INTRODUCTION

The dopamine (DA) hypothesis of schizophrenia, which attributes the positive schizophrenic symptoms to an excess of dopaminergic neurotransmission, was formulated on the basis of observations that typical antipsychotic drugs block D2-like DA receptors in vitro (Carlsson, 1988) with affinities that correlate with clinical potency (Seeman and Lee, 1975). PET studies show that the treatment of schizophrenic patients with neuroleptics at clinically relevant doses leads to occupancy of D2-like DA receptors in the range of 65–90% in the case of haloperidol and other typical neuroleptics and 20–40% in the case of clozapine, an atypical neuroleptic (Farde et al, 1992).

For several reasons, the antipsychotic effect of neuroleptics cannot be entirely attributed to the acute pharmacological blockade of postsynaptic D2-like DA receptors. First, DA receptor blockade occurs shortly after the administration of these compounds (Nordström et al, 1992), while the onset of antipsychotic effects is delayed by a period ranging from days to weeks (Johnstone et al, 1978). Second, similar striatal DA receptor occupancies by neuroleptics have been observed in responders to neuroleptic drug treatment and in nonresponders (Wolkin et al, 1989). Thus, the blockade of DA D2 receptors may be necessary, but cannot be sufficient to account fully for antipsychotic drug action. This discrepancy led to the suggestion that the antipsychotic effect of neuroleptics may arise in conjunction with the delayed depolarization block of the presynaptic neuron and reduced DA synthesis capacity, a phenomenon which occurs in experimental animals treated chronically with neuroleptics (Grace and Bunney, 1986). Several positron emission tomography (PET) studies with radiolabeled L-DOPA analogues have reported elevated activity of the enzyme DOPA decarboxylase (DDC) in the striatum of untreated patients with schizophrenia (Hietala et al, 1995; Lindström et al, 1999; Reith et al, 1994), indicating the presence of elevated capacity for DA synthesis within the terminals of DA neurons (Cumming and Gjedde, 1998). However, the transition of DA neurons to a partially inactivated state of depolarization block has not been demonstrated in schizophrenic patients under treatment with neuroleptics. Therefore, we used the PET tracer [18F]fluoro-L-DOPA (FDOPA) in conjunction with compartmental modeling to test the hypothesis that subchronic treatment with a typical neuroleptic decreases the capacity for DA synthesis in the brain of patients with schizophrenia.

METHODS

The study was approved by the local ethics committee.

Subjects

Nine male patients meeting DSM-IV criteria for schizophrenia, who were acutely admitted to the hospital, gave written, informed consent to participate in the study. The age of the subjects ranged from 27 to 47 years (mean±SD: 37.6±6.0 years). Except for their psychiatric disorder, they were healthy with respect to medical history, physical examination, 12-lead electrocardiogram (ECG), electroencephalogram (EEG), fasting clinical laboratory tests, and magnetic resonance imaging (MRI) of the brain. Among the inclusion criteria was the requirement that patients were free from any psychoactive medication for at least 6 months prior to the PET study. Three patients were entirely drug naïve, the remainder had been treated with various doses of oral neuroleptics. Two PET scans with FDOPA were performed, one at baseline and one after treatment with haloperidol for a period of 20–45 days (mean±SD: 35.3±7.6 days). To ensure that truly subchronic, adaptive effects of haloperidol were measured, the second PET scan was scheduled no earlier than 30 days after initiation of treatment in all but one patient. In this single patient, the second scan had to be performed earlier for logistical reasons. All patients improved during this course of treatment; haloperidol dose was selected so as to yield the optimal clinical benefit while minimizing extrapyramidal symptoms. The mean daily dose on the day of the second PET scan was 8.9±4.5 mg (range 1–15 mg).

Psychometric Rating Scales

Psychopathological symptoms and EPS were derived by trained raters on the day of the first PET scan, prior to initiation of haloperidol, and again on the day of the second PET scan by means of the following standard instruments: Positive and Negative Syndrome Scale (PANSS, Kay et al, 1987), Clinical Global Impression (CGI, Guy, 1976), Extrapyramidal Symptoms Rating Scale (ESRS; akinesia score, Chouinard et al, 1980) and the Barnes Akathisia Rating Scale (BARS, Barnes, 1989).

