In vivo veritas, the next frontier for functionally selective GPCR ligands

Methods xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Methods journal homepage: www.elsevier.com/locate/ymeth

In vivo veritas, the next frontier for functionally selective GPCR ligands Jean Martin Beaulieu ⇑ Dept. Psychiatry and Neuroscience, Faculty of Medicine, Laval University, Canada

a r t i c l e

i n f o

Article history: Received 2 July 2015 Received in revised form 22 August 2015 Accepted 24 August 2015 Available online xxxx Keywords: G protein coupled receptor Functional selectivity Animal models Beta-arrestin Cell signaling

a b s t r a c t The realization that G-protein coupled receptors (GPCR) engage several cell signaling mechanisms simultaneously has led to a multiplication of research aimed at developing biased ligands exerting a selective action on subsets of responses downstream of a given receptor. Several tools have been developed to identify such ligands using recombinant cell systems. However the validation of biased ligand activity in animal models remains a serious challenge. Here we present a general strategy that can be used to validate biased ligand activity in vivo and supports it as a strategy for further drug development. In doing so, we placed special attention on strategies allowing to discriminate between G-protein and beta-arrestin mediated mechanisms. We also underscore differences between in vitro and in vivo systems and suggest avenues for tool development to streamline the translation of biased ligands development to pre-clinical animal models. Ó 2015 Elsevier Inc. All rights reserved.

1. Introduction The realization that G-protein coupled receptors (GPCR) can engage several cell signaling mechanisms simultaneously, or under different conditions, has led to a multiplication of research aimed at developing receptor ligands exerting a selective action on subsets of responses downstream of a given receptor [1–3]. Such biased or functionally selective ligands, could in theory activate or antagonize therapeutically relevant signaling pathways without causing unwanted side effects by perturbing other receptor functions [4,5]. Considering the already central position of GPCRs as pharmaceutical targets [6], biased ligands hold great promise for the development of cleaner and more effective pharmaceutical treatments of animal and human diseases. However, this exciting potential has also increased the complexity of GPCR drug development. The prevalent model of GPCR signaling in the late 20th century was based on the postulate that a given GPCR is coupled preferentially to a single type of G protein. This working model allowed for relatively straightforward drug discovery via the identification of ligands showing good selectivity (e.g. binding) and efficacy (e.g. activation of G protein) in recombinant cell based systems [1]. Such an approach was then complemented by investigation in animal models to demonstrate among

Abbreviations: GPCR, G protein coupled receptors; bArr1, beta-arrestin 1; bArr2, beta arrestin 2; D2R, D2 dopamine receptor; GSK3, glycogen synthase kinase 3. ⇑ Address: 2601, Chemin de la Canardière, Québec City G1J 2G3, Canada. E-mail address: [email protected]

other parameters, the bioavailability, metabolism, efficacy and non-toxicity of a given drug candidate. This simple view has since been rendered more complex by the discovery that GPCRs can couple to more than a single G protein. Furthermore, beta arrestins (bArr1 and bArr2) have also been shown to exert dual functions downstream of several GPCRs. On the one hand bArr1 and bArr2 are negative regulators of G protein coupling and participate in receptor internalization [7–9]. On the other hand, these proteins also act as scaffolds for kinases and phosphatases thus constituting bona fides mediators of GPCR signaling [10–12]. The interactions between these dual functions of bArr remain unclear. In addition, G protein and bArr mediated signaling cascades demonstrate different temporal dynamics in cell systems, which are exacerbated in vivo. The realization of the complexity of GPCR signaling has led to the exponential development of elegant cell-based and mathematical models to study biased ligand activity [13–15]. These methods often rely on measurements of G protein activation, second messenger production or bArr recruitment, alone or in combination, to quantify different dimensions of pharmacological efficacy [1]. Methodological aspects of these methods are covered in several excellent articles published in this issue of Methods. While powerful tools exist to identify biased ligands in vitro, the validation of biased activity in vivo remains more hazardous. In addition to traditional parameters such as toxicity or bioavailability, several factors can potentially complicate the use of biased GPCR ligands in vivo. For instance, a given biased ligand may be processed into a non-biased active metabolite. The action of a

http://dx.doi.org/10.1016/j.ymeth.2015.08.018 1046-2023/Ó 2015 Elsevier Inc. All rights reserved.

Please cite this article in press as: J.M. Beaulieu, Methods (2015), http://dx.doi.org/10.1016/j.ymeth.2015.08.018

