Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

In Parkinson's disease, the progressive degeneration of nigrostriatal dopaminergic pathways is associated with diverse motor symptoms, including rigidity, tremor, bradykinesia, and postural instability (Jenner, 1995). In addition, patients present, often precociously, sensory and cognitive-attentional deficits together with depressed mood (Jenner, 1995). Despite increasing interest in neuroprotective strategies, Parkinson's Disease is principally treated by administration of the dopamine (DA) precursor L-dihydroxyphenylalanine (L-DOPA) (Bezard et al., 2001). However, there is evidence, albeit contentious, that L-DOPA exacerbates damage to dopaminergic neurons (Zou et al., 1999). Furthermore, L-DOPA displays variable pharmacokinetics, elicits dyskinesias and autonomic side effects, poorly improves certain motor symptoms, is largely ineffective against cognitive and mood deficits, and loses efficacy upon prolonged administration (Bezard et al., 2001). Abrupt transitions between "on" and "off" phases are particularly distressing to patients (Jenner, 1995). In light of these observations, the management of Parkinson's Disease by drugs directly stimulating postsynaptic DA receptors is of interest (Jenner, 1995; Montastruc et al., 1999). Such dopaminergic agents possess neuroprotective properties, mediated by both dopaminergic (autoreceptor) and nondopaminergic mechanisms (Zou et al., 1999) and elicit less marked dyskinesia (Uitti and Ahlskog, 1996; Rascol et al., 2000). Furthermore, they may improve mood and cognitive function (Weddell and Weiser, 1995; Nagaraja and Jayashree, 2001). In addition to adjunctive therapy, recent studies support the long-term efficacy of dopaminergic agonists in monotherapy, thereby delaying the introduction of L-DOPA (Rascol et al., 2000). Nevertheless, "sleep-attacks", sedation, and both psychiatric and cardiovascular side effects complicate utilization of dopaminergic agonists (Friedman and Factor, 2000).

The above-mentioned panoply of desirable and undesirable actions varies among antiparkinson agents (Uitti and Ahlskog, 1996). Such differences likely reflect contrasting patterns of interactions at sites other than dopamine D₂ receptors (Uitti and Ahlskog, 1996). D₃ receptors are of particular interest, although it remains controversial as to whether their engagement contributes to therapeutic and/or psychiatric and motor side effects (Millan et al., 2000b; Joyce, 2001). Activation of D₄ receptors does not, on the other hand, participate in the improvement of Parkinson's Disease (Newman-Tancredi et al., 1997; Oak et al., 2000). Although D₁ receptor agonists display antiparkinson activity in experimental models, their clinical efficacy upon long-term administration remains uncertain, and their stimulation is not obligatory for therapeutic activity (Jenner, 1995; Gulwadi et al., 2001). Furthermore, the relative roles of D₁ versus closely related D₅ sites remain unclear (see Discussion).

Inasmuch as 1) Parkinson's Disease is aggravated by degeneration of locus coeruleus-derived adrenergic and raphe-derived serotonergic pathways (Brefel-Courbon et al., 1998; Jellinger, 1999); and 2) adrenergic and serotonergic mechanisms modulate dopaminergic transmission, motor behavior, mood, and cognitive function (Meneses, 1999; Millan et al., 2000c), it is important to consider potential actions of antiparkinson agents at adrenoceptors (ARs) and 5-HT receptors. Although surprisingly little information is available, talipexole and 6,7-dihydroxy-N,N-dimethyl-2-ammotetralin (TL99) are known to possess agonist properties at native ₂-ARs (Horn et al., 1982; Meltzer et al., 1989). In contrast, blockade of ₂-ARs by piribedil reinforces frontocortical adrenergic, dopaminergic, and cholinergic transmission and favorably influences mood and cognitive-attentional function (Millan et al., 2000c, 2001a; Maurin et al., 2001; Nagaraja and Jayashree, 2001; Gobert et al., 2002). In addition to antagonist actions at ₂- and ₁-ARs, bromocriptine reveals pronounced affinity for 5-HT_1A receptors (McPherson and Beart, 1983; Jackisch et al., 1985; Uitti and Ahlskog, 1996). Other antiparkinson agents known to recognize 5-HT_1A and/or 5-HT_2A receptors are lisuride, terguride, and roxindole (Jackson et al., 1995; Uitti and Ahlskog, 1996).

