A Dual Component Analysis Explains the Distinctive Kinetics of cAMP Accumulation in Brown Adipocytes*

The mechanism behind the distinctive non-Michaelis-Menten, bell-shaped kinetics of cAMP accumulation in brown adipocytes (which underlies the similar kinetics of UCP1 and β1-adrenoreceptor gene expression) was investigated. A theoretical dual component analysis indicated that the observed dose-response curves could be constructed as the resultant of a stimulatory and an inhibitory component. Experimentally, inhibition of the α1-component of the norepinephrine response revealed the underlying existence of a much larger stimulatory β3-component which displayed monophasic Michaelis-Menten kinetics. The inhibitory α1-component (which was also monophasic but had a 2-fold higher EC50) was mediated via an increase in [Ca2+] i ; the protein kinase C pathway was not involved. The [Ca2+] i increase which resulted in massive inhibition of cAMP accumulation was very low: <100 nm. The [Ca2+] i signal stimulated a calmodulin-controlled phosphodiesterase, possibly PDE-1. The acquirement of this specific interaction pattern between β- and α1-adrenergic stimulation was thus part of the differentiation program of the brown adipocytes. It was concluded that an array of synergistic or inhibitory α1/β interactions occur in the adrenergic regulation of this cell type which is unique in its dependence upon adrenergic stimulation for cellular proliferation, differentiation, and metabolic function.

When stimulated with norepinephrine, brown adipocytes express the gene for the uncoupling protein-1 (UCP1) 1 (1)(2)(3)(4). However, the kinetics of the induction of the expression of this gene do not adhere to simple, monophasic Michaelis-Menten kinetics. Rather, a distinctive bell-shaped response is observed, with norepinephrine concentrations of Ϸ0.1 M being optimal for induction of expression, and higher, as well as lower, concentrations being less efficient (3,4).
The observation of non-Michaelis-Menten kinetics such as these may easily be considered an experimental artifact and/or interpreted as an indication of "overstimulation" of the relevant receptor, leading to an acute "desensitization" of the response. Because of difficulties in determining kinetic parameters in non-Michaelis-Menten-responding systems, the downward part of such dose-response curves may often be ignored (or perhaps even omitted) in the presentation of data.
However, in a theoretical analysis, Rovati and Nicosia (5) demonstrated that one conceivable possibility for generation of bell-shaped dose-response curves would be that these curves represent the resultant of interacting stimulatory and inhibitory components. They also implied that in such systems, large underlying stimulatory components, masked by the inhibitory component, could be revealed.
Following this proposal, we have here attempted to resolve the bell-shaped dose-response curve for norepinephrine into dual components. As the kinetics of UCP1 gene expression have been demonstrated to mirror those of norepinephrineinduced cAMP accumulation in these cells (6), we have, de facto, examined whether specific receptors and intracellular mediators could interact to create the unusual kinetics of cAMP accumulation.
We conclude that a Rovati/Nicosia model fully explains the kinetics of the control of cAMP levels in brown adipocytes and that through the elimination of the inherent inhibitory component, a strikingly high but otherwise masked potential for stimulation of cAMP accumulation via ␤ 3 -receptors may be revealed.

EXPERIMENTAL PROCEDURES
Cell Isolation-Brown fat precursor cells were isolated from 3-4week-old male mice of the NMRI strain, principally as described by Néchad et al. (7). Tissue was combined from the cervical, interscapular, and axillary depots and incubated in a Hepes-buffered solution (7), containing 200 units/ml crude collagenase type 2 (Sigma).
Cell Culture-Cells were routinely cultivated in 12-well plates (growth area 3.83 cm 2 /well) as described earlier (3,6,8) in 1 ml of a culture medium consisting of Dulbecco's modified Eagle's medium ((Flow) 1 ϫ liquid without glutamine, 4 mM glutamine (Flow) added) supplemented with 10% newborn calf serum (Flow), 4 nM insulin (Actrapid Human, Novo), 10 mM Hepes (Flow), and 50 IU of penicillin, 50 g of streptomycin, and 25 g of sodium ascorbate (Kebo) per ml, at 37°C in a water-saturated atmosphere of 8% CO 2 in air in a Heraeus CO 2 -auto-zero B5061 incubator. The medium was completely exchanged with fresh prewarmed medium on day 1 (when the cultures were first washed with 2 ml of prewarmed Dulbecco's modified Eagle's medium) and on days 3 and 6 (without wash).
cAMP Determinations-On day 6 or 7, as indicated, antagonists, PMA, or phosphodiesterase inhibitors were added 5-7 min, and cirazoline or ionomycin 1 min, before (other) agonist addition. After a further 20 -25 min, the culture medium was aspirated, 0.8 ml of 95% ethanol added, the cells scraped off, and the suspension transferred to Eppendorf tubes. The wells were washed with 0.5 ml of 70% ethanol, and the combined suspensions were dried in a Speedvac centrifuge. The dried samples were dissolved in 0.3-1 ml of the Buffer 1 provided with the Cyclic AMP [ 3 H]Assay System from Amersham Pharmacia Biotech and centrifuged at 14,000 rpm, 10 min. Two 50-l aliquots of the supernatants of every sample were analyzed according to the description in the assay system. For every concentration of any agent in each experiment, two wells were used. Thus, in each experiment, each value used in later calculations is the mean of the four measurements of cAMP. The results were routinely normalized in each series of experiments by setting the value of the cAMP response for 0.1 M NE (or 0.1 M isoprenaline or 1 M CGP-12177 where relevant) to 100%.

