INTRODUCTION

Atypical antipsychotic drugs, characterized by relatively high affinities for 5-HT2A serotonin receptors and lower affinities for D2-dopamine receptors (Meltzer et al, 1989), include clozapine, olanzapine, risperidone, quetiapine, and ziprasidone, and have been of considerable value in the treatment of schizophrenia (Meltzer, 1999a). The prototypical atypical antipsychotic drug clozapine has become the treatment of choice for treatment-resistant and treatment-intolerant schizophrenia and related disorders because of its greater efficacy (Kane et al, 1988; Meltzer et al, 1989) and lack of extrapyramidal sideeffects. Clozapine treatment also substantially decreases suicidality among patients with schizophrenia (Meltzer, 1999b; Meltzer and Okayli, 1995), and it may improve negative symptoms, cognition, and tardive dyskinesia. Other atypical antipsychotic drugs (eg olanzapine, risperidone, quetiapine, and ziprasidone) have also been reported to improve cognition, negative symptoms, and minimize the risks of tardive dyskinesia, when compared with typical antipsychotic drugs (Meltzer and McGurk, 1999; Meltzer et al, 1999).

Atypical antipsychotic drugs may also have serious side effects, including sedation, orthostasis, constipation and, in the case of clozapine, a substantial risk of agranulocytosis. Several studies have now demonstrated that a number of typical and atypical antipsychotic drugs may induce both short- and long-term weight gain (Taylor and McAskill, 2000; McIntyre et al, 2001; Wetterling, 2001). Recently, the related side effects of hyperlipidemia, hyperglycemia, and hypertension have attracted considerable attention. In fact, a recent study (Fontaine et al, 2001) suggested that the magnitude of weight gain and its attendant increases in morbidity and mortality may greatly diminish the positive effects of those atypical antipsychotic drugs that produce these effects to the greatest extent, that is, olanzapine and clozapine.

The mechanism(s) responsible for antipsychotic-drug-induced weight gain are not understood. Altered glucose homeostasis and metabolism as well as increased food intake have been proposed as mechanisms responsible for antipsychotic-drug-induced weight gain. Typical and atypical antipsychotic drugs have a complex pharmacology, interacting with a number of serotonergic (eg 5-HT1A, 5-HT2A, 5-HT2C, 5-HT6 and 5-HT7 (Roth et al, 1992, 1994, 1998), dopaminergic (eg D2, D3, and D4 (Seeman and Lee, 1975; Seeman et al, 1997; Van Tol et al, 1991), histaminergic (eg H1 (Peroutka et al, 1980) and H4 (Nguyen et al, 2001)), adrenergic, and muscarinic acetylcholine receptors (Bolden et al, 1992; Zeng et al, 1997). In mice, targeted deletion of some of these receptors can have effects on body weight; for example, 5-HT2C receptor knockout mice are obese (Tecott et al, 1995), whereas m3-muscarinic receptor knockout mice are lean (Yamada et al, 2001). The results with 5-HT2C receptor and m3-muscarinic receptor knockout mice imply that the interactions of antipsychotic drugs with these receptors may be responsible for weight gain. Others have suggested that interactions of typical and atypical antipsychotic drugs with H1-histamine receptors, or their relative affinities for the D2 and 5-HT2A receptors, may be responsible for antipsychotic-drug-induced weight gain (eg see Wetterling, 2001).

To determine which receptor(s) are most likely to be responsible for antipsychotic-drug-induced weight gain, we determined the affinities of 17 selected typical and atypical antipsychotic drugs for 12 human or rodent cloned receptors using the resources of the National Institute of Mental Health Psychoactive Drug Screening Program (NIMH-PDSP). We then attempted to correlate these receptor affinities with short-term weight gain data derived from a previous meta-analysis of the literature (Allison et al, 1999). The most robust predictor of a drug's propensity to induce weight gain was its affinity for the H1 histamine receptor. Our findings predict that newer atypical antipsychotic drugs that have low H1-histamine receptor affinities (eg ziprasidone, aripiprazole) will have low weight gain liabilities. These results also suggest that centrally active drugs with relatively higher H1-histamine receptor affinities should be used with caution because of the propensity of these drugs to induce weight gain and the potential for subsequent hyperlipidemia, hypertension, and hyperglycemia.

Methods

Materials

Typical and atypical antipsychotic drugs were obtained as previously described (Rauser et al, 2001), with the following exceptions: molindone was obtained from Research Diagnostics, Inc. (Flanders, NJ), whereas aripiprazole was a kind gift of Richard Mailman (University of North Carolina, Chapel Hill, NC). Cloned receptor preparations were obtained via the resources of the NIMH-PDSP as previously detailed (Rothman et al, 2000), with the exception of the human H1-histamine receptor, which was cloned via PCR amplification of ‘Quick-Clone’ human cDNA (Clontech) and subcloned via NotI adaptors into the pUNIV-SIG expression vector (Kroeze Roth, unpublished vector sequence). The entire coding region was sequenced by automated dsDNA sequencing (Cleveland Genomics, Cleveland, OH) to verify that PCR-induced mutations had not occurred.

