Main

Ubiquitination is a rapid and reversible post-translational modification that is involved in numerous aspects of biology. The 76-residue polypeptide ubiquitin fulfils essential functions in eukaryotes through its covalent attachment to other intracellular proteins1,2. Historically, the best-characterized role for this modification is the targeting of proteins for degradation by the 26S proteasome after the transfer of an ubiquitin chain of at least four units, referred to as polyubiquitination3. More recently, the addition of a single ubiquitin to one (monoubiquitination) or multiple (multiubiquitination) protein sites has been described4. In these cases, the post-translational modification serves as a reversible signal involved in a variety of key cellular functions including intracellular signaling, subcellular localization and protein-protein interactions.

Despite the growing interest in ubiquitination, inspired by the diversity of cellular functions that appear to be influenced by this process, the lack of tools for monitoring its dynamic regulation limits studies in this field. Western blot analysis using antibodies to ubiquitin allows the detection of protein ubiquitination. Although this assay can be used to determine whether a protein is ubiquitinated, it cannot be easily adapted to study the dynamic nature of the ubiquitination and deubiquitination cycles. To monitor the ubiquitination process in living cells, we took advantage of a bioluminescence resonance energy transfer (BRET) method, which allows real-time detection of protein-protein interactions in vivo. This technique is based on the nonradiative transfer of energy between a luminescent energy donor (here, Renilla luciferase, Rluc) and a fluorescent energy acceptor (here, green fluorescent protein, GFP). This is a system of choice for monitoring both constitutive and regulated intermolecular interactions because of the strict dependence on molecular proximity (<100 Å) between the donor and acceptor molecules for energy transfer5.

It was recently proposed that β-arrestin 2 (β-arrestin), a versatile protein involved in the regulation of G protein–coupled receptor (GPCR) signaling6, is ubiquitinated in response to receptor activation7. Based on the stability of their interaction with β-arrestin, two broad classes of GPCRs are distinguishable8,9: GPCRs known as class A receptors (for example, the β2-adrenergic receptor, β2AR) interact only transiently with β-arrestin after their activation, whereas activated class B receptors (for example, the V2-vasopressin receptor, V2R) form stable complexes with β-arrestin. Recent findings10 are suggestive of a link between the dynamics of β-arrestin ubiquitination and the nature of its interaction with the activated receptor. A stable interaction correlates with sustained β-arrestin ubiquitination, whereas a transient interaction results in short-lived ubiquitination. Using β-arrestin as a model ubiquitinated protein, we report here that BRET allows monitoring of the changes in the ubiquitination state of a protein in real time.

Results

Ubiquitination monitored by BRET in living cells

As ubiquitination results from the covalent attachment of ubiquitin to specific protein substrates, BRET should be adaptable to monitoring this post-translational modification in living cells. The covalent attachment of a GFP-ubiquitin fusion protein to a substrate fused to Renilla luciferase (Rluc) should bring the Rluc and GFP moieties within a BRET-permissive distance, resulting in a detectable BRET signal. To determine whether we could use BRET to study ubiquitination of β-arrestin, we designed N-terminally tagged Rluc–β-arrestin and GFP-ubiquitin (GFP2-Ubi) fusion protein constructs. The close proximity of several GFP2 moieties could cause quenching or interference phenomena that could lead to a decrease in the BRET signal, and indeed, we observed this (Supplementary Fig. 1 online). To limit the formation of poly(GFP2-Ubi) chains11, we created a mutant GFP2-Ubi protein by changing Lys48 and Lys63, the two residues through which linkages occur, to alanine residues.

