HSP22, a new member of the small heat shock protein superfamily, interacts with mimic of phosphorylated HSP27 ((3D)HSP27).

Most of the members of the superfamily of mammalian small heat shock or stress proteins are abundant in muscles where they play a role in muscle function and maintenance of muscle integrity. One member of this protein superfamily, human HSP27, is rapidly phosphorylated on three serine residues (Ser(15), Ser(78), and Ser(82)) during cellular response to a number of extracellular factors. To understand better the role of HSP27, we performed a yeast two-hybrid screen of a human heart cDNA library for HSP27-interacting proteins. By using the triple aspartate mutant, a mimic of phosphorylated HSP27, as "bait" construct, a protein with a molecular mass of 21.6 kDa was identified as an HSP27-binding protein. Sequence analysis revealed that this new protein shares an overall sequence identity of 33% with human HSP27. This protein also contains the alpha-crystallin domain in its C-terminal half, a hallmark of the superfamily of small stress proteins. Thus, the new protein itself is a member of this protein superfamily, and consequently we designated it HSP22. According to the two-hybrid data, HSP22 interacts preferentially with the triple aspartate form of HSP27 as compared with wild-type HSP27. HSP22 is expressed predominantly in muscles. In vitro, HSP22 is phosphorylated by protein kinase C (at residues Ser(14) and Thr(63)) and by p44 mitogen-activated protein kinase (at residues Ser(27) and Thr(87)) but not by MAPKAPK-2.

The superfamily of mammalian small heat shock or stress proteins (sHSP) 1 in humans consists of the known members HSP27, ␣B-crystallin, ␣A-crystallin, HSP20, HSPB2, HSPB3, and cvHSP (1)(2)(3)(4)(5). Throughout the animal, plant, and microbiotic kingdoms, members of this protein superfamily share the so-called ␣-crystallin domain in their C-terminal part, whereas other parts of the sequence (N-terminal halves and extreme C-terminal tails) are more variable. Some of the sHSPs are known to be phosphoproteins. ␣B-crystallin is phosphorylated at three sites by the protein kinases p44/42 MAPK (Erk1/2), MAPKAPK-2, and probably by PKA (6,7), whereas HSP20 possesses at least one PKG/PKA phosphorylation site (8). The best studied is HSP27, which is phosphorylated at two (mouse) or three (human) phosphorylation sites by MAPKAPK-2 (9 -11). Phosphorylation of HSP27 is an early and prominent event in many cells when stimulated by a variety of mitogenic and stress factors. For example, contraction of smooth muscle cells in response to bombesin or endothelin-1 involves phosphorylation of HSP27 (12,13). Similarly, Sertoli cells respond to germ cells (14) and pancreatic acinar cells to cholecystokinin (15) by rapid phosphorylation of HSP27. Another characteristic of sHSPs is their ability to form supramolecular structures (complexes). The basic structural units of sHSP complexes are dimers that are formed by interaction of segments of the ␣-crystallin domain (16). The formation of structures of higher orders, such as tetramers, octomers, etc., appears to require additional sites of interaction that have been located to the N-terminal region of the molecules (17). Besides homo-oligomeric complexes, sHSPs can also form hetero-oligomeric complexes if different sHSP species are present (16,18,19). sHSP complexes have a dynamic structure with subunits exchanging rapidly between complexes as has been determined by fluorescence resonance energy transfer measurements (18). The molecular mass of cellular sHSP complexes varies over a wide range (50 -1000 kDa), with complexes of increasing size being predominant under stress conditions. Complex formation is believed to be a key feature in regulating the activity of sHSPs in vivo, although details are not understood (20). Also, the reported in vitro activities of HSP27, chaperoning and inhibition of actin polymerization, depend positively or negatively, respectively, on complex formation, which itself appears to depend, to a certain extent, on the degree of phosphorylation (21)(22)(23)(24).
In recent years, evidence has been obtained for the crucial role of some of the sHSPs in muscle function. A point mutation in the ␣B-crystallin gene causes a severe desmin-related cardiomyopathy in humans (32), and HSPB2 binds and activates the myotonic dystrophy protein kinase, an enzyme that when absent results in myotonic dystrophy (4). Whereas in most cell types HSP27 and ␣B-crystallin have a diffuse cytosolic location, the situation in striated muscles is different. Both proteins have been found in specific locations such as the I-band (33) and Z-disc (34) of myocytes, and most importantly, they quickly relocate between the I-band, Z-disc, and the cytosol upon disease-related, ischemic, or thermal stress (33)(34)(35). The highly specific location of sHSPs in myocytes and the stressrelated relocation of sHSPs led us to hypothesize that HSP27 interacts with other proteins in muscle. It is also hypothesized that the observed relocation of HSP27 involves a regulatory mechanism able to respond to physiological signals.
In an effort to identify proteins that interact with HSP27 in muscle cells, we screened a human heart cDNA library by the yeast two-hybrid method. In order to obtain data related to the regulation of the binding preference of HSP27 in dependence on its phosphorylation, we used as "bait" proteins both human wild-type HSP27 ( wt hHSP27) and a mimic of phosphorylated human HSP27 in which the three serines phosphorylated by MAPKAPK-2 (Ser 15 , Ser 78 , and Ser 82 ) had been substituted by aspartate ( 3D hHSP27). Here we report the identification, molecular cloning, and characterization of a novel sHSP we designated HSP22 and that is expressed in striated and smooth muscle tissues. This protein binds preferentially to 3D hHSP27 rather than to wt hHSP27. Thus, HSP22 matches the requirements of the initial hypothesis. It is a new binding partner of HSP27 occurring in muscles, and the HSP27-HSP22 interaction is regulated by phosphorylation of HSP27.

