Arrestins: Introducing Signaling Bias Into Multifunctional Proteins

Abstract

Arrestins were discovered as proteins that bind active phosphorylated G protein-coupled receptors (GPCRs) and block their interactions with G proteins, i.e., for their role in homologous desensitization of GPCRs. Mammals express only four arrestin subtypes, two of which are largely restricted to the retina. Two nonvisual arrestins are ubiquitous and interact with hundreds of different GPCRs and dozens of other binding partners. Changes of just a few residues on the receptor-binding surface were shown to dramatically affect GPCR preference of inherently promiscuous nonvisual arrestins. Mutations on the cytosol-facing side of arrestins modulate their interactions with individual downstream signaling molecules. Thus, it appears feasible to construct arrestin mutants specifically linking particular GPCRs with signaling pathways of choice or mutants that sever the links between selected GPCRs and unwanted pathways. Signaling-biased “designer arrestins” have the potential to become valuable molecular tools for research and therapy.

1. WILD-TYPE ARRESTINS: TOO MANY FUNCTIONS

The first arrestin, visual (systematic name arrestin-1^a), was discovered for its ability to specifically bind light-activated phosphorylated rhodopsin and suppress G protein activation.^1,2 The first nonvisual arrestin, originally called β-arrestin, as it preferred β2-adrenergic receptor (β2AR) over rhodopsin (systematic name—arrestin-2), was discovered due to similar function: suppression of G protein activation by β2AR.^3,4 The second nonvisual arrestin, arrestin-3, was also discovered as a protein that specifically binds active phosphorylated G protein-coupled receptors (GPCRs),^5–7 whereas cone-specific arrestin-4 was cloned by homology.^8,9 It was shown in all cases that prior phosphorylation of GPCRs by GPCR kinases that specifically target active receptors¹⁰ was a prerequisite for high-affinity arrestin binding.¹¹ For some time, it appeared that the only function of all arrestins was the binding to active phosphorylated GPCRs, which precludes receptor interactions with G proteins.^12,13 Thus, arrestins play a critical role in the second step of GPCR homologous desensitization.¹⁴ The findings that nonvisual arrestins bind major components of the internalization machinery of the coated pit clathrin¹⁵ and clathrin adaptor AP2¹⁶ were first viewed as an extension of their desensitizing activity because the removal of GPCRs off the plasma membrane reduces cellular response. However, the discoveries that receptor-bound arrestin facilitates the activation of c-Src,¹⁷ as well as of MAP kinases JNK3¹⁸ and ERK1/2,¹⁹ suggested that many more interactions were likely. In fact, each nonvisual arrestin was found to interact with more than a hundred diverse proteins²⁰ and the list of their binding partners keeps growing. Among other things, arrestins were implicated in recruiting signaling proteins to microtubules,²¹ in cytochrome C release during apoptosis,²² focal adhesion dynamics,²³ activation of small GTPases,^24–27 glucose maintenance,²⁸ and many other cellular functions,^29,30 including even heterologous desensitization of GPCRs phosphorylated by second messenger-activated kinases.^31,32

Structurally, all four vertebrate arrestins are quite similar: elongated two-domain molecules, with the long axis of no more than 90 Å and the short axis of ~30–35 Å ^33–38 (Fig. 1). Even taking into account their conformational flexibility,³⁹ arrestins are not big enough to accommodate more than 4–5 binding partners at the same time.⁴⁰ This is particularly true considering that many signaling proteins interact with receptor-bound arrestins, where one side of the molecule is engaged, and therefore shielded, by the GPCR⁴¹ (Figs. 1 and 2). Thus, there must be distinct interaction sites for each individual partner, and its binding likely excludes the binding of many others.

Fig. 1.

The structure of arrestin. Two views are shown: (A) side view; (B) view down the concave sides of both domains (based on the crystal structure of bovine arrestin-2 solved with the resolution of 1.9 Å; PDB ID: 1G4M³⁵). The molecule is colored by secondary structure, as follows: light blue, β-strand; green, β-turn; red, α-helix. The long axis of the arrestin molecule is smaller than 90 Å, whereas its short axis is ~30–35 Å. N- and C-domains, their concave sides, as well as the central crest on the receptor-binding side of the molecule are indicated. The N-terminus (N-t) and C-terminus (C-t) of the part resolved in crystal are also indicated. Note that several N-terminal residues, the C-terminus extending beyond β-strand XX, as well as the linker between the C-domain and β-strand XX are not resolved in any arrestin structures.

Fig. 2.