Radioligand Binding Assays

All binding assays were as previously described (Glennon et al, 2000; Rothman et al, 2000), primarily using cloned human receptor preparations. On-line protocols are available at http://pdsp.cwru.edu. Initial screening assays were performed at 10 μM concentration with quadruplicate determinations, and the percent inhibition of total specific binding was quantified. For drugs that caused >50% inhibition at 10 μM, full dose–response studies were performed using drug concentrations spanning 5–6 orders of magnitude. Typically, specific binding represented >90% of total binding. Ki values were calculated using GraphPad Prism (GraphPad Software, San Diego, CA, USA). All Ki values represent the mean of 3–4 separate determinations.

Statistical Analysis

Short-term weight gain data were derived from a meta-analysis of the literature (Allison et al, 1999). For all receptor–drug pairs, binding affinities (Ki's) in nanometers were converted to their log values; a maximum Ki of 10 000 nM (log Ki=4.0) was used for low-affinity interactions. All statistical analyses were carried out with the XLstat version 5.0 add-on package to Microsoft Excel from Addinsoft (Paris, France). For all analyses, P-values of less than 0.05 were considered significant.

RESULTS

Typical and Atypical Antipsychotic Drugs have a Complex Pharmacology

Initial screening assays were performed with a large number of typical and atypical antipsychotic drugs. In preliminary studies, we discovered that selected typical and atypical antipsychotic drugs interacted with virtually every biogenic amine receptor tested, in confirmation of many prior studies (see on-line database at http://pdsp.cwru.edu/pdsp.asp for comparison with published studies). We therefore narrowed our studies to a selected group of biogenic amine receptors that have been most closely linked with weight gain or loss.

Comparison of Weight Gain with Selected Biogenic Amine Receptor Affinities

Since reliable weight gain data are available for only a subset of antipsychotic drugs (Allison et al, 1999), we more closely examined this set of drugs, which includes both typical and atypical antipsychotics. The typical antipsychotic drugs chosen for study were chlorpromazine, perphenazine, trifluoperazine, thioridazine, thiothixene, fluphenazine, haloperidol, molindone, and pimozide. The atypical antipsychotic drugs chosen were clozapine, olanzapine, loxapine, sertindole, risperidone, ziprasidone, quetiapine, and aripiprazole. Of importance for the present study, the typical antipsychotic drugs tested included those reported to induce substantial weight gain (chlorpromazine, perphenazine, thioridazine, thiothixene), those that induce little weight gain (fluphenazine, haloperidol), and those reported to induce weight loss (molindone, pimozide) in short-term studies (see Allison et al, 1999). Additionally, the atypical antipsychotic drugs included those that are reported to induce substantial weight gain in short-term studies (clozapine, olanzapine, quetiapine), those that induce moderate weight gain (risperidone), and those reported to induce minimal weight gain in short-term studies (ziprasidone, aripiprazole). We then examined the affinities of these receptors at a subset of 12 cloned biogenic amine receptors.

As can be seen graphically in Figure 1, and in Table 1, the tested drugs had a relatively complex pharmacology with substantial affinities for nearly all tested receptors. For the most part, high H1-histamine receptor affinities were associated with drugs that cause weight gain, whereas drugs that induce little or no weight gain had low H1-histamine receptor affinities. It is interesting to note that three atypical antipsychotic drugs that do not induce substantial short-term weight gain (ziprasidone, aripiprazole, risperidone) have relatively high affinities for 5-HT2C serotonin receptors, in confirmation of recent studies showing that two of these drugs (risperidone, ziprasidone) are potent 5-HT2C inverse agonists (Rauser et al, 2001). These results imply that neither 5-HT2C receptor affinity nor 5-HT2C receptor inverse agonist activity reliably predicts atypical antipsychotic drug-induced weight gain.

Figure 1
figure 1

Surface plot of affinities of 17 typical and atypical antipsychotic drugs for 12 receptors. Drugs are in order of their propensity to induce weight gain, from the least on the left to the most on the right.

Table 1 Weight gain and binding data for typical and atypical antipsychotic drugs as used in this study

The Spearman correlation coefficients (a two-tailed nonparametric test was used, since the data deviate significantly from normality) of the relationship between the propensity of the drugs examined to induce weight gain and their affinities at the 12 receptors tested are given in Table 2. For 10 of 12 receptors, there was a negative relationship between Ki and weight gain; for the D2 and 5-HT7 receptors, there was a positive correlation between Ki values and weight gain. Only the correlations between weight gain and H1, α1A, 5-HT2C, and 5-HT6 receptor affinities were statistically significant.