RESULTS
In the present experiments, precursors of brown adipocytes were isolated from mice and cultured under conditions leading to confluence on days 6 or 7. The cells are then fully differentiated, in that they contain numerous fat droplets and are able to express high levels of the brown fat-specific mitochondrial uncoupling protein (UCP1) in response to norepinephrine stimulation (3,4). It is in cells like this that the unusual and distinctive kinetics for norepinephrine-induced UCP1 expression (3,4,6) and ␤ 1 -adrenoreceptor expression (11) have been observed, as well as the underlying similar kinetics of cAMP accumulation (6). The aim of the present investigation has been to obtain an explanation for these kinetics.
An Unusual Norepinephrine Dose-Response Curve-The characteristic dose-response curve for cAMP generation induced by norepinephrine (NE) is presented in Fig. 1A. As seen, in agreement with Ref. 6, the response did not adhere to simple Michaelis-Menten kinetics; rather, a very distinctive semi-bellshaped dose-response curve was obtained. In control experiments, it was confirmed that the same relative relationships for the cAMP levels induced by 0.1, 1, and 10 M NE were observed at any time point up to 90 min after NE stimulation (when the cAMP levels were still markedly elevated over unstimulated levels) and that the response peaked at Ϸ20 min for all NE concentrations (not shown). Therefore, in the following experiments, the cAMP levels were determined 20 min after NE stimulation.
A possible explanation for a bell-shaped dose-response curve would be the presence of an inhibitory component, activated by high concentrations of NE. Following the theoretical suggestions of Rovati and Nicosia (5) and Equation 2 derived therefrom (see "Experimental Procedures"), we have analyzed in Fig.  1B the kinetics of the NE dose-response curve and found that it could be resolved into two hypothetical components: a stimulatory component with an EC 50 of 35 nM and an inherent ability to increase the cAMP level to 250% of that maximally observed experimentally, and an inhibitory component with a somewhat higher IC 50 value of 100 nM and an inherent ability to reduce the cAMP level by 200%. Through the interaction of these hypothetical components (the Resultant curve in Fig. 1B), a very good description of the experimental results ( Fig. 1A) was obtained. Although it was thus possible with this theoretical analysis to describe the observation by introducing two counteracting components, the question to be addressed was whether the two components postulated here had a true existence.
The Nature of the Inhibitory Component-In mature brown adipocytes, the stimulatory component of the cAMP response to NE is mediated via ␤ 3 -adrenergic receptors (6), and undoubtedly ␤ 3 -activation would be the stimulatory component in the above theoretical analysis.
The hypothetical inhibitory component could be mediated via the same ␤ 3 -receptors that mediate the stimulatory component or via other adrenergic receptors. The activated ␤ 3 -receptors in themselves could induce the inhibition, since released ␤␥-subunits of the G s (or G i ) protein can inhibit certain adenylyl cyclase subtypes (12)(13)(14). If this effect were to occur at high intensity of ␤ 3 -adrenoreceptor activation, it could explain the curve shape but it would not be possible to block only the inhibitory component by the use of a subtype-selective adrenergic antagonist. However, if other adrenergic receptors were involved, the correct subtype-selective adrenergic antagonist should be able to transform the biphasic dose-response curve into a simple, monophasic curve with the appearance of the hypothetical stimulatory component in Fig. 1B.
The obvious candidate for such an inhibitory receptor is the ␣ 2 -adrenergic receptor that mediates its inhibitory effect directly on adenylyl cyclase via G i proteins. We therefore examined whether the ␣ 2 -antagonist yohimbine would be able to eliminate the inhibitory component of the NE response. As seen in Fig. 1C, ϩYoh, this was not the case: although there was a clear augmenting effect of yohimbine on the maximum response to NE, the inhibitory component was still very evident.
We therefore also examined whether the ␣ 1 -antagonist prazosin could influence the dose-response curve. As seen in Fig.  1C, ϩPraz, this was indeed the case: prazosin not only doubled the maximal response (as did yohimbine) but it also eliminated the inhibitory component and converted the bell-shaped curve into a typical Michaelis-Menten curve.
In Fig. 1D, we have used the data from Fig. 1C to characterize the components predicted by the theoretical analysis in Fig.  1B. As seen, the experimentally obtained values, indicating a stimulatory ␤ 3 -component with an EC 50 of 34 nM and a large increase in cAMP up to 231% of that normally seen, plus an inhibitory ␣ 1 -component with an IC 50 of 74 nM and inducing an inhibition of 170%, were very close to those predicted by the theoretical analysis. It was therefore concluded that the inhibitory component was not inherent to the ␤ 3 -receptors and was not mediated in the classical way via ␣ 2 -receptors. Instead, this inhibition was unexpectedly mediated via ␣ 1 -receptors.