Residues responsible for the receptor specificity of arrestins. The residues in arrestin-2 (PDB ID: 1G4M) in positions identified as important for the receptor preference of different arrestin proteins are shown as CPK models colored as follows: light blue, those identified by reduced mobility of the spin label in receptor-bound arrestins using EPR (Val70, Leu71, Leu73, Val167, Leu191, Ser234, Thr246, Tyr249); green, those identified by site-directed mutagenesis and receptor binding in vitro and in live cells (Leu48, Glu50, Arg51, Asp240, Cys251, Pro252, Asp259, Thr261); dark blue, those identified by both of these methods (Leu68, Tyr238); magenta, those identified by direct contact with the receptor in the crystal structure of the arrestin-1 complex with rhodopsin (Lys138, Asn245, Ala247, Gln248). Note that all receptor-binding residues identified by different methods are localized on the concave sides of the two arrestin domains. Many of these are localized in the central crest of the receptor-binding side of arrestins.

This notion implies two things: (a) targeted manipulation of individual sites can bias arrestin toward or away from particular partners; and (b) when these specific sites are identified, smaller elements with limited functional capabilities extracted from multifunctional arrestin proteins can be used as molecular tools to promote or suppress certain branches of signaling.

2. RECEPTOR SPECIFICITY: SELECTIVE TARGETING OF CERTAIN GPCRs?

Mammals express hundreds of different GPCRs,⁴² but only four arrestins.⁴³ However, evolution shows that an arrestin protein can be specific for a particular receptor. For example, arrestin-1 demonstrates remarkable preference for rhodopsin over other GPCRs.^3,44 Experimental swapping of various elements⁴⁴ and individual residues⁴⁵ between arrestin-1 and nonvisual arrestin-2 identified a limited set of side chains that determine receptor preference. Mutagenesis of every residue in arrestin-1^46,47 coupled with analysis of the crystal structure of the arrestin-1 complex with rhodopsin^48,49 (Fig. 2) revealed several additional residues that play important roles in GPCR binding and/or directly contact the receptor, and therefore might contribute to receptor specificity. Collectively, these studies identified more than a dozen “suspects.” Placing every one of the 20 possible residues in each position would require an enormous number of combinations that could not be tested experimentally within a reasonable period of time. Luckily, the analysis of arrestin evolution^29,43 shows that there were very few different residues in each of the critical positions, which substantially decreases the number of necessary mutations and combinations. The mutations of these “receptor discriminator” residues were actually introduced into the most promiscuous of the two nonvisual subtypes, arrestin-3.³⁸ Mutants were tested for their binding to β2AR, M2 muscarinic receptor, D1 and D2 dopamine receptors, and Y1 and Y2 neuropeptide Y receptors.^50–52 The results were encouraging. Replacement of just two residues enhanced arrestin-3 preference for some receptors over others ~50–60-fold.⁵¹ Certain mutations selectively affected arrestin interactions with inactive (predocking) and activated GPCRs.^50,52 The good news is that these data suggest that the construction of nonvisual arrestins with relatively narrow receptor specificity is feasible. The bad news is that testing even a few dozen arrestin mutants with ~400 nonodorant GPCRs expressed in every mammalian species is hardly realistic. Thus, this work should specifically target GPCRs with known disease-causing gain-of-function mutations.^53,54

3. MAKING ARRESTINS PHOSPHORYLATION INDEPENDENT

Receptor binding induces a global conformational change in arrestins,⁵⁵ which includes the movement of the two domains relative to each other^{48,49,56–59} (Fig. 3). In all arrestins relatively few intramolecular interactions hold the molecule in its basal conformation. The most important among these are the polar core, an arrangement of five charged residues in the center of the molecule between the two domains,^34,61 and a hydrophobic three-element interaction that involves β-strand I and α-helix in the N-domain and β-strand XX of the C-terminus.^34,62 Destabilization of either of these interactions in the visual arrestin-1 ^61–65 or in nonvisual arrestins ^66–69 yields “enhanced” mutants that bind the receptor more readily and demonstrate high binding even to unphosphorylated GPCRs.^64,66–70 Therefore, these enhanced arrestins have the potential to suppress excessive signaling of GPCR mutants and, specifically, compensate for the defects in receptor phosphorylation. Thus, the construction of receptor-specific arrestins should focus on these enhanced mutants and GPCR subtypes where gain-of-function mutations cause human disorders.^53,54 This work should follow the existing precedent of partial compensation of excessive signaling caused by the defects in rhodopsin phosphorylation by enhanced phosphorylation-independent version of visual arrestin-1, achieved in vivo in rod photoreceptors.^71,72 After all, the approach that yields measurable compensation in rods with single photon sensitivity and subsecond signaling shutoff is likely to yield near-complete compensation in any other GPCR-driven signaling system, all of which are a significantly less sensitive and much slower.

