Identification, purification, and characterization of GRK5, a member of the family of G protein-coupled receptor kinases.

A novel member of the family of G protein-coupled receptor kinases (GRKs), named GRK5, has been cloned from bovine taste epithelium. The cDNA sequence predicts a 590-amino acid protein with high overall similarity to rhodopsin kinase. GRK5 mRNA is found most abundantly in lung, heart, retina, and lingual epithelium, but is expressed very little in brain, liver, kidney, or testis. GRK5 expressed in Sf9 cells was purified to apparent homogeneity. GRK5 major autophosphorylation sites were mapped to Ser484 and Thr485. Purified GRK5 phosphorylates rhodopsin in a light-dependent manner and beta 2-adrenergic receptor in an agonist-dependent manner and phosphorylates the C-terminal tail regions of both receptor proteins. GRK5 possesses neither a CAAX motif specifying protein prenylation like rhodopsin kinase nor similarity to the G protein beta gamma-subunit binding domain of beta-adrenergic receptor kinases. GRK5 phosphorylation of rhodopsin or beta 2-adrenergic receptor is not stimulated by G protein beta gamma-subunits. The GRK5 protein does not undergo agonist-dependent translocation from cytosol to membranes as do beta-adrenergic receptor kinase and rhodopsin kinase, but rather appears to associate with membranes constitutively. GRK5 thus appears functionally similar to other characterized GRKs, but has distinct regulatory properties which may be important for its cellular function.

* This work was supported in part by National Institutes of Health in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

The nucleotide sequence(s) reported in this paper has been submitted to the GenBank-IEMBL Data Bank with accession numbeds) U01206.
these two authors contributed eauallv to this work and may be cited in either order.  1 and 2); GRK, G proteincoupled receptor kinase; G329-rhodopsin, rhodopsin treated with endoprotease h p -N to remove the C-terminal 19 amino acids; ROS, rod outer segments; bp, base paids); kb, kilobase(s); PCR, polymerase chain reaction; HPLC, high performance liquid chromatography; PAGE, polyacrylamide gel electrophoresis. with an arrestin protein, which prevents the further coupling of the receptor to G proteins, and thus results in a desensitized state (4, 5).
The G protein-coupled Receptor Kinases (GRKs) which mediate tEs activation-dependent phosphorylation of receptors have only recently been characterized at the molecular level (6). Two widely distributed "0-adrenergic receptor kinase" isoforms, PARK1 and flARK2, have been cloned and shown to phosphorylate P z A R (7,8). Rhodopsin kinase has recently been cloned from the retina and shown to phosphorylate rhodopsin (9). However, neither the PAR& nor rhodopsin kinase exhibits specificity limited to the receptors for which they were named, but can phosphorylate diverse G protein-coupled receptor types (10)(11)(12)(13), although the extent of this diversity is not yet clear.
Quite recently, three additional rhodopsin kinase-related sequences have been reported. One, named IT11 (or GRK41, was cloned positionally due to proximity to the Huntington's disease locus (14). The GRK4 message is abundant mainly in testis, but the functional properties and receptor specificity of this putative kinase are not yet known (14). The GRK5 and GRK6 kinase cDNAs were recently cloned from human heart, and the expressed enzymes were shown to phosphorylate rhodopsin in a light-dependent manner (15, 16).
GRKs mediate homologous desensitization of receptors in many tissues, although particular GRKs appear to predominate in specific tissues. Rhodopsin kinase plays a role in inactivation of visual stimulation (17). Similarly, PARK2 has recently been implicated in receptor desensitization in the olfactory system (18,19). As part of an effort to characterize G protein-coupled receptor kinases which may play a role in adaptation of taste responses, we screened taste tissue to determine which receptor kinases are present and identified an unknown rhodopsin kinase-like sequence (which has recently been cloned from human heart (15)). We report here on the cDNA sequence, tissue distribution, purification, and functional properties of this G protein-coupled receptor kinase, GRK5.

EXPERIMENTAL PROCEDURES
Polymerase Chain Reaction-Oligonucleotide primers were synthesized corresponding to two absolutely conserved stretches of G proteincoupled receptor kinase catalytic domains (7-9) which contain residues not found in most other protein kinases: G(EVK)GGFGE (5"CACCG-GCTCGAGGGIMGIGGIGGI"YGG1GA) and G(YIFMAF"APV (5'-GC-CCl"l'CTCGAGGACYTCIGGIGCCATRWAICC).2 First strand cDNA was synthesized from 1 pg of poly(A) RNA using oligo(dT) primer and Superscript Moloney murine leukemia virus reverse transcriptase (Life Technologies, Inc.) (20). cDNA from 10 ng of bovine circumvallate papillae poly(A) RNA was amplified using 500 n~ primers, 200 p concentration of each dNTP, 1.5 m~ MgCl,, and 2.5 units of Tbq DNA polymerase, and 1 x buffer (Promega) for 35 cycles at 95 "C for 1 min, 55 "C for 1 min, and 72 "C for 3 min (20). The 500-base pair product was gel-purified and subcloned into pBS I1 (Stratagene) usingXho1 sites in Nucleotide codes are: I = inosine, M = A or C, Y = A or T, R = A or G, and W =Aor T. the primers. Clones were classified using unique restriction sites and DNA sequencing.
cDNA Cloning"Oligo(dT)-and random-primed cDNA libraries were prepared in A g t l O using the Superscript cDNA kit (Life Technologies, Inc.), and screened by hybridization to the 500-bp PCR fragment using standard techniques (20). Individual virus inserts were subcloned as EcoRI fragments into pBS I1 (Stratagene) and sequenced on both strands by chain termination using Sequenase T7 DNA polymerase (U. S. Biochemicals). Clones 7.1 and 19.1 (Fig. IA) were used to create the full-length cDNA by ligation at the unique NcoI site at base 860. Sequence analysis was performed using the Geneworks software (Intelligenetics) and the GCG GAP program for sequence similarity (21).