PET Data Acquisition

For 24 h prior to the PET scans, patients were maintained on a diet low in protein in order to minimize the variance in the competition between dietary large neutral amino acids and the tracer FDOPA for facilitated diffusion across the blood–brain barrier (Stout et al, 1998). Decarboxylation of FDOPA in the peripheral tissue was blocked by oral administration of the peripheral DDC inhibitor carbidopa (Merck Sharp & Dome, 2 mg/kg body weight) 1 h prior to injection of the FDOPA (Cumming et al, 1993). Emission recordings from the first two patients were obtained with a GE 4096+ whole-body scanner with 15 planes, an in-plane resolution of 5–6 mm, and an axial resolution of 6–7 mm. All other studies were performed with a Siemens ECAT EXACT whole-body PET scanner with a field of view of 16.2 cm in 47 planes, in interplane spacing of 3.375 mm and an axial resolution of 5.4 mm FWHM. After a brief attenuation scan, a sequence of 28 emission frames lasting a total of 128 min was recorded. Frame length increased progressively from 1 to 10 min.

A mean of 153 MBq FDOPA (SD: 52 MBq, range: 95–300 MBq) was injected intravenously as a bolus into a cubital vein. A series of 23 blood samples were collected from a radial artery at the following intervals after FDOPA injection: 8 × 30 s, 1 × 1 min, 2 × 2.5 min, 2 × 5 min, and 10 × 10 min. Blood samples were immediately centrifuged and the total plasma radioactivity counted by gamma spectroscopy. The fractions of untransformed FDOPA and its major plasma metabolite O-methyl-FDOPA were measured in nine plasma samples by HPLC (Cumming et al, 1993). The fractions at each of the 23 sample times were estimated by fitting multiexponential functions to the nine fractions measured by HPLC and the continuous plasma input functions were calculated.

After the first PET scan, a T1-weighted 3D gradient echo magnetic resonance scan with 1.5 mm slice thickness and 128 slices was acquired from each patient The MRI was resliced according to the ac-pc-line, which was identified in a midsagittal plane. The summed PET images were then coregistered with the resliced MRI to permit anatomical positioning of volumes of interest (VOIs) using the automatic image registration (AIR) algorithm (Woods et al, 1992). For caudate, putamen, and thalamus, VOIs were generated from polygonal regions of interest (ROIs) on all MRI planes containing the respective tissue types. VOIs for the cortical areas were created from ten consecutive MRI slices.

PET Data Analysis

A constrained two-compartment model assuming irreversible trapping in the brain of decarboxylated metabolites was fitted to the timeradioactivity curves (TACs) recorded in brain (Gjedde et al, 1991; Huang et al, 1991). The temporal changes in the plasma concentrations of FDOPA and its principal metabolite O-methyl-FDOPA were used to calculate the apparent first-order rate constant for the formation of O-methyl-FDOPA (kD0, min−1), corresponding to the whole-body activity of catechol-O-methyltransferase, and the fractional rate constant for the elimination of O-methyl-FDOPA from circulation ( k − 1 μ , min − 1 ) , corresponding to the renal elimination rate (Cumming et al, 1993). In order to test for possible treatment effects on the peripheral pharmacokinetics of FDOPA (see Cumming et al, 1999), the magnitudes of these rate constants were calculated for each subject in the baseline condition and after haloperidol treatment.

In brain regions such as the cerebellum or the occipital cortex that are nearly devoid of DDC activity, the concentration of FDOPA in the brain tissue at any time after intravenous injection is determined by the convolution of the arterial input with the unidirectional blood–brain clearance of FDOPA (KD1, ml g−1 min−1) and the fractional rate constant for the diffusion of FDOPA from brain back to plasma (kD2, min−1). The ratio of these two terms, KD1/kD2, corresponds to the tracer blood–brain distribution volume at equilibrium (VeD, ml g−1). In brain regions containing DDC, FDOPA is decarboxylated at rate constant kD3 (min−1), corresponding to the local activity of DDC (Cumming et al, 1993). The product [18F]fluorodopamine is retained in synaptic vesicles, resulting in the progressive accumulation of radioactivity in the caudate and putamen, which are richly endowed with DDC. Here, the solution of the two-compartment model is facilitated by constraining the magnitude of VeD to the individual observation obtained in the cortical reference region. This approach allows stable estimates of the term of interest, the local relative activity of DOPA decarboxylase kD3 (Cumming and Gjedde, 1998; Gjedde et al, 1991; Huang et al, 1991). Thus, the parameters to be estimated in the ROIs were the apparent plasma volume occupied by FDOPA Vo (ml g−1), the local magnitude of KD1, and the magnitude of kD3.