Fig. 3.

Activation-induced conformational changes in arrestins. The most notable conformational changes upon arrestin activation are shown here using arrestin-3 as an example. These include the release of the arrestin C-tail (magenta; magenta arrow shows the direction of the movement) anchored to the body of the molecule via the three-element interaction in the basal state (PDB ID: 3P2D³⁸), the movement to the receptor of the finger loop and helix formation in it (red; red arrow shows the direction of the movement), as well as the rotation of the two domains relative to each other (N-domain, gray; C-domain, teal). Rearrangements of several loops are also shown: 139-loop (middle loop in nonvisual arrestins; dark blue; dark blue arrow indicates the direction of movement); interdomain hinge (yellow with yellow arrowhead in a circle), one of the connectors between the N-domain and the α-helix (green with green arrowhead in a circle), register-shifted β-strand XI (dark brown), and the lariat loop (orange with orange arrow; note that in the active structure (PDB ID: 5TV1⁶⁰) it is not resolved). In Movie 1 in the online version at https://doi.org/10.1016/bs.pmbts.2018.07.007 arrestin-3 transitions from basal to active conformation are shown. The elements are color coded, as indicated earlier, except that the C-tail is not shown and register-shifted β-strand XI is shown in magenta.

4. CHANNELING THE SIGNALING TOWARD OR AWAY FROM SELECTED PATHWAYS

Arrestins have been implicated in the activation of numerous signaling proteins, including protein kinases c-Src,¹⁷ JNK3,¹⁸ and ERK1/2¹⁹ that play important roles in cell proliferation and death. Although arrestin-dependent signaling was long considered to be G protein independent,³⁰ recent studies showed that arrestin-mediated signaling in many cases (possibly always, although generalizations are dangerous) depends on the activation of G proteins.^73,74 In fact, this is biologically understandable, as the previous model of arrestin-dependent signaling did not explain how the upstream-most kinases, such as c-Raf1 in case of ERK1/2, or ASK1 in case of JNK3, were activated. These kinases are not activated by arrestin-dependent scaffolding and must come to the scaffold already active. Therefore, their activation might require G protein-mediated signaling. As c-Src and ERK1/2 usually transduce prosurvival and proproliferation signals, whereas JNK family kinases mediate antiproliferative, sometimes proapoptotic signaling, selective channeling of arrestin-mediated signaling in one direction, but not the other, appears a tempting way of telling the cell to live or die in its own language. Targeted enhancement of pro- and antisurvival signaling has therapeutic potential in degenerative disorders and cancer, respectively.

However, to attempt this we need to understand the molecular mechanisms involved. Originally the activation of both ERK1/2 and JNK3 was reported to be the result of arrestin binding to the receptor.^18,19 Subsequent experiments confirmed this in the case of ERK1/2—this kinase has fairly low affinity for free arrestins⁷⁵ and gets activated by phosphorylation only upon receptor activation.^76,77 It turned out that JNK3 can be activated by free arrestin-3,^75,77–79 and even by its receptor binding-deficient mutant^75,77 in cells. Experiments with purified arrestin-3 and kinases in vitro left no doubt that neither GPCRs nor any other proteins are necessary.^80,81 Since receptors were not required, it made sense to identify JNK3-binding elements in arrestin-3 that did not contain the GPCR-binding parts.⁸² Interestingly, one of these JNK3-binding peptides, containing the N-terminal residues 1–25, effectively facilitated JNK3 activation in vitro and in cells⁸³ (Fig. 4). This short peptide (smaller than 3 kDa) is so far the smallest known scaffold of any three-tier MAP kinase activation cascade. The smallest scaffold previously described was MP1 (~13.5 kDa), which appears to act in complex with its partner p14, which is not related by sequence, but very similar by 3D structure, making the whole scaffold ~27 kDa.⁸⁴ The other known MAP kinase scaffolds in yeast (Ste5p and Pbs2p) and mammals (KSR1, JIP1/2/3/4/5, and arrestin-2/3) range in size from 45 to 147 kDa (reviewed in⁸⁵). Interestingly, homologous arrestin-2-derived peptide, that differs from its arrestin-3-derived counterpart only in five positions (Fig. 4), does not facilitate JNK3 activation,⁸³ recapitulating the functional differences between parental full-length arrestin-2 and arrestin-3 proteins. The ability of the arrestin-3-derived mini-scaffold to suppress cell proliferation and/or support other JNK3-dependent biological functions needs to be tested. Mechanistically, its identification is an important breakthrough, as this is only the second example of a monofunctional element extracted from a multifunctional arrestin protein. The first one was the arrestin C-terminus, carrying both clathrin- and AP2-binding sites, that was shown to out-compete the receptor–arrestin complexes from coated pits, thereby specifically inhibiting arrestin-dependent GPCR internalization.⁸⁶ Two other arrestin-3-based tools were also developed: a V343T point mutant with severely impeded ability to activate JNK3⁷⁹ and a receptor binding-deficient multiple mutant that interacted with JNK3 but did not facilitate its activation.⁷⁷ The latter was even shown to act as a dominant-negative silent scaffold, suppressing arrestin-3-dependent JNK3 activation in cells.⁷⁷ As it was found to bind not only JNK3 but also upstream kinases ASK1 and MKK4, the most logical explanation of its dominant-negative action is that it recruits these kinases away from productive scaffolds, thereby suppressing signal transduction in the cascade. However, both of these are unlikely to be clean monofunctional tools, as they contain the whole arrestin-3 molecule, and therefore very likely retain many of its functions unrelated to the JNK signaling.