The purpose of the present studies was to consolidate these fragmentary data by evaluating the actions of 14 dopaminergic agonists (antiparkinson agents) at multiple classes of monoaminergic receptor. In addition, actions at muscarinic (M₁) sites and histamine (H)₁ sites were evaluated in light of 1) their role in the control of motor behavior, mood, and cognition (Bacciottini et al., 2001; Brown et al., 2001); 2) alterations in histaminergic and cholinergic transmission in Parkinson's Disease (Jellinger, 1999; Anichtchik et al., 2000); and 3) the use of anticholinergic agents for management of refractory tremor (Wilms et al., 1999). The strategy adopted was as follows. First, using competition binding assays, drug affinities were determined at recombinant, stably transfected, human receptors as well as at rat _2D-ARs¹ and at native H₁ receptors. Second, to facilitate analysis of the extensive database and comparisons of drug profiles, a correlation matrix was constructed: data were subjected to principle components analysis (PCA) and then drugs were classified by hierarchical (cluster) analysis in accordance with their overall homology. This innovative multivariate approach to drug comparisons has the advantage that it is not founded upon specific hypotheses requiring testing via post hoc, inferential statistics. Rather, by fully and simultaneously exploring total variance, it permits the objective identification and interpretation of hidden patterns not revealed by visual inspection or drug-by-drug/receptor-by-receptor comparisons (Krzanowski, 2000; Millan et al., 2000a; Carlsson et al., 2001). Third, as described in the accompanying articles (Newman-Tancredi et al., 2002a,b), efficacies of antiparkinson agents were determined at (the majority of) monoaminergic receptor subtypes incorporated into these multivariate analyses.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Determination of Drug Affinities. Procedures used for the determination of drug affinities have been described in detail previously (Newman-Tancredi et al., 1997; Millan et al., 2001a). They are summarized in Tables 1, 2, and 3. Isotherms were subjected to nonlinear regression analysis by use of the program PRISM (GraphPad Software, San Diego, CA) to yield IC₅₀ values. These were subsequently transformed into K_i values according to the Cheng-Prussof equation K_i = IC₅₀/(1 + L/K_d), where L corresponds to the radioligand concentration and K_d to its dissociation constant.

View this table: [in this window] [in a new window]   TABLE 2 Competition binding conditions at recombinant, human h1-, h2-, and h-adrenoceptors and at rat r2-adrenoceptors

Multivariate Analysis: Principal Components Analysis. The database used for multivariate analysis comprised the affinities ("parameters") for 14 drugs at (21) separate receptors ("variables") indicated in Tables 4, 5, and 6. (pK_i values of <5.0 were considered as 5.0 for these analyses.) Because all parameters are intrinsically equivalent, they were not transformed ("standardized"): pK_i values are the negative logarithmic expression of affinities. After construction of a correlation matrix (Pearson product-moment coefficients) across all parameters (Table 7), the database was subjected to PCA (Krzanowski, 2000) using SPAD-3, a computer program developed by the Centre International de Statistiques et d'Informatiques Appliquées (St. Mandé, France). This generates a "multidimensional space" of 21 axes from the database, with all axes mathematically "perpendicular" to each other. The first axis, principal component (PC)1, represents the linear combination of all parameters (affinities) in the data set that accounts for the maximal possible variance. Correspondingly, loading values (correlation coefficients) in Table 8 indicate the contribution of individual parameters to PC1. Successive axes (PC2 onwards) account for progressively less variance. PCs 1 to 3, which accounted for a substantial majority of variance (see Results), were two-dimensionally represented in scatter diagrams ("biplots") upon which drugs were superimposed together with the parameters underlying their dispersion.

Multivariate Analysis: Hierarchical Cluster Analysis. After PCA, and likewise exploiting SPAD-3, drugs were hierarchically ("cluster") classified in accordance with their overall homology to yield a binary dendrogram (Krzanowski, 2000). Using an "agglomeration" algorithm, the array of drugs was progressively ("hierarchically") fused into subclusters and clusters until it comprised a single group. With the formation of each successive cluster, the loss of "objective function value" (information) was constrained as much as possible, that is, the intragroup compared with intergroup variance was minimized. The length of bars between pairs of drugs in the two-dimensional dendrogram reflects their dissimilarity, that is, the shorter the distance, the more closely related the pairs of drugs. Nodes on the dendrogram therefore represent the consecutive aggregation of two individual elements (drugs or drug clusters).