EXPERIMENTAL PROCEDURES
Plasmid Constructs-In order to create two-hybrid bait vectors, the cDNAs encoding the human (h) wt hHSP27 and 3D hHSP27 were amplified by PCR using the vectors pBS-wt hHSP27 and pBS-3D hHSP27 (obtained from L. A. Weber, University of Nevada, Reno) as template. The sense and antisense primers were 5Ј-cagccatcatgaccgagcgccgcgtc-3Ј (restriction site BspHI underlined) and 5Ј-caactcgaggtggttgctttgaactttatttg-3Ј (restriction site XhoI underlined), respectively. The PCR products (ϳ730 bp) were digested with the restriction enzymes BspHI and XhoI and fused in frame with the Gal4-DNA binding domain of the vector pAS2-1 (CLONTECH) using the restriction sites NcoI and SalI. The resulting bait constructs, designated pAS2-1-wt hHSP27 and pAS2-1-3D hHSP27, were used in the two-hybrid screening.
Full-length hHSP22 cDNA was constructed on the basis of the isolated "prey" plasmid pACT2-hHSP22␦NT (see below) and of the ESTs 874 288, 1 579 386, and 1 962 578. Two synthetic oligonucleotides (5Ј-gatccatggctgacggtcagatgcccttctcctgccactacccaagccgcctgcgccgagaccccttccggg-3Ј and 5Ј-agtcccggaaggggtctcggcgcaggcggcttgggtagtggcaggagaagggcatctgaccgtcagccatg-3Ј) were hybridized yielding the doublestranded DNA fragment F1 (representing the 5Ј-end of hHSP22 cDNA) with the upstream BamHI-compatible end 5Ј-gatc . . . and the downstream HinfI-compatible end . . . tga-5Ј. From the vector pACT2-hHSP22␦NT a 576-bp HinfI fragment (F2) was excised containing most of the translated sequence of hHSP22. Fragments F1 and F2 were ligated yielding F3, which contained the complete sequence of the translated part of hHSP22 cDNA. F3 was inserted into the prokaryotic expression vector pRSET.B (Invitrogen) using the restriction sites BamHI and EcoRI and taking advantage of the fact that the downstream HinfI-compatible end of F3 ( . . . tta-5Ј) could be ligated at low frequency to the EcoRI end of the vector. The resulting construct pRSET.B-hHSP22 was used for expression of hHSP22 in Escherichia coli and for PCR (see below).
In order to create an eukaryotic expression vector, the full-length hHSP22 sequence was amplified from pRSET.B-hHSP22 by PCR using the primers 5Ј-cggaattcaacatggctgacggtcagatgcccttct-3Ј (restriction site EcoRI underlined) and 5Ј-agccctcgagcttcgaatcaaagaagcccta-3Ј (restriction site XhoI underlined). The PCR fragment was digested with EcoRI and XhoI and cloned into the vector pcDNA3.1 (Invitrogen) yielding the construct pcDNA3.1-hHSP22 which was used for transfection of COS-7 cells and for PCR.
In order to create a full-length hHSP22 cDNA in the vector pACT2 (CLONTECH), the hHSP22 sequence was amplified from the construct pcDNA3.1-hHSP22 by PCR using the primers 5Ј-gccagtgtggccgtggtggccatggctgacggtcagatgcccttctcctgc-3Ј (restriction site SfiI underlined) and 5Ј-ccacaagagctcatctcaggtacaggtgacttcctg-3Ј (restriction site Ecl136II underlined). The PCR fragment was digested with SfiI and Ecl136II and fused in frame with the Gal4 transcription activation domain of the vector pACT2 (using the restriction sites SfiI and SmaI) yielding the construct pACT2-hHSP22 which was used for two-hybrid experiments.
All plasmid constructs were verified by sequencing. Two-hybrid Screening of the Human Heart cDNA Library-The human heart cDNA library (representing the cDNA of three human hearts, CLONTECH) in the prey vector pACT2 with the Gal4-transcription activation domain was propagated as described in the manufacturer's instructions. Two-hybrid screening (covering ϳ1.6 ϫ 10 6 independent clones for each bait construct) was performed by transforming the yeast strain Y190 first with the bait plasmids (pAS2-1-wt hHSP27 or pAS2-1-3D hHSP27), followed by a second transformation with the library plasmids. Primary positive clones were obtained with nutrition deficiency (-Trp,-Leu,-His) selection and the reporter gene expression assay (␤-galactosidase lift assay) according to the manufacturer's instructions (CLONTECH). The primary positive colonies were depleted of the bait vectors and mated with yeast strain Y187 carrying the vector pAS2-1 with the unrelated bait ␥-interferon cDNA as insert. This procedure permits identification and elimination of false positive clones. From the remaining "true" positive clones the plasmids were recovered, and proteins interacting with wt hHSP27 or 3D hHSP27 were identified by sequencing the inserts, either full-length cDNA or N-terminally truncated fragments. To establish the correct sequence of hHSP22, both strands of the isolated plasmid pACT2-hHSP22␦NT were sequenced three times.
Two-hybrid Experiments-In order to test interaction between hHSP22 and different forms of HSP27, yeast strain Y190 was transformed first with the plasmid pACT2-hHSP22. In a second transformation step the yeast was transformed with the plasmid pAS2-1wt hHSP27, pAS2-1-wt rHSP27 (25), or pAS2-1-3D hHSP27. Small scale transformation and selection procedures were as described in the manufacturer's instructions (CLONTECH).
