PL in any of the three brain regions studied. However, schizophrenics showed a trend toward higher occipital SM than controls (p=0.081) that survived post-hoc testing.

The resonances from PL classes display complex shapes or splitting due to the presence of molecular species (differing alkyl side chains at the sn-1 and/or sn-2 positions) with slightly different chemical shifts (24, 25). Here each resonance band was fit as the sum of 1 to 4 individual lines (Figure 1). These individual lines represent potentially identifiable molecular species or unresolved clusters of species. As seen before for PS (24), molecular species are well resolved under the present solution conditions. Assignments for these resolved PS species have been suggested previously (24), but are not of particular interest here. Because these individual resonances are generally not resolved (except for PS and partly for PE_p), we do not report these individual molecular species results. However, statistical comparisons by t-test for these component resonances revealed no differences between schizophrenics and controls.

In order to compare with previous results in peripheral tissue, we summed PE and PC subclasses to yield total PE and total PC measures (Table 1), respectively, which we subjected to MANOVA analysis. Total PE was significantly lower for schizophrenics than for either controls (p=0.047) or psychiatric controls (p=0.017) in frontal cortex, and lower than for psychiatric controls (p=0.022) in temporal cortex.

Because the ratio SM/PE in erythrocytes was found to be sensitive to psychopathology (5, 7), we include the ratio SM/total PE in our analysis (Table 1). This ratio trended (0.05<p<0.10) higher for schizophrenics in all three brain regions.

In the previous study (17) of aqueous PL metabolites in these samples, a males-only comparison yielded a significant result different from that of the total analysis. Hence a comparison of the schizophrenic and control groups which included only males (15 S, 20 C) was made. There were no statistically significant differences at p<0.05 for any PL. Total frontal PE headgroups trended lower, and temporal SM higher, in male schizophrenics relative to controls.

Correlation of PL and PL-Metabolite Concentrations

We also performed Pearson correlations for selected PLs from the present work with related PL metabolites from a previous study (17) of these same samples in each of the three brain regions for the schizophrenic group. Specifically, correlations were examined for ethanolamine-based PLs [LPE, PE_aa, PE_p, PE, and total PE (all PE headgroups)] with pe and gpe. A similar analysis was performed for choline-based PLs with pc and gpc. Total occipital PE correlated with gpe (r=0.645, p=0.002) and total occipital PC inversely correlated with gpc (r=−0.638, p=0.002).

Discussion

Phospholipid Classes and Subclasses

Studies on erythrocytes and platelets suggest that alterations in membrane PL composition may play a role in the pathophysiology of schizophrenia (4, 5). Early findings have been summarized (5, 6). These findings were mixed, but generally suggested decreases in PC and PE, and increases in PS and SM in schizophrenics relative to controls. Decreases in PC, PE, and PS were seen for skin fibroblasts of schizophrenics (28). Ponizovsky et al. (7) found increased SM, decreased PE, and a correlation of the ratio of these PLs with symptomatology, in erythrocytes from schizophrenic patients. Perhaps the most consistent findings are reduced PE and increased SM in peripheral measures of PL concentrations (5, 29). Results are also mixed for PI in peripheral tissues (30–32).

Yao et al. (33) found significantly lower concentrations of PC and PE from the caudate in schizophrenia. Trends were also noted for a reduction in total PL content, and for increases in PI and SM. Schmitt et al. (34) saw reduced PC and SM, increased PS, and unchanged PE, PI, and LPC in left thalamus postmortem for schizophrenia.

Because of its central role in PL metabolism, considerable attention has been directed to the enzyme phospholipase A₂ (PLA₂). Gattaz and coworkers found increased PLA₂ activity and the expected product, LPC, in serum or platelets from drug-naive schizophrenics (8, 35, 36).

In a preliminary ³¹P NMR study (15) we found increased PI, total PL content, and PC molecular species with one saturated and one unsaturated acyl chain in frontal cortex for schizophrenics relative to controls. No differences were seen for other PLs, including LPC and LPE. Unlike the present study, that preliminary study was performed in a single-phase, aqueous-detergent solution, which permitted absolute quantitation (see above).