Radar Plots. "Radar" representations of binding profiles at certain key receptors were constructed to further visualize similarities and differences in drug binding profiles.

Drugs. Pramipexole dihydrochloride, piribedil hydrochloride, and ropinirole were synthesized by Institut de Recherches Servier (Paris, France). Lisuride maleate and terguride were donated by Schering (Berlin, Germany); bromocriptine, ()-quinpirole, pergolide, and TL99 were purchased from Sigma/RBI (Natick, MA); apomorphine hydrochloride was purchased from Sigma (St. Quentin Fallavier, France); and roxindole was donated by Merck (Darmstadt, Germany) and talipexole (BHT-920) by Boehringer Ingelheim GmbH (Ingelheim, Germany). Cabergoline was obtained from Farmitalia Carlo Erba (Rueil-Malmaison, France). Quinelorane dihydrochloride was a gift from Eli Lilly & Co. (Indianapolis, IN).

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

General Comments. In view of the large number of drugs and binding sites examined, a detailed text description of all (~300) interactions cannot be presented below. Full data are shown in Tables 4 to 7.

Dopamine hD_2S, hD_2L, hD₃, and hD₄ Receptors. There was a substantial (5000-fold) range in drug affinities at hD_2S receptors (Table 4).² For example, the affinity of cabergoline was very pronounced compared with that of pramipexole. At hD_2L receptors (which possess a 29 amino acid insert in the third intracellular loop), affinities of drugs likewise varied broadly and were similar to those at hD_2S sites. There was likewise marked (~1000-fold) variability in drug affinities at hD₃ sites. The ratio of drug affinities at hD₃ compared with hD_2L and hD_2S sites differed considerably from modest (e.g., cabergoline and pergolide) to pronounced (e.g., pramipexole). The variation in drug affinities at hD₄ receptors was also striking (~400-fold). Certain drugs displayed considerably higher affinities at hD₄ versus hD_2S/hD_2L sites (such as apomorphine), whereas others showed modest differences (such as piribedil) or a marked preference for hD_2S/hD_2L sites (such as bromocriptine).

Dopamine hD₁ and hD₅ Receptors. Drug affinities for hD₁ sites were substantially lower than for hD_2S and hD_2L receptors: apomorphine and cabergoline showed the least and most pronounced difference, respectively (Table 4). There was ~200-fold variability in drug affinities at hD₁ sites with certain agents, including pramipexole, displaying negligible affinity. For all drugs manifesting significant affinity for hD₁ receptors, affinities were higher at hD₅ receptors. This difference was mild for certain drugs, such as bromocriptine, and pronounced for others, such as pergolide.

h_1A-, h_1B-, and h_1D-ARs. For each drug, affinities at h_1A-, h_1B-, and h_1D-ARs were similar (Table 5). Affinities varied over a ~10,000 range from negligible (e.g., pramipexole) through intermediate (e.g., apomorphine) to pronounced (e.g., bromocriptine).

View this table: [in this window] [in a new window]   TABLE 5 Affinities (pKi values) of antiparkinson agents at recombinant human h1-, h2-, and h-adrenoceptors and at native, rat r2D-ARs Data are pKi values (minus log Ki values) and, unless indicated, are expressed as means ± S.E.M. values of at least three independent determinations carried out in triplicate.

r_2D-, h_2A-, h_2B-, and h_2C-ARs. There was considerable (>10,000) variability in drug affinities for h_2A- and r_2D-ARs, ranging from quinerolane (negligible) to lisuride (very high), with piribedil, talipexole, and several other drugs showing intermediate values (Table 5). Only cabergoline revealed (slightly) higher affinity for r_2D- versus h_2A-ARs. No drug clearly differentiated h₂-AR subtypes, although pramipexole showed negligible affinity for h_2C- versus h_2A- and h_2B-ARs. Bromocriptine and roxindole were the only drugs to show similar or weaker affinities at h_2A- compared with h₁-AR subtypes.

