Characterization of the G protein-coupled receptor kinase GRK4. Identification of four splice variants.

A novel human G protein-coupled receptor kinase was recently identified by positional cloning in the search for the Huntington's disease locus (Ambrose, C., James, M., Barnes, G., Lin, C., Bates, G., Altherr, M., Duyao, M., Groot, N., Church, D., Wasmuth, J. J., Lehrach, H., Housman, D., Buckler, A., Gusella, J. F., and MacDonald, M. E. (1993) Hum. Mol. Genet. 1, 697-703). Comparison of the deduced amino acid sequence of GRK4 with those of the closely related GRK5 and GRK6 suggested the apparent loss of 32 codons in the amino-terminal domain and 46 codons in the carboxyl-terminal domain of GRK4. These two regions undergo alternative splicing in the GRK4 mRNA, resulting from the presence or absence of exons filling one or both of these apparent gaps. Each inserted sequence maintains the open reading frame, and the deduced amino acid sequences are similar to corresponding regions of GRK5 and GRK6. Thus, the GRK4 mRNA and the GRK4 protein can exist as four distinct variant forms. The human GRK4 gene is composed of 16 exons extending over 75 kilobase pairs of DNA. The two alternatively spliced exons correspond to exons II and XV. The genomic organization of the GRK4 gene is completely distinct from that of the human GRK2 gene, highlighting the evolutionary distance since the divergence of these two genes. Human GRK4 mRNA is expressed highly only in testis, and both alternative exons are abundant in testis mRNA. The four GRK4 proteins have been expressed, and all incorporate [3H]palmitate. GRK4 is capable of augmenting the desensitization of the rat luteinizing hormone/chorionic gonadotropin receptor upon coexpression in HEK293 cells and of phosphorylating the agonist-occupied, purified beta2-adrenergic receptor, indicating that GRK4 is a functional protein kinase.

G protein-coupled receptor kinases (GRKs) 1 are a family of serine/threonine protein kinases that phosphorylate G proteincoupled receptor proteins (1,2). Receptors are phosphorylated by GRKs in vitro on multiple serine and threonine residues on intracellular loops and/or the carboxyl-terminal tail. GRKphosphorylated receptors are bound by arrestin proteins, leading to the functional removal of the receptor from the signaling pathway by preventing further receptor coupling to G proteins (3). Agonist-occupied, activated G protein-coupled receptors are phosphorylated by GRKs, but inactive receptor proteins are not recognized by these kinases as substrates (1,2). The GRKs, along with arrestin proteins that bind to phosphorylated receptors and prevent coupling to G proteins, mediate homologous desensitization of hormonal responses (1).
Six GRKs have been identified by purification and by cloning. Rhodopsin kinase (GRK1) in the visual system phosphorylates light-bleached rhodopsin on its carboxyl-terminal tail (1,3). Arrestin bound to phosphorylated rhodopsin prevents the further activation of transducin, dampening activation of retinal cGMP phosphodiesterase (3). The "␤-adrenergic receptor kinases," GRK2 and GRK3, were identified by their functional ability to phosphorylate ␤ 2 -adrenergic receptors and are widely distributed throughout the body (2). GRK2 and GRK3 have been shown to phosphorylate a variety of G proteincoupled hormone and neurotransmitter receptors (2). Together with the somatic ␤-arrestin proteins, GRK2 and GRK3 have been implicated in the uncoupling of receptors from their respective G proteins, a mechanism leading to homologous desensitization (1,3).
A putative member of the family of G protein-coupled receptor kinases was recently identified by a positional cloning strategy, due to the proximity of its gene to the Huntington's disease locus (4). Originally named IT11, this sequence has been renamed GRK4 in an effort to systematize the nomenclature of these enzymes (1). GRK4 is most similar in sequence to mammalian GRK5 (5) and GRK6 (6) and to Drosophila GPRK2 (7). Together, these sequences define a subfamily of GRK enzymes that is distinct from the more extensively characterized rhodopsin kinase and ␤-adrenergic receptor kinase subfamilies. Both GRK5 and GRK6 have been functionally expressed and shown to phosphorylate several G protein-coupled receptors (5, 6, 8 -10). Functional activity of the GRK4 kinase has not yet been reported, and GRK4 is the least well understood member of the GRK family. Sallese et al. (11) have recently described an alternatively spliced exon encoding 32 amino acids in the amino-terminal region of human GRK4. In this paper, we define the structure of the human GRK4 gene and show that the GRK4 gene transcript undergoes extensive alternative splicing to generate four distinct forms of GRK4 mRNA that encode four forms of the GRK4 protein. We demonstrate the functional ability of GRK4 to augment desensitization of the LH/CG receptor in transfected cells and to phosphorylate the agonistoccupied, purified ␤ 2 -adrenergic receptor.