To substantiate this interpretation, the dual nature of NE action was mimicked by combining a ␤ 3 -selective agonist (BRL-37344 or CGP-12177) with an ␣ 1 -selective agonist (cirazoline). In the brown adipocytes studied here, ␤ 3 -agonists induce a monophasic cAMP response (6) and they were used here at saturating concentrations. As expected, the ␤ 3 -agonists increased the cAMP level (Table I), and also as expected, this response was not affected by the ␣ 1 -antagonist prazosin. The ␣ 2 -antagonist yohimbine in itself exhibited a tendency to an unexpected inhibitory effect. However, in accordance with the model presented above, activation of ␣ 1 -receptors with cirazoline led to a significant suppression of the cAMP accumulation induced by either of the ␤ 3 -selective agonists, which could be prevented by the ␣ 1 -selective antagonist prazosin. Thus, this result is in accordance with the implication that it is through ␣ 1 -receptors that the inhibitory component of the responses to NE is mediated.
The Response to Isoprenaline also Includes an Inhibitory ␣ 1 -Component-The agonist isoprenaline is generally considered to be a selective ␤-agonist. Thus, isoprenaline is not expected to stimulate ␣ 1 -receptors, and its dose-response curve would therefore not be expected to exhibit the ␣ 1 -inhibitory component identified above. However ( Fig. 2A), although the maximal response to isoprenaline stimulation (66 pmol/well)

-selective agonists
To brown adipocyte cultures, 10 M of the indicated antagonist(s) were added 5 min before the indicated ␤ 3 -agonists and, where indicated, 1 M cirazoline 1 min before the ␤ 3 -agonists. The cultures were harvested 20 min after ␤ 3 -agonist addition and cAMP levels were determined. Results are mean Ϯ S.E. of five experiments (three experiments with yohimbine). In each series, the effect of 0.1 M BRL-37344 or 1 M CGP-12177 alone was set to 100% (which corresponded to 96 Ϯ 11 pmol of cAMP per well for BRL-37344 and 68 Ϯ 17 for CGP-12177). The data were analyzed for significant effect of further agents on the response to the ␤ 3 -agonist alone.
was much higher than that to NE (25 pmol/well), presumably because of a much lower activation of ␣ 2 -receptors, the doseresponse curve for isoprenaline was nonetheless bell-shaped, although the apparent inhibitory component was smaller than that of norepinephrine (both absolutely and relatively). A theoretical component analysis was therefore also performed for isoprenaline ( Fig. 2B), indicating again that the response could be explained as the resultant of counteracting components. Yohimbine was unable to transform the bell-shaped curve to a simple Michaelis-Menten curve, whereas prazosin was able to accomplish this (Fig. 2C). The response to isoprenaline thus also contained an inhibitory ␣ 1 -component. From the analysis of the experimental data ( Fig. 2D), it is evident that at experimental concentrations where isoprenaline may be expected to be selective for ␤-receptor interaction (ϳ1 M), it displayed a prominent ␣ 1 ϭeffect in these cells.
An Unabridged cAMP Response Is Only Observed with ␤ 3 -Selective Agonists-The combined implication of the above analyses of the responses to norepinephrine (Fig. 1) and to isoprenaline (Fig. 2) is that these agonists to different degrees interact with ␣ 1 -receptors and that this can explain both the differences in shape of their dose-response curves and most of their apparently different efficacy. Based on the results presented in Figs. 1B and 2B, and Table I it was calculated that the inhibitory ␣ 1 -component led to a 74% reduction of the response to 1 M NE and to a 39% reduction of the response to 1 M isoprenaline; no inhibitory ␣ 1 -component was found in the response to CGP-12177 or BRL-37344.
Activation of Protein Kinase C Does Not Mediate the Inhibition-To understand the cellular mechanism of the ␣ 1 -inhibition, we proceeded to identify the second messenger responsible. As in other systems, stimulation of ␣ 1 -adrenoreceptors in brown fat cells produces two second messengers, inositol 1,4,5trisphosphate (15,16) and diacylglycerol; the inositol 1,4,5trisphosphate releases Ca 2ϩ from intracellular stores and [Ca 2ϩ ] i is increased (10,17,18), and the diacylglycerol activates protein kinase C (PKC) (19).
To investigate the possible participation of the diacylglycerol-PKC pathway, we stimulated the cultured adipocytes by the phorbol ester PMA (12-O-tetradecanoylphorbol-13-acetate); this had in itself no detectable effect (Table II). If PKC activation was responsible for the inhibitory effect of ␣ 1 -stimulation, PMA would be expected to inhibit selective ␤ 3 -agonist-induced cAMP accumulation, just as did ␣ 1 -stimulation (Table I). However, PMA did not inhibit, but rather enhanced the cAMP accumulation. The nature of this PMA-elicited enhancement of ␤ 3 -adrenoreceptor-induced cAMP accumulation was not further investigated but could conceivably be related to reported stimulation of certain adenylyl cyclase subtypes by PKC (20,21). As a similar enhancing effect of PMA was observed also when the cells were stimulated with isoprenaline or NE (Table  II), it is unlikely that this mechanism is actually activated when the cells are responding to adrenergic stimulation.