h₁- and h₂-ARs. Only four ligands (lisuride, terguride, bromocriptine, and roxindole) displayed significant affinity for h₁-ARs (Table 5). Bromocriptine and roxindole showed similar affinity at h₂-ARs, whereas lisuride and terguride revealed higher affinity for h₂- compared with h₁-ARs. Compared with h₂-ARs, affinities at h₁- and h₂-ARs were relatively weak for all drugs, and only lisuride and terguride approached affinities seen at h₁-ARs.

h5-HT_1A Receptors. There was substantial (10,000-fold) variation in drug affinities at h5-HT_1A receptors varying from quinerolane (<5.0) to roxindole (9.9) (Table 6). Several other agents, such as bromocriptine and apomorphine, showed marked affinity for h5-HT_1A sites although others, such as piribedil and talipexole, showed only modest affinities.

h5-HT_1B and h5-HT_1D Receptors. Several drugs, including piribedil and talipexole, failed to recognize h5-HT_1B receptors, although others, such as cabergoline and pergolide, displayed modest affinities (Table 6). At structurally related h5-HT_1D receptors, affinities were generally elevated compared with h5-HT_1B receptors, notably for cabergoline and pergolide, whereas these sites were not recognized by piribedil and talipexole.

h5-HT_2A, h5-HT_2B, and h5-HT_2C Receptors. For closely related h5-HT_2A, h5-HT_2B, and h5-HT_2C receptors, binding profiles were generally comparable (Table 6). However, certain agents, including cabergoline and pergolide, showed substantially lower affinity for h5-HT_2C versus h5-HT_2A and h5-HT_2B receptors. No drug clearly differentiated h5-HT_2A from h5-HT_2B sites. There was marked (>100-fold) variability in drug affinities at h5-HT_2A, h5-HT_2B, and h5-HT_2C receptors in each case: piribedil, talipexole, and pramipexole, for example, showed low affinities compared with apomorphine, bromocriptine, and, in particular, cabergoline and pergolide.

H₁ Receptors. Affinities of drugs at H₁ receptors varied from negligible (for example, apomorphine, bromocriptine, piribedil, and pramipexole) to modest (for example, terguride and lisuride) (Table 6).

Interrelationship among Binding Sites: Correlation Matrix. An important and general feature of the correlation matrix was the lack of negative correlation coefficients, that is, in no case was high affinity at one receptor associated with low affinity at a second site (Table 7). This feature was reflected in numerous statistically significant correlation coefficients among pairs of receptors (Table 8). Some were unsurprising, such as between hD_2S and hD_2L receptors, among h₁-AR and h₂-AR subtypes and between h5-HT_2A and h5-HT_2C receptors. More notably, there were high correlation coefficients between affinities at hD_2S and hD_2L receptors on the one hand, and h5-HT_2A, h5-HT_2B, and h5-HT_2C receptors on the other. Furthermore, hD₅, h5-HT_2C, and h5-HT_2A sites were well correlated. Similarly, high correlation coefficients were observed between hD₂ receptors and ₁- and ₂-AR subtypes. On the other hand, affinities at hD₄ receptors were poorly correlated with affinities at hD_2S, hD_2L, and hD₃ receptors, as well as other sites with the exception of H₁ receptors. H₁ receptors were themselves distinguished by relatively poor (and, in certain cases, nonsignificant) correlation coefficients with other receptor types.

Principle Components Analysis. Application of PCA to the pK_i values revealed that almost 90% of variance could be accounted for by three axes: PC1, PC2, and PC3 (Figs. 1, 2, and 3; Table 8). That is, a reduction of "dimensionality" from 21 to a "subspace" of three axes preserved almost the entire variance in the data. This permitted the construction of bidimensional "biplots" (1/2, 1/3, and 2/3) upon which both the drugs and the variables that contributed to their dispersion could be projected (Figs. 1-3).