MATERIALS AND METHODS
Amplification of cDNAs-DNA was amplified by the polymerase chain reaction using standard conditions (12). Each 100-l reaction contained 1 ϫ Thermo buffer, 200 M each dNTP, 500 nM each oligonucleotide primer, and 10 ng of first-strand cDNA template. Reactions were assembled except for polymerase and heated for 5 min at 95°C, and 2.5 units of Taq DNA polymerase (Promega) were added. Reactions were cycled 35 times for 1 min at 95°C, 1 min at 60°C, and 3 min at 72°C, followed by a final 10-min extension at 72°C.
The amino-terminal coding region of the GRK4 cDNA was amplified from human testis first-strand cDNA using the primers 5Ј-ctcctcggtctcgcagaatcc (in the 5Ј-untranslated region) and 5Ј-aggtagttatgggcaactcta (antisense to RVAHNYL), while the carboxyl-terminal coding region was amplified using the primers 5Ј-aacatgggatccccccctttctgtcctgat (encoding PPFCPD) and 5Ј-gcaccggaattctcagcattgcttgggttc (antisense to EPKQC*). Product DNA bands were cut from agarose gels and TA subcloned into the pCRII vector (Invitrogen) or digested with BamHI and EcoRI (underlined sites) and subcloned into pBSII (Stratagene).
Amplification of Genomic DNA (Introns)-Human genomic DNA or individual cosmid DNAs were amplified by a long-accurate PCR technique (13). Taq and Tli DNA polymerases (Promega) were mixed 20:1 by volume, and PCR was carried out using standard buffer and nucleotide conditions as described above. Reactions were cycled 35 times for 1 min at 95°C, 1 min at 55-65°C, and 10 min at 72°C, followed by a final 10-min extension at 72°C. Reactions were performed with a series of primers spanning individual introns and their flanking exons as determined by DNA sequencing. The GRK4 gene was contained on the overlapping cosmids 8C10A12 (14), L142C5 (15), BJ56w4-3 (16), and BJ56 (17).
DNA Sequencing-Double-stranded sequencing of cDNAs was carried out by chain termination using Sequenase 2 (Amersham Corp./ U. S. Biochemical Corp.) and specific primers. For direct sequencing from cosmid DNAs, cycle sequencing was performed using radiolabeled oligonucleotide primers and the fmol TM sequencing kit (Promega). Primers located randomly throughout the cDNA sequence were used to identify intron locations, and specific primers were prepared to sequence through each exon and into the exon/intron junction.
Northern Blotting-Nylon filters containing 2 g of human tissue poly(A) RNAs transferred from agarose gels were obtained from CLON-TECH (human MTN I and MTN II). Blots were hybridized (12) overnight with the random prime-labeled IT11A clone (4), which lacks both alternative exons II and XV. The blot was washed in 0.2 ϫ SSC buffer at 65°C and exposed to x-ray film for 5 days at Ϫ70°C. Blots were then stripped of probe and rehybridized at 55°C with end-labeled antisense 48-mer oligonucleotides specific for the amino-terminal (bases 306 -353) and carboxyl-terminal (bases 1839 -1886) alternative exons. Oligonucleotide probes were washed in 2 ϫ SSC buffer at 55°C and exposed to x-ray film for 3 days at Ϫ70°C.
Antiserum and Immunoblotting-A fusion protein of glutathione Stransferase and the carboxyl-terminal 115 amino acids of GRK4␣ (the long form including exon XV) was constructed by inserting the amplified GRK4 DNA fragment (see above) into the BamHI and EcoRI sites of the pGEX-2T vector (Pharmacia Biotech Inc.). Protein was expressed by induction of Escherichia coli strain NM522 bearing the pGEX-GRK4-CT plasmid with 500 M isopropyl-1-thio-␤-D-galactopyranoside and growth for 2 h at 30°C, and the fusion protein was purified using glutathione-Sepharose as described (5). Antiserum was raised in two rabbits using this purified protein as immunogen (5).