2

J.M. Beaulieu / Methods xxx (2015) xxx–xxx

ligand affecting the recruitment of bArr to a GPCR may also affect the coupling of this receptor to G proteins. Furthermore, interaction of a GPCR with other membrane proteins in a natural environment can conceivably affect receptor coupling and the biased selectivity of a compound in vivo [16,17]. This article will provide an overview of a general approach that can guide the validation of biased ligand activity in animal models and a base to support further drug development. In doing so, I will focus mostly on situations involving biased activity targeting bArr versus G protein mediated signaling. I will also identify major limitations of current approaches and suggest how new research tools may simplify in vivo validation of biased ligands activity. 2. A general approach to study biased ligand activity in vivo 2.1. General principles, seven basic criteria In a perfect world, biased ligand development should be undertaken with a knowledge of which receptor signaling mechanism represents a valid therapeutic target. Several parameters such as GPCR interaction partners and drug metabolism may also affect the functional selectivity of a GPCR ligand in vivo. Furthermore, functional selectivity in vivo and relevance to disease treatment should be demonstrated to support further drug development and clinical trials. In order to achieve this, I propose that validation of biased ligand activity as a valid drug target and of biased ligands as promising drug candidates should follow seven basic criteria. These would allow to establish that a given compound has biased activity in vivo and that such activity could be functionally implicated in the therapeutic effects of drugs targeting a given GPCR in vivo. Seven basic criteria: (1) The targeted cell signaling mechanism must be affected in disease or by less selective clinically effective drugs. (2) Modulation of the targeted signaling mechanism must be dependent on the targeted receptor in vivo. (3) Modulation of the targeted cell signaling mechanism by the receptor must be dependent of a specific effector in vivo. (4) Biological effects of clinically effective drugs must be dependent of the targeted cell signaling mechanism. (5) Modulation of the targeted cell signaling mechanism must replicate relevant drug actions in models with predictive validity. (6) Biased ligands must engage targeted signaling mechanism specifically downstream of the targeted receptor. (7) Biased ligands must replicate relevant drug action in models with predictive validity. The following subsections constitute an explanation of the basis of these seven criteria and propose approaches allowing to establish that a given compound meets these in vivo. For the sake of simplicity, development of compounds targeting bArr2 mediated signaling downstream of the dopamine D2 receptor (D2R) [18] will be used as a working example (Fig. 1). This modality of dopamine receptor signaling has been shown to result in the formation of a signaling complex comprised at least of bArr2, the kinase Akt and protein phosphatase 2A [11]. The formation of this complex results in an inactivation of Akt following D2R activation. Research projects have targeted this modality of D2R signaling for the development of new antipsychotics [19–21]. 2.1.1. The targeted cell signaling mechanism must be affected in disease or by less selective clinically effective drugs One of the postulates supporting the development of biased GPCR ligands is that such compounds would allow to separate

therapeutically positive from negative consequences of GPCR signaling. This implies that relevant target GPCRs are known for a given condition and that a biased ligand would make acting on this GPCR more effective and/or safe. In order for this to occur, it could be important to establish that a given signaling mechanism may be implicated in disease causation or that it may participate in the effect of drugs that are already effective to treat a given condition, which is our criterion #1. Establishing contribution to disease causation may be difficult and constitutes a separate endeavor. That being said, information about cell signaling mechanisms relevant to a given pathology may already be available from existing literature. In contrast, establishing the effects of a drug on a subset of signaling responses in vivo is much more feasible. Both chronic and acute administration of a selection of drugs known to act on a therapeutic mechanism should be considered for this stage. However, acute treatment should be tested first since chronic effects may result from long-term adaptations several steps downstream from the action of a ligand on its receptor. For acute administration, it is important to keep in mind that different cell signaling responses may occur in response to different drug doses and that bArr and G protein mediated signaling display different temporal dynamics. Moreover, one should also consider that in contrast to simple in vitro systems GPCR ligands in vivo compete with a natural agonist and may affect natural regulatory mechanisms thus complicating data analysis. More information on this is provided in Section 2.2. To address these limitations one approach is to identify measurable cell signaling outcomes that are associated to different effectors for the same GPCR. In the case of D2R, one can use pAkt (Th308) as a readout for bArr2 mediated and DARPP32 (Th34) for Gi/cAMP mediated signaling mechanisms (Fig. 1). Specific guidance for the design of in vivo cell signaling experiments interrogating protein phosphorylation is provided in Section 3. However, cell signaling readouts used for this type of evaluation may differ across different systems and may also include measurement of second messenger levels or other responses that would be specific to a given type of GPCR signaling mechanism. Independently of the experimental setup, cell signaling readouts will have to be evaluated at different drug doses and over different time periods. As a rule of thumb, the response of G protein mediated signaling events to receptor agonists or antagonists generally occurs in a matter a minutes (e.g. 5–30 min) following drug administration, conditional on the pharmacokinetics. This can be much slower (e.g. 30–120 min) for events depending of bArr. It is thus probable that the acute response of different cell signaling mechanisms to a non-biased GPCR ligand will occur at different time points in vivo. Measurement of signaling outcomes at a single time point may thus provide an illusion of functional selectivity and that should be avoided. In order to minimize animal use, a simple strategy can be to perform a time course study using a superoptimal dose of each drug. When an optimal time point has been identified for each GPCR signaling mechanism a top-down dose response study can then be conducted for these time points only. 2.1.2. Modulation of the targeted signaling mechanism must be dependent of the targeted receptor in vivo Therapeutically effective drugs often act on several GPCRs at the same time. This is particularly true in psychopharmacology. For instance the second-generation anti-psychotic clozapine has been shown to activate Akt in vivo [22]. This action of clozapine may result from its antagonistic action on D2R. However, clozapine is also an antagonist of 5HT2A receptors [23], which have been shown to regulate Akt mediated cell signaling in vivo [24]. This exemplifies the need to establish that therapeutically effective drugs modulate the pathway of interest by acting on the GPCR

Please cite this article in press as: J.M. Beaulieu, Methods (2015), http://dx.doi.org/10.1016/j.ymeth.2015.08.018