Table 2 Correlation between propensity of 17 drugs to induce weight gain and affinity (log Ki) for 12 receptors

It has been amply demonstrated that the ratio of an antipsychotic drug's affinity for the 5-HT2A and the D2 receptors is predictive of its atypicality (Meltzer et al, 1989). We therefore examined the relationship between the propensity of drugs to induce weight gain and the ratio of affinities for the various receptors tested and the affinity for the D2 receptor (Table 3). When analyzed in this fashion, the affinities of eight of the 11 receptors examined were correlated to weight gain; however, the strongest relationship was again seen for the H1, α1A, and 5-HT2C receptors. Interestingly, no association was evident correlating weight gain with 5-HT2A/D2 affinity ratios, as has been recently suggested by others (eg Wetterling, 2001). In subsequent analyses, only the log Ki values, and not the ratio data, were used.

Table 3 Correlation between propensity of 17 drugs to induce weight gain and the ratio of the affinity for 11 receptors to the affinity for D2 receptors

In order to examine the data more fully, a principal components analysis was performed (Table 4 and Figure 2). Initially, it was observed that there was a statistically significant relationship among many of the receptor affinities; this led to the computation of a correlation matrix based on the nonparametric Spearman correlation coefficient (Table 4). Because many of the variables studied were significantly correlated, a principal components analysis, which essentially ‘uncorrelates’ a set of correlated variables, was carried out. As can be seen in Figure 2, the 17 drugs studied could be segregated by this analysis into those that induce considerable weight gain and those that induce little or no weight gain. Drugs that induce weight gain tended to have low values on principal component 1, and high values on principal component 2. Principal component 1 explained 40% of the total variation in the data matrix, and principal component 2 explained an additional 18%. Principal components are composites of all the variables in a data matrix, and are mutually orthogonal in multidimensional space; the contributions of the variables to each principal component are weighted in such a fashion as to make them mutually orthogonal. The variables most closely correlated with principal component 1 were the affinities at the H1, α2A, α2B, 5-HT2A, 5-HT2C, and 5-HT6 receptors, and with principal component 2 they were weight gain and affinities for the D2, 5-HT1A, and 5-HT7 receptors (data not shown).

Table 4 Correlation matrix based on Spearman's nonparametric coefficient showing correlation between all pairs of variables examined
Figure 2
figure 2

Principal components analysis of weight gain and receptor affinity data. Drugs that induce weight gain are shown in bold face. The horizontal axis is the first principal component, which explains 41% of the total variance in the data matrix, and the vertical axis is the second principal component, which explains 17% of the total variance in the data matrix.

The preceding analysis suggested that it might be possible to separate the drugs studied into two groups, those with a propensity to induce weight gain and those lacking this propensity, based on their affinities for particular receptors. To further investigate this possibility, a discriminant functions analysis was carried out (Figure 3). This type of analysis maximizes the separation between two a priori groups, and determines which of the variables contribute most to this separation. This is then followed by an a posteriori reclassification of the groups to test whether the discriminant function can, in fact, adequately separate the two a priori groups. In order to exclude the trivial outcome that the weight gain data themselves would be the best predictor of the propensity to induce weight gain, weight gain data were excluded from this analysis. From Figure 3, it is clear that the drugs that induce weight gain can be discriminated from those that do not based only on receptor affinities. The variables most highly correlated to the discriminant axis were the affinities for the H1, and α1A receptors (Table 5). In the a posteriori reclassification of the drugs, loxapine and fluphenazine, which do not induce weight gain, were classified with the group of drugs that does induce weight gain. Thus, 15 of 17 drugs (88.2%) were correctly classified by the discriminant functions analysis. We then attempted to classify the drugs by their affinities for individual receptors; only the H1 receptor affinities correctly classified as many as 15 of the 17 drugs, with loxapine and sertindole being misclassified. For loxapine, this could be a reflection of how the weight gain data were extracted from the study of Allison et al (1999), or possibly the relatively small sample size for loxapine in that study. Taken together, however, our analysis suggests that receptor binding affinities can predict a drug's propensity to induce weight gain, and that the affinity for the H1-histamine receptor is the best single predictor of that propensity.

Figure 3
figure 3

Discriminant function analysis of weight gain and receptor affinity data. The axis represents the discriminant function, which maximizes the separation between the two groups of drugs, those that do not induce weight gain (closed circles) and those that do induce weight gain (open circles). Weights of the variables that comprise the discriminant function are given in Table 5.