Expression in COS Cells and SP Cells-Prior to functional assays, the 5'-and 3"untranslated regions of the full-length cDNA were removed by cassette PCR, utilizing the unique PstI (base 333) and KpnI (base 1765) restriction sites. An EcoRI site, BglII site, and consensus ribosome binding sequence (GCCGCCACC) were inserted immediately prior to the start ATG codon for the new 5'-cassette, and an XbaI site was inserted immediately after the stop codon for the new 3' cassette. Individual 5' and 3' end cassettes were amplified with Taq DNA polymerase (20) using the full-length cDNA as template. Purified products were digested with EcoRIIPstI (5') or with KpnIIXbaI (3') and ligated sequentially (3' first) with the PstIor KpnI-digested central fragment of the cDNA. The new 5' and 3' ends were sequenced to verify the amplified sequences. The resulting EcoRIIXbaI fragment was inserted into pcDNAI (Invitrogen) and used to transfect COS-7 cells by the DEAE-dextran method (23). Cells were harvested 48 h after transfection, along with control cells transfected with pcDNAI vector only, and cell membranes and supernatants were prepared as described previously for rhodopsin kinase (23).
The full-length BglIIIXbaI fragment of GRK5 was subcloned into the pVL1392 baculovirus shuttle vector, and co-transfected with a mutant baculovirus carrying a lethal deletion into Sf9 cells following the manufacturer's protocol (BaculoGold kit, Pharmingen). A single recombinant baculovirus plaque was amplified to produce a viral stock.
Purification of GRK5 from S p CellsSf9 cells (1 x 106/ml) were infected at a multiplicity of infection of 0.1 with the recombinant GRK5 baculovirus and grown for 2 days. Cells were harvested by centrifugation, and the cell pellet rinsed with phosphate-buffered saline. The cell pellet from 2 liters of culture was resuspended in 50 ml of 20 m HEPES, 2 m~ EDTA, pH 7.2 (buffer A) supplemented with a mixture of protease inhibitors (2 pg/ml aprotinin, 10 pg/ml benzamidine, 10 pg/ml leupeptin, 1 p g / d pepstatin A, 20 pg/ml phenylmethylsulfonyl fluoride, 10 pg/ml soybean trypsin inhibitor) and stored at -70 "C. Cell pellets from 2-liter cultures were thawed, supplemented with fresh protease inhibitors, and homogenized with 10 strokes of a tight glass Dounce homogenizer on ice. All subsequent manipulations were performed at 4 "C, and all buffers contained protease inhibitors as above. The homogenate was spun at 43,000 x g for 20 min. The resulting pellet was rehomogenized with 50 ml of buffer A with 20 m~ NaCl and spun as before. The two supernatants were pooled and passed through a 10-ml column of S-Sepharose (Pharmacia) at a flow rate of 1 d m i n . Most proteins failed to bind to the resin. The column was washed with 50 ml of buffer A with 20 m~ NaCl and eluted with a linear 100-ml gradient of 20-750 m~ NaCl in buffer A. Fractions of 1 ml were collected and assayed for rhodopsin phosphorylating activity in the light. Fractions at approximately 50C600 m~ NaCl containing activity and the fewest protein contaminants were pooled (20 ml), diluted with buffer A to below 100 m~ NaC1, and applied to a 10-ml column of heparin-Sepharose (Pharmacia) at 1 ml/min. The column was washed with 50 ml of buffer A with 150 rn NaCl and eluted with a linear 100-ml gradient of 15&1500 m NaCl in buffer A with 0.02% Triton X-100. Fractions of Preparation of GSTIGRK5 C-terminal Fusion Protein and Antisera-The C-terminal fragment of GRK5 was expressed in Escherichia coli as a fusion protein with glutathione S-transferase. Briefly, oligonucleotide primers corresponding to bases 1561-1578 (encoding PPFWD) and the inverse complement of bases 1930-1947 (encoding STGSS[stop]) were used to amplify the C-terminal fragment from the GRK5 cDNA by PCR as described above. The amplified fragment was inserted in-frame into the BamHI and EcoRI sites of pGEX-2T vector (Pharmacia) using restriction sites built into the 5' ends of the primers. The resulting pGEX-GRK5CT plasmid was grown in NM522 strain E.
coli. Bacteria were treated with isopropyl-P-D-thiogalactopyranoside to induce expression of the glutathione S-transferasdGRK5 fusion protein, which was purified using glutathione-Sepharose (Pharmacia) as described for glutathione S-transferasdPARK C-terminal fusion proteins (23). The purified glutathione S-transferasdGRK5 fusion protein was used to immunize rabbits as previously described for PARK isozymes (24).