To initiate the energy transfer between Rluc and GFP2 in cells coexpressing Rluc–β-arrestin and GFP2-Ubi, we used the luciferase substrate DeepBlueC coelenterazine, which readily crosses the plasma membrane. Hydrolysis of DeepBlueC coelenterazine leads to the emission of light with a spectrum overlapping the excitation spectrum of GFP2, thus allowing BRET2 (Fig. 1). Coexpression of a constant level of Rluc–β-arrestin with increasing concentrations of GFP2-Ubi led to an increase in the BRET2 signal (Fig. 2a). The energy transfer increased hyperbolically as a function of the GFP2-Ubi expression level, indicative of a specific interaction between the two proteins12. To confirm that the signal specifically reflects the ubiquitination process, we used either a GFP2-tagged ubiquitin mutant that is unable to take part in the ubiquitination process as a result of the replacement of its last two glycines by alanine residues (GFP2-UbiAA) or GFP2 alone as negative controls. Coexpression of Rluc–β-arrestin with GFP2-UbiAA or GFP2 led to weaker signals that increased linearly rather than hyperbolically with the increase in the fluorescence/luminescence ratio, most likely reflecting random collision (bystander BRET) between Rluc–β-arrestin and these two GFP2 constructs. Western blot analysis of lysates from cells expressing GFP2, GFP2-UbiAA or GFP2-Ubi corroborated the lack of integration of GFP2-UbiAA into proteins (Fig. 2b). The smeared band detected above 63 kDa in lysates from cells expressing GFP2-Ubi most likely reflects ubiquitination of multiple proteins, and the sharp bands observed in lysates from cells expressing GFP2-UbiAA probably correspond to monomeric and multimeric forms of the GFP2-tagged ubiquitin. The BRET2 signal detected between Rluc–β-arrestin and GFP2-Ubi did not result from the ubiquitination of the Rluc moiety itself, as no signal was detected when only GFP2-Ubi and Rluc were coexpressed (Fig. 2a). To determine whether the GFP2-Ubi protein can covalently modify β-arrestin, we immunoprecipitated Myc-tagged β-arrestin and detected GFP2-Ubi–immunoreactive species, which probably correspond to a ubiquitinated β-arrestin population (Fig. 2c). Taken together, these results demonstrated that energy transfer between GFP2-Ubi and Rluc–β-arrestin reflects a nonrandom clustering of these two proteins and the covalent ubiquitination of β-arrestin in vivo.

Figure 1: Schematic representation of the β-arrestin ubiquitination BRET2 assay.
figure 1

Upon degradation of its catalytic substrate (DeepBlueC coelenterazine), the Rluc fused to the N terminus of β-arrestin emits blue light with an emission peak at 395 nm. If the β-arrestin is ubiquitinated, a nonradiative transfer of energy will occur between Rluc and the GFP2 fused to the N terminus of ubiquitin, resulting in re-emission of light with an emission peak at 510 nm.

Figure 2: Specific BRET2 signal between Rluc–β-arrestin and GFP2-Ubi.
figure 2

(a) HEK 293 cells were cotransfected either with constant concentrations of Rluc–β-arrestin, Myc-tagged V2R and HA-tagged β2AR DNA constructs and increasing concentrations of plasmids encoding GFP2-Ubi (), GFP2-UbiAA () or GFP2 (), or with a constant concentration of Rluc and increasing concentrations of GFP2-Ubi DNAs (▪). The data shown represent pooled individual readings obtained from five to eight independent experiments. The curves were fitted using a nonlinear regression equation, assuming a single binding site (GraphPad Prism). (b) Lysates obtained from mock-transfected HEK 293 cells or cells transfected with pGFP2, pGFP2-UbiAA or pGFP2-Ubi were resolved by SDS-PAGE, and the attachment of ubiquitin to proteins was assessed by western blot analysis using an antibody to GFP. Numbers on the right indicate molecular weight of protein markers. The blot shown is representative of two independent experiments. (c) Immunoprecipitation with an antibody to Myc was performed on lysates prepared from cells expressing pGFP2-Ubi or pGFP2-Ubi and pCMV-Myc–tagged β-arrestin. GFP2-Ubi protein associated with the immunoprecipitated β-arrestin was revealed as in b with an antibody to GFP, whereas the amount of β-arrestin immunoprecipitated was controlled with an antibody to Myc.

Detection of agonist-induced ubiquitination

In a recent study, ex vivo western blot analysis suggested that GPCR activation modulates β-arrestin ubiquitination7,10. To assess whether BRET can detect receptor-regulated ubiquitination in living cells, we monitored the transfer of energy between Rluc–β-arrestin and GFP2-Ubi in the presence of coexpressed β2AR and V2R. Activation of each receptor with its selective agonist, isoproterenol (ISO) for the β2AR and arginine-vasopressin (AVP) for the V2R, substantially increased the BRET2 signal (Fig. 3a). In contrast, ligands did not have an effect on the bystander BRET2 observed between GFP2-UbiAA and Rluc–β-arrestin (Fig. 3b), confirming the selectivity of the agonist-promoted ubiquitination observed with GFP2-Ubi. This increase in BRET2 signal was dose dependent, and the agonist concentrations yielding a half-maximal response (EC50) were 140 and 1.8 nM for ISO and AVP, respectively, consistent with the affinity of these ligands for their receptors (Fig. 3c).