Conversely, we attempted to inhibit the PKC pathway during adrenergic stimulation. The addition of the PKC inhibitor added 5-7 min before the indicated concentration of isoprenaline, and cAMP levels determined 20 min later. In each series, the effect of 0.1 M isoprenaline alone was set to 100%, corresponding to 66 Ϯ 9 pmol of cAMP/well. Points are mean ϩ S.E. of five experiments. D, component analysis of the experimental data in C. The points are identical to those for isoprenaline ϩ prazosin in Fig. 1C but are here analyzed for adherence to simple Michaelis-Menten kinetics, yielding the indicated values. The µ points were calculated as the difference between the isoprenaline points and the "isoprenaline ϩ prazosin" points in Fig. 1C; they thus represent the inhibitory component. The inhibitory component was also analyzed for adherence to Michaelis-Menten kinetics, yielding the indicated values.  7) and an inhibitory component (V max ϭ 50%, IC 50 ϭ 500 nM, h ϭ 1). C, effect of ␣-antagonists on the dose-response curve for isoprenaline-stimulated cAMP accumulation. 10 M yohimbine (Yoh) and/or prazosin (Praz) were chelerythrine (1 M) before addition of NE or isoprenaline led to a decrease in the responses to these agents by ϳ10%, i.e. the opposite of what would be expected if PKC mediated the inhibitory effect (not shown). Furthermore, long-term pretreatment with PMA, a treatment that is expected to desensitize PKC (22) and earlier indirectly shown to do so in these cells (10), did not lead to an augmentation in the cAMP response to ␤-agonists (Table II). Thus, the PKC pathway was not involved in mediation of the inhibitory ␣ 1 -component.
Ca 2ϩ Is Necessary for the Inhibitory ␣ 1 -Response-To examine whether the ␣ 1 -induced inhibitory effect was mediated via an increase in cytosolic Ca 2ϩ levels, we manipulated the intracellular Ca 2ϩ response during adrenergic stimulation.
NE stimulation of cultured brown adipocytes led to an elevated intracellular Ca 2ϩ level (Fig. 3) (10). To investigate whether this was a necessary component in the inhibitory effect, we examined the effect of incubation conditions intended to diminish the increase in [Ca 2ϩ ] i . Changing to a Ca 2ϩ -free medium led to an approximate halving of the norepinephrineinduced [Ca 2ϩ ] i response (Fig. 3); this was associated with an increase in the level of norepinephrine-induced cAMP accumulation (Table III). To further diminish the [Ca 2ϩ ] i levels, we added the Ca 2ϩ chelator BAPTA to the nominally Ca 2ϩ -free medium. A further reduction in [Ca 2ϩ ] i response was observed (Fig. 3), and a higher cAMP accumulation (Table III). To fully eliminate the increase in cytosolic Ca 2ϩ levels, we preincubated the cells with the permeable Ca 2ϩ chelator BAPTA/AM. As seen, no NE-induced increase in [Ca 2ϩ ] i was then observed (Fig. 3), and a further increase in NE-induced cAMP accumulation was found. (The cell-permeable non-chelator "half-BAPTA/AM," which is not able to chelate Ca 2ϩ , had only a limited effect on NE-induced [Ca 2ϩ ] i and did not influence the NE-induced cAMP response ( Fig. 3; Table III).) Thus, an increase in [Ca 2ϩ ] i was clearly a necessary step in the mediation of the inhibitory ␣ 1 -component.
Correspondence between Adrenergic Effects on [Ca 2ϩ ] i and on Inhibition of cAMP Accumulation-If an increase in [Ca 2ϩ ] i was fully responsible for the mediation of the inhibition, there should be good correspondence between the relative ability of adrenergic agonists to increase [Ca 2ϩ ] i and their ability to inhibit cAMP accumulation. In Fig. 4A, we exemplify the effect of different adrenergic agonists on [Ca 2ϩ ] i levels in these cells. In Fig. 4, B-D, dose-response curves for these effects are presented as a means from a series of experiments.
As seen, NE increased [Ca 2ϩ ] i levels markedly, with an EC 50 value of about 0.2 M, and a maximal induction of more than micromolar Ca 2ϩ concentrations (Fig. 4B). At concentrations Ն1 M, isoprenaline had a clear, although modest, effect on [Ca 2ϩ ] i levels (Fig. 4C) (this effect could be blocked by 1 M prazosin but not by propranolol, indicating that it was indeed mediated via ␣ 1 -receptors; not shown). CGP-12177, however, even at 10 M, was fully without effect on [Ca 2ϩ ] i (Fig. 4A). It is clear that this relationship is fully parallel to the inherent ability of each of these compounds to self-inhibit cAMP accumulation through ␣ 1 -pathways: CGP-12177 had no such effect (Table I), isoprenaline some (Fig. 2), and NE much greater (Fig.  1).
Effect of Increase in [Ca 2ϩ ] i on cAMP Accumulation-A further, critical step in the demonstration that [Ca 2ϩ ] i mediates the inhibitory component in the control of cAMP level in cultured brown adipocytes would be to experimentally increase the intracellular levels of Ca 2ϩ . Two methods were used for this. One was through activation of the receptor-mediated pathway, i.e. through stimulation of ␣ 1 -receptors by the ␣ 1selective agonist cirazoline. The other one was through the use of the Ca 2ϩ -ionophore ionomycin.