Autophosphorylation Assays-Purified enzymes were assayed for autophosphorylation by incubation in a 25-pl volume with 100 1.1~ [-y-32PlATP (500-2000 dpm/pmol) in the presence of assay buffer (20 m~ Tris-HC1, pH 7.5, 2 m EDTA, 10 m~ MgC12, 1 m~ dithiothreitol, 10 pg/ml protein kinase A inhibitor-(1424)-amiide peptide) for 15 min at 24 "C. Assays were terminated by addition of Laemmli sample buffer and separated on SDS-polyacrylamide gels. Gels were fixed, stained with Coomassie Blue, and dried for autoradiography with x-ray film. Incorporated disintegrationdmin were quantified by Cerenkov counting of protein bands cut from the dried gel.
For determination of autophosphorylation sites in GRK5, 3 nmol of purified enzyme was incubated with 100 p~ ATP (10 dpdfmol) as described, but the reaction was terminated by passage through a Bio-Spin-30 column (Bio-Rad) equilibrated with 20 m "is, pH 7.5, to remove free ATP. The eluate was added to 2 pg of sequencing grade endoproteinase Asp-N (Boehringer Mannheim) and digested overnight at room temperature. The resulting peptides were separated by sequential reverse-phase HPLC using an Aquapore OD-300 column eluted with a linear gradient of 0.1% trifluoroacetic acid (A) to 0.095% trifluoroacetic acid in acetonitrile (B). Peptides were detected by absorbance at 210,220,260, and 280 nm and radioactivity by Cerenkov counting of fractions. Peptides were sequenced using a gas-phase amino acid sequencer (Applied Biosystems model 477A). Phosphoamino acid analysis was performed on acid-hydrolyzed peptides by TLC on cellulose using isobutryric acid, 0.5 M ammonium hydroxide ( 5 3 v/v) (25), and autoradiography of the TLC plate.
Receptor Phosphorylation Assays-For receptor phosphorylation, enzyme was incubated as in autophosphorylation assays, except that incubations also contained urea-washed rod outer segment membranes containing rhodopsin or phospholipid vesicles reconstituted with p,adrenergic receptor. For rhodopsin, assays were incubated at 24 "C in the dark or light as indicated. For &-adrenergic receptor, assays were incubated at 30 "C and contained freshly made 100 p~ (-)-isoproterenol or (=)-propranolol as indicated. Rod outer segment membranes were prepared from dark-adapted bovine retinas and stripped with urea as described (26). Washed ROS membranes were treated with endoprokinase Asp-N overnight at 24 "C to produce truncated G329-rhodopsin (27). Recombinant human @,-adrenergic receptors were purified from membranes of baculovirus-infected Sf9 cells and reconstituted into phospholipid vesicles as previously described (28). G protein p-y-subunits were purified from bovine brain cortex (29) and transducin pysubunits from bovine retina (28).
Danslocation-Phosphorylation Assay-The ability of GRK5 to bind to ROS was determined as described (26) by preincubation of 5 pmol of GRK5 enzyme with 25 pmol of rhodopsin in ROS membranes in translocation assay buffer (50 m phosphate, pH 7.4, 100 m~ NaCI, 1 m MgC12, 5 m dithiothreitol, 10 pg/ml protein kinase A inhibitor-( 14-24)amide peptide, and the mixture of protease inhibitors, but not containing ATP). Incubations were performed at 24 "C for 5 min in the dark or light, and reactions were pelleted at 100,000 x g. The supernatant fractions were removed and supplemented with 25 pmol of rhodopsin, while pelleted membranes were resuspended in the original volume of assay buffer (without ATP). Reactions were incubated in the light for 5 min prior to addition of [y3,P1ATP, and the incubations continued for 10 min in the light. Reactions were stopped with SDS-PAGE sample buffer and separated on 12% gels to determine rhodopsin phosphorylation.
Peptide Phosphorylation Assay-The PARK and rhodopsin kinase substrate peptide RRREEEEESAAA (30) was incubated in a 100-pl volume with autophosphorylation assay buffer, containing 500 pg/ml bovine serum albumin, for 15 min at 30 "C. Assays were termi-nated by spotting triplicate 25-pl aliquots onto 2 x 2-cm P81 phosphocellulose  filter squares (Whatman) and washing 5 times for 20 min in 75 m~ phosphoric acid. Radioactivity of filters was determined by Cerenkov counting, and nonspecific counts were determined by the absence of substrate peptide. Data Analysis-Experimental procedures were performed at least three times unless otherwise indicated. Curve-fitting was performed using the SigmaPlot program using single experiments, and parameters from replicate experiments were averaged.

RESULTS
Degenerate primers corresponding to kinase catalytic domain sequences conserved only among known members of the G protein-coupled receptor kinase family were used to amplify a 500-bp product from bovine circumvallate papillae cDNA by PCR. Of 24 clones analyzed, 15 were PARK1 (GRK2) and 1 was PARK2 (GRK3). No rhodopsin kinase (GRK1) clones were observed, but 8 clones encoding a novel rhodopsin kinase-like sequence were obtained. A screen of 2 x lo6 recombinant viruses of a n unamplified circumvallate cDNA library using this 500-bp fragment as a probe led to the identification of 5 independent clones (Fig. lA).