J.M. Beaulieu / Methods xxx (2015) xxx–xxx

3

Fig. 1. Modalities of dopamine D2 receptor signaling (D2R). (A) Stimulation of D2R leads to activation of G proteins (Gai/o) resulting in an inhibition of adenylyl cyclase and modulation of cAMP-dependent PKA (canonical G-protein-dependent signaling). Recruitment of bArr to receptors results in two distinct processes: the termination of Gprotein-dependent signaling and the formation of an internalization complex comprising bArr 1 and/or bArr2, adaptor protein 2 (AP2), clathrin and other intermediates. Formation of the internalization complex leads to receptor internalization through clathrin-mediated endocytosis. The recruitment of bArr2 also results in the formation of a signaling complex that comprises at least bArr2, PP2A and Akt. The formation of this complex results in the deactivation of Akt by PP2A and the subsequent stimulation of GSK3-mediated signaling. (B) Different kinetics of G-protein-mediated and bArr2-mediated dopamine receptor signaling following the administration of amphetamine. Shown are the two waves of signaling responses involved in slow dopamine synaptic transmission. In the first wave of responses, G-protein-mediated signaling induces a rapid and transient change in the phosphorylation of direct or indirect PKA targets such as DARPP32 or cAMP-response-element-binding protein (CREB). A second wave of signaling is mediated by the Akt–bArr2–PP2A complex and results in a more progressive and longer-lasting response (adapted from: [5]).

against which one is generating new biased ligands, which is our criterion #2. One simple way to achieve this is to use mice lacking the receptor of interest. In our example, clozapine and other antipsychotics should not affect Akt phosphorylation in D2R knockout mice. However, receptor knockout animals are not always viable and, when they are, they may be subject to compensatory changes that may complicate data interpretation. It is thus important to complement this approach by using acute administration of selective receptor ligands to examine the activation of the mechanism of interest in normal animals as well as in receptor knockout when these are available. Of course, studies involving selective ligands should comprise both dose response and time course analyses as described in the previous subsection (see Section 2.1.1).

2.1.3. Modulation of the targeted cell signaling mechanism by the receptor must be dependent of a specific effector in vivo After establishing that a potentially therapeutic pathway of interest is modulated by a given GPCR it is important to determine the mechanism by which this pathway is regulated in vivo. Interestingly, different dynamics of bArr and G protein mediated signaling can provide some information. Indeed, events showing longer delays of activation should stand a better chance of being bArr dependent while faster responses should be more probably associated with G proteins. However, a more direct proof for the involvement of bArr or G protein dependent signal transduction should also be obtained. If bArr1 or bArr2 is involved, this can be done using knockout mice for either of these genes. If alternative G protein coupling is involved, determination of which second messenger is implicated will provide element of responses.

Please cite this article in press as: J.M. Beaulieu, Methods (2015), http://dx.doi.org/10.1016/j.ymeth.2015.08.018

4

J.M. Beaulieu / Methods xxx (2015) xxx–xxx

In either case, demonstration of a role for bArr or G protein in a given cell signaling response might prove complicated. For instance, a mechanism may be mediated by both bArr1 and bArr2. Since mice lacking both bArr are not viable, new research tools (see Section 4) would be needed to fully tackle such a case. Another potentially difficult situation would involve the regulation of a signaling pathway by both G protein and bArr mediated mechanisms. Activation of the extracellular signal regulated kinase (Erk) by GPCR represents such a case [25]. Under these conditions, development of biased ligands may be one of the best avenues and continuation of compound development should provide a response to criterion #3 at a later stage in vivo. 2.1.4. Biological effects of clinically effective drugs must be dependent of the targeted cell signaling mechanism Following the previous criterion, it should be possible to identify conditions under which a potentially therapeutically relevant signaling cascade is not engaged anymore by modulating the activation of a given GPCR. One would predict that under such conditions, a therapeutic agent that engages this signaling cascade should lose its efficacy in tests with predictive validity for therapeutic action. In contrast, the drug should conserve its ability to induce unwanted adverse effects, therefore stressing out the need to develop biased ligands as better alternatives. As an example of this, antipsychotics have been postulated to exert their therapeutic effects by antagonizing bArr2 mediated D2R signaling while extrapyramidal side effects of these drugs would be mediated by Gai mediated responses [5]. Thus, administration of antipsychotics to bArr2-KO mice should still induce catalepsy, which is often employed as a model for the unwanted effects of these drugs. In contrast, effects of antipsychotics on other behaviors used to model therapeutic responses [26] should be considerably attenuated in these same mice. 2.1.5. Modulation of the targeted cell signaling mechanism must replicate relevant drug actions in models with predictive validity In line with the previous criterion, acting directly on the targeted signaling cascade should also replicate, at least partially, the effects of the drugs for which a biased ligand alternative is being sought. In some cases this can be achieved using kinase or phosphatase knockout mice and/or pharmacological inhibitors. Furthermore, the effects of direct pathway modulation should be independent of the GPCR and its immediate transduction mechanisms since it occurs downstream of receptor activation. As an example, stimulation of D2R leads to the activation of GSK3 resulting from Akt inactivation. In line with this pharmacological inhibitors of GSK3, replicate some of the behavioral effects of D2R antagonists in vivo [27]. However, while these effects depend on bArr2 when acting on D2R activity, GSK3 inhibitors are equally effective in bArr2-KO mice and normal littermates [28]. 2.1.6. Biased ligands must engage targeted signaling mechanism specifically downstream of the targeted receptor After validating a dimension of the signaling responses occurring downstream of a given GPCR as a valid drug target, it is important to verify that biased ligands targeting this signaling modality are effective and selective in vivo. Present in vitro tools allow for the precise quantification of parameters such as G protein activation and bArr1 or bArr2 recruitment following receptor activation therefore allowing to detect incomplete biased ligand activity. In vivo approaches are less precise. This is due in part to inherent variability in living organisms and to the need to rely on robust, yet less sensitive measurement assays. That being said, present means of in vivo studies are sufficient to detect full or strongly biased activity.