Affinity-purified GRK4 antipeptide antibody raised against a peptide sequence from GRK4␦ was obtained from Santa Cruz Biochemicals. The antigen peptide (IPWQNEDCLTMVPSEKEVEP) is interrupted in the GRK4␣ and GRK4␤ sequences by the addition of exon XV, and this antibody was found to recognize only GRK4␥ and GRK4␦ (data not shown).
Immunoblotting was performed as described (5), using the crude GRK4-CT antiserum at 1:2500 dilution or affinity-purified antibody at 1 g/ml. Briefly, protein samples were separated by SDS-PAGE on 10% acrylamide gels, transferred to nitrocellulose membranes, and blocked by incubation with 3% bovine serum albumin in phosphate-buffered saline. Primary antiserum in 3% bovine serum albumin was allowed to bind for 1 h and washed, and the membrane was incubated for 1 h with goat anti-rabbit IgG-alkaline phosphatase conjugate (Pierce) for 1 h. Avidin-alkaline phosphatase conjugate was included to detect biotinylated molecular weight standards (Bio-Rad). After washing, the membrane was developed using Western blue reagent (Promega).
Expression of GRK4 Forms-The full-length GRK4 variants were reconstructed from amplified amino-and carboxyl-terminal halves of the cDNA, which were engineered to introduce BamHI and NotI sites to the immediate 5Ј-and 3Ј-ends, respectively, of the cDNA. These short and long GRK4 halves were joined in all four combinations by ligation at the unique internal BspEI site and inserted into pSL1180 (Pharmacia Biotech Inc.). All amplified clones were confirmed by DNA sequencing. All four forms of cDNA were inserted into pVL1393 (Pharmingen) using the BamHI and NotI sites and into pRK5 (18) using XmaI and SpeI/XbaI.
Recombinant GRK4 baculoviruses were produced by cotransfection of Sf9 cells with pVL1393 vectors bearing individual GRK4 inserts and BaculoGold baculovirus DNA (Pharmingen). Expression of GRK4 forms in Sf9 cells was induced by infection of 1.5 ϫ 10 6 cells/ml with the appropriate recombinant baculovirus at a multiplicity of infection of 5. Cells were grown for 48 h after infection; harvested by centrifugation; and stored frozen in 20 mM HEPES, pH 7.2, 5 mM EDTA, and 20 mM NaCl supplemented with a mixture of protease inhibitors (2 g/ml aprotinin, 10 g/ml benzamidine, 4 g/ml leupeptin, 1 g/ml pepstatin, and 100 M phenylmethylsulfonyl fluoride) (buffer A).
Whole Cell Desensitization Assays-HEK293 cells were maintained in Dulbecco's modified Eagle's medium containing 10% bovine serum. Cells were transfected by the calcium phosphate precipitation method (12) with 25 g of rat LH/CG receptor cDNA (20) in the pcDNAI vector in the presence of 12.5 g of pRK5 vector or pRK5 containing individual GRK cDNAs. Cells were split into 12-well plates (2 ϫ 10 5 cells/well) 1 day after transfection and labeled by incubation with 2 Ci/ml [ 3 H]adenine for 24 h. Cells were assayed for total cAMP accumulation in presence of human CG for 15 min. cAMP was purified using Dowex and Alumina chromatography (21) and measured as percent conversion of [ 3 H]ATP to [ 3 H]cAMP. LH/CG receptor expression was measured using 125 I-labeled LH (2 ng/ml) binding to whole cells in the absence and presence of excess (2 g/ml) human CG (22). Receptor levels were determined to be within 2-fold for all transfection groups in all experiments.
Partial Purification of GRK4␣-Sf9 cells from 250 ml of culture infected with GRK4␣ baculovirus, frozen in 10 ml of buffer A, were thawed and lysed by 10 strokes of a Dounce homogenizer. Crude lysates were spun at 1000 ϫ g to separate unlysed cells and nuclei. The lysate was supplemented with deionized Lubrol PX to 1% (w/v) and NaCl to 100 mM and rotated for 1 h at 4°C. The lysate was spun at 100,000 ϫ g for 1 h. The detergent-solubilized fraction was diluted 10-fold in buffer A with 100 mM NaCl to reduce Lubrol to 0.1%, filtered through a 0.2-m Nalgene filter, and passed over a Mono-Q HR 5/5 column (Pharmacia Biotech Inc.). GRK4␣ protein was eluted with a 100-ml gradient of 100-1000 mM NaCl in buffer A plus 0.1% Lubrol. Fractions containing immunoreactive GRK4 eluted at 400 -500 mM NaCl and were pooled for use in functional assays.