The 2.5-kb sequence contains a 1770-bp open reading frame preceded by 174 bp of 5'-untranslated sequence and followed by 592 bp of 3'-untranslated sequence ending in a short poly(A) tract (Fig. 1B). The assignment of the first ATG codon as the initiator methionine is based on the high degree of similarity of the deduced amino acid sequence of the N-terminal domain to those of other receptor kinases and the presence of an upstream, in-frame stop codon. The open reading frame encodes a 590-amino acid protein with a predicted molecular mass of 68 kDa. The central protein kinase catalytic domain contains several features characteristic of the G protein-coupled receptor kinase subfamily, including the sequence DLG in kinase subdomain VI1 (9, 31), and is highly similar to the catalytic domains of rhodopsin kinase (57% identity, 78% similarity) and the recently identified GRK4 or I T l l kinase (79% identity, 91% similarity) (9,14). The 138-amino acid C-terminal domain is short like those of rhodopsin kinase and GRK4 kinase: but contains neither a CAAX motif specifying protein prenylation like rhodopsin kinase (32) nor any similarity to the G protein Py-subunit binding region identified in PARK1 and PARK2 (23).
Comparison of the GRK5 deduced protein sequence with those of the previously identified mammalian G proteincoupled receptor kinases indicates that this sequence is the bovine homolog of the recently identified human GRK5 receptor kinase (99% similarity) (15). GRK5 is closely related to the GRK4 kinase (80.7% similarit?) (14) and to rhodopsin kinase (68.4% similarity) (91, but shares only 58.6% similarity with PARK1 and PARK2 (7,8). An additional receptor kinase recently identified in human neutrophils and human heart, termed GRKG (16, 331, is also highly similar to the GRK5 sequence (83.6% similarity). These six sequences can thus be grouped into three distinct subfamilies: the GRK5 receptor kinase, together with the GRK4 and GRKG kinases, form one subfamily, while PARK1 and PARK2 form a second subfamily.
Rhodopsin kinase is more closely related to the GRK4/GRK5/ GRKG subfamily than to the two PARKS, but in light of its limited tissue distribution as well as its distinct mechanism of regulation and intracellular localization involving C-terminal farnesylation and carboxymethylation, rhodopsin kinase is best considered as a distinct subfamily. Of the known nonmammalian receptor kinases, the Drosophila GPRK2 sequence4 is a member of the GRK4/GRKS/GRK6 subfamily and shares 79.6% similarity with the GRK5 sequence, while the Drosophila GPRKl sequence (57.2% similarity) encodes a member of the PARK subfamily (34).
The GRK5 receptor kinase PCR fragment was obtained as part of an effort to identify taste-specific G protein-coupled receptor kinases. To ascertain whether GRK5 might represent such an enzyme, its mRNA distribution was determined by Northern blotting (Fig. 2). Equivalently high mRNAlevels were observed in both taste (circumvallate and fungiform papillae) and adjacent non-taste lingual epithelial poly(A) RNAs, indicating that expression of the GRK5 receptor kinase is not unique to taste tissues nor apparently enriched there. Highest expression was evident in lung, heart, and retina, while little or no expression was detected in brain cortex, kidney, liver, and The I T l l (GRK4) cDNA clone (14) contains two regions of alternative splicing leading to four potential forms of the protein, with the longest sequence containing a 46 codon insertion in the N-terminal region after codon 17, and a 32 codon insertion in the C-terminal region after codon 483 of the reported sequence (R. T. Premont, unpublished observation). Replacement of these additional sequences into the reported sequence results in a more complete alignment of GRK4 to GRKB and GRK6, and similarity to GRK5 was computed using this longest form of the GRK4 kinase.
The "5' untranslated region" of the Drosophila GPRK2 sequence (34) has significant nucleotide and amino acid similarity to the GRK5 sequence in equivalent positions (with the addition of a single nucleotide), indicating that the GPRK2 clone may not be complete (R. T. Premont and W. J. Koch, unpublished observation). However, since the sequence remains incomplete, similarity to the GRK5 kinase was computed only from the reported start methionine. testis. Two messages, a major band of 2.8 kb and a minor species of 7 kb, were observed in all cases, perhaps due to the use of alternative polyadenylation signals. This distribution is in agreement with the human GRK5 mRNAdistribution, which was also found to be high in skeletal muscle and placenta (15).
To examine the functional properties of the GRK5 kinase, the cDNA was expressed in Sf9 cells from the strong baculovirus polyhedrin promoter. Cell supernatants from infected cells exhibited a markedly increased ability to phosphorylate rhodopsin in rod outer segment membranes in the presence but not absence of light (data not shown; 15). Therefore, the expressed GRK5 enzyme was subjected to protein purification according to protocols devised for the purification of PARK enzymes. Sf9 cell cytosolic extracts were sequentially chromatographed on S-Sepharose and on heparin-Sepharose to yield a 400-fold enrichment of the GRK5 kinase. This preparation is substantially (>80%) pure as assessed by Coomassie Blue staining of SDS gels. One pg each of purified preparations of recombinant bovine GRK5, rhodopsin kinase, and PARK1 purified from baculovirus-infected Sf9 cells, separated by SDS-PAGE and stained with Coomassie Blue, are shown in Fig. 3A.