Figure 3: Detection of GPCR-induced β-arrestin ubiquitination.
figure 3

(a,b) Cells cotransfected with a constant concentration of Rluc–β-arrestin, Myc-tagged V2R and HA-tagged β2AR plasmids and increasing concentrations of GFP2-Ubi (a) or GFP2-UbiAA (b) constructs were unstimulated () or stimulated with 10−5 M ISO () or 10−6 M AVP () for 120 s before BRET measurement. DeepBlueC coelenterazine was added in the continuous presence of the agonists and the BRET measurements were taken 20 s after the addition of luciferase substrate. The data shown represent pooled individual readings obtained from three independent experiments. (c) Cells expressing the same construct as in a and displaying a GFP2-Ubi/Rluc–β-arrestin expression ratio of 0.1 ± 0.005 (assessed by measuring total fluorescence and luminescence signals) were stimulated with increasing concentration of ISO (left) or AVP (right) for 120 s; BRET measurements were then collected as in a. The curves shown represent the mean ± s.e.m. of three independent experiments.

Concomitant detection of ubiquitination and recruitment

It has been suggested that receptor activation may regulate the ubiquitination of β-arrestin and its recruitment to GPCRs10,13 in a coordinated fashion. To monitor these two events simultaneously in the same cells, we took advantage of the distinct spectral properties of two luciferase substrates that allow energy transfer with two different fluorescent proteins. Whereas the Rluc-catalyzed oxidation of DeepBlueC coelenterazine leads to BRET2 between Rluc and GFP2, the use of coelenterazine h leads to a distinct emission spectrum that promotes BRET1 between Rluc and a variant of GFP, the yellow fluorescent protein (YFP)14,15,16. We divided a culture of cells coexpressing Rluc–β-arrestin, GFP2-Ubi and V2R-YFP into two samples. In one sample we monitored the β-arrestin ubiquitination by assessing the BRET2 between Rluc–β-arrestin and GFP2-Ubi, and in the second sample we measured BRET1 between Rluc–β-arrestin and V2R-YFP to evaluate the recruitment of β-arrestin to the receptor (Fig. 4). Stimulation with AVP resulted in an increase in both BRET2 and BRET1 signals, reflecting concomitant ubiquitination (Fig. 4a) and recruitment of β-arrestin (Fig. 4b). Each of the BRET signals detected was specific to the interaction monitored, as no substantial transfer of energy occurred between Rluc and GFP2 in the presence of coelenterazine h or between Rluc and YFP in the presence of DeepBlueC coelenterazine.

Figure 4: Concomitant detection of β-arrestin ubiquitination and its recruitment to the activated receptor.
figure 4

(a,b) Cells cotransfected with Rluc–β-arrestin, V2R-YFP and GFP2-Ubi were stimulated with 1 μM AVP or vehicle alone for 10 min. DeepBlueC coelenterazine (BRET2, a) or coelenterazine h (BRET1, b) were then added in the continuous presence of the agonist and BRET measurements collected 20 s after the addition of the luciferase substrate to monitor Rluc–β-arrestin ubiquitination or its recruitment to V2R-YFP, respectively. Bar graphs shown represent the mean ± s.e.m. of three independent experiments. The lack of cross-talk between BRET1 and BRET2 in the same samples was verified by measuring the BRET signal between Rluc–β-arrestin and GFP2-Ubi in the presence of coelanterazine h for BRET2 and between Rluc–β-arrestin and V2R-YFP in the presence of DeepBlueC for BRET1. Virtually no cross-talk signal was observed (right side of each panel). Statistical significance of the difference was assessed using paired Student's t-test. *P < 0.05.

Real time monitoring of distinct ubiquitination kinetics

As described above, one can distinguish two GPCR classes based on their profile of interaction with β-arrestin. Because the ubiquitination state of β-arrestin seems intimately linked to its interaction profile with receptors10, we followed the kinetics of agonist-promoted β-arrestin ubiquitination in real time in cells coexpressing β2AR and V2R. Both ISO and AVP induced rapid and similar increases in the BRET-detected ubiquitination of β-arrestin that reached a peak at 2 minutes after activation (Fig. 5). However, although it remained stable for at least 10 min after V2R activation (Fig. 5b,c), the BRET signal returned progressively toward basal values after activation of the β2AR (Fig. 5a,c). This reduction in BRET signal most likely reflects a deubiquitination process and not a degradation of β-arrestin. As BRET is a ratiometric measurement (GFP2 emission divided by Rluc emission), degradation of Rluc–β-arrestin should induce a decrease in luminescence and lead to an equivalent reduction in the transfer of energy to the acceptor without affecting the signal. These results, therefore, confirm that in living cells, the nature of the activated receptor determines the dynamics of β-arrestin ubiquitination.