A general strategy to achieve this is to repeat the approaches described in criteria 1–3. Briefly, evaluation of cell signaling responses evoked by the ligand should be measured for different doses and post-injections times for each pathway. Contributions of the receptor and of specific signaling intermediates should then be established. For example the following steps could be undertaken, in the case of a ligand targeting selectively bArr2 mediated D2R signaling. Step 1: measurement of the effect of the ligands on Akt and DARPP32 phosphorylation at different doses and times postadministration. Step 2: measurement of the effect of the ligands on these same signaling readouts in D2R-KO mice. Step 3: evaluation of these same parameters in bArr2-KO mice. Overall, the perfect ligands should affect Akt but not DARPP32 phosphorylation in normal animals while having no effect on Akt in D2R and bArr2-KO animals. 2.1.7. Biased ligands must replicate relevant drug action in models with predictive validity Obviously, the ideal biased ligand should replicate the action of a non-biased ligand in regards to the effects relevant to therapeutic action. In contrast the biased ligand should not exert an effect in tests used to model adverse effects. To continue with the example of a compound that would antagonize selectively bArr2 mediated D2R signaling, the expected effects would be as follows; first, the biased compound would be expected to antagonize the effects of amphetamine on forward locomotion. Second the compound should not induce catalepsy. Of course more than two behavioral or physiological responses should ideally be considered and full validation of the compound action should be implemented in rodents and non-human primate models. 2.2. Limitations and particularities of in vivo pharmacology The criteria described above present a general strategy to validate the development of functionally selective GPCR ligands and to validate their efficacy and selectivity in vivo. However, when proceeding to this type of experiments, it is important to keep in mind some major differences between in vitro and in vivo model systems of GPCR signaling. These differences are not specific to biased ligands. Yet, their sometimes counterintuitive impact on data interpretation is sufficient to warrant mention. The first major difference is that any GPCR ligand in vivo generally acts in the presence of a natural agonist. This is particularly important for the validation of biased antagonists and partial agonists. In a cellular system under serum starvation, stimulation with an agonist will be needed to evaluate a neutral antagonist activity. This is not the case in vivo since the agonist is naturally present. Therefore, simple administration of the antagonist will exert an effect. The presence of a natural agonist also complicates the study of partial agonists in vivo. Indeed, stimulation of a GPCR with a partial agonist in a recombinant system that is not stimulated otherwise will result in the incomplete activation of cell signaling. In contrast, treatment of an animal with a partial agonist may result in a reduction of cell signaling responses due to competition with the naturally present full agonist. Another difference between recombinant systems and animal models is that GPCR ligands will activate homeostatic mechanisms in vivo. This is particularly important when dealing with ligands that act on GPCR having both autoreceptor and heteroreceptor activities. For example, an agonist of D2R will on the one hand mimic the effect of dopamine on behavior by acting on postsynaptic receptors and limit the release of dopamine itself by acting on pre-synaptic ones. Such potential mixed effects should be taken into consideration when analyzing the biological actions of new receptor ligands in animal models.

Please cite this article in press as: J.M. Beaulieu, Methods (2015), http://dx.doi.org/10.1016/j.ymeth.2015.08.018

J.M. Beaulieu / Methods xxx (2015) xxx–xxx

Finally, biased ligands aimed at modulating bArr mediated functions can potentially lead to further complications when interpreting data. Indeed, most existing screening assays for biased bArr mediated activity measure the effect of compounds on bArr recruitment to GPCR and do not provide actual information about bArr mediated signaling. As such, an agonist that would be biased toward bArr recruitment may at the same time stimulate this form of signaling while also potentiating the inactivation of G protein coupling. While this is still speculative, a recent study of the effects of bArr2 expression on the outcome of L-DOPA treatment supports the relevance of considering bArr/Gprotein activation ratios for the development of novel therapeutics [29]. Indeed in a mouse model of Parkinson disease, expression of bArr2 was shown to be essential for the therapeutic action of L-DOPA, potentially by mediating the regulation of Akt by D2R. Furthermore, bArr2 deficiency also potentiated L-DOPA induced dyskinesia, which has been associated to G-protein mediated dopamine receptor signaling responses. In contrast, overexpression of bArr2 in rodent and primate models of Parkinson disease reduced dyskinetic symptoms. In line with these results it is possible that by limiting bArr2 recruitment to D2R some biased ligands of these receptors may also affect G protein mediated signaling by preventing the desensitization of this modality of receptor signaling.

3. General principles for in vivo measurement of relative changes in protein phosphorylation Most protocols for the evaluation of biased ligand activity in cells rely either on reporter systems to access second-messenger levels, G-protein activation or bArr recruitment. Most of these tools are not available for work in tissue. Different imaging methods can be used to monitor second messenger production, in vivo. Non-invasive imaging using transgenic mice expressing Ca+ sensors such as Red-Aequorin [30] or cAMP sensors such as pGL3-LuciferasePKA [31] can present an interesting avenue with good dynamic responsiveness. Alternatively, relying on second messenger driven expression of luciferase can also be considered for whole animal imaging using CCD camera in the far-red/near-infrared spectrum. However these latter approaches have much less fast dynamic responsiveness and may thus not be well suited for all use. Finally, methods relying on noninvasive imaging often require for animals to be anesthetized, which may not be compatible with all experimental questions. Furthermore, no method is currently available to image bArr mediated event selectively in vivo. As an alternative to imaging, older biochemical approaches can also be used in freely moving awakened animals. For instance, measurement of inositol triphosphate (IP3) by binding displacement has been used with success as a proxy to quantify Gq mediated signaling in vivo [32]. More proximal to the receptor, measurements of G-protein recruitment using the GTPcS method can be used on membrane preparations from drug treated mice [33,34]. Finally, direct measurements of signaling protein such as Erk2 and Akt phosphorylation/activation using immunoblot analysis (Western blot) also constitute interesting tools to characterize the signaling footprint of a GPCR ligand in vivo [11]. As an alternative to Western blot, dot-blot can also be used. However this is often not suitable since several commercially available primary antibodies for phospho-protein often detect phospho-epitope on other proteins of a different molecular weight. Furthermore alternative approaches such as alpha-screen or ELISA are seldom used for such measurement due to an incompatibility of these methods with denaturing agents (see Section 4) that are used to maximize the preservation of protein phosphorylation during extraction [35].