To verify that this purified protein represents the GRK5 kinase, a gel containing 100 ng of each of these three GRK proteins was immunoblotted using an anti-GRK5 antiserum (Fig. 3B). This polyclonal serum was raised against a glutathione S-transferase fusion protein containing the C-terminal 128 residues of GRK5. Based on sequence similarities, this serum is predicted to also recognize GRK4 and GRKG proteins, but not rhodopsin kinase or the two PARKS. The antibodies readily detect the purified recombinant GRK5 protein, but fail to recognize equivalent amounts of purified PARK1 or rhodopsin kinase proteins.
The ability of the purified GRK5 enzyme to phosphorylate rhodopsin in rod outer segment membranes was characterized. Phosphorylation reactions were incubated in the light for various times to determine the linearity of the enzymatic activity. Rhodopsin phosphorylation by GRK5 appears linear over 20 min (Fig. 4A), while kinase autophosphorylation is maximal within 5 min (not shown). No phosphorylation of rhodopsin occurs when incubations are performed in the dark, where rhodopsin is in the inactive state. This receptor activation dependence of kinase activity toward receptor substrates is a hallmark of the G protein-coupled receptor kinases. Incubation of the rod outer segment preparation without added GRK5 leads to virtually no rhodopsin phosphorylation, indicating the absence of endogenous rhodopsin kinase or other GRKs in the membrane preparations (not shown).
The ability of the GRK5 kinase to interact with light-activated rhodopsin was measured by determining the K , of GRK5 for rhodopsin phosphorylation. Various concentrations of rod outer segment membranes were incubated with purified GRK5 in the light, and rhodopsin phosphorylation was quantified Both PARK1 and rhodopsin kinase have been shown to be inhibited by polyanions such as heparin (40,41). The ability of heparin to inhibit GRK5 kinase phosphorylation of rhodopsin was assessed (Fig. 4C). Heparin inhibited the light-dependent phosphorylation of rhodopsin by GRK5 with an IC50 of 46 * 10 &mI. This is intermediate between the heparin IC50 for PARK1 (770 ng/ml) and for rhodopsin kinase (1 mg/ml) (40,41).
The PARK1 and PARK2 enzymes have been shown to be activated by heterotrimeric G protein fly-subunits, due to Pybinding to a domain in the C-terminal tail of the kinase and concomitant localization of the kinase to the membrane (28,42). Rhodopsin kinase, which lacks this Py-binding domain, is not activated by fly-subunits (26). Purified GRK5 and PARK1 were assayed for their ability to phosphorylate rhodopsin in the absence or presence of flysubunits purified from bovine brain or from bovine retina (Fig. 5). Rhodopsin phosphorylation by  (-) or presence (+) of 100 I~M brain G protein (Goy) or retinal (T@J py-subunits as indicated. Reactions were separated by SDS-PAGE on a 10% gel and dried for autoradiography. Phosphorylated rhodopsin bands are shown. PARK1 is markedly enhanced by added brain py-subunits and less stimulated by transducin py-subunits as previously demonstrated (281, but rhodopsin phosphorylation by GRK5 is not significantly altered by either form of py-subunits. In our studies, GRK5 and PARK1 have nearly an equivalent ability to phosphorylate rhodopsin in the absence of added py-subunits (60-70 nmol of phosphate/min/mg of protein), but PARK1 plus brain fly-subunits has approximately 10-fold higher activity than GRK5.
The ability of the GRK5 kinase to phosphorylate the G,coupled &-adrenergic receptor was also assessed, using recombinant human P2AR affinity-purified from Sf9 cell membranes and reconstituted into phospholipid vesicles (Fig. 6). GRK5 phosphorylates the P2AR in an isoproterenol-dependent manner, indicating that GRK5 recognizes the agonist-activated form of the PzAR as substrate. Further, addition of G protein py-subunits has no stimulatory effect on GRK5 phosphorylation of the &AR. Interestingly, GRK5 has an approximately 7-fold higher ability to phosphorylate PzAR than does PARK1 in the absence of added py-subunits, while PARK1 plus brain py-subunits is approximately 2-fold better than GRK5.
Incubation of rhodopsin kinase with Mg-ATP leads to the rapid and extensive autophosphorylation of the enzyme at Ser488 and Thr489 (351, while the PARK enzymes are autophosphorylated only to low stoichiometry (38) and do not contain phosphorylatable serine or threonine residues in these cognate positions (7,8). The GRK5 sequence does share Ser484/Thr4"5 in the cognate positions with rhodopsin kinase, and GRK5 extensively autophosphorylates (see Figs. 6 or 8A). To define the sites of GRK5 autophosphorylation, 3 nmol of purified enzyme was incubated with [y3'P]ATP for 15 min, and the autophosphorylated protein was digested with endoproteinase Asp-N. Peptide fragments were separated by HPLC, yielding one major radioactive peak (67% of counts) and two minor peaks (20% and 12% of counts) (Fig. 7A). The most radioactive fraction (No. 50) was subjected to a second HPLC step with a more shallow gradient which resolved a radioactive peptide peak containing 45% of the applied counts and a broad peak of radioactivity apparently containing several peptides (Fig. 7B). This major peak (fraction 77) was subjected to amino acid sequencing and was determined to correspond to the predicted proteolytic fragment of residues 479 to 491, DIEQFSTVKGVNL. Phosphoamino acid analysis of this peptide fraction indicated that under these hydrolysis conditions, most of the 32P label was associated with phosphothreonine (70%) and only 30% with phosphoserine. The less radioactive peak (fraction 80) had a dried for autoradiography and Cerenkov counting.  as shown. B , radioactivity in HPLC eluate fractions above was monitored by Cerenkov counting. Radioactivity is plotted for each fraction. C, the major peak of radioactivity (fraction 50) was subjected to a second HPLC separation using a more shallow gradient 1%B/10 min, with the elution profile a t 220 nm for I-min fractions (175 p1) as shown in the smooth trace and radioactivity as shown in the hatched burs. Abscissa marks indicate the beginning and end of individual fractions. Peptides of fraction 77 also subjected to acid hydrolysis and TLC phosphoamino recovered from fractions 77 and 80 were microsequenced, and an aliquot acid analysis as described under "Experimental Procedures." mixed sequence which included the serine-rich C-terminal55residue fragment. Therefore, the major autophosphorylation sites on GRK5, Ser484flhr4a5, are conserved with those of rhodopsin kinase. Additional minor sites remain to be defined.