Cirazoline, just as NE, led to a rapid increase in [Ca 2ϩ ] i levels. The increase induced by both NE and cirazoline displayed Michaelis-Menten kinetics (Fig. 4, B and D), but the EC 50 for cirazoline was, as expected, lower than that for NE (35 and 200 nM, respectively). The maximal level reached after NE stimulation was consistently higher than that reached after selective stimulation of the ␣ 1 -receptors with cirazoline; as pure ␤-stimulation and an increase in cAMP with forskolin did not elevate [Ca 2ϩ ] i (not shown), a complex interaction is indi-   cated but this has not been further studied here. The induced level was not stable but successively declined (Fig. 4A), although it remained elevated over the basal level for more than 40 min. Further analysis of the decline in [Ca 2ϩ ] i indicated that it could not be fitted to a single exponential function. However, it could be analyzed as being composed of two exponential functions (not shown) (or it could be described mathematically as a power function of time, i.e. as basal ϩ ⌬[Ca 2ϩ ] i -max ⅐ t c , where t is time and the constant c is 0.6 -0.8).

TABLE III Effect of Ca 2ϩ elimination on NE-induced cAMP accumulation
The Ca 2ϩ elevation initiated by ionomycin tended to be stable with time. The effect of ionomycin was not saturable and may consist of two components: a Michaelis-Menten part and a linear part (Fig. 4D). This may relate to the Ca 2ϩ coming from two sources: intracellular and extracellular (cf. studies by Mason and Grinstein (23) in lymphocytes). Ionomycin concentrations of 1-10 M induced increases in [Ca 2ϩ ] i within the physiological range induced by NE.
We thereafter analyzed the effects of cirazoline-or ionomycin-induced Ca 2ϩ increases on the cAMP levels in the brown adipocytes. The analysis was performed with CGP-12177 as the agonist since this agent does not in itself alter [Ca 2ϩ ] i (Fig. 4A). The effect of cirazoline on agonist-stimulated cAMP accumulation is shown on Fig. 5A. The effect was dose-dependent, and, in agreement with expectations and the single-dose experiment in Table I, cirazoline strongly inhibited the CGP-12177-induced response. When ionomycin was used to increase [Ca 2ϩ ] i levels, an even more pronounced effect was observed. This agent drastically reduced (and at high levels fully eliminated) cAMP accumulation (Fig. 5B). Thus, an increase in [Ca 2ϩ ] i , even when artificially induced, was sufficient to fully mimic the inhibitory ␣ 1 -component of the adrenergic response.
The dose dependence for the effect of ionomycin on ⌬[Ca 2ϩ ] i was then used to estimate the quantitative relationship between ⌬[Ca 2ϩ ] i and cAMP accumulation. Using the dose-response curve for the effect of ionomycin on ⌬[Ca 2ϩ ] i levels (Fig.  4D), we replotted the inhibition of cAMP levels (Fig. 5B) as a function of the corresponding, maintained ⌬[Ca 2ϩ ] i (Fig. 5C). As seen, very low increases in ⌬[Ca 2ϩ ] i , of only about 30 nM, induced an inhibitory effect on the cAMP level, and at levels Ն1 M, cAMP accumulation was totally inhibited.
The relationship between ⌬[Ca 2ϩ ] i and cAMP accumulation was also plotted when cirazoline was the [Ca 2ϩ ] i -increasing agent (Fig. 5C). Apparently, a Ca 2ϩ elevation induced by the ␣ 1 -agonist was only one-third as effective as a Ca 2ϩ elevation induced with ionomycin in bringing about inhibition of cAMP accumulation (IC 50 was about 300 and 80 nM Ca 2ϩ , respectively). This difference can, however, be understood if the different kinetics of the Ca 2ϩ responses to these agents are taken into account (Fig. 4A). The dose-response curves for the increases in Ca 2ϩ (Fig. 4D) are based on the peak response. As the ⌬[Ca 2ϩ ] i for ionomycin is fairly stable with time, the peak level is a good approximation of the mean level during the 20-min incubation   Fig. 4A and is about a factor 3 lower than the maximal values (this proved to be valid for both cirazoline and NE). Thus, this kinetic factor of 3 may fully explain the apparent 3-fold difference between the cirazoline and the ionomycin curves in Fig. 5C, and an increase in [Ca 2ϩ ] i is equally effective in inhibiting cAMP accumulation whether it originates via a receptor-mediated pathway or is induced by a Ca 2ϩ ionophore.