5

As a general rule, biochemical approaches can be robust but often have limited temporal and spatial resolutions. Moreover, post-mortem signal decay is often critical thus requiring rapidity of dissection [36]. Furthermore, the method chosen for euthanasia can also be critical since common approaches such as CO2 asphyxiation and overdose of anesthetic have been shown to preclude some types of measurements [36,37]. Overall these different methods all have advantages and limitations. However, a shared characteristic of all of them is a need for statistical power in order to take into consideration the inherent greater variability of measurements obtained from living animals. The total numbers of animals needed for a given technique could vary and should be established for each experimental system. In all cases it is important that statistical variance be conserved over data analysis. Therefore, common cell culture practices like running several times different conditions with an effective (n) of 1 per condition and then pooling independent immunoblot experiments are not usable. Indeed, by normalizing each condition to a single control for each experiment this approach eliminates variance within the control condition. As one of several possible practical approaches, we now present a detailed method to evaluate ligand selectivity downstream of D2R in the mouse striatum. To achieve this, we use phosphorylation of DARPP32 on Th32 as a reporter for the modulation of cAMP production by Gai mediated signaling. Phosphorylation of Akt on Th308 or GSK3b on Ser9 can be used as reporters for bArr2 mediated signaling. To demonstrate the involvement of D2R, experiments have to be carried out in D2R-KO mice and WT littermates. Involvement of bArr2 can be established using bArr2KO animals (Fig. 2A–C). 4. Materials and methods 4.1. Animal models Mice lacking bArr1 or bArr2 are available at the Jackson Laboratory (Bar Harbor, ME, www.jax.org). bArr2-KO mice are available under two lines (stock numbers: 011130 and 023852), the second line is at 99.90% congenic on a C57Bl5J background. bArr1-KO mice are available under stock number 011131. Additional bArr1 and bArr2-KO mice lines are also available from the International mouse consortium (www.mousephenotype.org). Mice lacking D2R are also available from the Jackson Laboratory (stock number: 003190). Alternatively, conditional D2R-KO mice are also available (stock number: 020631). For each condition, 5–10 animals are to be used. 4.2. Acute drug administration Compounds should be solubilized in an appropriate vehicle. For hydrophilic compounds sterile saline should be used as a vehicle. For hydrophilic compounds dimethyl sulfoxide (DMSO) can be use as a solvent. However, high concentration of DMSO can be toxic for mice. If this is to be a problem, compounds can be administered as a suspension in tween-20. Suspension is prepared by grinding the compound in 5–10 ll tween-20, completing to final volume in sterile distilled water followed by sonication until the suspension is homogenous. Compounds are administered either by intra-peritoneal (i.p.) or sub-cutaneous (s.c.) injection. To reduce toxicity, s.c. injections are favored for compounds solubilized in DMSO. Mice within a cohort are injected alternatively either with vehicle or compound. A delay of 3 min is left between each mouse in order to allow time for tissue collection at the same time post-injection for all the mice of a given cohort. As a note of caution, expression of some signaling molecules such as GSK3 is

Please cite this article in press as: J.M. Beaulieu, Methods (2015), http://dx.doi.org/10.1016/j.ymeth.2015.08.018

6

J.M. Beaulieu / Methods xxx (2015) xxx–xxx

Fig. 2. Measurement of GSK3b phosphorylation on Ser9 in the mouse striatum. (A) Example of an immunoblots stained using rabbit anti-pGSK3b and mouse anti-total GSK3 and detected with respective goat anti-rabbit (IR-Dye 680, green) and anti-mouse (IR-Dye 800, red) labeled secondary antibodies. Signal was acquired using an Odyssey Imager (Licor Biotechnology). (B) Quantification of pAkt (Thr308) and pDARPP-32 (Thr34), in extracts prepared from the striatum of D2R-KO, bArr2-KO mice and their respective WT littermates following acute treatment with haloperidol (1 mg/kg i.p.). pDARPP-32 was measured at 15 min and pAkt at 120 min post injection. (C) Column data-point graph showing the modulation of GSK3b phosphorylation in response to acute D2R antagonist haloperidol in WT and bArr2-KO mice. Mice (n = 5 per group) were treated with haloperidol (1 mg/kg, i.p.) and euthanized 2 h after drug administration. The relative change was calculated from the ratio of phospho-dependent signal to phospho-independent signal for each sample. Since WT and bArr2-KO mice were measured on different blots, results were normalized to the average ratio of vehicle treated mice for each genotype. Under such a presentation, bArr2 and WT mice are not directly comparable as illustrated by the broken x-axis. However, data allow to conclude that haloperidol does not affect GSK3b phosphorylation in bArr2-KO mice. (D) Time course of Akt phosphorylation (Th308) as measured by Western blot analysis at 0, 30, 60, 90, and 120 min after amphetamine injection (2 mg/kg, i.p.) in WT and bArr2-KO mice. For B and C data are mean ± SEM. *p 6 0.05, **p 6 0.001 double tailed Student T-test (n = 10 mice per data point). For D, data are mean ± SEM. ***p 6 0.001. One-way ANOVA.