Phosphorylation of rhodopsin by rhodopsin kinase occurs on serine and threonine residues in the C-terminal tail of the protein (431, and enzymatic removal of the last 19 amino acids of the tail using endoproteinase Asp-N eliminates these phosphorylation sites from the remaining truncated rhodopsin fragment (27). Rod outer segment membranes were digested over-night with endoprotease Asp-N to produce truncated G329rhodopsin by removal of the last 19 amino acids (27). This results in the increased mobility of the truncated rhodopsin in SDS gels. Incubation of this truncated rhodopsin with GRK5 kinase in the light does not result in any phosphorylation of the G"29-rhodopsin protein (Fig. 8A). The ability of GRK5 to interact with the truncated G329-rhodopsin appears unimpaired as assessed by augmentation of peptide phosphorylation activity (see below). Alternatively, if rhodopsin previously phosphorylated by GRK5 (or rhodopsin kinase or PARK11 is then digested with endoproteinase Asp-N, the radioactivity is lost from the rhodopsin band (but not entirely lost, due to incomplete digestion) but fails to appear in the G329-rhodopsin fragment (Fig. 8B). Therefore, phosphorylation of rhodopsin by GRK5, PARK1, or rhodopsin kinase occurs predominantly if not entirely within the C-terminal 19 amino acids of the opsin protein.
Phosphorylation of the P z A R by PARK1 occurs on the Cterminal tail of the receptor, since enzymatic removal of the tail (up to membrane span 7) using carboxypeptidase Y eliminates phosphorylation (441, while mutagenesis of the tail region to remove potential phosphorylation sites reduces the ability of the receptor to undergo homologous desensitization (45). If P2AR previously phosphorylated by GRK5 (or rhodopsin kinase or PARK1) is then digested with carboxypeptidase Y, the radioactivity is lost from the remaining receptor fragment (Fig. 8C), indicating that phosphorylation by GRK5, PARK1, or rhodopsin kinase occurs predominantly or entirely on the C-terminal tail of the &-adrenergic receptor. Thus, for two distinct receptor proteins, each of these three kinases appears to phosphorylate the same or overlapping sets of serine and threonine residues.
The ability of both PARK1 and rhodopsin kinase to phosphorylate a series of synthetic peptide substrates has been characterized (27,30,46). The peptide RRREEEEESAAA has been shown to be substrate for both PARK1 and rhodopsin kinase (30,46). Phosphorylation of this peptide by GRK5 was linear only over 20-25 min (not shown), in contrast to PARK1 and rhodopsin kinase which give linear activity over at least 2 h (30). The GRK5 enzyme was incubated for 15 min with various concentrations of the RRREEEEESAAA peptide to characterize its phosphorylation (Fig. 9A). GRK5 was determined to have a K, of 1.4 2 0.5 mM for this peptide with a calculated V,,, of 3 nmol of phosphatelmidmg of protein for this enzyme preparation. These values are intermediate between those previously determined for PARK1 and rhodopsin kinase, with K , values of 1 mM and 2 mM, respectively (30).
Peptide substrates for G protein-coupled receptor kinases are generally quite poor (having about 1000-fold lower K,,, than activated receptor). However, the peptide phosphorylation activity of PARK1 and rhodopsin kinase can be augmented by the presence of activated (but not inactive) receptor in the assay (27,461, since binding of the kinase to an activated receptor appears to stimulate the kinase catalytic activity. Assays containing 1 mM RRREEEEESAAA peptide were incubated in the presence of rhodopsin in the light or in the dark to determine whether binding to rhodopsin stimulates the peptide phosphoryltransferase activity of the GRK5 enzyme (Fig. 9B). Rod outer segment membranes containing native rhodopsin as well as ROS membranes containing endoprotease Asp-N-treated rhodopsin were tested. Compared to the absence of rhodopsin, addition of rhodopsin or truncated G329-rhodopsin stimulated the phosphorylation of the peptide 2-3-fold when incubations were performed in the light. In the dark, addition of either rhodopsin preparation had no effect. Thus, like rhodopsin kinase and PARK, the GRK5 enzyme appears to undergo an activation step upon binding to an activated receptor, which is gel and dried for autoradiography. Standards shown are the same as in Fig. 3. B, endoproteinase Asp-N digestion of GRK-phosphorylated rhodopsin. Purified GRK5, PARK1, or rhodopsin kinase (20 pmol) was incubated with 100 pmol of rhodopsin for 10 min a t 24 "C in the light in a 100-p1 reaction. Phosphorylated membranes were pelleted by centrifugation at 100,000 x g, washed with 20 mM Tris, pH 7.5, and resuspended in 60 pl of the same buffer. An aliquot of 25 pl was incubated with 1 pl of 80 pg/ml endoproteinase Asp-N for 2 h a t 24 "C, while another 25 pl was untreated. Reactions were separated by SDS-PAGE on a 12% gel and dried for autoradiography. Different exposure times were required to equalize the apparent signals of the three kinases, and this experiment was repeated twice. C, carboxypeptidase Y digestion of GRK-phosphorylated &AR. Purified GRK5, PARK1, or rhodopsin kinase (20 pmol) was incubated with 20 pmol of reconstituted PzAR in the presence of 100 PM isoproterenol for 10 min a t 24 "C in a 100-pl reaction. Phosphorylated vesicles were pelleted by centrifugation a t 100,000 x g, washed with 50 mM phosphate, pH 6.5, and resuspended in 60 pl of the same buffer. An aliquot of 25 p1 was incubated with 1 pl of 2500 unitdml carboxypeptidase Y for 2 h at 24 "C, while another 25 pl was untreated.