Surprisingly, repeated-measures MANOVA revealed no overall differences among the groups. However, differences found in previous studies were never particularly large or encompassing. Taken as a whole, they were quite mixed. Here we detect few differences among the PL class and subclass compositions of schizophrenics, psychiatric controls, and controls. When considering only schizophrenics and controls, repeated-measures MANOVA showed a trend for the effect of subject group on PL composition. Schizophrenics showed reduced frontal total PE relative to both normal and psychiatric controls, and a trend toward higher occipital SM than controls. Reduced frontal total PE and increased SM in schizophrenics are consistent with a number of previous studies in peripheral tissue (5), and the observations in caudate by Yao et al. (33). In erythrocytes, the derivative ratio SM/PE was found to be sensitive to the severity of psychopathology as measured by BPRS scores (5, 7). Here the ratio SM/total PE trended higher for schizophrenics relative to controls for all three brain regions.

As in our preliminary study (15), we observe low levels of LPC and LPE, which do not differ between schizophrenics and controls. Primarily because of their low levels, LPC and LPE are difficult to measure by ³¹P NMR. They can also be generated by hydrolysis during sample workup. Although absolute quantitation of lysophospholipids is problematic in the solvent system used here, the concentrations appear to be substantially lower as a fraction of total PC than those reported by Pangerl et al. (8) for platelets. Our observations again provide no evidence for increased PLA₂ activity or LPC in schizophrenic brain.

Although primarily directed at schizophrenia, the phospholipid hypothesis has been studied and invoked in the context of all psychiatric illnesses (2). Thus a psychiatric control group was included in this study to confirm that any differences seen between the schizophrenic and control groups were specific to schizophrenia. Because of its smaller size and mixed diagnosis, the psychiatric control group is less useful than the normal control group for comparison to the schizophrenic group. However, our results for the psychiatric control group are consistent with the primary results comparing schizophrenics to a normal control group. Interestingly, the major difference we see between schizophrenics and normal controls, that for total PE in frontal cortex, is also seen between schizophrenics and psychiatric controls. This suggests that this reduction in total PE may be specific to schizophrenia.

Both frontal and temporal cortices are implicated in schizophrenia (37). Occipital cortex was chosen as a control region because it has not been strongly implicated in schizophrenia. However, the phospholipid hypothesis of schizophrenia postulates global changes in PL composition and metabolism (1, 2), which are also suggested by changes in peripheral PL measures in schizophrenia (4–6). Thus our observation of trends toward higher SM and SM/total PE in occipital cortex is not inconsistent with the phospholipid hypothesis.

It is reasonable to expect PL metabolism and concentrations to be altered by antipsychotic medications (38). Acute neurotoxic effects of antipsychotic medications have been observed (38). Our schizophrenic subjects were old and had been on (primarily first-generation) antipsychotic medications for many years. In functional neuroimaging measures in schizophrenics, changes with age, and presumably with antipsychotic treatment, have been characterized as somewhat similar to the effects of normal aging (37). Although anatomic changes are often seen after years of both disease burden and antipsychotic treatment, the effects on membrane PL composition have not been well studied. Over six months of treatment in schizophrenics, Schmitt et al. (39) saw fluctuating, but statistically significant, changes in platelet PC, PE, and LPC relative to the neuroleptic-free baseline. No changes were observed in brain PL fatty-acid composition in rats treated with either a typical or atypical antipsychotic medication (40). Very recently the effects of atypical antipsychotics on lipid metabolism were studied using metabolomic mapping (41). Atypical antipsychotics raised plasma PE levels that had been suppressed in schizophrenic patients (41).

Unfortunately, here we can shed no light on the effects of antipsychotic medication. The 11 schizophrenics whose medication status was known definitively had all been on a variety of first-generation antipsychotics. However, the other 9 schizophrenics were likely on a variety of first-generation antipsychotics also.

It is not unreasonable to expect external factors such as diet or smoking to affect PL composition, although less so for brain than for peripheral measures. Most schizophrenics are heavy smokers, which may influence brain PL composition. Hibbeln et al. (9) found that smoking status and dietary intake predicted erythrocyte membrane fatty-acid composition for schizophrenic patients. Unfortunately, we do not know either the smoking or dietary status of our subjects, whether schizophrenics or controls.

Although the schizophrenic and control groups are matched for PMI, it is still desirable to minimize PMI. The average PMIs for our schizophrenic and control groups were about 4 hours, which is relatively short for a study of human postmortem tissue. We previously determined that PL concentrations do not vary with PMI for up to 18 hours (24). The evidence suggests that any PL-concentration changes with PMI will be nonexistent or minor for the schizophrenic and control groups. For our psychiatric control group the average PMI was greater, but even so we found no changes in PL concentrations with PMI. Metabolite concentration did not correlate with PMI in the unmatched PC group.