Table 5 Coordinates of the variables used for the discriminant functions analysis in Figure 3

DISCUSSION

The major finding of this paper is that H1-histamine receptor affinity is significantly correlated with short-term weight gain when a large number of typical and atypical antipsychotic drugs are examined. These results imply that antipsychotic drugs with relatively high H1 receptor affinities are likely to induce short-term weight gain. These results also suggest that the next generation of atypical antipsychotic drugs should be screened for H1-histamine receptor affinities and that drugs with relatively high H1-histamine receptor affinities should be avoided. It has been known for several decades that (1) centrally active drugs with high affinities for H1-histamine receptors can induce weight gain, (2) some antipsychotic drugs are potent H1-histamine receptor antagonists, and (3) antipsychotic drugs can induce weight gain. Thus, it is perhaps not surprising that H1-histamine receptor affinities predict weight gain for both selected typical and atypical antipsychotic drugs. What is surprising, however, is that H1 receptor affinity was the only variable that reliably predicted weight gain for the typical and atypical antipsychotic drugs tested.

Several lines of evidence have suggested, in fact, that other receptors might be responsible for atypical antipsychotic-drug-induced weight gain, particularly the 5-HT2C receptor. Thus, for instance, mice with targeted deletions of the 5-HT2C receptor are obese (Tecott et al, 1995). Additionally, many typical and atypical antipsychotic drugs have high 5-HT2C receptor affinities (Canton et al, 1990; Roth et al, 1992), and many antipsychotic drugs are potent 5-HT2C inverse agonists (Herrick-Davis et al, 2000; Rauser et al, 2001). Finally, 5-HT2C knockout mice are resistant to the anorectic effects of fenfluramine (Vickers et al, 1999), although chronic administration of potent and selective 5-HT2C antagonists does not induce weight gain in rats (Wood et al, 2001). The present study clearly demonstrates that 5-HT2C affinity does not predict weight gain among this group of typical and atypical antipsychotic drugs. However, this does not rule out the possibility that polymorphisms of the 5-HT2C receptor (Lappalainen et al, 1995; Reynolds et al, 2002), or that altered expression of various editing isoforms of the 5-HT2C receptor (Burns et al, 1997), might differentially contribute to variations in weight gain for individual subjects. Indeed a recent study (Reynolds et al, 2002) demonstrated that a polymorphism in the 5-HT2C gene was highly associated with weight gain induced by either chlorpromazine or risperidone. Interestingly, both chlorpromazine and risperidone are rather weak inverse agonists at h5-HT2C-INI receptors (Rauser et al, 2001), so it is unlikely that direct interaction of these drugs with 5-HT2C receptors is important for the weight gain induced by both drugs. Instead it is more likely that the 5-HT2C polymorphism predisposes individuals to weight gain via an unknown mechanism.

The mechanism(s) by which H1-histamine antagonism might induce weight gain are currently unknown, although prior studies have amply demonstrated that H1-histamine receptor antagonism increases feeding in rodents whereas H2-histamine antagonism does not (Sakata et al, 1988; Fukagawa et al, 1989). Additionally, depletion of neuronal histamine increases feeding (Menon et al, 1971; Sakai et al, 1995). Interestingly, other psychoactive compounds with high H1-histamine receptor affinities, for example, amitryptiline (Altamura et al, 1989), have been associated with significant weight gain. Finally, H1-knockout mice are relatively resistant to the anorectic actions of leptin, and are prone to obesity when placed on high-fat diets (Masaki et al, 2001a, 2001b). Taken together, these results imply that H1-histamine receptors modulate feeding behavior via a leptin-dependent mechanism.

It is also clear that mechanisms other than H1-histamine receptor blockade can also induce weight gain. Thus sulpiride, a selective D2/D3 antagonist, has virtually no affinity for H1-histamine receptors (Roth et al, in preparation), yet it induces significant long-term weight gain among individuals with schizophrenia (Taylor and McAskill, 2000). Similarly, haloperidol and fluphenazine have relatively low H1-histamine receptor affinities, yet, when given in depot formulations, have been reported to induce substantial weight gain (Taylor and McAskill, 2000). In this regard, it is interesting to note that the discriminant functions analysis predicts that fluphenazine will induce weight gain. Thus, factors independent of H1-histamine receptor affinity may contribute to weight gain induced by typical and atypical antipsychotic drugs.

Taken together, these results clearly indicate that antipsychotic drugs with high H1-histamine receptor affinities are associated with significant weight gain. These results are consistent with the evidence that two of the newer atypical antipsychotic drugs, ziprasidone and aripiprazole, are associated with less short-term weight gain. Finally, our results suggest that since the H1-histamine receptor is the most likely molecular target responsible for atypical antipsychotic-drug-induced weight gain, high H1-histamine receptor affinity should be avoided in the next generation of multireceptor atypical antipsychotic drug candidates.