RESULTS
Splice Variants of GRK4 -Comparison of the deduced amino acid sequence of human GRK4 (4) with that of bovine GRK5 (5) and other G protein-coupled receptor kinases indicated the absence of two stretches of residues in GRK4 that are conserved in other members of this family of kinases. One apparent gap of 32 amino acid residues occurred after Gln 17 in the amino-terminal domain of GRK4, while the second gap of 46 amino acids occurred after Glu 515 in the carboxyl-terminal domain (numbering based on the longest (GRK4␣) sequence reported here). To determine whether these gaps were present in the mRNA encoding GRK4, PCR amplification across both of these regions was performed using cDNA prepared from human testis poly(A) RNA from two patient samples. For both cDNA samples, each of the amino-and carboxyl-terminal reactions yielded product bands of sizes consistent with the reported sequence (containing the apparent gaps), as well as larger bands consistent with a sequence having these gaps filled (data not shown). Surprisingly, the longer products were more abundant in all reactions.
All product bands were subcloned and sequenced. Comparison of the DNA sequences of the short and long PCR-amplified clones indicated that the longer clones were analogous to the shorter clones and to the previously reported GRK4 sequence, with additional nucleotide sequences occurring precisely at the predicted gaps, for each of the amino-and carboxyl-terminal products. These insertions each maintained the open reading frame, and deduced amino acid sequences of these insertions corresponded well to sequences found in GRK5 and GRK6 (Fig.  1). The amino-terminal insertion is identical to that reported by Sallese et al. (11).
GRK4 Genomic Structure-The presence of alternative internal sequences in the amino-and carboxyl-terminal regions of GRK4 suggested the existence of four distinct forms of GRK4 mRNA and thus of GRK4 protein. These variants could arise from alternative splicing of a single GRK4 gene transcript at two distinct sites. To determine whether these inserted sequences lie on distinct exons within the GRK4 gene, the exon/ intron organization of the human GRK4 gene was determined.
Gene fragments corresponding to all exons and all exon/intron junctions were sequenced directly from individual cosmid DNAs using specific radiolabeled oligonucleotide primers. The size of each intron was determined by polymerase chain reaction using pairs of primers from the flanking exons and the corresponding cosmid DNA as template. A schematic diagram of the GRK4 gene structure is shown in Fig. 2, and the precise exon/intron junction sequences are summarized in Table I Four GRK4 Forms-We have called the longest form of GRK4, with both the amino-and carboxyl-terminal alternative exon sequences, GRK4␣. The deduced GRK4␣ protein sequence, which is essentially colinear with GRK5 and GRK6, has 578 amino acids and a predicted molecular mass of 66.5 kDa. The next shorter form, lacking only the amino-terminal alternative exon (32 codons), is GRK4␤ (546 amino acids, 62.9 kDa), while the variant lacking only the carboxyl-terminal alternative exon (46 codons) is GRK4␥ (532 amino acids, 61.2 kDa). GRK4␥ is identical to the splice variant described by Sallese et al. (11), who called it GRK4A. The shortest variant, missing both alternative exons, is GRK4␦ (500 amino acids, 57.6 kDa). This is the originally described GRK4 sequence and has been called IT11 (4) and GRK4B (11).
Several differences were noted between the sequences obtained here and the originally reported sequence (4), and several polymorphisms were noted within the samples analyzed here. All numbering refers to the longest (GRK4␣) form reported here. Gly 562 was observed in all cDNA clones from two patient cDNA samples, in amplified genomic fragments from an independent patient, and in cosmid BJ56, rather than Asp as described previously. The original IT11A clone (4) was resequenced in this region and found to encode Asp 562 as reported.
In one patient cDNA sample, all clones contained Leu 61 , Val 142 instead of Arg 61 , Ala 142 , which was found in all other cDNA and genomic DNA samples. Cosmid L142C5 encodes Ile 247 instead of Val. Cosmid BJ56 carries a silent C 3 T polymorphism within Ser 519 . Both cDNA samples, one genomic DNA sample, and cosmid BJ56 all encode Ala 486 , while cosmid BJ56w4-3 encodes Val 486 , the one previously reported polymorphism in the human GRK4 sequence (4).