affected by circadian regulation. To ensure replicability it is recommended to carry out experiments at the same circadian time. Haloperidol and amphetamine (Fig. 2) were obtained from Tocris Bioscience (Minneapolis, MN). 4.3. Euthanasia and tissue collection Mice are euthanized every 3 min following order of injection. To prevent signal decay, animals are euthanized by cervical dislocation followed by decapitation and immersion of the head in liquid nitrogen for a period of 5–7 s. Striata are dissected on a ice cold surface with a delay of less then 1 min post euthanasia, placed in a conic 1.5 ml microfuge tube, flash-freeze in liquid nitrogen and stored at 80 °C. Time between euthanasia and freezing of tissue sample should be below 1.5 min. 4.4. Protein extraction Frozen tissue samples are kept on dry ice until homogenization. Samples are homogenized in conic minifuge tubes using a motorized pellet pestle (Sigma–Aldrich, Oakville, Ontario). Homogenization is carried out in a volume of 100 ll of a 1% sodium dodecyl sulfate (SDS) solution at 95 °C. Homogenate is placed on a heater

block at 95 °C and 5 ll is taken for protein quantification. Homogenate is then completed with a volume of 100 ll 2 Laemmli sample buffer (Bio-Rad, Hercules, CA) containing 5% (v/v) bmercaptoethanol (Sigma–Aldrich) and boiled for 5 min. Protein concentration in the homogenate is determined using a DCprotein assay (Bio-Rad). Phospho-protein signal generally demands a relatively large amount of protein (25 or 50 lg) to ensure detectability. When calculating protein concentration one should keep in mind that an additional dilution factor must be included to account for Laemmli sample buffer.

4.5. Immunoblots Protein extracts are separated on 10% SDS/PAGE Tris–glycine gels (Invitrogen, Burlington, Ontario) and transferred to nitrocellulose membranes. Since most SDS–PAGE gels contain 8–20 wells, this limits the number of samples that can be tested at the same time. Furthermore, samples that are run on different gels may lead to different signal intensities due to variations in transfer. This thus limits the number of conditions that can be tested on a same gel. For example, one 12 well gel allows to compare five vehicle treated mice with five drug treated mice, the two extra wells being used for molecular weight standards. This disposition allows to preserve

Please cite this article in press as: J.M. Beaulieu, Methods (2015), http://dx.doi.org/10.1016/j.ymeth.2015.08.018

J.M. Beaulieu / Methods xxx (2015) xxx–xxx

statistical variance within both groups. If more mice are needed or if one wishes to evaluate more than two conditions at once, one solution can be to use larger gels. Alternatively, several gels can be cut and the regions corresponding to the molecular weight of interest can be transferred to a single membrane to allow samples run on different gels to be compared directly. Following transfer blots are stained overnight at 4 °C with primary antibodies. Immune complexes are revealed using appropriate secondary antibodies. 4.6. Primary antibodies Anti-phospho-GSK3b (Ser9) polyclonal antibody, was from Cell Signaling Technology (Beverly, MA). Anti-total-GSK3a/b clone 0011-A monoclonal antibody and anti-phospho-Akt (Thr308) polyclonal antibody, were from Santa Cruz Biotechnology (Santa Cruz, CA). Anti-total-Akt monoclonal antibody, was from Biosources (San Diego, CA). Anti-phospho DARPP-32 (Thr34) polyclonal antibodies, was from Phospho-Solutions (Aurora, CO). The anti-totalDARPP32 monoclonal antibody was from BD Transduction Laboratories (Lexington, KY).

7

5. Results and discussion Immunoblot analysis shows that the neutral D2R antagonist haloperidol increases the phosphorylation of Akt (Th308) and DARPP-32 (Th34) albeit at different time post injection in WT mice (Fig. 2B). This effect of haloperidol does not require coadministration of an agonist. Furthermore it is mediated by D2R since haloperidol did not affect Akt and DARPP-32 phosphorylation in D2R-KO mice. Haloperidol still affects DARPP-32 phosphorylation in bArr2-KO mice (Fig. 2B). However effects of haloperidol on Akt and GSK3b (Ser9) phosphorylation are absent in bArr2-KO mice. This indicates that haloperidol is not a biased D2R ligand as it affects both bArr and G protein mediated modalities of D2R signaling. In contrast, when testing a biased antagonists affecting bArr2 mediated signaling one would expect to observe effects on Akt and GSK3 phosphorylation and no effect on DARPP-32 phosphorylation. Furthermore, effects on Akt and GSK3b should not occur in bArr2-KO mice. The time course presented in Fig. 2D illustrates that inhibition of Akt in response to the dopamine releasing agent amphetamine displays a slow kinetic and is dependent of bArr2 as it was shown previously [11].