Reactions were separated by SDS-PAGE on a 12% gel and dried for autoradiography. Different exposure times were required to equalize the apparent signals of the three kinases, and this experiment was repeated twice. Assays were terminated by spotting triplicate 25-pl aliquots onto Whatman P81 filters and immediately washing in phosphoric acid. Specific phosphorylation was calculated as the difference with a reaction containing no peptide. Kinetic parameters were derived from double-reciprocal plots using the Sigma-plot program. B, effect of rhodopsin and G"2"-rhodopsin on peptide phosphorylation activity of GRK5. Purified GRK5 (1 pmol) was incubated with 1 m~ RRREEEEE-SAAA peptide in the absence or presence of 25 pmol of rhodopsin or G""9-rhodopsin for 15 min a t 30 "C in the light or dark, as indicated. This experiment was repeated twice. independent of the ability of the receptor to be a productive kinase substrate.
Previously characterized GRK enzymes undergo activated receptor-dependent "translocation" from cytosolic fractions to membranes. That is, in the presence of inactive receptors, rhodopsin kinase and PARK1 appear predominantly soluble, but, upon addition of agonist to P2AR or light to rhodopsin, the kinases appear associated with the membrane (1, 6). Membrane association of rhodopsin kinase requires the post-translational farnesylation of the kinase (26), while membrane association of the PARKS requires the additional presence of G protein &-subunits in the membrane (26,28). Since GRK5 appears to neither bind py-subunits nor to have sequences directing post-translational protein prenylation, it was of interest to determine whether GRK5 also undergoes an activated receptor-dependent translocation from cytosol to membranes. GRK5 enzyme was added to ROS membranes in the dark or light for 5 min, and reactions were centrifuged to separate soluble from membrane fractions. The partitioning of GRK5 between these two fractions was assessed by subsequent rhodopsin phosphorylation assays (Fig. 10). In marked contrast to rhodopsin kinase and PARK1, the majority of GRK5 activity appears membrane-associated even in the absence of rhodopsin activation, and light activation fails to influence the amount of kinase associated with the ROS membranes. Interestingly, COS cells contain a native GRK5-like immunoreactive band which appears in the membranes but not the cytosol, but overexpression of GRKS leads to significant accumulation of the enzyme in the cytosolic fraction (data not shown). Thus, the native GRK5 protein may be constitutively associated with the membrane and perhaps with receptors.

DISCUSSION
Similarities to Other GRKs-The GRK5 kinase, like other characterized G protein-coupled receptor kinases, is capable of phosphorylating rhodopsin and &AR, but only when the receptors are in the (light or hormone) activated state. Activated rhodopsin (or a non-substrate truncated rhodopsin) is also capable of stimulating the catalytic activity of GRK5 toward peptide substrate, presumably by binding to noncatalytic sites of the enzyme, as has been demonstrated for both rhodopsin kinase (27) and PARK1 (46). Removal of the C-terminal of rhodopsin with endoproteinase Asp-N or of the P2AR with carboxypeptidase Y eliminates GRK5 phosphorylation sites on these receptors. Whether the GRK5, rhodopsin kinase, and PARK1 sites on the C termini of these two receptor proteins are distinct, overlapping, or identical is not yet known. The similarity of the substrate sites on these two proteins, however, indicates that these three GRK enzymes may recognize the same activated conformational determinants in the receptors and may all serve equivalent cellular roles in marking the receptors for arrestin proteins.
Like rhodopsin kinase, GRK5 is extensively autophosphorylated. This autophosphorylation is quite rapid and appears unaffected by the presence of light-activated rhodopsin (Fig.   8 A ) or by agonist activation of the P2AR (Fig. 6, upper band 1. Ser4R4 and Thr4R5 in GRK5 are the cognate residues of the major autophosphorylation sites identified on rhodopsin kinase (27) and also appear to be the major autophosphorylation sites in GRK5. The minor rhodopsin kinase autophosphorylation site a t Se? is not conserved in the GRK5 sequence. Additional radioactive peptide fragments of GRK5, which have not yet been further characterized, may represent minor autophosphorylation sites. Autophosphorylation of rhodopsin kinase has been reported to play a functional role in regulating kinase interaction with receptors (47); whether GRK5 autophosphorylation plays a similar role is not yet known.