GRK4 mRNA Distribution-Based on PCR/Northern analysis of human cDNA libraries, GRK4 mRNA was reported to have a widespread, but low level, distribution in human tissues (4,11). The original IT11A clone was obtained from a human brain cDNA library (4). However, Northern analysis of six baboon tissue mRNAs indicated a significant level of GRK4 mRNA only in testis (4). To more fully characterize the pattern of GRK4 mRNA expression in man, Northern analysis was performed using mRNAs from 16 human tissues. Using the IT11A clone (4) as a probe, a 2.5-kb band was observed only in human testis poly(A) RNA (Fig. 3A). An additional weak band was evident at 4.5 kb in brain and skeletal muscle mRNAs. This 4.5-kb band is distinct from those of all known GRK4related GRKs and may represent an additional GRK4-like family member or alternatively processed GRK4 mRNA. Thus, GRK4 mRNA is highly abundant in testis, in agreement with the distribution and message size in baboon (4).
The IT11A clone used as probe above contains neither alternative exon. To ascertain whether these alternative exons are also present in the mRNA in significant levels, Northern blots were probed using antisense oligonucleotide probes specific for exons II and XV. As shown in Fig. 3B, exon II and XV probes each recognized a 2.5-kb mRNA from human testis. The ready detectability of both amino-and carboxyl-terminal alternative exons in testis mRNA indicates that they are highly abundant within the GRK4 mRNA pool, implying that the longest mRNA form may be the most abundant. The two oligonucleotide probes recognized no other RNA bands in 15 other human tissues tested (data not shown). The similarity in size among the GRK4 splice variant mRNAs is not surprising as only small internal sequences are alternatively spliced.
Expression of Four Forms of GRK4 -The four variants of the GRK4 cDNA were each reconstructed for expression studies. An antiserum prepared to the last 115 amino acids of GRK4␣ was used to assess the expression and subcellular localization of the four GRK4 proteins. It has been reported that GRK5 and GRK6 proteins are associated with cell membranes (5,19), in contrast to the cytosolic localization of GRK1, GRK2, and GRK3 (1, 2). Lysates from GRK4 baculovirus-infected Sf9 cells were separated into cytosolic and particulate fractions, and equivalent volumes were separated by SDS-PAGE. Anti-GRK4 antiserum detected each GRK4 protein form, while no immunoreactivity was apparent in uninfected Sf9 cells (Fig. 4). For all forms, substantially more immunoreactivity was seen in the membrane fraction than in the soluble fraction. The sizes of the four recombinant GRK4 proteins were in accord with their predicted molecular masses, from 58 to 66 kDa. Similar sizes and membrane association were seen in transfected COS cells (data not shown). Thus, GRK4, like GRK5 and GRK6, appears to have significant basal membrane association.
Palmitoylation of GRK4 Proteins-It has recently been demonstrated that GRK6 is palmitoylated on carboxyl-terminal cysteine residue(s) and that this lipid modification is important for the membrane localization of the GRK6 protein (19). GRK5, GRK2, and GRK1 are not palmitoylated (19). The ability of GRK4 proteins to incorporate [ 3 H]palmitate was determined, as GRK4 shares one carboxyl-terminal cysteine residue with GRK6 (but is not found in GRK5) and ends with a unique carboxyl-terminal cysteine residue. COS cells expressing the individual GRK4 forms were labeled with [ 3 H]palmitate or with a mixture of [ 35 S]cysteine and [ 35 S]methionine. GRK4 proteins from labeled cells were immunoprecipitated using anti-GRK4 antiserum and separated by SDS-PAGE. A fluorograph is shown in Fig. 5. Like GRK6, all four GRK4 proteins incorporated [ 3 H]palmitate. GRK4 palmitoylation is likely to play a role in the membrane association of the kinase, as has been shown for GRK6 (19). 2 Functional Receptor Desensitization by GRK4 Forms-To assess the functional ability of the GRK4 variants to desensitize G protein-coupled receptors, we chose to examine the effects of GRK4 on potentially colocalized receptors. Using immunoblotting to identify the native GRK4 protein forms in testis and mature sperm has been unsuccessful due to a prominent nonspecific band of 65 kDa in all rabbit preimmune and immune sera tested (data not shown). Mature sperm did not appear to contain large amounts of GRK4 protein, although substantial GRK3 was detected, in agreement with a previous report (24). In the absence of finer localization of GRK4 expression within testis, it seemed reasonable to expect expression in Leydig and/or Sertoli cells, where the enzyme would have access to either the LH/CG or follicle-stimulating hormone receptors, respectively, as potential substrates.