Isolation of Mouse and Rat HSP22 cDNAs from Heart RNA Preparations-Total RNA was prepared from mouse and rat hearts with TRIZOL (Life Technologies, Inc.), and the HSP22 cDNAs were amplified by RT-PCR. The primers were designed on the basis of alignment with hHSP22 cDNA of several mouse and rat ESTs as follows (nucleotides complementary to the cDNAs are in uppercase): mouse sense, 5Ј-ttgagctccAAGCTCCGACCAACATCATGGCTGA-3Ј (restriction site SacI underlined); mouse antisense, 5Ј-tttgccatGGCTGACGTCTTAG-GAACAGGTGA-3Ј (restriction site NcoI underlined); rat sense, 5Ј-TGGCTCgagCTCTCTGAGCCTCTGTTTC-3Ј (restriction site XhoI underlined); and rat antisense, 5Ј-GGAattcAGAAGGACCAAGGCT-GACGTC-3Ј (restriction site EcoRI underlined). The RT-PCR products were digested with the restriction enzymes SacI and NcoI (mouse) and XhoI and EcoRI (rat), ligated into the vector pRSET.A (Invitrogen) using the corresponding restriction sites, and processed for DNA sequencing.
Expression of hHSP22 in E. coli, Purification of Recombinant hHSP22, and Development of a Specific Antibody-The vector pRSET.B-hHSP22 was used to transform the E. coli strain BL21(DE3) (Invitrogen). Synthesis of hHSP22 was induced by isopropyl-␤-thiogalactopyranoside, and the recombinant His-tagged protein was purified on a nickel column according to the manufacturer's instructions. The recombinant hHSP22 was used to raise a specific antibody in sheep. One mg of protein was mixed with Freund's complete adjuvant and injected into a sheep. Two weeks after a boost with another milligram of the protein in Freund's incomplete adjuvant, serum was collected from the sheep. The IgG fraction of the immunoglobulins was isolated using protein G-Sepharose (Sigma). The antibody was tested using transfected COS-7 cells expressing hHSP22 (cf. Fig. 4).
Expression of hHSP22 in COS-7 Cells-COS-7 cells (ATCC) were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. Transient transfection was performed using FuGENE 6 (Roche Molecular Biochemicals) according to the manufacturer's protocol. Briefly, 24 h prior to transfection, 5 ϫ 10 4 cells were seeded into 35-mm Petri dishes. Transfection was with 3 l of FuGENE 6 and 1 g of plasmid DNA dissolved in 100 l of serum-free medium. Cells were harvested 48 h after transfection and processed for Western blotting using the anti-hHSP22 antibody.
Electrophoretic Methods and Western Blotting-SDS-PAGE, two-dimensional-PAGE, and IEF-PAGE were performed according to standard procedures (36 -38). For Western blotting, proteins were extracted from COS-7 cells in suitable sample buffer (depending on the electrophoretic system), separated on SDS gels, two-dimensional gels, or IEF gels and transferred onto a PVDF membrane as described (25,38). For detection of hHSP22, the purified anti-hHSP22 IgG fraction was used at a dilution of 1:10.000. Visualization was with a secondary anti-sheep antibody conjugated to horseradish peroxidase using the ECL detection system (Amersham Pharmacia Biotech).
Identification of Phosphorylation Sites by Mass Spectrometry-20 g of recombinant hHSP22 was phosphorylated with either PKC or p44 MAPK (Erk1) as described above. After incubation, the reaction was stopped by addition of 500 l of ethanol (Ϫ20°C). The protein was pelleted by centrifugation (14,000 ϫ g for 10 min), washed with 70% (v/v) ethanol, reduced, and alkylated with iodoacetamide prior to overnight digestion at 37°C with trypsin (Promega) in 2 M urea, 0.1 M ammonium bicarbonate, pH 8.5. Peptides were desalted using a C18 ZipTip (Millipore) and eluted into 3 l of 70% (v/v) acetonitrile, 3% (v/v) formic acid. The eluate was placed into a static nanoflow probe tip (Micromass) and introduced into a Micromass Q-TOF mass spectrometer using a Z-spray ionization source. Comparison of a theoretical hHSP22 tryptic digest to the observed spectra was used to identify putative phosphopeptides. Putative phosphopeptides were analyzed by MS/MS to identify sites of phosphorylation, using collision voltages over the range of 20 -35 V.

Identification of hHSP22 and Cloning of Its Full-length
cDNA-In order to identify HSP27-binding proteins using the yeast two-hybrid system, hHSP27 cDNA was fused with the Gal4-DNA binding domain of the vector pAS2-1 and used as bait. We used both the full-length human wild-type HSP27 ( wt hHSP27) and a triple mutant of this protein in which the serine residues at positions 15, 78, and 82 were substituted with aspartate residues in order to mimic phosphorylation ( 3D hHSP27). After transformation of the yeast, the expression of these fusion proteins was verified by Western blotting using anti-HSP27 monoclonal antibody (not shown). After a second transformation of the yeast with the library plasmid representing the cDNA of three human hearts fused with the Gal4 transcription activation domain in the vector pACT2, a total of ϳ1.6 million yeast colonies each were screened. After eliminating false positive colonies by yeast mating, 17 colonies were obtained as true positives. By using wt hHSP27 as bait, h␣Bcrystallin (5 times) and hHSP20 (2 times) were identified as binding proteins. By using 3D hHSP27 as bait, hHSP27 (6 times), h␣B-crystallin (2 times), human mitochondrial cytochrome oxidase subunit III (1 time), and a novel protein (1 time), later designated hHSP22, were identified as binding proteins. With the exception of mitochondrial cytochrome oxidase subunit III, all identified proteins, including hHSP22 (see below), are members of the superfamily of sHSPs.