View larger version (15K): [in this window] [in a new window]   Fig. 1.   Principle component analysis of drug interactions at multiple binding sites examined herein: axes (principal components) 1 and 2. The database (pKi values) comprised 14 drugs and 21 binding sites indicated in Tables 4 to 6. These values (a total of 294) were simultaneously analyzed in a "polydimensional" space, as described under Materials and Methods. Top (scatter diagram), plots the distribution of each parameter (specific binding sites) as a function of its contribution to the first two principle components (axes), which accounted for 76.27 and 6.53% of variance, respectively. Reflecting extensive, positive "covariance" indicated in Table 8, all parameters (vectors) clustered in the same direction along axis 1. Within the plot, the "circle of correlation" is depicted. The degree of correlation among specific parameters is revealed by projection of vectors onto this circle, and by the angle between pairs of vectors, which is inversely correlated to their degree of correlation (Table 8). For example, in the bottom left quadrant, all three 1-AR subtypes are highly intercorrelated, whereas they are poorly correlated to hD4 receptors located in the upper quadrant (Table 8). The tips of the vectors, when vertically or horizontally projected onto PC1 and PC2, respectively, yield loading values indicated in Table 8. Bottom, drugs are superimposed upon PCs 1 and 2. S, 5-HT.

View larger version (15K): [in this window] [in a new window]   Fig. 2.   Principle component analysis of drug affinities at multiple binding sites examined herein: axes (principal components) 1 and 3. Top, distribution of each parameter (specific binding sites) as a function of principle components (axes) 1 and 3, which accounted for 76.27 and 5.4% of variance, respectively. Bottom, drugs are superimposed upon PCs 1 and 2. See the legend to Fig. 4 and Results for further details. S, 5-HT

View larger version (14K): [in this window] [in a new window]   Fig. 3.   Principle component analysis of drug affinities at multiple binding sites examined herein: axes (principal components) 2 and 3. Top, distribution of each parameter (specific binding sites) as a function of principle components (axes) 2 and 3, which accounted for 6.53 and 5.4% of variance, respectively. Bottom, drugs are superimposed upon PCs 2 and 3. See the legend to Fig. 4 and Results for further details. S, 5-HT.

Hierarchical (Cluster) Analysis of Global Drug Homology. From the dendrogram of overall drug homology generated by analysis of the total database, several drug clusters not apparent from inspection of the biplots could be recognized (Fig. 4). It is pragmatic to comment the dendrogram in a direction opposite to that of its mathematical construction, and the most striking separation between drugs was at the first "node", which yielded two major subdivisions.

View larger version (16K): [in this window] [in a new window]   Fig. 4.   Hierarchical (cluster) classification of overall drug homology based upon binding profiles at all receptors. The database (pKi values) exploited for generation of the dendrograms is given in Tables 4 to 6. Note that the distance between two individual drugs is inversely proportional to their overall homology. That is, drugs situated adjacently present very similar binding profiles, whereas those widely separated show substantially different binding profiles. The chemical class of each drug is indicated.

Radar Plots. The radar representations of Fig. 5 complement the dendrogram in exemplifying similarities and differences among various drugs at specific receptor types discussed above.

View larger version (20K): [in this window] [in a new window]   Fig. 5.   "Radar" representations of the binding profiles of (closely related) pairs of drugs at certain receptor types. In each radar plot, the two drugs depicted show broadly similar profiles. Affinities are indicated in pKi values. Thus, the greater the distance from the central node of the radar plot, the higher the drug affinity for a specific binding site. Plots of quinpirole and quinelorane are not shown for reasons of space: they presented very similar profiles close to those of pramipexole and ropinirole.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

hD_2S and hD_2L Receptors. Although benzamides display contrasting affinities at hD_2L compared with hD_2S receptors, other classes of antagonist show similar affinity; likewise, all agonists examined to date (including several antiparkinsonian agents) revealed comparable affinities for these sites (Leysen et al., 1993). The present observations extend such reports in demonstrating similar affinities of numerous antiparkinson drugs at hD_2S versus hD_2L sites. Such information is important because 1) D_2S versus D_2L sites present differential patterns of post-translational processing, coupling, regulation, and localization; 2) D_2S autoreceptors modulate DA release and may contribute to neuroprotective properties of antiparkinson agents; and 3) postsynaptic D_2S and (predominant) hD_2L sites, perhaps via contrasting interactions with D₁ receptors, differentially control motor function (Zou et al., 1999; Usiello et al., 2000).