Is the Increase in [Ca 2ϩ ] i Induced by Norepinephrine Respon-sible for the Inhibition of cAMP Accumulation?-The relevant physiological question is really whether the observed increase in [Ca 2ϩ ] i during adrenergic stimulation can quantitatively explain the inhibition observed in response to norepinephrine addition. We therefore tested whether the relationship between the increase in [Ca 2ϩ ] i and the level of cAMP can be considered the same as that above, also when [Ca 2ϩ ] i is increased through the innate stimulation elicited by NE. As seen in Fig. 6A, cirazoline, which may further increase [Ca 2ϩ ] i , further inhib- ited NE-stimulated cAMP accumulation, although only to a fairly small degree. Also when [Ca 2ϩ ] i was increased by ionomycin, cAMP accumulation could be practically fully inhibited (Fig. 6B). From this curve, it should be possible to relate the physiologically (NE-) induced [Ca 2ϩ ] i increase to the degree of inhibition observed, provided that the initial functional level of ⌬[Ca 2ϩ ] i prior to ionomycin addition was known. The effective level of [Ca 2ϩ ] i was estimated, as indicated above, from the area under the curve, to be about one-third of the peak level, i.e. about 300 nM. From Fig. 1B and calculations therefrom, it was clear that at this NE concentration, there was already a 74% inhibition of the cAMP response. With these corresponding [Ca 2ϩ ] i and cAMP values as starting points, a full dose-response curve for the effect of ⌬[Ca 2ϩ ] i on cAMP levels could be constructed (Fig. 6C). As seen, the values originating from CGP-12177 and NE data could be superimposed. Thus, not only qualitatively but also quantitatively, the inhibitory ␣ 1 -component could be described as a simple monophasic effect of Mediation of the Ca 2ϩ Effect on cAMP Accumulation-To identify the link between the increase in [Ca 2ϩ ] i and the inhibition of cAMP accumulation, we used the fact that the level of cAMP is determined by the balance between two processes, one synthetic (mediated by adenylyl cyclase(s)) and one degradative (mediated by cyclic nucleotide phosphodiesterase(s) (PDE)). To allow for a distinction between these two sites of action, experiments were performed in which adenylyl cyclase was stimulated with the direct activator forskolin (in the brown adipocytes studied here, forskolin does not affect the basal level of Ca 2ϩ or the NE-induced Ca 2ϩ response (10)). These experiments were first carried out in the presence of the general phosphodiesterase inhibitor IBMX, i.e. under conditions under which all observed effects of additions can be assumed to be due solely to effects on adenylyl cyclase. As seen (Fig. 7A), the addition of forskolin led to the expected increase in cAMP. When [Ca 2ϩ ] i was increased by the addition of 1 M ionomycin, only a minor difference (ϳ20%) in the kinetic constants of cAMP accumulation was registered (Fig. 7A). This effect of Ca 2ϩ on adenylyl cyclase activity was thus too small to account for the large inhibition of cAMP accumulation observed above (Ϸ80% according to Fig. 5B).
We therefore investigated whether the effect could be mediated via activation of a phosphodiesterase. Experiments were therefore performed in the absence of IBMX, i.e. when the phosphodiesterases were active. A different pattern was then observed (Fig. 7B). The maximal cAMP level achieved by forskolin was now only half that in the presence of phosphodiesterase inhibitor, implying the presence of a high basal phosphodiesterase activity in these cells (principally in accordance with Ref. 24)). When [Ca 2ϩ ] i was increased by the addition of 1 M ionomycin, the cAMP level was further decreased by a factor of 2 (Fig. 7B). As the inhibitory effect of Ca 2ϩ was only evident when the phosphodiesterases were active (compare Fig. 7, A and B), it was clear that activation of a phosphodiesterase was responsible for the inhibition of the cAMP accumulation by Ca 2ϩ (and thus, by NE).
Which Phosphodiesterase Mediates the Effect?-Not all phosphodiesterases are Ca 2ϩ sensitive. Out of at least seven distinct PDE gene families which generate more than 20 different isozymes, only the subtypes II, III, and IV have so far been positively identified in brown fat (25)(26)(27) but none of these PDEs are generally accepted to be Ca 2ϩ stimulated; this is only the case for PDE I (CaM-PDE), which has not as yet been identified in brown adipose tissue. However, if CaM-PDE is the PDE responsible for the bell shape of the dose-response curve for NE, specific inhibition of this PDE should transform the bell-shaped curve to a normal Michaelis-Menten-type curve. The effect of several PDE inhibitors on the cAMP dose-response curve for NE was therefore studied.
Three selective PDE inhibitors, 8-MM-IBMX, OPC-3911, and Ro 20-1724, i.e. the inhibitors of subtypes I, III, and IV, respectively, were tested. At the concentrations used, the inhibitors should inhibit only the indicated PDE subtypes (28 -30). The inhibitors of PDE III and IV, i.e. OPC-3911 and Ro 20-1724, failed to convert the bell shape of the dose-response curve into normal kinetics, despite the fact that they considerably increased the maximal cAMP-response (Fig. 8A). Only 8-MM-IBMX, the selective inhibitor of the CaM-PDE (PDE-I), converted the dose-response curve to a normal Michaelis-Menten type (Fig. 8A). This thus indicated that the inhibition was mediated via the PDE-I subtype.
