Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

The neuropeptide corticotropin-releasing factor (CRF) acts as a neurotransmitter in the central nervous system (CNS) to regulate the autonomic and behavioral responses to stress in addition to its role as a secretagogue in the hypothalamic-pituitary-adrenal (HPA) axis where it regulates the neuroendocrine stress response (Owens and Nemeroff, 1991). CRF-containing neurons are distributed in cortical, limbic, and brain stem nuclei in the rat (Swanson et al., 1983) and primate (Foote and Cha, 1988; Lewis et al., 1989; Bassett and Foote, 1992), with those CRF neurons in the amygdala and hypothalamus projecting into the medullary noradrenergic nuclei implicated in the CNS response to stress. The two CRF receptors in the CNS, CRF₁ and CRF₂, are heterogeneously located, perhaps indicating unique roles for each in modulating neuronal systems (Chalmers et al., 1996). Indeed, the discovery of urocortin, a second endogenous ligand for these receptors possessing a 10-fold greater affinity for CRF₂, and several congeners, as well as a distribution to neuroanatomical sites preferentially expressing CRF₂ mRNA, suggests a division into CRF-CRF₁ and urocortin-CRF₂ systems (Skelton et al., 2000b).

There is ample evidence from animal models that modulation of central CRF systems play a role in stress responses. For example, central administration of CRF in rats produces behaviors associated with anxiety (Sutton et al., 1982; Dunn and Berridge, 1990), whereas CRF antagonists act as anxiolytics (Heinrichs et al., 1992; Rassnick et al., 1993). Similarly, in nonhuman primates, central CRF administration produces symptoms of behavioral despair (Kalin, 1990). Adverse early experience paradigms, developed to model mood and anxiety disorders in rodents and primates, produce altered CRF neuronal systems with changes persistent into adulthood (Sanchez et al., 2001).

Alterations in CRF systems have also been found in patients diagnosed with mood and anxiety disorders. Patients with major depression (Nemeroff et al., 1984) and post-traumatic stress disorder (Bremner et al., 1997; Baker et al., 1999) exhibit elevated concentrations of CRF in cerebrospinal fluid. Similar to animal models of adverse early experience, childhood stressors such as abuse seem to cause persistent CRF system alterations that may contribute to psychopathology in adulthood (Heim et al., 2000; Kaufman et al., 2000). Moreover, elevated cerebrospinal fluid CRF concentrations are normalized after treatment with antidepressants (Debellis et al., 1993; Veith et al., 1993) or electroconvulsive therapy (Nemeroff et al., 1991). These results have lead to the investigation of CRF receptor antagonists as potential novel therapeutic agents (Owens and Nemeroff, 1991).

There is definitive evidence of HPA axis dysregulation in various bipolar disorders (Manji and Lenox, 2000), and although there is no evidence to support changes in CSF CRF concentrations in euphoric mania (Berrettini et al., 1987; Banki et al., 1992; Risch et al., 1992), central CRF systems are likely hyperactive in dysphoric or mixed mania because such patients exhibit marked HPA axis hyperactivity. Moreover, a strong relationship exists between bipolar disorders and childhood abuse (Pribor and Dinwiddie, 1992; Levitan et al., 1998; Agid et al., 1999; Hyun et al., 2000), which, as noted above, leads to lasting changes in CRF neuronal systems. Therefore, there are multiple lines of investigation to study CRF systems in bipolar disorder and its treatment.

Lithium is still considered the first line of treatment for manic depressive illness and is approved by the U.S. Food and Drug Administration for both acute and maintenance treatment. Divalproex sodium, a common formulation of valproic acid (VPA), is often effective in the significant population of patients nonresponsive to lithium or unable to tolerate its side effects (Nemeroff, 2000), and although approved for acute treatment only may be more effective in duration of mania prophylaxis and posses greater antidepressant effects than originally thought (Bowden et al., 1994, 2000). Moreover, VPA seems to be more effective in treating specific subsets of bipolar disorder, including rapid-cycling, mixed bipolar disorder, and bipolar II disorder (Calabrese and Delucchi, 1990; Calabrese et al., 1992; McElroy et al., 1992; Bowden, 1998a; Keck and McElroy, 1998).