HSP22 Is a Member of the Superfamily of sHSPs-The identified cDNA consisted of the sequence between positions 71 and 1511 as shown in Fig. 1A. A BLAST search of nucleotide and protein data bases did not reveal any known protein with this sequence, although similarity with sHSPs was evident. However, several human EST sequences (cf. "Experimental Procedures") were found with 95% or greater identity, most of them derived from various muscle tissues. Alignment with these ESTs suggested that the novel sequence was 5Ј-truncated and also permitted the establishment of the complete cDNA sequence. The open reading frame has a translation start codon in a Kozak context, an in-frame translation stop codon, and, after an extended 3Ј-non-translated stretch, several polyadenylation signals (Fig. 1A). Since all ESTs extending into the 5Ј-direction have an in-frame stop codon 15 bp upstream of the translation start codon (not shown), a further 5Ј-extension of the translated sequence can be excluded. Translation of this sequence results in a putative protein composed of 196 amino acids with a calculated molecular mass of 21.6 kDa and a calculated pI of 5.0. Alignment of this protein with hHSP27 reveals ϳ33% overall identity with the highest similarity in the C-terminal half of the molecule, the ␣-crystallin domain. (cf. Fig. 1D). From these data it is evident that the novel protein belongs to the superfamily of mammalian sHSPs, and consequently we have named it HSP22. While this work was ongoing, other groups submitted the identical protein sequence to GenBank TM (GenBank TM accessions AAF09481, NP055180, and AAD55359), and in one report (39) the protein was described (see "Discussion"). Based on the sequence of hHSP22 we identified a number of ESTs derived from several mouse and rat tissues. Alignment of these ESTs permitted us to define the sequences of mouse (mHSP22) and rat HSP22 (rHSP22) cDNAs. According to these sequences, primers comprising the complete translated parts were designed, and the mouse and rat HSP22 cDNAs were amplified by single step RT-PCR using mouse and rat heart RNA preparations. The sequences of the mHSP22 and rHSP22 cDNAs are shown in Fig. 1, B and C, respectively. The translated part of the hHSP22 cDNA is 87.1 and 88.5% identical with the mHSP22 and rHSP22 cDNA, respectively. The rodent HSP22 cDNAs are 94.8% identical. Translation of these cDNAs results in the amino acid sequences for mHSP22 and rHSP22 as shown in Fig. 1D. hHSP22 and mHSP22 share 94.4% sequence identity, with 11 different amino acid residues, whereas the rodent sequences share 98.5% identity, with 3 different amino acid residues.
hHSP22 Interacts with 3D hHSP27 (Mimic of Phosphorylated hHSP27)-The fact that HSP22 was identified by two-hybrid screening using 3D hHSP27 as bait raised the question of whether it interacts preferentially with this form of the protein or if it also interacts with wt hHSP27. This question was addressed in a two-hybrid experiment comparing wt hHSP27 and 3D hHSP27 for their ability to bind to hHSP22. As an additional control, wt rHSP27 was included in this experiment. As shown in the control plate in Fig. 2, hHSP22 alone does not permit growth of the yeast on nutrient-deficient medium. Subsequent transformation with either of the bait constructs ( wt rHSP27, wt hHSP27, and 3D hHSP27) activates the reporter gene hisϩ thus permitting growth. In a second step, the colonies were tested for the activation of the second reporter gene galϩ. As shown in Fig. 2, neither of the wild-type proteins permitted activation of this reporter gene, whereas 3D hHSP27, in the presence of hHSP22, does activate the galϩ reporter gene. We conclude from these data that hHSP22 interacts preferentially with the phosphorylated form of hHSP27. Since both wt HSP27 species activate the reporter gene hisϩ, a residual interaction between wt HSP27 (both human and rat) and hHSP22 may occur, although in two-hybrid experiments usually two proteins are considered as interacting partners only if they activate both reporter genes.
Expression of HSP22 in Human Tissues-Analysis of abundance of hHSP22 mRNA (Northern blotting) reveals a single signal in several tissues with a size of ϳ1.8 kilobase pairs (Fig.   3). The highest expression of hHSP22 was found in skeletal and smooth muscles, heart, and brain. Expression was moderate in cervix, prostate, lung, and kidneys, whereas virtually no signal was seen in ovaries, testis, liver, pancreas, and spleen.
hHSP22 Isoforms and Dimers-The translated part of the hHSP22 sequence was inserted into the eukaryotic expression vector pcDNA3.1 and used for transient transfection of COS-7 cells. Protein extracts of COS-7 cells expressing hHSP22 were analyzed by SDS-PAGE and two-dimensional PAGE followed by Western blotting using a specific antibody raised against the recombinant protein. As predicted, hHSP22 could be detected in transfected COS cells on SDS gels as a single band with an FIG. 1. cDNA and amino acid sequences of human, mouse, and rat HSP22. A, complete sequence of hHSP22 cDNA as obtained from the plasmid pACT2-hHSP22␦NT isolated from the corresponding yeast colony (positions 71-1511) and from analysis of aligned ESTs (cf. "Experimental Procedures") (positions 1-70). B, sequence of mHSP22 cDNA as obtained from sequencing of the RT-PCR product from a mouse heart mRNA preparation. C, sequence of rHSP22 cDNA as obtained from sequencing of the RT-PCR product from a rat heart mRNA preparation. Translation start and stop codons are in boldface type, and polyadenylation signals are underlined by dots. Sequence elements constituting Kozak motifs are underlined. D, multiple alignment of hHSP22, mHSP22, and rHSP22 with hHSP27. Identical amino acid residues are highlighted with black if they occur in at least three sequences. Similarity in the remaining sequence or in at least three sequences is highlighted with gray (as similar residues are only considered as follows: D/E, N/Q, K/R, I/L/V, A/G, S/T and F/W/Y). E, potential and identified in vitro phosphorylation sites of hHSP22 for PKC, proline-directed protein kinases (e.g. the p44 MAPK Erk1), CK-2, and MAPKAPK-2. Potential phosphorylation sites for PKC fitting into the general motif (R/K)X 0 -2 (S/T)X 0 -2 (R/K) or its parts are underlined. Potential phosphorylation sites for proline-directed protein kinases fitting into the minimal motifs (S/T)P or PX(S/T)X(R/K) are highlighted in gray. Potential phosphorylation sites for CK-2 fitting into the general motif (S/T)XX(D/E) are double underlined. The potential phosphorylation site for MAPKAPK-2 fitting into the general motif HyXRXXS is indicated with strikethrough. ؉ indicates the identified in vitro phosphorylation sites for PKC (Ser 14 and Thr 63 ), and * indicates the identified in vitro phosphorylation sites for p44 MAPK (Ser 27 and Thr 87 ). apparent molecular mass of ϳ25 kDa, whereas no signal was obtained in control COS-7 cells (Fig. 4A). When the same sample was analyzed on two-dimensional gels, the 25-kDa species was separated into 1 major and 2 minor isoforms with isoelectric points of ϳ4.3, 4.6 (major form), and 4.9 (Fig. 4B). Additionally, two signals were obtained at an apparent molecular mass of ϳ50 kDa indicating the existence of hHSP22 dimers similar to what has been described for other sHSPs such as hHSP27 (40).