Fig. 4.

Mini-scaffold facilitating JNK3 activation. The structure of the arrestin-3-derived T1A peptide, encompassing residues 1–25 (only residues 5–25 are resolved in the crystal structure, PDB ID: 3P2D), are shown. This peptide facilitated JNK3 activation in vitro and in cells, apparently scaffolding ASK1-MKK4/7-JNK3 cascade.⁸³ The residues that differ from closely related arrestin-2, which does not promote JNK3 activation (Ser14, Pro15, Cys17), are shown as ball-and-stick models, with the atoms colored, as follows: gray, carbon; red, oxygen; blue, nitrogen; yellow, sulfur. The conformation shown is of the T1A peptide in the context of fully folded basal arrestin-3 (PDB ID: 3P2D³⁸), which is not necessarily the conformation of this peptide when it acts as a scaffold for the ASK1-MKK4/7-JNK3 cascade.

Other signaling-biased arrestin-based molecular tools were developed in which individual functions were disabled by mutagenesis. Arrestin-2-R307A mutant was shown to be defective in c-Raf1 binding,⁷⁶ whereas arrestin-2-D26A/D29A double mutant did not bind MEK1.⁸⁷ As both c-Raf1 and MEK1 are upstream kinases in the ERK1/2 activation cascade, these mutants failed to promote ERK1/2 activation. Several arrestin-3 mutants that do not bind phosphodiesterase but appeared to retain other functions, including GPCR binding, were also described.⁸⁸ Collectively, these data prove that individual functions of arrestins can be eliminated, yielding signaling bias. Relatively short peptides that effectively bind clathrin and AP2⁸⁶ or facilitate the activation of JNK3⁸³ show that, in some cases, monofunctional elements of arrestin proteins can be created and used for targeted manipulation of cellular signaling.

5. “DESIGNER ARRESTINS”: A ROADMAP OR A DREAM?

Chimera construction,^38,44,89 site-directed mutagenesis,^{45–47,50–52,90} peptide interaction studies,⁹¹ EPR,^21,92 NMR,⁹³ and X-ray crystallography^48,49,56,94 all suggest that the receptors engage the concave sides of both arrestin domains (Figs. 1 and 2). With few exceptions (such as calmodulin⁹⁵ and microtubules²¹), nonreceptor partners bind to the other side of the molecule or the C-terminus released upon receptor binding, although the binding sites were fully or partially localized, out of many dozens of proteins that bind arrestins,²⁰ only for clathrin,¹⁵ AP2,¹⁶ c-Raf1,⁷⁶ MEK1,⁸⁷ PDE,⁸⁸ and JNK3.^82,83 In general, any protein that binds the arrestin–receptor complex has to engage the elements that are not shielded by the receptor⁴¹ (Figs. 1 and 2). Experimental evidence shows that targeted mutations in distinct parts of arrestins can modulate receptor preference and the interactions with downstream signaling proteins. The next obvious step would be to combine these two to construct true “designer arrestins” that link the receptor we want to the signaling pathway of our choosing. Potential scientific and therapeutic utility of such mutants is obvious. Although there might be certain limitations of what can be constructed, due to apparent allosteric connections between the receptor-binding sites and elements where downstream signaling proteins dock,^59,60 only experimental work in this direction can reveal what is possible. Even a limited set of designer arrestins channeling the signaling from particular GPCRs to defined pathways would greatly enrich our toolbox used in research and medicine.

ABBREVIATIONS

ASK1

apoptosis signal-regulating kinase 1

EPR

electron paramagnetic resonance

ERK

extracellular signal-regulated kinase

GPCR

G protein-coupled receptor

JNK

c-Jun N-terminal kinase

MAPK

mitogen-activated protein kinase

NMR

nuclear magnetic resonance