We therefore chose to examine the ability of GRK4 proteins to modulate signaling by the rat LH/CG receptor transiently expressed in HEK293 cells. Individual GRKs were cotrans-fected with LH/CG receptor cDNA, and cAMP accumulation was assessed after stimulation of [ 3 H]adenine-labeled cells with human CG (Fig. 6). In the absence of added GRK cDNAs, human CG produces a dose-dependent cAMP accumulation (data not shown). In the presence of GRK2 cDNA, a GRK that is known to be active on many G protein-coupled receptors, human CG-induced cAMP accumulation is reduced by 38% at 200 ng/ml human CG. This loss of stimulation reflects GRKmediated processes occurring in the continued presence of the hormone, which occur naturally but are augmented in the presence of exogenous GRK2 (9,25). Transfected GRK4␣ induces a degree of desensitization (36% decrease at 200 ng/ml human CG) that is comparable to that achieved by transfected GRK2. Similar results were observed at 50 ng/ml human CG (data not shown).
Individual GRK4 variant cDNAs were each cotransfected with LH/CG receptors to assess their functional ability. Human CG-stimulated cAMP accumulation was reduced in the presence of GRK4␣, GRK4␤, and GRK4␦ to a similar extent, with 34 -43% reduction at 200 ng/ml human CG (equivalent to that seen with GRK2). In contrast, GRK4␥ produced a more modest 20% loss of LH/CG receptor signaling at 200 ng/ml human CG. Similar results were observed at 50 ng/ml human CG for ]methionine/cysteine mixture. Immunoprecipitated GRK4 proteins were separated by SDS-PAGE, and dried gels were exposed to x-ray film for 1 day. Lower panel, COS-7 cells transfected with individual GRK4 cDNAs were labeled with [ 3 H]palmitate. Immunoprecipitated GRK4 proteins were separated by SDS-PAGE, and the gel was treated with fluor. The dried gel was exposed to x-ray film for 2 months. A representative experiment (of three) is shown. GRK4␣, GRK4␤, and GRK4␦, although GRK4␥ did not appear significantly different from the control at this concentration (data not shown). Thus, all four GRK4 variants appear capable of decreasing signaling through the G protein-coupled LH/CG receptor.
GRK4 Kinase Activity in Vitro-Extracts from cells expressing GRK4 proteins showed no activity above background for the agonist-dependent phosphorylation of the G protein-coupled receptor rhodopsin or ␤ 2 -adrenergic receptor (data not shown and Ref. 11). This is similar to the weak activity of GRK6, but is in contrast to the readily detectable activity of GRK1, GRK2, GRK3, and GRK5. To circumvent this difficulty, the activity of recombinant GRK4␣ was assessed after partial purification from baculovirus-infected Sf9 cells. The high association of expressed GRK4 with the particulate fraction prompted the use of detergent to remove enzyme from the membrane. Crude cell lysates were extracted with Lubrol, and the extract was passed over a Mono-Q column to purify and concentrate the GRK4 protein. The activity of the eluted GRK4␣ protein was assessed using purified ␤ 2 -adrenergic receptor as substrate (Fig. 7). As the lipid composition of vesicles has recently been shown to have a great effect on the activity of several GRKs (23,26,27), 3 GRK4␣ was tested using the ␤ 2 -adrenergic receptor reconstituted into vesicles containing 100% phosphatidylcholine as well as 95% phosphatidylcholine plus 5% PIP 2 . Using purified receptor reconstituted into phosphatidylcholine vesicles, GRK4␣ was unable to phosphorylate the ␤ 2 -adrenergic receptor in the presence of isoproterenol or propranolol. Given receptor vesicles containing PIP 2 , GRK4␣ phosphorylated the ␤ 2 -adrenergic receptor in the presence of isoproterenol, but not in the presence of propranolol. This agonist-dependent receptor phosphorylation is the hallmark of the GRKs and indicates that GRK4 is indeed an active GRK enzyme. Like other GRKs (23), 3 GRK4 exhibits a requirement for PIP 2 in membranes containing substrate receptors. It is also apparent that GRK4 does not undergo autophosphorylation, despite the presence of a serine residue cognate to the autophosphorylation site mapped in GRK1 and GRK5 (5,28).  