The quantitative analysis of dynamic emission sequences obtained with FDOPA is based upon the compartmental model established for the quantification of cerebral glucose metabolism, in which the rate constant k3 there corresponds to the phosphorylation of glucose analogues by the hexokinase reaction (Sokoloff et al, 1977). An additional complication of FDOPA modeling is presented by the peripherally formed metabolite O-methyl-FDOPA, which becomes the predominant radiolabeled compound in the blood and brain during prolonged experiments (Cumming et al, 1993). The entry of O-methyl-FDOPA into brain requires additional physiological constraints in the compartmental model for estimating the DDC activity. In the present model, the two amino acids O-methyl-FDOPA and FDOPA are assumed to have a single common distribution volume, and are assumed to have a fixed permeability ratio, q, of 1.5, based upon the mean of several estimates (Cumming and Gjedde, 1998). The sensitivity of the estimation of kD3 to the constrained values of q and VeD was investigated in the monkey brain (Léger et al, 1998).

In our study, individual estimates of VeD in occipital cortex were used to constrain fitting of the same model to time radioactivity recorded in caudate and putamen during 45 min of tracer circulation. This procedure has been shown to optimize the precision and accuracy of the estimation of kD3 when the eventual elimination of decarboxylated metabolites from the brain is neglected (see Figure 3 in Cumming and Gjedde, 1998). For thalamus and cortical VOIs, 60 min emission recordings were used in order to improve the estimation of kD3 in brain regions of low DDC activity. However, this claim has not been completely examined in the literature.

Statistical Analyses

Means and standard deviations were calculated for psychopathological and kD3 values. Shifts from baseline to end point were analyzed by means of Wilcoxon matched-pairs signed-ranks tests. Spearman rank correlations were calculated for treatment-related changes in kD3 and psychopathology or EPS side effects, respectively. To account for baseline values, proportional change scores ((end point−baseline)/baseline × 100) were computed for kD3, CGI, and PANSS, while absolute changes in BARS and EPS scores (end point−baseline) were used. In all analyses, the two-tailed level of statistical significance was set at α=0.05. Owing to the exploratory character of the study, no adjustment for multiple testing was performed.

RESULTS

The mean magnitude of the FDOPA distribution volume VeD in occipital cortex was 0.60±0.11 ml g−1 in the baseline condition and 0.70±0.22 ml g−1 after haloperidol treatment. This increase in VeD failed to reach statistical significance (p=0.051). Also, the change in VeD was not significantly correlated to change in kD3. The unidirectional blood–brain clearance of FDOPA (KD1, ml g−1 min−1) ranged from 0.028 ml g−1 min−1 in the left caudate nucleus to 0.038 ml g−1 min−1 in the left putamen in the baseline condition; haloperidol was without effect on the magnitude of KD1 in any region. Also, haloperidol did not alter kD0, the apparent rate constant for the formation of the O-methylated FDOPA metabolite in the plasma (pretreatment, mean±SD: 0.011±0.002 min−1; post-treatment: 0.011±0.002 min−1) or k - 1 μ .

In the baseline PET study, no significant differences in the magnitude of kD3 between hemispheres were detectable, although estimates tended to be higher in the left striatum (Table 1). kD3 decreased significantly with treatment in all brain regions investigated except for the right caudate (Table 1, Figure 1). Baseline kD3 in the cerebral cortex VOIs was low, but also decreased significantly with neuroleptic treatment (Table 1). Thalamic kD3, which was significantly higher than in cortex, was also significantly reduced. The reduction in kD3 was most pronounced in the left caudate (Table 1, Figure 1). There was no statistically significant correlation between treatment duration and the magnitude of change in kD3. However, the single subject who was scanned after only 20 days treatment, had very small decreases (eg lowest decrease in left caudate) or even increases in kD3.