Phosphorylation of hHSP22-According to data published previously (39), HSP22 is a phosphoprotein in vivo (see "Discussion"). PROSITE analysis of the hHSP22 sequence reveals a number of potential phosphorylation sites for several protein kinases, including PKC and CK-2. Additionally, hHSP22 may contain phosphorylation sites for MAPKAPK-2 using the phosphorylation site motif HyXRXXS and for proline-directed protein kinases such as p44 MAPK (Erk1) using the phosphorylation site motif PXS/TP or parts of it (Hy indicates any hydrophobic amino acid, and X indicates any amino acid) (see "Discussion").
To study potential phosphorylation of hHSP22, we tested these four protein kinases for their ability to phosphorylate HSP22 in vitro. For that, we inserted the hHSP22 sequence into the bacterial expression vector pRSET (Invitrogen), expressed hHSP22 in E. coli, and purified the protein as described under "Experimental Procedures". The purified recombinant hHSP22 was incubated with the four protein kinases, and the phosphorylated hHSP22 was analyzed by two different methods. Incorporation of 32 P into hHSP22 was determined using [␥-32 P]ATP as phosphate donor. As shown in Fig. 5A, hHSP22 is phosphorylated by PKC, CK-2, and p44 MAPK (Erk1), whereas it is not phosphorylated by MAPKAPK-2. For control purposes, recombinant hHSP27 was tested under the same conditions. As expected, hHSP27 is phosphorylated by MAPKAPK-2, whereas no signal was obtained using the other protein kinases. As a second assay, isoelectric focusing was used to analyze phosphorylation of hHSP22. As shown in Fig.  5B, hHSP22 is phosphorylated to a substantial extent by PKC and p44 MAPK (Erk1) (as is seen by the acidic shift of the bands), whereas CK-2 phosphorylates only a minor fraction of hHSP22.
Analysis of hHSP22 Sites Phosphorylated by PKC and p44 MAPK (Erk1)-To identify in vitro hHSP22 phosphorylation sites, tryptic digests of PKC-and p44 MAPK (Erk1)-treated recombinant hHSP22 were introduced into a high resolution Q-TOF mass spectrometer by nanoelectrospray ionization. Several peptide ions were observed with masses 80 Da greater than theoretical predictions, indicating possible phosphopeptides. Each of these tryptic peptides was analyzed by MS/MS for peptide identification and detection of phosphorylation sites (Fig. 6).
In the case of PKC-phosphorylated hHSP22, a doubly charged ion of m/z 584.8 corresponding to a putative phosphopeptide, residues 56 -65 (LSSAWPGTLR), was fragmented. As shown in Fig. 6A, fragmentation of this peptide revealed an extensive y ion series (y 5-9 ) of fragment masses that readily established Thr 63 as a site of phosphorylation. This was supported by the satellite y 5-9 ion series detailing the loss of 98 Da, characteristic of collision-induced dissociation ␤-elimination of phosphoric acid (41). Furthermore, mass spectroscopy evidence from chemical derivatization of hHSP22 that targets phosphoryl groups supports the assignment of Thr 63 as a prominent site of PKC-catalyzed phosphorylation of hHSP22 (data not shown). 2 We also observed ions b 2 , b 4 , and b 5 , consistent with unmodified amino acids at Ser 57 /Ser 58 2. HSP22-HSP27 interaction. In a first transformation step, yeast strain Y190 was transformed with the plasmid pACT2-hHSP22 which does not permit growth on the selective medium (-Trp,-Leu,-His) (C, control). In a second transformation step, three batches of these yeast cells were transformed with the plasmid pAS2-1, containing either wt hHSP27, wt rHSP27, or 3D hHSP27 as insert. All three constructs, in the presence of pACT2-hHSP22, activate the hisϩ reporter gene thus permitting growth on the selective medium (upper row). Thereafter, colonies were transferred to a nitrocellulose membrane and processed for detecting the activation of the second reporter gene galϩ (the ␤-galactosidase activity) by the colony lift gal assay. In the presence pACT2-hHSP22, only the 3D hHSP27 construct activates galϩ, whereas both human and rat wt HSP27 constructs do not do so (lower row).

FIG. 3. Tissue distribution of HSP22 mRNA in human organs (Northern blots).
A ϳ600-bp-long 32 P-labeled cDNA probe was used for hybridization of separated mRNAs isolated from human tissues (blots from Invitrogen). Note the strong expression of hHSP22 in all three types of muscles and in brain. kb, kilobase pairs. FIG. 4. Expression of hHSP22 in COS-7 cells. A, expression of hHSP22 in transfected COS-7 cells was analyzed by SDS-PAGE followed by Western blotting using a specific anti-hHSP22 antibody. hHSP22 has an apparent molecular mass of ϳ25 kDa. B, expression of hHSP22 in transfected COS-7 cells was analyzed by two-dimensional PAGE followed by Western blotting. Expressed hHSP22 can be separated into three isoforms with isoelectric points (pI) of ϳ4.3, 4.6 (major isoform) and 4.9. Additionally, two dimeric forms of hHSP22 with a molecular mass of ϳ50 kDa can be detected under these conditions. Visualization of hHSP22 on the PVDF membranes was performed using ECL and a horseradish peroxidase-coupled secondary antibody. 1230.5). In addition, we observed y ions with the loss of phosphoric acid, y 3 (m/z 341.2), y 4 (m/z 504.3), y 9 (m/z 1132.5), and y 10 (m/z 1263.5). These data suggest that Ser 14 is a site for PKC phosphorylation. Phosphorylation of Ser Ϫ1 was ruled out upon the evidence of ions b 2 (m/z 219.1), b 4 (m/z 405.2), and b 6 (m/z 590.3) that show no addition of phosphate. The preliminary assignment that Ser 9 was not a phosphorylated residue was evident by the absence of phosphate from the b 12 and b 13 ions (data not shown).