7. In vitro phosphorylation of the ␤ 2 -adrenergic receptor by GRK4␣. One pmol of ␤ 2 -adrenergic receptor (␤2AR) reconstituted into phospholipid vesicles (either 100% phosphatidylcholine (PC) or 5% PIP 2 and 95% phosphatidylcholine) was incubated with partially purified GRK4␣ in the presence of 100 M isoproterenol (ISO) or 100 M propranolol (PRO). Reactions were separated by SDS-PAGE, and dried gels were exposed to x-ray film. A representative experiment (of three) is shown. to that of the GRK2 gene, but containing significantly longer average intron sequences. 4 Comparison of the structure of the GRK4 gene with those of GRK2 and GRK3 reveals that all intron locations in the GRK4 gene are distinct from those in the GRK2 and GRK3 genes. Even the exons composing the conserved catalytic domain are completely distinct, which was unexpected for such closely related protein kinases. However, cloning of GRK cDNAs highly similar to GRK2 (dGPRK1) and GRK4 (dGPRK2) from Drosophila melanogaster (7) clearly demonstrates that the precursors of the GRK2 and GRK4 genes diverged before insects diverged from the lineage leading to chordates. GRK2 and GRK4 may be seen as models for their two subfamilies of GRKs, and it seems probable that the organization of the genes for the GRK4-related GRK5 and GRK6 enzymes will be more similar to that of the GRK4 gene. GRK1 (rhodopsin kinase) is intermediate in similarity between GRK2 and GRK4, so it will be interesting to compare the structure of the GRK1 gene with those of GRK2 and GRK4.
The GRK4 gene transcript undergoes alternative splicing to yield four distinct forms of GRK4 mRNA. This occurrence of alternative splicing is unique among the known GRK enzymes. The four GRK4 mRNAs arise from the presence or absence of exon II and/or exon XV. These alternative exons maintain the open reading frame and encode four GRK4 proteins that differ in the presence or absence of 32 amino acids in the aminoterminal region (11) or 46 amino acids in the carboxyl-terminal region. The locations of these alternative exons within the GRK4 protein suggest that they may play important functional roles in regulation of the enzyme, while the catalytic domains of all GRK4 forms are identical.
In this work, we have demonstrated that GRK4 is an active protein kinase with the activated receptor substrate recognition expected of a GRK enzyme. GRK4␣ phosphorylated purified ␤ 2 -adrenergic receptor in the presence of isoproterenol, but only using vesicles that contained PIP 2 . GRK4 appeared essentially inactive at phosphorylating ␤ 2 -adrenergic receptors reconstituted into pure phosphatidylcholine vesicles. GRK5 autophosphorylation (26) and activity, 3 GRK6 activity, 3 and GRK2 activity (23,27) all exhibit similar lipid cofactor requirements. PIP 2 -dependent activity appears to be an under-appreciated common feature of GRKs, although its physiological significance remains unexplored. The previous inability to measure the activity of GRK4␥ (11) may be due to use of a soluble cell fraction rather than detergent-extracted proteins, use of rhodopsin as a substrate, or lipid composition of the rod outer segment membranes.
The carboxyl-terminal domain of GRKs is involved in their subcellular localization. For GRK1 (rhodopsin kinase), the primary amino acid sequence encodes a CAAX box that directs post-translational farnesylation, proteolysis, and carboxymethylation (30). GRK1 is a cytosolic enzyme in the rod outer segment, but translocates to the membrane upon light activation of rhodopsin. A GRK1 mutant that lacks a CAAX motif fails to be prenylated and exhibits a deficit in translocation (31). GRK2 and GRK3 also appear cytosolic in unstimulated cells, but translocate to the cell membrane upon agonist activation of G protein-coupled receptors (31). This translocation is due to association of GRK2 with free G protein ␤␥-subunits and PIP 2 through a carboxyl-terminal region that includes a pleckstrin homology domain (23,32,33). Removal of this carboxylterminal region renders GRK2 unable to translocate to activated receptors (33). GRK4 contains neither a CAAX motif for protein prenylation nor a G protein ␤␥-subunit-binding domain.