Table 1 Magnitudes of k3D (mean±SD, min−1) in Striatal, Thalamic, and Cortical Regions in Nine Schizophrenic Patients at Baseline and after Haloperidol Treatment as well as Percentage Change with Treatment in the Respective Regions
Full size table
Figure 1
figure 1

Scatter plots of individual changes (in % from baseline) in kD3 (min−1) in striatal and thalamic regions. Subjects scanned on the GE 4096+ camera are indicated with a black circle: —, sample median.

Full size image

Mean baseline and post-treatment psychopathological and extrapyramidal side effect (EPS) rating scores are presented in Table 2. There were significant improvements in all psychopathological rating scores after haloperidol treatment. EPS scores were significantly increased by haloperidol (Table 2). At baseline, we could not detect any statistically significant correlation between kD3 in any region and psychopathological or side effect rating scales, respectively. The same was true, when the follow-up condition was investigated separately or when both conditions were analyzed together. However, when the change between pre- and post-treatment kD3 was correlated with change in psychopathology or side effects, respectively, we could indeed detect statistically significant relations. The change in thalamus kD3 was highly significantly correlated with percent change in the PANSS negative subscore (RS=0.93; p<0.0005), but not with change in the PANSS positive subscore (RS=−0.18; p=0.651). Moreover, there were positive, although less robust, correlations between thalamic kD3 and CGI (RS=0.74; p=0.023) as well as BARS rating scores (RS=0.71; p=0.032).

Table 2 Treatment-Related Changes in Psychopathological and Side Effect Rating Scores
Full size table

DISCUSSION

Soon after the introduction of chlorpromazine, the first neuroleptic drug, Carlsson and Lindqvist discovered that acute treatment of rats with antipsychotics increases DA metabolite concentrations in the brain and stimulated the activity of tyrosine hydroxylase in the rat striatum suggesting that pharmacological blockade of presynaptic DA ‘autoreceptors’ increased the rate of DA synthesis and release in the brain (Carlsson and Lindqvist, 1963). This claim was substantiated by observations that neuroleptics increase the basal firing rate of DA neurons and block the amphetamine-induced inhibition of the depolarization of DA neurons (Grace, 1992; Grace and Bunney, 1986). However, these effects of neuroleptics are obtained acutely, and therefore seem insufficient to account for the onset of antipsychotic effects, and might indeed be considered paradoxical. Furthermore, the clinical response can be delayed by several weeks, whereas potentiation of DA transmission by neuroleptics occurs immediately. Subsequent electrophysiological studies showed that chronic administration of neuroleptics decreased the number of spontaneously active DA neurons in the rat substantia nigra, a condition known as depolarization block. Grace hypothesized that the antipsychotic effect of neuroleptic drugs occurs in conjunction with the development of this depolarization block (Grace, 1992). Others have stressed the importance of block in mesolimbic, but not striatal neurons for obtaining therapeutic effects (Chiodo and Bunney, 1983; White and Wang, 1983), whereas depolarization block of nigrostriatal neurons may underlie the extrapyramidal side effects (White and Wang, 1983). The present documentation of a haloperidol-induced decrease of DA synthesis capacity is in good agreement with observations of amphetamine-evoked changes in DA D2 receptor occupancy. Amphetamine-induced changes in [123I]iodobenzamide binding in the striatum suggest the presence of a supernormal amphetamine-releasable DA pool in acutely psychotic patients, which was normalized with medication (Laruelle et al, 1999).

In the present study, we find that a course of treatment with haloperidol producing significant clinical improvement was associated with a decline in kD3, the DDC activity, calculated by fitting of a constrained compartment model to the recorded TACs, in several brain regions. This finding was robust, occurring in eight out of nine patients (Figure 1). Although we found no correlation between kD3 and treatment duration, it is remarkable that the single subject who for logistical reasons had to be scanned 20 days after initiation of haloperidol, demonstrated the mildest decrease in kD3 or even increases in some regions. In this patient, 20 days treatment may have been too short to observe a decrease in kD3.