Figure 5: Real-time assessment of receptor stimulated β-arrestin ubiquitination.
figure 5

Cells cotransfected with Rluc–β-arrestin, Myc-tagged V2R, HA-tagged β2AR and GFP2-Ubi (GFP2-Ubi to Rluc–β-arrestin ratio of 0.1 ± 0.005) constructs, were incubated in the presence of DeepBlueC coelenterazine immediately before stimulation with 10−5 M ISO (; a,c) or 10−6 M AVP (; b,c). BRET2 measurements were then collected immediately every 10 s in the continued presence of the agonists for 10 min. (a,b) The real-time agonist-promoted increases in BRET are represented as the difference in the absolute BRET values obtained in the presence and in the absences of ISO or AVP. (c) Compilation of three independent experiments carried out as in a and b. The data shown represent the mean ± s.e.m. and are expressed in percentage of the maximal BRET2 signal obtained for each curve.

Discussion

The aim of this study was to adapt an energy transfer assay to the detection of protein ubiquitination in living cells. Using β-arrestin as a model, we demonstrated that BRET can specifically detect both basal and dynamically regulated ubiquitination processes. As compared to western blot analysis, the use of BRET presents several advantages: (i) because the BRET assay is carried out in living cells, it avoids the possible signal alterations that could result from cell lysis, protein solubilization or any purification steps before western blot analysis; (ii) the ability to follow the evolution of the BRET signal in real time allows one to capture the dynamic nature of the ubiquitination process; (iii) one can monitor changes in ubiquitination levels resulting from specific treatments in one population of cells using BRET, whereas western blot experiments require distinct cellular pools; and (iv) because of the ratiometric nature of the BRET technique, a reduction in signal truly represents a decrease in ubiquitination and does not result from the degradation of the ubiquitinated protein target.

As the BRET method is a proximity-based assay that was previously used to detect noncovalent protein-protein interactions, it was important to validate that BRET detected between Rluc–β-arrestin and GFP2-Ubi truly reflected the covalent attachment of the ubiquitin moiety. This was achieved by comparing the BRET signals obtained between Rluc–β-arrestin and either GFP2-Ubi, GFP2-UbiAA or the GFP2 moiety alone. The BRET signal obtained between Rluc–β-arrestin and GFP2-Ubi increased hyperbolically with increasing concentration of GFP2-Ubi before reaching a plateau once the covalent attachment of ubiquitin to Rluc–β-arrestin attained the maximal steady state. In contrast, the smaller, linear BRET signal obtained between Rluc–β-arrestin and GFP2-UbiAA or GFP2 most likely resulted from bystander BRET12. Because it cannot be incorporated into proteins, the concentration of free GFP2-UbiAA available for random collision should be equivalent to that reached when the GFP2 moiety alone is expressed, with both being much higher than that attained with GFP2-Ubi for equivalent GFP2 constructs. It follows that the bystander BRET component of the hyperbolic signal observed between Rluc–β-arrestin and GFP2-Ubi should then be much smaller than that observed with GFP2-UbiAA or GFP2. The lower signal detected between the Rluc moiety alone and GFP2-Ubi compared to that monitored between Rluc–β-arrestin and GFP2-UbiAA or GFP2 confirms this prediction. This makes GFP2-UbiAA a stringent control that allows one to distinguish between protein-protein interactions that are due to random collision and those due to covalent ubiquitination of a specific substrate.

Both BRET and immunoprecipitation approaches indicated that β-arrestin undergoes basal ubiquitination to a significant extent. These results are somewhat different from those of a published report7,10 in which ubiquitination of β-arrestin was observed almost exclusively as a result of receptor activation. The specific reasons for this difference are not known, but it probably results from differences in the experimental conditions used, such as receptor expression level or the overexpression of exogenous ubiquitin protein.