Preliminary experiments from our laboratory suggest that subchronic (7-day) treatment with VPA alters CRF neuronal systems (Stout et al., 2001). In the present study, we sought to more thoroughly test the hypothesis that the mood-altering effects of chronic (2 weeks) VPA may be partially attributed to its effects on CRF neuronal systems, and to compare these effects to those of lithium. The current results support and extend our previous findings that mood stabilizers do alter neuronal CRF systems, albeit perhaps via unique mechanisms.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Animals and Dosing Paradigm

Male Sprague-Dawley rats (150-175 g on arrival; Charles River Laboratories, Raleigh, NC) were housed two per cage with food and water available ad libitum in an environmentally controlled animal facility with a 12-h light/dark cycle. Animals were weighed and handled every other day throughout the course of the experiment. Two days after arrival, animals in the drug groups were switched from standard rat chow to formulated chow containing either 22.5 g/kg valproic acid for 2 weeks or 1.2 g/kg lithium carbonate for 1 week followed by 2.4 g/kg lithium carbonate for 2 weeks. Because step-up dosing over 3 weeks was required for lithium treatment compared with 2 weeks at a single dose for valproic acid, unique control groups were maintained for either 2 or 3 weeks. Animals in the lithium control and treatment groups were also provided with 0.9% saline ad libitum. Doses were chosen after a series of dose-finding experiments in which serum drug concentrations were determined after 1-week treatment (data not shown). At the end of the dosing period, the rats were killed by decapitation between 9:00 AM and 10:30 AM. Trunk blood was collected for measurement of ACTH and corticosterone as well as drug concentrations; the brains were rapidly removed, frozen on dry ice, and stored at 80°C.

Radioimmunoassays

ACTH Radioimmunoassay. Trunk blood was collected on ice in EDTA-containing glass tubes and centrifuged for 10 min at 1500g at 4°C. ACTH was measured in duplicate plasma samples by a two-site immunoradiometric assay (Nichols Diagnostics, San Juan Capistrano, CA) with a coefficient of variation of 5% and sensitivity (blank ± 2 S.D.) of 1 pg/ml.

Corticosterone Radioimmunoassay. Trunk blood was collected on ice in polycarbonate tubes and centrifuged for 10 min at 1500g at 4°C. Corticosterone was assayed in duplicate serum samples by double antibody radioimmunoassay (ICN Pharmaceuticals, Costa Mesa, CA) with a coefficient of variation of 6% and sensitivity (blank + 2 S.D.) of 1.2 ng/ml.

Drug Concentrations

Serum drug concentrations were determined using radioimmunoassay for valproic acid and an ISE instrument (Beckman Coulter, Inc., Fullerton, CA) with an ion-specific electrode for Li⁺ by the Emory Medical Laboratories.

In Situ Hybridization

Serial coronal sections (20 µm) of the rat brains were prepared on a cryostat at 18°C, thaw-mounted onto SuperFrost Plus slides under RNase-free conditions, and stored with Humi-Cap desiccant capsules at 80°C until assayed. CRF, urocortin, CRF₁, and CRF_2A mRNA in situ hybridization was performed as described in detail previously (Skelton et al., 2000a).

CRF Receptor Autoradiography

Serial coronal sections (20 µm) of the rat brains were prepared on a cryostat at 18°C, thaw-mounted onto SuperFrost Plus slides under RNase-free conditions, and stored with Humi-Cap desiccant capsules at 80°C until the assays. CRF₁ and CRF_2A receptor binding autoradiography was performed as detailed previously (Skelton et al., 2000a).