The relative activity of GRK5 and PARK1 against rhodopsin and pzAR have been compared. If PARK1 alone phosphorylates rhodopsin with a relative activity of 1, addition of G protein Py-subunits increases the activity to 10-20; GRK5 phosphorylates rhodopsin with a relative activity of 1. Similarly, if PARK1 alone phosphorylates P2AR with a relative activity of 1, addition of G protein By-subunits increases this activity to 10-20; GRK5 phosphorylates &AR with a relative activity of 7-8.
Thus, relative to PARK1 plus Pysubunits, GRK5 is better able to phosphorylate the P2AR than rhodopsin. The lower apparent activity of GRK5 toward either substrate compared to fully (By) activated PARK1 may be intrinsic or may be due to the lack of some ancillary factor for GRK5 (such as By-subunits for PARK enzymes).
The GRK5 mRNA appears to have an extensive somatic distribution. The mRNA encoding the related GRKG kinase is even more widely distributed (16, 33), while the GRK4 mRNA is found primarily in testes (14). Thus, various G protein-coupled receptors are expected to be natural substrates for the GRK5 enzyme and other members of this GRK subfamily. The range of potential receptor substrates has not been determined for any of the receptor kinases, but the availability of diverse cloned receptors and kinases should facilitate this mapping process. Conversely, the involvement of PARK enzymes in hormonal desensitization has been assessed using heparin as a "specificD PARK inhibitor in permeabilized cells (481, but these concentrations of heparin would also be expected to inhibit GRK5 activity. More recent work using PARK subtype-specific antibodies to inhibit the enzymes has shed light on the role of P A R K 2 in olfactory desensitization (18,19). A similar approach may be useful to elucidate the role of GRK5 in hormonal desensitization in various tissues, such as heart. Two possibilities exist for GRK specificity: either there are distinct differences in receptor substrate specificity among GRKs or GRK subfamilies or all GRKs are essentially functionally equivalent. Clearly, G W can phosphorylate the & A R , as can GRKG (16). Both enzymes then qualify to be called "PARKS." PARK1 activated by Py-subunits (presumably the native circumstance) is better able to phosphorylate the BAR than is GRK5, but not by a very great extent. Given the divergence between these two kinases, it is perhaps surprising that differences in substrate recognition domains do not produce a larger difference in phosphorylating activity toward a given receptor substrate. However, only a miniscule fraction of the known G protein-coupled receptors have been tested as GRK substrates, and distinct differences may well exist. Nevertheless, from the initial receptor phosphorylation site mapping experiments described here, GRK5 appears to phosphorylate essentially the same sites on rhodopsin and the P z A R as do PARK1 and rhodopsin kinase and would thus be expected to lead to the same desensitizing effect on receptor coupling to G proteins.
Differences from Other GRKs-The GRK5 enzyme, however, appears to have some aspects to its regulation which are distinctly different from those of previously characterized GRKs.
Rhodopsin kinase and both PARK isozymes have been shown to be cytosolic enzymes which translocate to the membrane in response to receptor activation (by light or by hormone) (1,23). This translocation is associated with stimulation of kinase activity, presumably by making the activated receptor substrate available to the enzyme without altering the catalytic activity of the kinase (28). Translocation of rhodopsin kinase requires the C15 farnesyl modification of the enzyme at its C-terminal CAAX motif (26), while translocation of both PARK isozymes is dependent on interaction of a C-terminal domain with prenylated G protein By-subunits released upon receptor activation of coupled G proteins (23,421. The GRK5 sequence (and the GRK4 and GRKG sequences) has no CAAX motif for direct prenylation, nor does it have sequences similar to the P y binding region defined on PARK1 and /3ARK!2 (23). The activity of GRK5 to phosphorylate rhodopsin or the. P2AR is not appreciably affected by the Py-subunits of G proteins from brain or retina, nor does the glutathione S-transferase fusion protein containing the C-terminal portion of GRK5 appear to bind to py-~ubunits.~ Assays for the activated receptor-dependent translocation of GRK5 from solution to ROS membranes surprisingly demonstrate that the kinase appears to associate with the membranes constitutively, even in the absence of receptor activation. Native GRK5-like immunoreactive proteins are present in several cell lines, including COS, and these proteins are associated primarily with the membrane fractions.6 This pattern of agonist-independent enzyme localization to the membrane is similar to that seen with a mutant rhodopsin kinase modified with the C20 geranylgeranyl lipid rather than the native C15 farnesyl lipid (26). GRK5 also appears to have a higher basal (antagonist-occupied receptor) activity than PARK1 alone (Fig.   6), but is similar to the activity of PARK1 plus Pr in the presence of propranolol.6 One possible explanation for this higher basal activity is that the GRK5 enzyme is already close to its substrates on the membrane, as is the "activated" flARK1.p~ complex. This implies that no other factors are likely to play a role in localizing the GRK5 enzyme to the membrane, as do Py-subunits for the PARK enzymes. It is interesting to note that the last 46 residues of the GRK5 kinase are predominantly positively charged, hydroxyl-or amine-containing amino acids. Whether this strongly basic domain (PI = 11.3) plays a role in kinase localization is not yet known. However, the mechanisms responsible for regulation of GRK5 (and related kinases) access to membrane receptor substrates are quite distinct from the recently defined PARK and rhodopsin kinase paradigms.