In Fig. 6C, a doubly charged peptide of m/z 1073.9 corresponding to a putative phosphopeptide of p44 MAPK (Erk1)treated hHSP22, residues 79 -97 (FGVPAEGRTPPPF-PGEPWK), was fragmented by MS/MS. A sufficient number of y ions revealed the identity of the peptide; however, the y ion series did not extend through the only potential phosphorylation site at Thr 87 . Nonetheless, several distinct b ions were detected, in particular, b 8 at m/z 814.4 and b 9 at m/z 995.4. The mass observed for b 8 is consistent with residues 79 -86, whereas the mass for b 9 corresponds to the addition of a phosphoryl group at Thr 87 . This assignment was supported by the presence of the intense b 9 Ϫ98 Da ion (m/z 897.5). Furthermore, the [M ϩ 2H Ϫ H 3 PO 4 ] 2ϩ and the [M ϩ 2H Ϫ PO 3 ] 2ϩ ions that indicate peptide phosphorylation were also detected in the spectrum. Fig. 6D shows the MS/MS spectrum obtained for a triply charged tryptic fragment 19 -29 (DPFRDSPLSSR) at m/z 452.8 of p44 MAPK (Erk1)-treated hHSP22. An informative series of y ions (y 1 -y 5 ) as well as the y ions minus phosphoric acid (y 3 -y 5 ) were detected by the Q-TOF. We also detected a peak at m/z 420.2, consistent with the [M ϩ 3 H Ϫ H 3 PO 4 ] 3ϩ ion providing further evidence of a phosphorylation site within this peptide. Ser 24 was ruled out as a site of phosphorylation because we observed the y 3 Ϫ H 3 PO 4 and the y 4 Ϫ H 3 PO 4 ions that indicate Ser 27 or Ser 28 are the sites of phosphorylation. In addition, the presence of the b 6 ion (m/z 718.3) supports our conclusion that residue Ser 24 is not phosphorylated. In the case of this peptide, defining the site of phosphorylation by MS/MS was hampered due to the consecutive arrangement of potential phosphorylation sites (Ser 27 /Ser 28 ). Unfortunately, the b 9 or b 10 ions that would unambiguously determine which serine residue was modified were not ob-served. However, since we detected y 2 , and Ser 27 (and not Ser 28 ) would be part of the phosphorylation site motif PX(S/ T)X(R/K) which is occasionally observed for proline-directed serine/threonine protein kinases, it is highly likely that Ser 27 is the actual phosphorylation site (cf. "Discussion"). DISCUSSION A number of observations suggest that sHSPs, which may constitute 1% or more of total muscular protein (2), play a major role in function and differentiation of muscles and in the maintenance of their integrity under stress conditions. Striking arguments for the involvement of sHSPs in normal muscle function comes from genetic diseases. Autosomal dominant desmin-related cardiomyopathy (32) is caused by the missense mutation R120G in the ␣B-crystallin gene resulting in an abnormal organization of the intermediate filament protein desmin, a known interacting partner of ␣B-crystallin. Myotonic dystrophy is caused by a genetic defect in which GTC repeats in the 3Ј-untranslated region of the gene coding for the myotonic dystrophy protein kinase are unstably transmitted resulting in a lack of this enzyme. However, the complex pattern of symptoms is not simply a matter of lack of this enzyme (42), and the activation and protection of this protein kinase by HSPB2 (4) is an indication that both proteins are involved in protection of muscles under stress conditions. Also, the differentiation of multipotent embryonic carcinoma cells P19 into functional cardiomyocytes requires the p38/HSP27 pathway (43). Both, HSP27 and ␣B-crystallin were shown to protect cardiomyocytes from adverse anoxic conditions, a model system for pathological ischemia (44).
Studies of the intracellular localization of sHSPs in myocytes have also led to interesting observations. In early studies ␣Bcrystallin was found to localize in the Z-disc in adult cardiomyocytes (45), while later both HSP27 and ␣B-crystallin were found to localize in the I-band (33) and the cytosol (46,47). This difference in observations is probably resolved by the fact that sHSPs can re-localize quickly, depending on the physiological conditions. Pathological stress was reported to result in a depletion of HSP27 and ␣B-crystallin from the sarcomeres into the cytosol (33), whereas ischemia resulted in a relocation of ␣B-crystallin in the Z-disc and, with increasing severity, also in the I-band (34). In other experimental systems, ischemia and preconditioning caused a shift of HSP27, HSPB2, and ␣Bcrystallin from the cytosol to the sarcomeres (46,47). In neonatal cardiomyocytes, in which sHSPs have a basic cytosolic location, thermal stress shifted HSP27 and ␣B-crystallin into the myofibrils (35). Recently, phosphorylation and relocation of sHSPs were suggested to play a role in physiological phenomena such as ischemic preconditioning (48).