hD₃ and hD₄ Receptors. Comparisons of affinities at hD₃ versus hD₂ sites should be made cautiously in the light of multiple affinity states of the latter (Mierau et al., 1995; Coldwell et al., 1999; Perachon et al., 1999). Nevertheless, the high potency of all agents for hD₃ sites underscores their potential relevance to beneficial and/or undesirable properties of antiparkinson drugs (Millan et al., 2000b; Joyce, 2001). The modest correlation coefficients of hD₄ to hD_2S/hD_2L/hD₃ receptors indicate distinctive structure-activity relationships, in line with the discovery of many selective hD₄ receptor antagonists. Bromocriptine and piribedil displayed modest affinity, and antagonist properties (Newman-Tancredi et al., 2002a), at hD₄ receptors indicating, as discussed elsewhere, that their stimulation is not mandatory for clinical efficacy (Rondot and Ziegler, 1992; Jenner, 1995; Newman-Tancredi et al., 1997). Indeed, blockade of D₄ receptors may improve cognitive-attentional processing (Oak et al., 2000).

hD₁ and hD₅ Receptors. Surprisingly, affinities were well correlated between hD₁ and hD_2S/hD_2L sites, suggesting that structure-activity relationships are less distinct than might be imagined. Joint D₁/D₂ receptor stimulation may improve therapeutic efficacy for drugs such as apomorphine and pergolide (Jenner, 1995; Markham and Benfield, 1997; Perachon et al., 1999; Aizman et al., 2000). Functional interactions among (partially colocalized) D₁ and D₃ receptors are also of importance in the actions of L-DOPA and other antiparkinson agents (Karasinska et al., 2000; Joyce, 2001). Nevertheless, the low affinities of clinically effective drugs, such as pramipexole and ropinirole, at hD₁ sites support the notion that their engagement is not requisite for therapeutic efficacy. Although we corroborate the preference of apomorphine for hD₅ versus hD₁ sites (Sunahara et al., 1991; Demchyshyn et al., 2000), we found (~5-fold) higher affinities of apomorphine, bromocriptine, lisuride, and pergolide at hD₅ receptors compared with these studies. One factor underlying this difference may be the use of Chinese hamster ovary versus COS-7 cells. Indeed, Kimura et al. (1995) (using GH4C1 cells) similarly concluded that the cell line confers distinctive binding properties to hD₁ and hD₅ receptors. D₅ sites are of significance in several respects: 1) multivariate analyses revealed that hD₅ affinities discriminate antiparkinson agents; 2) D₅ receptors are situated on striatal dopaminergic, cholinergic, and GABAergic neurons (Ciliax et al., 2000); and 3) antisense probes against D₅ and D₁ receptors potentiated and inhibited, respectively, induction of rotation by D₁/D₅ agonists in unilateral substantia nigra-lesioned rats (Dziewczapolski et al., 1998). Differential modulation of motor function is supported by the contrasting phenotypes of mice lacking D₅ versus D₁ receptors and their distinctive patterns of localization (Sibley, 1999; Ciliax et al., 2000).

h₂ and h₁-ARs. The observations herein amplify isolated studies (Uitti and Ahlskog, 1996) of antiparkinson agents at r₁- and r₂-ARs in demonstrating that many recognize h₁- and h₂-AR subtypes. The present data thus complement reports of the weak (agonist) interaction of pramipexole with r₂-ARs (Mierau et al., 1995) and of actions of apomorphine and bromocriptine at hippocampal r₂-ARs (Jackisch et al., 1985). Furthermore, the high affinity of TL99 for h₂-AR subtypes amplifies observations with native r₂-ARs (Martin et al., 1983). Of particular interest, whereas talipexole behaves as an agonist at ₂-ARs (Meltzer et al., 1989), piribedil manifests antagonist properties. Correspondingly, in contrast to talipexole, piribedil reinforces corticolimbic adrenergic and cholinergic transmission (Millan et al., 2000c, 2001a; Gobert et al., 2002), actions contributing to its favorable influence upon cognitive function and mood (Brefel-Courbon et al., 1998; Bezard et al., 2001; Maurin et al., 2001; Nagaraja and Jayashree, 2001). Extending work with native ₁-ARs, bromocriptine, lisuride, terguride, and roxindole displayed high affinities at h₁-AR subtypes (McPherson and Beart, 1983; Uitti and Ahlskog, 1996). Potent blockade of ₁-ARs may interfere with the influence of antiparkinson agents upon motor performance and perturb cardiovascular function (Hieble et al., 1995; Millan et al., 2000).