Image Analysis

Images on film from in situ hybridization and receptor autoradiography assays were digitized with CCD-72 (DAGE-MTI, Michigan City, IN) image analysis system equipped with a camera (Nikon, Tokyo, Japan) using MCID (Imaging Research, Inc., St. Catherine's, ON, Canada) software. Optical densities were calibrated against ¹⁴C standards for in situ hybridization or ¹²⁵I standards for receptor autoradiography. Messenger RNA expression levels were calculated for distinct anatomical regions as defined by Paxinos and Watson (1986) in each brain slice by subtracting the neutral background density from the specific signal. The density of CRF receptor binding was calculated for distinct anatomical regions (Paxinos and Watson, 1986) as follows: CRF₁ receptor-specific binding = total binding  CRF_2A; CRF_2A receptor-specific binding = CRF_2A receptor-specific binding  nonspecific binding. For each animal, brain region, and assay four to eight individual measurements were averaged to produce a single value for that animal. Measurements made by two independent observers on slides coded to blind them to the dosing paradigm for each animal were indistinguishable in the final results.

Data Analysis

Three animals were eliminated from the control VPA group for all analysis because their ACTH concentrations at the time of sacrifice were greater than 3 standard deviations from the mean for unknown reasons. Therefore, the number of animals analyzed in each group was as follows: VPA control, n = 7; VPA, n = 10; lithium control, n = 10; and lithium, n = 12. For each brain region examined, the mean treatment group densitometric value was compared with the appropriate mean control group densitometric value via a two-tailed t test. The mean treatment group densitometric value and S.E.M. were then expressed as a percentage of the mean control group densitometric value, and are represented as vertical bars ± S.E.M. to allow straightforward comparison across different brain regions and assays.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

Weight Gain/Plasma Drug Levels. Animals in the VPA treatment group gained 56% as much weight as the control group (69 ± 6.3 versus 121 ± 8 3.0 g) over the 2-week treatment period. Animals in the lithium treatment group gained 42% as much as the control group (73 ± 5.8 versus 173 ± 2.8 g) over the 3-week treatment period. Animals in all groups gained weight continuously.

HPA Axis Activity. Chronic VPA treatment resulted in a mean ACTH concentration of 90.4 ± 24.7 pg/ml that was not significantly different from the VPA control group (54.3 ± 15.6 pg/ml; Fig. 1A). Similarly, the mean corticosterone concentration in the VPA treatment group was 22.0 + 5.7 ng/ml, not significantly different from the control group (41.0 + 20.9 ng/ml; Fig. 1A).

View larger version (19K): [in this window] [in a new window]   Fig. 1.   Chronic valproic acid and lithium do not alter ACHT or corticosterone. Each vertical column represents the mean ± S.E.M. (VPA control, n = 7; VPA, n = 10; lithium control, n = 10; lithium, n = 12). Treatment groups were compared with their control group by a two-tailed t test. A, VPA (22.5 g/kg valproic acid formulated chow for 2 weeks) versus control (standard rat chow for 2 weeks): ACHT, p = 0.28; corticosterone, p = 0.32. B, Li+ (1.2 g/kg lithium carbonate formulated rat chow for 1 week and 2.3 g/kg lithium carbonate formulated rat chow for 2 weeks) versus control (standard rat chow for 3 weeks): ACTH, p = 0.21; corticosterone, p = 0.07.

CRF/Urocortin mRNA Expression. CRF mRNA expression was measured in the bed nucleus of the stria terminalis (BNST), central nucleus of the amygdala (CeA), and paraventricular nucleus of the hypothalamus (PVN; Fig. 2A). VPA significantly decreased CRF mRNA expression in the PVN (17%), but did not significantly alter CRF mRNA in the BNST or CeA. Lithium significantly increased CRF mRNA expression in the CeA (28%), but did not significantly alter CRF mRNA in the BNST or PVN.

View larger version (13K): [in this window] [in a new window]   Fig. 2.   Chronic valproic acid and lithium differentially alter CRF mRNA expression as measured by densitometric analysis subsequent to in situ hybridization. Effects of VPA treatment on CRF and urocortin mRNA expression are represented by solid black columns. Effects of lithium treatment on CRF and urocortin mRNA expression are represented by diagonal hatched columns. Each vertical column represents the mean ± S.E.M. (VPA control, n = 7; VPA, n = 10; lithium control, n = 10; lithium, n = 12). A, CRF mRNA expression is increased in the CeA after chronic valproic acid treatment and decreased in the PVN after chronic lithium treatment. B, urocortin mRNA expression is not changed after chronic mood stabilizer treatment. *, p < 0.05; ***, p < 0.001.