Involvement of Calmodulin in the Mediation of the Inhibition-Mediation of the Ca 2ϩ effect on PDE I occurs through calmodulin (31). Thus, if it is directly through the formation of a Ca 2ϩ -calmodulin complex and its interaction with PDE-I that this phosphodiesterase is activated, antagonism of calmodulin should remove the inhibitory component of norepinephrine stimulation. In agreement with this, the potent antagonist of calmodulin, calmidazolium (IC 50 ϳ0.2 M for CaM-PDE-activity (32)) at a concentration of 3 M converted the bell-shaped curve to a Michaelis-Menten-type curve (Fig. 8B). DISCUSSION When stimulated with norepinephrine, brown adipocytes may exhibit non-Michaelis-Menten response kinetics in their responses. This is, e.g. evident for norepinephrine-induced UCP1 (3,4) and ␤ 1 -adrenoreceptor (11) gene expression in cultured brown adipocytes from mice. These kinetics of gene expression are reflections of the norepinephrine-induced cAMP levels, which thus also display these characteristic kinetics (6). Thus, to understand the regulation of UCP1 and ␤ 1 -adrenore- ceptor gene expression, the mechanism behind the distinctive non-Michaelis-Menten kinetics of cAMP accumulation should be understood.
The goal of the present investigation was to elucidate the mechanism behind this response pattern. The analysis was based on the theoretical model of this type of response pattern as being the result of a stimulatory and an inhibitory component, as presented in a generalized theory by Rovati and Nicosia (5). Experimentally, it was possible in the present study to substantiate the existence of these components. The elimination of the inhibitory component revealed that the stimulatory ␤-component was inherently much larger than normally observed. The inhibitory component was ␣ 1 -adrenergic in nature, with a relatively high EC 50 , explaining the bell-shaped kinetics of the resultant. The inhibition was intracellularily mediated via an increase in [Ca 2ϩ ] i ; the protein kinase C pathway was not involved. The Ca 2ϩ mediation was very sensitive, in that a ⌬[Ca 2ϩ ] i increase of less than 100 nM was sufficient to mediate the effect. The increased [Ca 2ϩ ] i stimulated a calmodulin-sensitive phosphodiesterase, possibly of the PDE-1 type, the activity of which was thus of sufficient magnitude to more than halve the outcome of full adenylyl cyclase activation. The acquirement of this process is apparently a part of the differentiation program in the brown adipocytes.
This dual component analysis thus fully explains the kinetics of norepinephrine-stimulated UCP1 and ␤ 1 -adrenoreceptor gene expression. Furthermore, as cAMP is a second messenger with a crucial role in the control of both acute thermogenesis (33), cell proliferation (8), apoptosis (34), and cell differentiation (3,6) in the brown adipocytes, knowledge of the exact mechanism of cAMP control may be of significance for understanding the control of other cellular processes and may help in devising pharmacological agents to control brown adipocyte recruitment and activity.
The Nature of the Two Components-In the dual component analysis, the stimulatory component was undoubtedly a regular ␤-adrenergic response, leading to stimulation of adenylyl cyclase and a monophasic increase in cAMP level. As a consequence of the analysis, it was realized that the inherent capacity of this system was much higher than that revealed by norepinephrine stimulation. In traditional investigations of adenylyl cyclase activity in plasma membrane preparations from brown adipose tissue, only monophasic responses to norepinephrine have been reported (35). Although this is thus valid for the isolated adenylyl cyclase system, it may, as understood from the present investigation, not be representative for the actual effect of norepinephrine on the integrated cAMP system, as the final and determining cAMP value is substantially influenced by the inhibitory component.
The inhibitory component was mediated by Ca 2ϩ stimulation of a cAMP phosphodiesterase. The interaction is therefore principally an example of negative cross-talk between intracellular signaling pathways, a phenomenon amply described over the last decades (36,37). However, in its classical forms, cross-talk occurs as interaction between the signaling pathways of two different extracellular stimuli (hormones). Norepinephrine may mediate the stimulatory or the inhibitory component in different systems. Thus, cAMP levels induced via ␤-adrenergic stimulation may be reduced due to the effect of other hormones, which may act via activation of phosphodiesterase, as is the case for insulin (38). Conversely, when other hormones stimulate cAMP accumulation, norepinephrine may interact negatively, via ␣ 1 -receptors and activation of a Ca 2ϩ /calmodulinsensitive phosphodiesterase (39).
However, the present observations fall outside these types of classical cross-talk, because NE activates both the stimulatory adenylyl cyclase component and the inhibitory component, and thus the cross-talk occurs between two pathways activated via the same activator. There is apparently only one earlier case in which norepinephrine has been implied to activate both pathways: heart cells (40). However, in those cells, the inhibitory ␣ 1 effect was small and the resultant curve was still a monophasic Michaelis-Menten relationship.
Thus, the uniqueness of the mechanisms reported here is related to the quantitative relationship between the two components: the inhibitory ␣ 1 -pathway has a high capacity but also a relatively high EC 50 , compared with the stimulatory pathway, and through this, it shapes the characteristic dose-response curve and substantially influences the maximal cAMP levels reached.