An important aspect of the BRET-based method is its ability to detect dynamic changes in the ubiquitination state of a protein and to monitor changes over time. The ubiquitination BRET assay allowed us to monitor distinct kinetics of β-arrestin ubiquitination resulting from the activation of different GPCRs. Activation of the β2AR induced a fleeting β-arrestin ubiquitination, whereas a long-lasting ubiquitination of the regulatory protein was observed after activation of the V2R. This is consistent with a previous study examining the ubiquitination status of β-arrestin at 1 and 15 min after receptor stimulation10. However, because the activated receptor recruits β-arrestin, the increase in ubiquitination BRET signal (Fig. 2a) could result from the ubiquitination of the receptor itself. This was not the case, as no BRET modulation occurred between β2AR-Rluc or V2R-Rluc and GFP2-Ubi after brief agonist stimulation (Supplementary Fig. 2 online). The data show that the ubiquitination BRET assay can faithfully reflect rapid and differential changes in the ubiquitination state of a specific substrate after specific stimuli.

The GFP2-Ubi construct used for the BRET experiments was designed to lack the two major sites involved in polyubiquitination in mammals (Lys48 and Lys63) to prevent the formation of GFP chains that could interfere with the energy transfer process. Despite the constraint that such a construct could impose on the system, ubiquitination of β-arrestin, which is believed to be polyubiquitinated7,10, could be readily observed. This is not surprising given that the process of polyubiquitination permits the attachment of the GFP2-Ubi through its terminal glycine either to β-arrestin targeted lysine(s) or to growing chains of ubiquitin. As the use of the GFP2-Ubi construct, lacking the polyubiquitination sites, could interfere with the properties of ubiquitinated proteins, we compared the ubiquitination BRET signal obtained with this construct to the signal obtained in the presence of a wild-type GFP2-Ubi protein (GFP2-UbiWT). Although a specific hyperbolic BRET signal was detected between Rluc–β-arrestin and GFP2-UbiWT, it was lower than the one obtained with GFP2-Ubi (Supplementary Fig. 1 online). Despite the lower absolute signal, similar BRET modulations were observed with both constructs after agonist stimulation (Supplementary Fig. 1 online). This indicates that GFP2-Ubi expression, per se, does not influence the nature of the results obtained but facilitates the detection of BRET signal changes. The fact that GFP2-UbiWT could also be used in the ubiquitination BRET assay makes the method more broadly applicable.

Besides its usefulness in directly monitoring protein ubiquitination, the BRET assay offers the possibility of concomitant detection of additional protein interactions involving the ubiquitinated protein. Taking advantage of two spectrally distinct coelenterazines, we demonstrated the feasibility of such dual detection by monitoring both ubiquitination of β-arrestin and its recruitment to the V2R in the same cells.

In addition to shedding new light on the dynamics of β-arrestin ubiquitination in living cells, our study represents proof of the principle that resonance energy transfer approaches allow the monitoring of ubiquitin attachment in real time. BRET and possibly fluorescence resonance energy transfer (FRET) could be used to study the ubiquitination of various proteins whose activity and fate are regulated by this post-translational modification. Studies of the growing number of ubiquitin-like proteins that regulate diverse cellular functions through their covalent attachment to specific protein substrates could also benefit from advances in ubiquitination BRET assays.

Methods

Plasmids.

To produce a wild type GFP2-Ubi construct (GFP2-UbiWT), we amplified the gene encoding human ubiquitin from a pQE32-Ubi plasmid and subcloned it into pGFP2-C1 (Perkin Elmer BioSignal). We then mutated by PCR lysines 48 and 63 in the ubiquitin moiety of the GFP2-UbiWT construct into alanines (GFP2-Ubi). Finally, we constructed the GFP2-UbiAA17,18 by mutating by PCR the two C-terminal glycines 75 and 76 into alanines. We previously described pcDNA3.1-RLuc–β-arrestin, pcDNA3.1-hemagglutinin (HA)–tagged β2AR, pcDNA3.1-Myc–tagged V2R and pRK5-V2R-YFP13,19,20,21. pCMV-Myc–tagged β-arrestin was a kind gift from S. Marullo's laboratory.

Cell culture and transfection.

HEK293 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 100 U/ml penicillin and streptomycin, 2 mM L-glutamine (Wisent) and cultured at 37 °C in 5 % CO2. We transfected the cells 24 h after seeding using the calcium phosphate precipitation method22.

BRET measurement.