4.7. Secondary antibodies 6. Conclusion 680 nm anti-rabbit IgG and 800 nm anti-mouse IgG IR dye labeled from Licor Biotechnology (Lincoln, NE) were used. 4.8. Signal detection The use of a dedicated gel scanner such as the Odyssey Imager (Licor Biotechnology) is strongly recommended at this stage. This gives a direct quantification of secondary antibody fluorescence and allows multiplexing (Fig. 2A). The presence of white areas on bands indicates signal saturation. As a note of caution, images obtained from the imager should not be considered quantitative and use of ‘‘densitometric analysis” in other software (e.g. Image J). Only the data provided by the Imager are quantitative. When using autoradiography and a luminescence based method, it is important to generate several exposures of each membrane in order to work within a linear range. Also when scanning autoradiograms it is important to disengage any image correction and not to save the resulting image in a compressed format (e.g. JPEG) so as not to compromise quantitative information. 4.9. Data analysis Several simple parameters are important for data analysis. First, the only valid loading control to monitor variation in protein phosphorylation is a signal coming from a phospho-independent antibody for the same protein. Using antibodies for actin or another house-keeping gene does not provide information about variations in phosphorylation. Second, a simple solution to normalize data to control is to divide all data obtained from a membrane by the average signal generated for the control group from the same blot. This is also valid for data from the control group itself, since the objective is to preserve their variance for subsequent statistical analysis using the appropriate test as defined by the effective, normal vs non-normal distribution and the number of groups being compared. Finally one should not be surprised if there is an overlap between individual control samples and test samples. This is normal and is a part of the variability associated to approaches involving whole animals. Data could be presented as histograms (Fig. 2B) or as column data-point graphs in order to represent and illustrate variability (Fig. 2C).

Validation of ligand functional selectivity in pre-clinical animal models is central to further validation and acceptance of this type of drug to the clinic. Yet, there is still a relative paucity of tools to study it in vivo. One of the biggest needs is to improve sensitivity as well as spatial and temporal resolutions for signal measurement. This might be achieved by developing new reporter mouse lines allowing for real-time non-invasive quantitative imaging of both G-protein and bArr mediated signaling events. Furthermore, development of new tools allowing for cell type selective validation of signaling responses is also needed. At the moment tools to demonstrate the implication of either bArr in vivo are mostly restricted to full knockout mice which may affect several intricate receptor systems and signaling modalities at the same time. The recent development of mice expressing GPCR that exhibit strongly biased signal transduction for either G-protein or bArr is a step in the right direction [38]. However such models do not allow differentiating between various functions of bArr. Complementing these with long awaited conditional bArr-KO mice or somatic and/or germinal genome editing methods allowing to inactivate specific signaling intermediates [39] or to prevent protein interactions would be welcomed additions. The development of such new tools will certainly speed up considerably the validation of biased GPCR ligands in vivo. Acknowledgements J.M.B. is supported by a Canada research Chair in Molecular Psychiatry. This work was supported by a discovery Grant form the National Council for Research in Engineering and Natural Sciences (NSERC). Funding agencies had no contribution to the content of this article. References [1] S. Galandrin, G. Oligny-Longpre, M. Bouvier, Trends Pharmacol. Sci. 28 (2007) 423–430. [2] J. Shonberg, L. Lopez, P.J. Scammells, A. Christopoulos, B. Capuano, J.R. Lane, Med. Res. Rev. 34 (2014) 1286–1330. [3] J.D. Violin, A.L. Crombie, D.G. Soergel, M.W. Lark, Trends Pharmacol. Sci. 35 (2014) 308–316. [4] S. Rajagopal, K. Rajagopal, R.J. Lefkowitz, Nat. Rev. Drug Discov. 9 (2010) 373– 386.

Please cite this article in press as: J.M. Beaulieu, Methods (2015), http://dx.doi.org/10.1016/j.ymeth.2015.08.018

8

J.M. Beaulieu / Methods xxx (2015) xxx–xxx

[5] J.M. Beaulieu, R.R. Gainetdinov, M.G. Caron, Trends Pharmacol. Sci. 28 (2007) 166–172. [6] J.P. Overington, B. Al-Lazikani, A.L. Hopkins, Nat. Rev. Drug Discov. 5 (2006) 993–996. [7] S.S. Ferguson, W.E. Downey 3rd, A.M. Colapietro, L.S. Barak, L. Menard, M.G. Caron, Science 271 (1996) 363–366. [8] J.L. Benovic, H. Kuhn, I. Weyand, J. Codina, M.G. Caron, R.J. Lefkowitz, Proc. Natl. Acad. Sci. U.S.A. 84 (1987) 8879–8882. [9] S.A. Laporte, R.H. Oakley, J. Zhang, J.A. Holt, S.S. Ferguson, M.G. Caron, L.S. Barak, Proc. Natl. Acad. Sci. U.S.A. 96 (1999) 3712–3717. [10] L.M. Luttrell, S.S. Ferguson, Y. Daaka, W.E. Miller, S. Maudsley, G.J. Della, Science 283 (1999) 655–661. [11] J.M. Beaulieu, T.D. Sotnikova, S. Marion, R.J. Lefkowitz, R.R. Gainetdinov, M.G. Caron, Cell 122 (2005) 261–273. [12] L.M. Luttrell, D. Gesty-Palmer, Pharmacol. Rev. 62 (2010) 305–330. [13] W. Stallaert, A. Christopoulos, M. Bouvier, Expert Opin. Drug Discov. 6 (2011) 811–825. [14] S. Costanzi, Trends Pharmacol. Sci. 35 (2014) 277–283. [15] Y. Namkung, O. Radresa, S. Armando, D. Devost, A. Beautrait, C. Le Gouill, S.A. Laporte, Methods (2015). [16] P. Su, S. Li, S. Chen, T.V. Lipina, M. Wang, T.K. Lai, F.H. Lee, H. Zhang, D. Zhai, S.S. Ferguson, J.N. Nobrega, A.H. Wong, J.C. Roder, P.J. Fletcher, F. Liu, Neuron 84 (2014) 1302–1316. [17] Y. Han, I.S. Moreira, E. Urizar, H. Weinstein, J.A. Javitch, Nat. Chem. Biol. 5 (2009) 688–695. [18] J.M. Beaulieu, R.R. Gainetdinov, Pharmacol. Rev. 63 (2011) 182–217. [19] N.M. Urs, P.J. Nicholls, M.G. Caron, Curr. Opin. Cell Biol. 27 (2014) 56–62. [20] B. Masri, A. Salahpour, M. Didriksen, V. Ghisi, J.M. Beaulieu, R.R. Gainetdinov, M.G. Caron, Proc. Natl. Acad. Sci. U.S.A. 105 (2008) 13656–13661. [21] J.A. Allen, J.M. Yost, V. Setola, X. Chen, M.F. Sassano, M. Chen, S. Peterson, P.N. Yadav, X.P. Huang, B. Feng, N.H. Jensen, X. Che, X. Bai, S.V. Frye, W.C. Wetsel, M. G. Caron, J.A. Javitch, B.L. Roth, J. Jin, Proc. Natl. Acad. Sci. U.S.A. 108 (2011) 18488–18493. [22] M.S. Roh, M.S. Seo, Y. Kim, S.H. Kim, W.J. Jeon, Y.M. Ahn, U.G. Kang, Y.S. Juhnn, Y.S. Kim, Exp. Mol. Med. 39 (2007) 353–360.