CRF Receptor mRNA Expression and Binding. CRF₁ mRNA expression and receptor binding were examined in the basolateral amygdala (BLA) and frontal/parietal cortices (F/P CTX). Chronic VPA treatment significantly increased CRF₁ mRNA expression in the cortex (26%), but did not in the BLA (Fig. 3A). In contrast, chronic VPA treatment significantly decreased CRF₁ receptor binding in the BLA (18%) and in the cortex (40%; Fig. 3B).

View larger version (30K): [in this window] [in a new window]   Fig. 3.   Chronic valproic acid and lithium differentially alter CRF receptor mRNA expression and receptor binding. Effects of VPA treatment on CRF1 and CRF2 mRNA expression are represented by solid black columns. Effects of VPA treatment CRF1 and CRF2 receptor binding are represented by solid gray columns. Effects of lithium treatment on CRF1 and CRF2 mRNA expression are represented by diagonal hatched columns. Effects of lithium treatment on CRF1 and CRF2 receptor binding are represented by horizontal hatched columns. Each vertical column represents the mean ± S.E.M. (VPA control, n = 7; VPA, n = 10; lithium control, n = 10; lithium, n = 12). A, CRF1 mRNA expression. B, CRF1 receptor binding. C, CRF2 mRNA expression. D. CRF2 receptor binding. *, p < 0.05; **, p < 0.01; ***, p < 0.001.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

We have shown that chronic administration of the mood stabilizers valproic acid and lithium alters CRF neuronal systems, although the effects of VPA seem more widespread. The precise mechanism(s) of action of these drugs in producing their therapeutic actions remains obscure, but it is likely that effects on various pathways combine to produce their therapeutic effects. Placing the current data in the framework of the growing body of evidence implicating CRF system pathology in a number of psychiatric conditions, and the stabilization of these systems after successful psychopharmacological treatment, clearly suggests that mood stabilizers, and VPA in particular, may act in part through their effects on CRF neuronal systems.

Eliminating stressful procedures is a primary concern when designing studies measuring indices of CRF neuronal activity. Using rat chow as the means of drug administration eliminated possible activation of CRF systems through repetitive injections or implantation surgery and subsequent manipulation of minipumps. Based on the administration of divalproex sodium in patients via sprinkle caps (Keck and McElroy, 1998) and the delivery of lithium in rat chow (Lambert et al., 1999; Yuan et al., 1999), preliminary trials were successfully conducted to test the feasibility of administering VPA in rat chow. Furthermore, initial concerns that a nonmetered dosing paradigm would result in a wide range of drug serum levels both between animals and diurnally proved to be unfounded because the range was actually quite limited (see Results).

In addition to using nonstressful methodology, a primary goal of this dosing paradigm was to achieve drug plasma concentrations that fell within an expected therapeutic range for these drugs in psychiatric patients. Preliminary trials and previous work (Stout et al., 2001) indicated that rather high doses of valproic acid were needed to achieve such levels due to the extremely short half-life of valproate in the rat (~17 min). Both valproate and lithium administration were associated with decreased weight gain compared with control groups over the course of the experiment. This is distinct from the weight gain currently experienced by patients treated with these agents and raises questions as to whether the drug treatment rendered the animals ill. However, drug-treated rats did exhibit weight gain albeit less than controls, and all animals seemed overtly healthy.

The mean serum VPA concentration of 27.4 µg/ml is below the reported therapeutic range of 45 to 100 µg/ml required for response in mania (Bowden et al., 1996) and as an anticonvulsant. However, serum concentrations ranging from 20 to 45 µg/ml have been reported to be effective in treating cyclothymia and bipolar II disorder (Jacobsen, 1993), subtypes of bipolar disorder reported to respond well to VPA treatment. Unpublished observations showed that the serum concentrations reported here for rats killed in mid-morning were ~2- to 2.5-fold higher in the late afternoon (~4:00 PM). The average lithium concentration of 1.0 mM was precisely within the recommended therapeutic serum concentration of 0.8 to 1.2 mM (Lenox and Manji, 1998).