Transfected cells were washed twice with PBS, detached in PBS with EDTA and resuspended in PBS with 0.1% glucose. We determined the cell number by measuring protein concentration using the DC protein assay kit (Bio-Rad). We obtained BRET measurements using two different sets of energy transfer partners. Energy transfer known as BRET2, between Rluc and GFP2, was determined after the addition of the Rluc substrate, DeepBlueC coelenterazine14 (Perkin Elmer), whereas BRET1, between Rluc and YFP, was assessed in the presence of another Rluc substrate, coelenterazine h15,16.

BRET2 monitoring of β-arrestin ubiquitination.

We distributed cells coexpressing Myc-tagged V2R, HA-tagged β2AR, Rluc–β-arrestin and GFP2-Ubi constructs in white 96-well Microplates (Corning) at a density of 100,000 cells per well. We collected BRET2 measurements 20 s after the addition of the luciferase substrate, DeepBlueC coelenterazine, at a final concentration of 5 μM. Readings were obtained using a modified TopCount NXT instrument (Packard Bioscience) that allows the sequential integration of the signals detected in the 370–450 and 500–530 nm ranges. The BRET2 signal corresponds to the ratio of the light emitted by the GFP2 (500–530 nm) over the light emitted by the Rluc (370–450 nm). We corrected the values by subtracting the background ratio value detected when the Rluc constructs were expressed alone. For titration experiments, the BRET signal was plotted as a function of the total fluorescence signal over the total luminescence signal (see below), a ratio used as an index of the concentration of GFP2-Ubi (or GFP2-UbiAA) expressed.

BRET1 monitoring of β-arrestin recruitment to the V2R receptor.

After dividing the cells coexpressing Rluc–β-arrestin, GFP2-Ubi and V2R-YFP into two samples, we initiated BRET1 between Rluc–β-arrestin and V2R-YFP by the addition of 5 μM coelenterazine h (the recruitment of β-arrestin to the receptor). We collected the readings at the appropriate emission wavelengths after the addition of 1 μM AVP using a Mithras LB 940 instrument (Berthold Technologies). We analyzed results as described above by calculating the ratio of the light emitted by the YFP (510–550 nm) to the light emitted by the Rluc (460–500 nm). We corrected the values by subtracting the background ratio value detected when the Rluc constructs were expressed alone.

Total fluorescence and luminescence measurements.

We dispersed cells in white 96 well plates with clear bottoms (Costar) at a density of 100,000 cells per well12. We measured total fluorescence of cells in a Fluorocount (Packard Bioscience) using an excitation filter at 400 or 485 nM and an emission filter at 510 or 530 nM in the case of GFP2 or YFP, respectively. We then incubated cells for 10 min with coelanterazine h to measure the total luminescence using a LumiCount instrument (Packard Bioscience). In the case of fluorescence, we subtracted the background value detected in untransfected cells to give a specific fluorescence value. In the case of luminescence, where the background was negligible, we used absolute values.

Immunoblotting and immunoprecipitation.

We solubilized cells expressing pcDNA3, GFP2-UbiAA or GFP2-Ubi in lysis buffer A containing 50 mM HEPES (pH7.5), 0.5 % Nonidet P-40, 250 mM NaCl, 2 mM EDTA, 10% (v/v) glycerol, 5 mM N-ethyl maleimide, 1 mM phenylmethylsulfonyl fluoride, 5 μg/ml leupeptin, 5 μg/ml aprotinin and 100 μM benzamidine. After separation of protein on SDS-PAGE and transfer to nitrocellulose, we revealed the immunoreactive bands using an antibody to GFP (Clontech), a horseradish peroxidase–conjugated antibody (Amersham) and the Western Lightning chemiluminescence Reagent Plus (Perkin Elmer). For immunoprecipitation experiments, we used cells expressing or not expressing pCMV-Myc–tagged β-arrestin 2 and pGFP2-Ubi and solubilized them in buffer A for 30 min with rocking at 4 °C. We then incubated soluble extracts with an antibody to Myc (9E10) and protein G–agarose beads overnight at 4 °C. Bound extracts were resolved by SDS-PAGE and transferred to nitrocellulose. We performed western blot analysis with an antibody to GFP, as described above, to reveal basal β-arrestin ubiquitination, and used A14 antibody to Myc (Santa Cruz) to detect the presence of Myc-tagged β-arrestin after its immunoprecipitation.

Note: Supplementary information is available on the Nature Methods website.