[23] B.L. Roth, D.J. Sheffler, W.K. Kroeze, Nat. Rev. Drug Discov. 3 (2004) 353–359. [24] A. Polter, S. Yang, A.A. Zmijewska, T. van Groen, J.H. Paik, R.A. Depinho, S.L. Peng, R.S. Jope, X. Li, Biol. Psychiatry 65 (2009) 150–159. [25] H. Wei, S. Ahn, S.K. Shenoy, S.S. Karnik, L. Hunyady, L.M. Luttrell, R.J. Lefkowitz, Proc. Natl. Acad. Sci. U.S.A. 100 (2003) 10782–10787. [26] A.J. Pijnenburg, W.M. Honig, J.M. Van Rossum, Psychopharmacologia 41 (1975) 87–95. [27] J.M. Beaulieu, R.R. Gainetdinov, M.G. Caron, Annu. Rev. Pharmacol. Toxicol. 49 (2009) 327–347. [28] J.M. Beaulieu, S. Marion, R.M. Rodriguiz, I.O. Medvedev, T.D. Sotnikova, V. Ghisi, W.C. Wetsel, R.J. Lefkowitz, R.R. Gainetdinov, M.G. Caron, Cell 132 (2008) 125–136. [29] N.M. Urs, S. Bido, S.M. Peterson, T.L. Daigle, C.E. Bass, R.R. Gainetdinov, E. Bezard, M.G. Caron, Proc. Natl. Acad. Sci. U.S.A. 112 (2015) E2517–E2526. [30] A. Bakayan, C.F. Vaquero, F. Picazo, J. Llopis, PLoS One 6 (2011) e19520. [31] J.O. Moskaug, H. Carlsen, R. Blomhoff, Mol. Imaging 7 (2008) 35–41. [32] I.O. Medvedev, A.J. Ramsey, S.T. Masoud, M.K. Bermejo, N. Urs, T.D. Sotnikova, J. M. Beaulieu, R.R. Gainetdinov, A. Salahpour, J. Neurosci. 33 (2013) 18125– 18133. [33] R.R. Gainetdinov, L.M. Bohn, T.D. Sotnikova, M. Cyr, A. Laakso, A.D. Macrae, G.E. Torres, K.M. Kim, R.J. Lefkowitz, M.G. Caron, R.T. Premont, Neuron 38 (2003) 291–303. [34] L.M. Bohn, L. Zhou, J.H. Ho, Methods Mol. Biol. 1335 (2015) 177–189, http://dx. doi.org/10.1007/978-1-4939-2914-6_12. PMID:26260601. [35] A.A. Fienberg, N. Hiroi, P.G. Mermelstein, W. Song, G.L. Snyder, A. Nishi, A. Cheramy, J.P. O’Callaghan, D.B. Miller, D.G. Cole, R. Corbett, C.N. Haile, D.C. Cooper, S.P. Onn, A.A. Grace, C.C. Ouimet, F.J. White, S.E. Hyman, D.J. Surmeier, J. Girault, E.J. Nestler, P. Greengard, Science 281 (1998) 838–842. [36] X. Li, A.B. Friedman, M.S. Roh, R.S. Jope, J. Neurochem. 92 (2005) 701–704. [37] M.S. Roh, T.Y. Eom, A.A. Zmijewska, P. De Sarno, K.A. Roth, R.S. Jope, Biol. Psychiatry 57 (2005) 278–286. [38] S.M. Peterson, T.F. Pack, A.D. Wilkins, N.M. Urs, D.J. Urban, C.E. Bass, O. Lichtarge, M.G. Caron, Proc. Natl. Acad. Sci. U.S.A. 112 (2015) 7097–7102. [39] P.D. Hsu, E.S. Lander, F. Zhang, Cell 157 (2014) 1262–1278.

Please cite this article in press as: J.M. Beaulieu, Methods (2015), http://dx.doi.org/10.1016/j.ymeth.2015.08.018