The Large Effect of ⌬[Ca 2ϩ ] i -The inhibitory component was mediated via an increase in [Ca 2ϩ ] i . The very low EC 50 for ⌬[Ca 2ϩ ] i for this effect, less than 100 nM, is remarkable. It is much lower than maximum norepinephrine-induced increases in [Ca 2ϩ ] i which may easily reach peak levels at least 10-fold higher, and there thus seems to be a high redundancy in the norepinephrine signal. However, considering the kinetics of the norepinephrine-induced [Ca 2ϩ ] i response, with its high negative time dependence, the high sensitivity may be understood as allowing for a sustained response: the phosphodiesterase system will remain activated until Ca 2ϩ levels approaching basal have been reached. This high sensitivity of the response also means that agents with effects on [Ca 2ϩ ] i that seem trivially small may still be of significance for the outcome. The effective increase, a ⌬[Ca 2ϩ ] i of Ͻ100 nM, seems low also when related to the basal level of [Ca 2ϩ ] i which has been estimated in this and most earlier investigations to be in the order of 200 nM. A technical complication may perhaps explain this. In the present investigation, as in many others, the R min used for the calibration of the fluorescence signal necessary for determination of [Ca 2ϩ ] i has been established with Mn 2ϩ , which functions to eradicate the fluorescence. This may lead to some overestimation of the basal level, and the relative increase in [Ca 2ϩ ] i may therefore be higher than anticipated from the values given here.
The high sensitivity of the system to ⌬[Ca 2ϩ ] i may also explain the somewhat unexpected results concerning isoprenaline. Isoprenaline is routinely used in receptor and receptor response studies as a selective ␤-agonist, and it is generally anticipated that the response does not include effects of ␣-receptors. The data presented here (Figs. 2 and 4) clearly demonstrate, however, that at commonly used isoprenaline concentrations, ␣-pathways are stimulated in brown adipocytes. In this respect, isoprenaline-induced responses are therefore principally different from those of selective ␤ 3 -agonists, for which we see no indication of an ␣ 1 effect. It is possible that this ␣ 1 -component may explain several recent observations on brown fat cells where the response to isoprenaline deviated from that to selective ␤ 3 -stimulation or direct adenylyl cyclase activation (41)(42)(43). That isoprenaline seems to influence the ␣ 1 -pathway to a larger extent in brown fat cells than in many other tissues may be due to the very high level of expression of ␣ 1 -receptors in brown adipose tissue, compared with the levels in other tissues (44). If this is parallelled by a similar high level of ␣ 1 -receptors, the functional affinity of brown fat cells for ␣ 1 -stimulation would be high, and agents with only a small ␣ 1 activity, such as isoprenaline, would manifest their ␣ 1 action particularly in brown adipocytes.
Physiological Relevance of the Inhibition-An important difference between brown adipocytes in situ and those studied here in culture is that the cultured cells are naive with respect to adrenergic stimulation, whereas the brown adipocytes in situ are exposed to chronic adrenergic stimulation of alternating intensity. However, this may even accentuate the situation described here, because chronic adrenergic stimulation increases the density of the ␣ 1 -receptors that mediate the inhibitory component (both gene expression levels (44,45) and receptor number (46,48)), whereas expression of the stimulatory ␤ 3 -receptors is down-regulated (47,48), as is the activity of some of the mediating steps (49).
Functionally, all processes that are rate-limited by cAMP levels will be expected to show bell-shaped dose-response curves, but this does not mean that all responses mediated via an increase in cAMP levels will show this pattern. Thus, processes saturated at fairly low cAMP levels will display monophasic kinetics even if the underlying cAMP response is biphasic. This may, e.g. be the case for thermogenesis where saturation apparently occurs already at levels of cAMP only 50% of those maximally induced by norepinephrine (24).
The ␤ 3 -adrenergic receptor has as one of its features, in comparison to ␤ 1 /␤ 2 -receptors, the absence in its primary structure of the serine residues that are involved in classical desensitization (50). Although this inability to classically desensitize was originally considered the feature of choice of ␤ 3 -receptors, it would seem that concerning this receptor, nature utilizes two alternative pathways to achieve a functional desensitization: both an intensive down-regulation of ␤ 3 -receptor gene expression (47,48), and the activation of an ␣ 1 -mediated inhibitory pathway, as demonstrated here. ␣ 1 /␤ 3 -Interaction in Brown Adipose Tissue Function-The function of brown adipose tissue is highly regulated via ␤-adrenergic processes, mediated through an increase in cytosolic cAMP levels. This is so concerning both the acute thermogenic response and stimulation of cell proliferation, apoptosis, and cell differentiation. Thus, based solely on the data reported here, ␣ 1 -stimulation should be associated with a suppressing effect on the function of brown adipose tissue.
However, a series of observations indicate a positive or synergistic interaction between ␣ 1 -and ␤-stimulation. Thus, ␣ 1stimulation enhances the ability of a given level of cAMP to induce thermogenesis in the cells (33), and ␣ 1 -stimulation may, at least in brown fat cells from certain species, have some thermogenic effects in itself (51). Furthermore, and perhaps most importantly, ␣ 1 -stimulation has additive or synergistic effects on the gene expression of a series of very significant genes for brown adipose tissue function, such as c-fos (10), thyroxine 5Ј-deodinase (52), lipoprotein lipase (53), and the uncoupling protein (UCP1) itself (3,4). 2 Thus, presently, it can only be concluded that cells of brown adipose tissue have an elaborate system for interaction between ␤and ␣ 1 -adrenergic signals, some positive, some apparently negative. We have clearly still not reached a sufficient depth of knowledge to allow us to understand the intricacies of the complex network found in this cell system, unique in its dependence upon adrenergic stimulation for cellular growth and function.