Because schizophrenia is generally considered to be a gray-matter disease, we decided to study pure gray-matter samples. White matter contamination should not significantly affect the present results. It is known that the PL headgroup and fatty acid compositions of gray and white matter in human brain are different (42). A 15% contamination of white in gray matter would, relative to a 0%-contaminated sample, result in about a 5% error at most in relative PL composition. Because samples from different groups were treated identically, the contamination level would be similar for the same anatomic regions across groups. For example, if there were a 15% contamination in a temporal-cortex sample from a control, a similar contamination is likely in a temporal-cortex sample from a schizophrenic. There is no reason to expect a systematic difference in the white-matter contribution between schizophrenics and controls. Ventricular enlargement and cortical atrophy have been studied extensively in schizophrenia (37), but are not always present. Since there is always some small level of white-matter contamination for both controls and schizophrenics, systematic differences, if present, would be small.

We are sensitive to the limitations surrounding the use of postmortem tissue. Lewis (14) describes a number of criteria characteristic of good studies employing postmortem tissue in psychiatric research. Our present study largely fulfills Lewis' criteria. Our samples are well characterized according to clinical diagnosis, neuropathology, postmortem factors, and toxicology. We have well characterized specimens pertinent to the disease, and that group is well matched to a control group. To the extent possible, relevant factors such as PMI have been matched and optimized. Postmortem studies often have perimortem confounds that cannot be eliminated. Comparison of matched groups compensates for this to some extent, as we have done here. Antemortem and some demographic factors are difficult to characterize, but should be relatively uniform.

Unfortunately, particularly for schizophrenia, patients are very likely to be elderly, institutionalized, and have a long history of treatment with antipsychotic medication. That is the case with our study. It is extremely difficult to obtain sufficient numbers of brains of younger, drug-naive, or unmedicated psychiatric patients. In most in vivo studies, the ages of chronic schizophrenia patients are typically under 50, whereas in our study the average age is about 71.

Since we did not see major changes in any individual PL subclass, it is appropriate to ask whether the ³¹P NMR method is sufficiently sensitive for relative quantitation. For the major resonances the signal-to-noise ratio is relatively high. Quantitative NMR is a longstanding and well understood technique that is suitable for this type and level of quantitation (27). It also remains possible that there are differences in molecular-species distributions between schizophrenics and controls, and that such differences may be detectable using an alternative ³¹P NMR analysis method with these same samples (15, 25). Other techniques may very well be applicable to this analysis. In particular, matrix-assisted laser desorption and ionization time-of-flight (MALDI-TOF) mass spectrometry seems promising (43). ³¹P NMR at higher magnetic field should also be valuable for better resolution and detection of minor PLs, PL subclasses, and molecular species.

Correlation of PL and PL-Metabolite Concentrations

The results of the present study do not appear to coincide with those of our previous study (17) of aqueous PL metabolites. Here we observe differences in total frontal PE and occipital SM in schizophrenia, whereas before we observed increased gpc in all three brain regions. Consistent with the implications of the membrane hypothesis of schizophrenia, we had predicted that PL concentrations would correlate with concentrations of related PL metabolites in schizophrenia in affected brain regions. For example, it might be expected that an increase in gpc would accompany a decrease in PC.

We performed Pearson correlations for selected PLs from the present work with related PL metabolites from the previous study for each of the three brain regions for the schizophrenic group. We did find the expected inverse correlation of gpc and PC in occipital cortex, but not in either frontal or temporal cortex, the regions that are primarily implicated in schizophrenia. The only other correlation found was the positive one between total occipital PE and gpe, which does not fit our prediction. The metabolite changes measured by in vivo ³¹P MRS in schizophrenia do not appear to reflect PL concentration changes, at least directly.

Conclusions

Surprisingly, repeated-measures MANOVA revealed no overall differences in PL composition among schizophrenics, normal controls, and psychiatric controls. Except for total PE in frontal cortex, there were no statistically significant differences between schizophrenics and controls for any of the major individual PL classes or subclasses. In addition, there was minimal correlation of the present results for the PLs themselves with results for the relevant PL metabolites obtained for these same samples. The inverse correlation of gpc and PC was seen in occipital cortex, but not in either frontal or temporal cortex. The metabolite changes measured by in vivo ³¹P MRS in schizophrenia do not appear to reflect PL concentration changes. The present results offer very little support for the phospholipid hypothesis of schizophrenia.