The effects of VPA and lithium on CRF neuronal pathways can be interpreted within an "antiparallel" stress system hypothesis, supported by several groups (Liebsch et al., 1999; Skelton et al., 2000a), which posits that there are two central CRF stress systems with opposing effects. CRF released from neurons in the CeA and PVN acts upon CRF₁ receptors in the cortices, BLA, locus coeruleus, and pituitary to coordinate the central nervous system response to stress. In contrast, urocortin released by terminals of neurons located in the Edinger-Westphal nucleus is thought to act on CRF₂ receptors in regions such as the LS and VMH to regulate stress-coping behaviors. In addition to more classical pharmacological (Radulovic et al., 1999) and maternal deprivation experiments (Ladd et al., 1996; Eghbal-Ahmadi et al., 1997), recent evidence from transgenic mice strongly supports this hypothesis. For example, mice overexpressing CRF exhibit anxiogenic behaviors (Heinrichs et al., 1997), whereas those lacking the CRF₁ receptor exhibit decreased basal and stress-induced anxiety (Timpl et al., 1998). In contrast, animals lacking the CRF₂ receptor exhibit increased anxiety-like behavior and decreased stress-coping behaviors (Bale et al., 2000; Coste et al., 2000; Kishimoto et al., 2000).

Placing our results in this framework, the major effects of VPA seem to be in the CRF-CRF₁ pathway as opposed to the urocortin-CRF₂ pathway. VPA decreased CRF mRNA expression in the PVN and ultimately decreased CRF₁ receptor binding. This suggests that VPA may mediate its therapeutic effects in part by ultimately dampening the overall tone of the CRF-CRF₁ pathway. Interpreted in this way, the effects of VPA would be similar to those of the benzodiazepine alprazolam. Previous results show that alprazolam alters CRF neuronal systems in accordance with the working hypothesis outlined above (Owens et al., 1989, 1991; Skelton et al., 2000a), that alprazolam and VPA both increase GABAergic neurotransmission (Illig et al., 2000), and are useful in the treatment of mania (Bowden, 1998b). Therefore, one might have predicted that the effects of VPA on CRF systems would be similar to those of alprazolam. Although we failed to detect an up-regulation of the urocortin-CRF₂ pathway as observed with alprazolam, and in fact found a decrease in CRF₂ mRNA, this does not preclude the possibility that the effects of VPA on the CRF-CRF₁ pathway are therapeutically relevant.

The effects of lithium on CRF neuronal systems were less pronounced than those of VPA. There was a significant increase in CRF mRNA expression in the CeA and significant decreases in CRF₁ mRNA expression in the BLA and cortices. However, no changes in binding for either receptor were observed. Although mania and anxiety are not necessarily regulated through the same neuronal pathways, an increase in the CeA, an important anxiogenic center for CRF, could be interpreted as increasing the tone of the CRF-CRF₁ pathway. Changes in receptor mRNA did not seem to translate into actual changes in receptor binding, obfuscating a clear interpretation of these data.

Bipolar disorders represent a group of complex disorders with noted changes in the CNS. Even within delineated bipolar disorder subtypes, there are subgroups of patients whose common clinical symptoms may be the result of distinct neuronal changes. This could explain, for example, why a significant subgroup of manic patients who are nonresponsive to lithium treatment show marked remission of symptoms when treated with VPA (Nemeroff, 2000). Moreover, patients are likely in different physiological states in distinct phases of the illness, as demonstrated by the study of rapid cycling patients (Juckell et al., 2000). Therefore, it is quite probable that mood stabilizers exert their therapeutic effects through many different neural pathways, depending on the state of the patient at a given time and the distinct pathways affected in that individual. We have shown that the mood stabilizers VPA and lithium do alter CRF neuronal systems in vivo, and based on these results posit that the therapeutic effects of VPA in particular may be mediated in part via these pathways. The continued development of tools to study CRF systems in vivo, either in animal models of affective disorders or the use of imaging techniques in bipolar patients (Soars et al., 2001), will greatly assist in delineating the complex effects of these pharmacological treatments for bipolar disorders.