h₁ and h₂-ARs. The finding that several drugs recognize h₁- and h₂-ARs is of interest. First, ₁/₂-ARs are excitatory to corticostriatal glutamatergic afferents (Niittykoski et al., 1999). Second, they activate dopaminergic, adrenergic, and serotonergic pathways in cortex and nucleus accumbens (Millan et al., 2000; Tuinstra and Cools, 2000). Third, stimulation of ₁/₂-ARs enhances cognitive function and improves mood (O'Donnell et al., 1994). Fourth, stimulation and blockade of central ₁/₂-ARs elicits and blocks tremor, respectively (Wilms et al., 1999).

h5-HT Receptors. Although all ligands showed some affinity for h5-HT_1A receptors, extending studies of native sites (Uitti and Ahlskog, 1996), marked differences among antiparkinson agents were seen at 5-HT₂ receptor subtypes. High affinities of lisuride, terguride, cabergoline, and pergolide at h5-HT_2A (and h5-HT_2C) receptors underpin studies showing that ergot-related compounds interact with native "5-HT₂" receptors (Beart et al., 1986; Uitti and Ahlskog, 1996; Markham and Benfield, 1997; Fariello, 1998). Interestingly, their marked serotonergic affinities were mimicked by the structurally distinct roxindole and apomorphine (Uitti and Ahlskog, 1996; Newman-Tancredi et al., 1999). Actions of antiparkinson drugs at 5-HT_2A/2C sites may, as discussed in the accompanying article (Newman-Tancredi et al., 2002b), influence motor function and mood.

H₁ and Muscarinic Receptors. Lisuride interacts with rat H₁ sites (Beart et al., 1986), an observation extended here to a further species and other drugs. Such actions at H₁ receptors are of potential importance. First, H₁ receptors modulate motor function, and inhibit and enhance striatal dopaminergic and cholinergic transmission, respectively (Bacciottini et al., 2001; Brown et al., 2001). Second, they influence arousal and cognition (Brown et al., 2001). Third, H₁ receptor blockade encourages sleep and elicits sedation, a troublesome symptom of treated and untreated parkinsonian patients (Brown et al., 2001; Friedman and Factor, 2000). Fourth, rats sustaining 6-hydroxydopamine lesions of the substantia nigra and Parkinson's Disease patients show an increase in striatal histaminergic innervation (Anichtchik et al., 2000). Reflecting functional interplay among dopaminergic and cholinergic networks in basal ganglia, muscarinic antagonists suppress tremor and dyskinesias provoked by L-DOPA, although side effects compromise their utilization (Wilms et al., 1999; Bezard et al., 2001). However, antiparkinson agents tested herein did not occupy cloned, human M₁ receptors (for all drugs, pK_i values of <6.0).

Hierarchical (Cluster) Analysis. High versus low affinities at multiple 5-HT and hD₅ receptors underpinned a major subdivision of agents into two groups. This association is intriguing because "selective" hD₁/hD₅ receptor ligands show pronounced affinity for h5-HT_2A and h5-HT_2C receptors (Millan et al., 2001b). Of drugs not interacting with serotonergic receptors, the data support experimental use of quinpirole and quinelorane as selective D₂-like receptor agonists. Furthermore, inasmuch as the receptor profiles of ropinirole and pramipexole were very similar, they should display common functional effects distinguishable from those of cabergoline and roxindole and from older agents such as bromocriptine and apomorphine. As regards piribedil and talipexole, which likewise recognized dopaminergic but not serotonergic receptors, it is important to emphasize their opposite antagonist and agonist properties at ₂-ARs, respectively; indeed, piribedil seems to be unique in simultaneously activating D₂/D₃ receptors and blocking ₂-ARs without markedly interacting with 5-HT receptors (Newman-Tancredi et al., 2002a,b). On the contrary, among ligands with pronounced serotonergic properties, cabergoline and roxindole were remarkably similar to pergolide and bromocriptine, respectively. Terguride and TL99, on the other hand, closely resembled lisuride and apomorphine, respectively. Certain closely related drugs possess similar structures, for example, pergolide and cabergoline. However, ropinirole/pramipexole and piribedil/talipexole presented similar binding profiles despite their chemical distinctiveness. Thus, chemical structure does not provide a satisfactory basis for prediction of receptor binding profiles.