In our study, the unidirectional blood–brain clearance of FDOPA (KD1, ml g−1 min−1) was not influenced by haloperidol treatment. Thus, the estimation of kD3 is isolated from the possible effects of drug treatment on the delivery of tracer to the brain. Also, haloperidol did not alter the plasma metabolism of FDOPA. However, we have documented a mild, but almost significant increase in distribution volume VeD, which is the ratio KD1/kD2. Since KD1 does not change at all in our study, the increase in VeD could possibly be attributed to a decrease in kD2, which is the fractional rate constant for the diffusion of FDOPA from the brain back to the plasma. Blood flow changes induced by haloperidol have been reported, but the literature is inconsistent in this regard. Most studies found blood flow changes in frontal cortex (Miller et al, 2001), basal ganglia (Miller et al, 1997), or hippocampus (Medoff et al, 2001), respectively, but not in occipital cortex, which served as reference region in our study, from which the equilibrium distribution volume VeD was determined. Thus, although we cannot totally exclude the possibility that the documented changes in kD3 are influenced by brain perfusion or tracer distribution, our study has the advantage that it actually measures these effects. Owing to numerous methodological factors, our present baseline estimates of kD3 (Reith et al, 1994), which represents the rate constant for the decarboxylation of FDOPA, cannot be compared directly with previously reported normative values obtained by other laboratories, and cannot readily be compared with estimates of the net clearance of FDOPA (Hietala et al, 1995) or [β-11C]DOPA (Lindström et al, 1999) to striatum KiD (ml g−1 min−1), which reflect the composite of perfusion, diffusion back to blood, and decarboxylation in nigrostriatal terminals. Consequently, estimates of the magnitude of Ki are weighted by the unknown equilibrium volume (VeD) of FDOPA in the brain tissue. Therefore, the effects of treatment or condition on the magnitude of Ki in the human brain cannot be unambiguously attributed to the altered rate of trapping of FDOPA in the brain (Hietala et al, 1995; Meyer-Lindenberg et al, 2002; Yatham et al, 2002). The present compartmental model explicitly evaluates the magnitude of VeD, and therefore isolates the parameter of interest, the activity of DDC, from the effects of perfusion and distribution.

FDOPA and other exogenous DDC substrates are tracers for the ultimate step in DA synthesis. Since concentrations of L-DOPA occurring in living brain are far below the saturating concentration for DDC, it is usually concluded that this enzyme occurs in vast excess and consequently cannot contribute to the regulation of DA synthesis. On the other hand, tyrosine hydroxylase, the penultimate enzyme in catecholamine synthesis, is nearly saturated with its main substrate tyrosine, and is consequently generally considered to be the sole rate-limiting step in DA synthesis (for a review, see Roth and Elsworth, 1995). However, in the living brain the DDC activity influences the fraction of substrate, which is directed to catecholamine synthesis rather than being exported from the brain. The presence of alternate fates for L-DOPA in vivo requires that regulation of DDC must also influence the rate of DA synthesis (Cumming and Gjedde, 1998; Gjedde et al, 1993). In support of this claim, the DDC activity in the striatum of living rats was potentiated by acute neuroleptic treatment and decreased by the DA presynaptic autoreceptor agonist apomorphine (Cumming et al, 1997). In an FDOPA/PET study, acute haloperidol markedly stimulated DDC activity in pig striatum (Danielsen et al, 2001). Whereas the functional state of DA autoreceptors may account for the acute effects of neuroleptics on DDC activity, the present finding of decreased DDC activity in human brain suggests that additional factors must influence DDC activity during chronic neuroleptic treatment.

It could be expected that decreased decarboxylation of FDOPA should predict decreased synthesis and turnover of endogenous DA in the brain. Concentrations of the DA metabolite homovanillic acid (HVA) measured in biological fluids have often served as an index of cerebral DA catabolism in humans. Thus, the baseline plasma HVA concentration and the initial increase occurring upon initiation of neuroleptic treatment both correlate with treatment response (Chang et al, 1988; Davila et al, 1988). Plasma HVA concentrations decrease with treatment in responders, whereas they are essentially unchanged in nonresponders (Bowers et al, 1984). Patients with predominantly negative schizophrenic symptoms show a lower initial increase in HVA concentrations and a flat subsequent decrease (Davila et al, 1988). However, much of the plasma HVA is actually derived from the synthesis in catecholamines in sympathetic organs, and so is an imperfect index of central dopaminergic transmission (Amin et al, 1992).