These facts strongly suggest an involvement of sHSPs in muscle function and led us to hypothesize that sHSPs must interact with other proteins in muscles. Additionally, there must be a regulatory mechanism, such as phosphorylation of HSP27, that would change the binding preference of these proteins and that could serve to explain the stress-related relocation. To test this hypothesis, we used the two-hybrid method to screen a human heart cDNA library for proteins interacting with hHSP27, using as bait both wild-type hHSP27 and a mutant that mimics phosphorylated hHSP27. Of the 17 positive yeast clones expressing HSP27-binding proteins, 16 were determined to be sHSPs (hHSP27 itself, h␣B-crystallin, hHSP20, and hHSP22). This was an expected result since the heart contains high amounts of these proteins that all contain the ␣-crystallin domain as the identified interacting site. The abundance of ␣B-crystallin (3-4 g/mg) and HSP20 (2 g/mg) in heart is ϳ7.5 and 4.5 times, respectively, higher than HSP27 itself (0.2-0.7 g/mg). This would explain, together with the FIG . 5. In vitro phosphorylation of hHSP22. A, incorporation of radioactive phosphorous into recombinant hHSP22 and hHSP27 by PKC, CK-2, p44 MAPK (Erk1), and MAPKAPK-2 (MK-2). The purified recombinant sHSPs were incubated with these protein kinases in the presence of [␥-32 P]ATP as described under "Experimental Procedures." After termination, the reaction products were separated by SDS-PAGE, and the dried gels were exposed to x-ray film. Note that under the reaction conditions PKC, CK-2, and p44 MAPK (Erk1) can phosphorylate hHSP22 (lanes 2) but not hHSP27 (lanes 1), whereas MAPKAPK-2 phosphorylates HSP27 but not hHSP22. B, phosphorylation of recombinant hHSP22 by CK-2, PKC, and p44 MAPK (Erk1) as analyzed by IEF-PAGE. The purified recombinant hHSP22 was incubated with these protein kinases in the presence of 1 mM ATP as described under "Experimental Procedures". After termination, the reaction products were separated by IEF-PAGE and transferred onto a PVDF membrane. hHSP22 isoforms were detected with a specific anti-hHSP22 antibody. Note that the control protein (incubated in the absence of protein kinase) focuses into two major isoforms. Under the reaction conditions, incubation with PKC and p44 MAPK (Erk1) phosphorylates hHSP22 by ϳ50% or more (detected as "acidic shift" of the bands toward the anode), whereas CK-2 phosphorylates hHSP22 only to a minor extent. ability of these sHSPs to form heterodimers, the high incidence with which these proteins were identified in the screening (1, 33, 49 -51). Similarly, the known tendency of HSP27 to form homodimers (16,40) explains the high incidence with which HSP27 itself was identified. More difficult to explain, however, is the different frequencies with which both bait constructs were found to bind to the other sHSPs in this screening. ␣B-Crystallin was identified 5 times versus 2 times, HSP20 2 times versus 0 times, and HSP27 0 times versus 6 times, using as baits wt HSP27 and 3D HSP27, respectively. These data may indicate that there is a general change in the binding preference of HSP27 upon phosphorylation. The possible consequences of this include alterations in the composition of sHSP complexes. This has been observed in several experimental systems (21,24,52,53). The preferential binding of 3D hHSP27 (as opposed to wt hHSP27) to hHSP22 strongly suggests such a change in binding preference upon HSP27 phosphorylation thus providing an argument for this view. In fact, these findings shed new light on the molecular function of HSP27 phosphorylation that may have consequences for the activity of sHSPs in a number of biological systems. For example, this shift in the binding preference among different sHSPs may well be the basis for the relocation of HSP27 and ␣B-crystallin observed in muscle cells and may be involved in responses such as ischemic preconditioning of hearts (48).
hHSP22 is most closely related to hHSP27 among the known human members of this protein superfamily according to a parsimony analysis (not shown). Similarity is highest in the ␣-crystallin domain with the common motifs VKTKDK, GKHEE, VDP, SLSPEG, and EAP (cf. Fig. 1D), and it is less pronounced in the N-terminal part with only a few common motifs (DPFRD and SLR) and in the C-terminal tail. Thus, HSP22 fits into the similarity pattern characteristic for this protein superfamily (2). The three MAPKAPK-2 phosphorylation sites of hHSP27 (Ser 15 , Ser 78 , and Ser 82 ) are not conserved in the hHSP22 sequence, and the only possible MAPKAPK-2 site at Ser 58 in the sequence LPRLSS (cf. Fig. 1E) is not functional (Fig. 5A), probably due to steric hindrance. In transfected COS-7 cells, hHSP22 exists in three isoforms (pI ϳ4.9, 4.6, and 4.3). The pI of the most basic isoform (4.9) is close to the predicted pI of 5.0. The more acidic isoforms may originate from modifications such as phosphorylation; however, to date the nature of these modifications is not known. Like other sHSPs studied, hHSP22 has a slightly higher apparent molecular mass than predicted (25 versus 22 kDa), and it has the ability to form homodimers (cf. Fig. 4B). Whether HSP22 also can form heterodimers with sHSPs other than 3D HSP27 and larger oligomers (complexes) is not known so far but seems reasonable to assume. On Northern blots, hHSP22 mRNA was detected in a single band of ϳ1.8 kilobase pairs which is slightly larger than the size predicted from the sequence shown in Fig. 1A, probably due to the addition of a poly(A) tail.
While this study was in progress, sequences identical with hHSP22 were submitted to GenBank TM (GenBank TM accession numbers AAF09481, NP055180, and AAD55359), and one report was published describing this sequence (39). In that study, the ␣-crystallin domain of hHSP22 was not recognized, as well as the similarity of this protein with the other sHSPs. hHSP22 (designated H11) was identified as a mammalian protein with weak similarity to ICP10, a herpes simplex virus type 2 ribonucleotide reductase. This protein was suggested to have a weak autophosphorylation activity similar to what has been reported earlier for ␣B-crystallin (54). The basis for the residual protein kinase activity of these sHSPs is the presence of a few (ϳ5) catalytic motifs in the ␣-crystallin domain, whereas a fully active protein kinase usually has 12 motifs. Despite this residual activity, protein phosphorylation or autophosphorylation is probably not the primary function of sHSPs including HSP22.