Principal Component Analysis. The compound nature of PC1, which accounted for 76% variance, reflects marked correlation among receptors. That is, with the exception of hD₄ and H₁ receptors, all receptor types made a pronounced contribution to PC1 (Table 8). In accordance with its generally high affinity, lisuride defined one extremity of PC1 in distinction to drugs of modest affinity, such as quinpirole, which migrated at the opposite limit. PC2 and PC3, nevertheless, proved discriminant in dissociating bromocriptine from TL99 based on low and high affinities, respectively, for H₁/hD₄ receptors versus ₁-ARs. Furthermore, cabergoline was located at one limit of PC3 on the basis of higher affinity for hD₅ and h5-HT_2A versus h₁/h₂-ARs and H₁ receptors, whereas lisuride (high affinity at ₁/₂-ARs and H₁ receptors) defined the opposite extremity. Thus, PCA identified several receptors contributing to diversity in the binding profiles of antiparkinson agents. Within this framework, PCA also provided insights into relationships among the drug themselves as a function of their affinities at above-mentioned and other sites. For example, reflecting their pronounced affinities for virtually all sites, lisuride and terguride comprised a subset of drugs (clustered together) when projected onto PC1, PC2, and PC3, whereas piribedil and talipexole were likewise adjacent to each other across all PCs, corresponding to their mixed dopaminergic-adrenergic profiles in the absence of serotonergic affinities.

General Discussion. Several general features of this novel multivariate approach should be evoked. First, although multivariate techniques have been used for evaluation of biochemical abnormalities in schizophrenia (Carlsson et al., 2001) and characterization of drug pharmacokinetic profiles (Ette et al., 2001), this is their first systematic utilization for characterization of drug receptor-binding profiles. Whereas pairwise drug/drug and site/site comparisons can be misleading, multivariate strategies simultaneously analyze the entire database in a multidimensional space permitting hypothesis-free exploration of similarities and differences as a function of overall binding profiles. Although both drugs and variables must be selected, substantial databases (as herein) minimize the risk that an involuntary "bias" may distort analyses. Second, a precondition for multivariate procedures is a homogeneous and extensive database incorporating many variables and drugs. Although onerous to generate, the database can be subsequently exploited for studies of other drugs under equivalent conditions. For example, in the search for antiparkinson agents presenting novel, binding profiles differing from known agents. Third, integration of in vivo parameters would be of considerable interest (Millan et al., 2000). Fourth, multivariate analyses assume that drugs behave in an identical manner at specific receptor types. This is appropriate for structure-activity relationships focusing on drug potency but neglects potential differences in efficacy. This important issue was addressed by investigations of coupling (Newman-Tancredi et al., 2002a,b), although currently, there is no solution to the integration of contrasting drug actions (agonist versus antagonist properties) into multivariate analyses. Fifth, similarities and differences in overall binding profiles of drugs provide a framework for interpretation of their contrasting functional profiles in vivo. Multivariate analyses facilitate, thus, predictions of the beneficial and deleterious actions of novel drugs and would be most appropriately performed before their therapeutic evaluation. Indeed, it would be of considerable interest to undertake direct therapeutic comparisons of drugs possessing contrasting receptorial profiles, for example, of antiparkinson agents behaving essentially as dopaminergic agonists compared with those displaying pronounced activity at serotonergic and/or adrenergic receptors. In focusing on specific parameters, such as dyskinesias, depressive symptoms, and memory, such studies could provide key clinical information concerning the drugs in question and, more generally, clarify the significance of particular classes of monoaminergic receptor in the control of motor function, cognition, and mood in Parkinson's disease.

Concluding Comments. This comprehensive, multivariate analysis of binding profiles of diverse antiparkinson agents revealed marked and unexpected heterogeneity, a conclusion amplified by efficacy studies (Newman-Tancredi et al., 2002a,b). These observations of similarities and differences among antiparkinson agents provide a framework for improved interpretation of their experimental and clinical actions, and for a more thorough understanding of the functional significance of individual classes of monoaminergic receptor in Parkinson's disease. A multivariate strategy could instructively be applied to other agents, such as antidepressants and antipsychotics, for which actions at multiple classes of receptor are likewise critical in determining their functional profiles.