Although left–right asymmetry in the baseline magnitude of kD3 did not reach significance, the decline in the magnitude of kD3 after haloperidol was significantly greater in the left caudate ( Table 1). Similarly, Hietala et al (1999) reported the presence of asymmetry in the FDOPA influx to caudate in normals, with the greater influx to the right side; this asymmetry was absent in untreated schizophrenic patients. Together, these findings suggest the presence of considerable asymmetry in the uptake of FDOPA is characteristic of caudate in normal subjects, and that haloperidol treatment preferentially reduced the FDOPA consumption on the abnormal left side of patients with schizophrenia ( Table 1). In a meta-analysis of numerous [123I]iodobenzamide/SPECT studies, higher D2 binding in the right striatum of healthy subjects is consistently reported (Larisch et al, 1998). Altered asymmetry in pre- and postsynaptic markers of DA might underlie the left-turning bias reported in chronic schizophrenic patients (Bracha et al, 1993; Lyon and Satz, 1991). While healthy control subjects consistently turn right to collect coin reinforcers, schizophrenic patients tend to turn left. However, this phenomenon has been attributed to a dopaminergic overactivity in the right hemisphere striatal or mesoprefrontal areas (Bracha et al, 1993). A correlation between excess [123I]IBZM binding in the left basal ganglia of schizophrenic patients was also demonstrated to correlate with stereotyped behavior (Pedro et al, 1994). However, with regard to the structure of the basal ganglia in untreated patients with schizophrenia, there are several reports showing either no change or decreases in volume, but studies demonstrating structural asymmetry in these structures are lacking (Corson et al, 1999; Gunduz et al, 2002; Keshavan et al, 1998; Shihabuddin et al, 1998,2001). Interestingly, Scheepers et al found a reduction in left caudate nucleus volume in patients who responded to clozapine treatment, but not in nonresponders (Scheepers et al, 2001).

We find that chronic haloperidol also decreased the magnitude of kD3 in the thalamus. The thalamus is the only structure in which the change of kD3 is related to the change in psychopathology. We find a very marked positive correlation between change in kD3 and improvement in negative symptoms; that is, patients with the strongest decrease in thalamic kD3 demonstrated the most pronounced decrease in the PANSS negative symptoms subscore. The thalamus seems to play a crucial role in the pathophysiology of schizophrenic symptomatology and is probably also an important site through which the clinical effects of antipsychotics are mediated (Carlsson et al, 1999b). It is an essential structure for controlling arousal and the transmission of sensory input—both from the outside world as well as from internal states of the body—to the cerebral cortex (Carlsson et al, 1999a). Thalamic volume has been reported to be reduced in most but not all of the available studies of schizophrenia, but an asymmetry in structure has not been shown (Andreasen et al, 1994; Arciniegas et al, 1999; Hazlett et al, 1999; Wolkin et al, 1998). The thalamus has a low, but significant dopaminergic innervation (Bouthenet et al, 1987; Farde et al, 1997). DA transmission in the ventral thalamus has been implicated in the mechanism for prepulse inhibition of the startle reflex (Young et al, 1995), which is abnormal in schizophrenia (Kumari et al, 2000). Moreover, the concentration of DA was elevated in subregions of the thalamus from patients dying with schizophrenia (Oke and Adams, 1987). However, little is known about the origin and function of dopaminergic projections to the thalamus. The present observations are suggestive that inhibition of catecholaminergic transmission in parts of the thalamus may mediate aspects of therapeutic response to neuroleptics. However, at present it can only be a smaller of speculation about the functional relation between thalamic DA metabolism and the negative symptoms of schizophrenia. Studies in larger patient samples are necessary to relate presynaptic dopaminergic function to psychopathology.

In an earlier PET/FDOPA study, we have reported that the NMDA antagonist amantadine potentiates DA synthesis capacity in the basal ganglia of normal humans (Deep et al, 1999). Our present findings could represent a linkage of therapeutic response to neuroleptics with reduced DA synthesis, consistent with the development of a drug-induced depolarization block of presynaptic DA neurons in human patients with schizophrenia. This lends support to Grace's hypothesis that an antipsychotic drug-induced blockade of DA synthesis and release need not necessarily lead to decreased extracellular DA levels, but instead serves to suppress the abnormal phasic DA release in response to external excitatory stimuli (Grace, 1992). Taken together, these results show that psychotic symptoms reflect increased responsiveness to DA, due in part to an elevated rate of DA synthesis that can be rectified by neuroleptic-induced depolarization block of DA neurons.