The expression of HSP22 varies in different human tissues, although the prevailing expression in all three types of muscles is evident (cf. Fig. 3). Hence, the expression pattern of HSP22 is similar to that of most of the other mammalian sHSPs that are, with the exception of ␣A-crystallin, abundant in muscles (2,5,19). In striated muscles, sHSPs have been found to form two types of complexes, with type I complex consisting of HSP27, HSP20, and ␣B-crystallin and type II complex consisting of HSPB2 and HSPB3 (19). In accordance with that, the type I sHSPs, HSP27, HSP20, and ␣B-crystallin, have been identified as binding partners for HSP27 in the two-hybrid screening in this study. Whether HSP22 is also a part of the type I complex is not known at present; it may also exist in a separate form, e.g. as monomer, dimer, or as a complex. The observed tendency of type I complexes to disassemble upon phosphorylation of HSP27 may be related to the fact that HSP27 shifts its binding preference toward HSP22. In general, sHSP complexes are highly dynamic structures that quickly exchange subunits (18). So far, no data on the dynamics or structure of complexes involving HSP22 are available.
There are some indications that hHSP22 is a phosphoprotein. In vivo, using transfected 293T cells, hHSP22 could be labeled with radioactive phosphorous, and the hydrolysis of this labeled hHSP22 revealed phosphoserine and phosphothreonine as phosphoamino acids (39). This fits with the in vitro data obtained in the present study with serine and threonine residues identified as being phosphorylated by PKC and p44 MAPK (Erk1) (Figs. 1E and 5).
PKC is a rather unspecific protein kinase requiring basic amino acid residues near the phosphoacceptor group. PKC uses the consensus motif of the general structure (R/K)X 0 -2 (S/ T)X 0 -2 (R/K) or its parts (S/T)X 0 -2 (R/K) and (R/K)X 0 -2 (S/T) (with up to two amino acid residues between the basic residues and the phosphoacceptor group) (55). Among the 16 potential phosphorylation sites of hHSP22 for PKC fitting into this general motif, two sites (Ser 14 -Arg,Thr 63 -Leu-Arg) were found to be phosphorylated. Both sites fit into the consensus motif and are conserved among the three HSP22 sequences studied (cf. Fig. 1D).
Proline-directed MAPKs generally use the consensus motif PX(S/T)P or frequently its part (S/T)P (56 -58). Among the three potential phosphorylation sites of hHSP22 with the minimal motif (S/T)P, one site (Thr 87 -Arg) was found to be phosphorylated by p44 MAPK (Erk1) in vitro. However, this site is not conserved among the HSP22 sequences studied (Asn in rat and Ser in mouse) in otherwise highly conserved sequences (cf. Fig. 1D). The significance of this is not known. The other identified phosphorylation site, Ser 27 , is positioned in the sequence Pro-Leu-Ser 27 -Ser-Arg. While Pro-Leu-Ser 27 would be part of the general motif PX(S/T)P, this minimal sequence is only occasionally used by proline-directed protein kinases. For example, the maturation promoting factor/cdc2 protein kinase has a modest preference for proline at position Ϫ2 using epidermal growth factor receptor-derived peptides as substrates (58). Furthermore, a basic residue at position ϩ2 (as the arginine in Pro-Leu-Ser 27 -Ser-Arg) has a strongly positive effect on proline-directed protein kinases (58). Thus, the sequence around Ser 27 meets the requirements for a potential phosphorylation site used by proline-directed protein kinases, although this type of motif deviates from the most frequently used motifs. Among the two potential phosphorylation sites for prolinedirected protein kinases fitting into the motif PX(S/T)X(R/K), one site was found to be phosphorylated. Ser 27 is conserved among the HSP22 sequences studied.
In future experiments, the in vivo relevance of these phospho-rylation sites as well as the involved signal transduction cascades will be elucidated. All four HSP22 phosphorylation sites identified in this study are positioned in the N-terminal part of HSP22 as is the case with the phosphorylation sites of the other musclerelevant sHSPs. According to the current state of knowledge, each sHSP seems to be phosphorylated by a unique set of protein kinases as follows: HSP27 by MAPKAPK-2 (9 -11), ␣B-crystallin by MAPKAPK-2, p44/p42 MAPK and PKA (6,7), and HSP20 by PKG/PKA (8). It may be speculated that the involvement of differential signaling pathways in phosphorylation of sHSPs is related to the control of the biological activity of these proteins or of the whole complexes. Despite considerable efforts, the functions of sHSPs are only poorly understood. As far as has been tested, sHSPs have the ability to protect cells from adverse conditions by preventing both apoptosis and necrosis (59). Although not understood in detail, it is likely that their stress-protecting properties are related to their in vitro chaperone-like activity (21) which may include their ability to stabilize and rearrange cytoskeletal elements (15,52,60,61). There are now many reports available describing the association of ␣B-crystallin, HSP27, and HSP20 with cytoskeletal elements such as actin, desmin, vimentin, laminin, and others (23, 29 -31, 61). Frequently, this association was found to stabilize the cytoskeletal structures, e.g. ␣B-crystallin and HSP27 stabilize actin filaments (29,52,61). It is not clear how the actin-stabilizing property of HSP27 may be related to its actin barbed end capping activity (which inhibits polymerization of actin in vitro) (23,28). However, the fact that stabilization of the microfilaments in vivo obviously requires phosphorylated HSP27 (or the corresponding D forms) appears to be consistent with the fact that in vitro phosphorylated HSP27 or derived peptides have reduced or absent actin polymerization inhibiting activity (15,23,52,62,63). Concerning the possible function of HSP22, no definitive conclusions can be drawn with the available data, although its function might be to bind to and modulate the activity of HSP27.