GRK5 and GRK6 also lack prenylation or ␤␥-subunit binding, but exhibit a significant degree of association with cellular membranes. GRK5 has a highly basic carboxyl-terminal sequence, which has been proposed to serve to anchor GRK5 to phospholipids in the membrane (5). GRK4 also has a basic carboxyl-terminal sequence, which is present only in the two variants containing exon XV (GRK4␣ and GRK4␥). The absence of this basic domain in GRK4␤ and GRK4␦ did not lead to striking losses in the apparent membrane association of these proteins. The carboxyl-terminal region of GRK6 has no basic region, but contains cysteine residues that have been identified as sites for the post-translational palmitoylation (19). Palmitoylated GRK6 is found only associated with cellular membranes (19). All four GRK4 proteins incorporate palmitate, presumably via their carboxyl-terminal cysteine residue or on an internal cysteine residue conserved with GRK6. While the functional role of GRK4 and GRK6 palmitoylation remains to be explored, this lipid clearly contributes to membrane localization of the proteins. In addition, palmitoylation and depalmitoylation are dynamic processes within the cell, so it is possible that palmitoylation of GRK4 and GRK6 may be regulated in a signal-dependent manner. Regulated palmitoylation and depalmitoylation have been observed for other signal transduction components (34,35).
Although the amino-terminal domain of the GRKs is thought to be involved in recognition by the kinase of the activated G protein-coupled receptor, much less is known about the role of this region of the kinases (1, 2). An antipeptide antibody raised to the GRK1 amino terminus has been shown to block recognition of activated rhodopsin (36). It remains possible that the amino-terminal splice variants of GRK4 differ in their receptor substrate preferences or activation-dependent recognition requirements. Testing these possibilities will require comparing the activity of GRK4 variants to recognize and phosphorylate several distinct receptor substrates.
Comparison of the ability of the four GRK splice variant proteins to augment desensitization of the LH/CG receptor in transfected cells revealed that all GRK4 forms are functional. The GRK4␣, GRK4␤, and GRK4␦ variants appeared as active as GRK2, while the GRK4␥ variant appeared weaker. This assay may be insensitive to subtle differences among the kinases, as the expression level of the various enzymes was not directly quantified. However, all of the GRK4 forms appear to have some activity, so the alternative splicing does not render any forms completely inactive. More quantitative comparisons among the four variants will require purification of all four enzymes to equivalent extents for assay against various substrate receptors.
The relative abundance of the four native GRK4 proteins is unknown. In the amplification of the GRK4 amino-and carboxyl-terminal regions, the longer variant cDNA bands were more abundant, suggesting a predominance of exon II-and exon XV-containing mRNAs. This is in contrast to the report of Sallese et al. (11), who observed that the short amino-terminal form of GRK4 was most common in brain cDNA. Direct immunochemical detection of GRK4 proteins in testis has proven unsuccessful, although specific affinity-purified antibodies should allow this question to be addressed. Affinity-purified antibodies raised against a GRK4␦ carboxyl-terminal peptide obtained commercially also failed to recognize native GRK4␥ and GRK4␦ proteins in testis membranes or cytosol (data not shown), supporting the hypothesis that the longest (GRK4␣) form may be the most abundant form in testis.
It is noteworthy that GRK4 mRNA is essentially limited to testis. Very low levels of GRK4 mRNA are present in several other tissues, including brain (4, 11). Of the known GRKs, only the retinal GRK1 has such a limited tissue distribution (37) and thus a defined native substrate receptor, rhodopsin. The widespread and generally overlapping distributions of GRK2, GRK3, GRK5, and GRK6 hamper efforts to understand their individual functions. Little is yet known about the receptor substrate preferences of GRKs, although certain receptors can be phosphorylated equivalently by many GRKs (8,9,38). The limited distribution of GRK4 may facilitate definition of the G protein-coupled receptors it regulates. Although it is not yet known which of the three main testis cell types express GRK4, several receptors are known to be present in each cell type (39). Leydig cells express LH/CG receptor and gonadotropin-releasing hormone receptor. Sertoli cells contain follicle-stimulating hormone receptor and glucagon receptor. The various germ cell stages express bombesin BB 3 receptors and a variety of olfactory-like receptors. In addition, several G protein-coupled receptor mRNAs are expressed at high levels in testis, although their cellular localizations have not been defined. These include the adenosine A 3 receptor, cannabinoid CB 1 receptor, and vasopressin V 1a receptor, as well as several "orphan" receptors. All of these receptors are potential substrates for GRK4. Further studies characterizing the function of GRK4 will focus on regulation of those receptors expressed in the testis cell type(s) determined to contain GRK4. Differential regulatory properties of the individual GRK4 variants may be more observable on those receptors to which the enzymes normally have access.