Muscle Develops a Specific Form of Small Heat Shock Protein Complex Composed of MKBP/HSPB2 and HSPB3 during Myogenic Differentiation*

Previously, we identified a new mammalian sHSP, MKBP, as a myotonic dystrophy protein kinase-binding protein, and suggested its important role in muscle maintenance (Suzuki, A., Sugiyama, Y., Hayashi, Y., Nyu-i, N., Yoshida, M., Nonaka, I., Ishiura, S., Arahata, K., and Ohno, S. (1998) J. Cell Biol. 140, 1113–1124). In this paper, we develop the former work by performing extensive characterization of five of the six sHSPs so far identified, that is, HSP27, αB-crystallin, p20, MKBP/HSPB2, and HSPB3, omitting lens-specific αA-crystallin. Tissue distribution analysis revealed that although each sHSP shows differential constitutive expression in restricted tissues, tissues that express all five sHSPs are only muscle-related tissues. Especially, the expressions of HSPB3, identified for the first time as a 17-kDa protein in this paper, and MKBP/HSPB2 are distinctly specific to muscles. Moreover, these sHSPs form an oligomeric complex with an apparent molecular mass of 150 kDa that is completely independent of the oligomers formed by HSP27, αB-crystallin, and p20. The expressions of MKBP/HSPB2 and HSPB3 are induced during muscle differentiation under the control of MyoD, suggesting that the sHSP oligomer comprising MKBP/HSPB2 and HSPB3 represents an additional system closely related to muscle function. The functional divergence among sHSPs in different oligomers is also demonstrated in several ways: 1) an interaction with myotonic dystrophy protein kinase, which has been suggested to be important for the maintenance of myofibril integrity, was observed only for MKBP/HSPB2; 2) a myotube-specific association with actin bundles was observed for HSP27 and αB-crystallin, but not for MKBP/HSPB2; and 3) sHSPs whose mRNAs are induced by heat shock are αB-crystallin and HSP27. Taken together, the results suggest that muscle cells develop two kinds of stress response systems composed of diverged sHSP members, and that these systems work independently in muscle maintenance and differentiation.

Heat shock and numerous other stress conditions lead to the rapid induction of several genes whose protein products, collectively called heat shock proteins (HSPs), 1 play protective roles in cell survival (1). Considering that muscles are frequently subjected to severe conditions caused by heat, oxidative, and mechanical stresses, especially during exercise (2), these HSPs may be especially important in this particular tissue. In fact, several HSPs, including HSP60, 70, and 90, have been shown to be induced after exhaustive exercise (3). In addition, the inducible isoform of HSP70 has been shown to be constitutively expressed in a certain type of skeletal muscle fiber (4), suggesting that muscle cells are chronically ready to respond to frequent stresses. However, there have been limited numbers of studies that focus on HSPs in muscle cells.
HSPs with low molecular masses of 15-30 kDa are called small heat shock proteins (sHSPs); they commonly share a homologous sequence of about 80 amino acids called the "␣crystallin domain" (5). Among mammalian sHSPs, HSP27 2 and ␣B-crystallin have been studied most intensely; they show chaperone-like activity in vitro (6 -8), and their expressions are induced in response to such diverse stimuli as heat shock, heavy metal exposure, and hypertonicity (9,10). They confer stress resistance when overexpressed in cultured mammalian cells (11)(12)(13). The most interesting feature of their cytoprotective roles is that they are believed to act to stabilize cytoskeletal structures such as actin stress fibers and intermediate filaments (12,14). Consistently, they are also suggested to interact with these cytoskeletal proteins in vitro (15)(16)(17). Therefore, considering that muscle cells develop specific cytoskeletal structures based on actin and intermediate filaments (desmin), such as myofibrils, it is reasonable to speculate that HSP27 and ␣B-crystallin also play important roles in muscle maintenance. In fact, several studies have pointed out that these two sHSPs show abundant constitutive expression in skeletal muscle and heart (18,19). Furthermore, their roles in organizing and protecting the myofibril structure have been suggested by the demonstration of their localization on specific sarcomeric structures such as Z-or I-bands, as well as their regulated expression during development or continuous contractile stimulation (20 -22). However, because of the relative ubiquitousness of their expression, most studies of HSP27 and ␣B-crystallin have been performed in non-muscle cells, and not enough attention has been paid to demonstrating the critical roles of sHSPs in muscle cells.
Previously, we identified a novel sHSP, MKBP, as a binding protein for myotonic dystrophy protein kinase (DMPK), the protein product of the gene responsible for myotonic dystrophy (23). This is the same protein as HSPB2, which was independently identified by Iwaki et al. (24) as the protein product of a gene located in the 5Ј upstream region of the ␣B-crystallin gene. We demonstrated that this MKBP/HSBP2 binds specifically to the kinase domain of DMPK, thus activating the kinase activity. MKBP/HSPB2 also shows a chaperone-like activity that protects the kinase from heat-induced inactivation. Importantly, the expression of MKBP/HSPB2, but not other sHSPs, is specifically up-regulated in the skeletal muscle of myotonic dystrophy patients as if to compensate for the reduced amount of DMPK. Together with the fact that DMPK knock-out mice develop a late-onset, progressive myopathy (25), these findings led us to propose that this kinase is involved in a stressresponse system in muscle cells by being a specific target of MKBP/HSPB2 (23). Furthermore, because MKBP/HSPB2 itself is localized not only at the neuromuscular junction where DMPK is concentrated, but also at the Z-band of myofibrils, we speculated that MKBP/HSPB2 also contributes to the maintenance of myofibril integrity by interacting directly with myofibrils independent of DMPK. Interestingly, in muscle cytosol, MKBP/HSPB2 forms an oligomeric complex separate from that containing HSP27 and ␣B-crystallin. Therefore, our previous work, which provided the first example of a correlation between a particular muscular dystrophy and a stress-response system in muscle cells, also suggested the presence of muscle sHSPs systems that are more complicated than previously expected.
Recently, the number of known mammalian sHSPs has rapidly risen to six (23, 24, 26 -28), but half of them have not been sufficiently characterized. Especially, there has been no work comparing them on the same basis. In this study, in an attempt to analyze the suggested sHSP systems in muscle cells, we carried out studies to characterize the five sHSPs expected to be expressed in muscles, that is, HSP27, ␣B-crystallin, p20, MKBP/HSPB2, and HSPB3. We first raised an antibody against HSPB3, the nucleotide sequence of which has been reported to be a possible novel mammalian sHSP (27,28), and identified the protein product as a 17-kDa protein. We then used antibodies and cDNAs to examine the tissue distribution, transcriptional regulation, mutual interactions, and cellular localization of all five sHSPs in cultured muscle cells. Based on the results, we suggest that the sHSP family may have diverged to form two independent chaperone systems specific for muscle cells.

EXPERIMENTAL PROCEDURES
cDNAs-Rat p20 and human HSPB3 cDNAs were amplified from a rat 3Y1 cell cDNA library (29) and a human skeletal muscle cDNA library (CLONTECH Laboratories, Inc., Palo Alto, CA), respectively, by PCR using appropriate synthetic oligonucleotide primers flanking the ORF of each protein. The preparation of the cDNAs for MKBP/HSPB2, ␣B-crystallin, and HSP27 has been described previously (23). For the expression of HSPB3 in COS1 cells, we constructed an expression vector, HSPB3/SRD, by inserting a 0.64-kb fragment of an HSPB3 cDNA clone (GenBank accession number U15590) containing 0.17 kb of the 5Ј-untranslated region and the whole ORF up to the predicted stop codon into an SRD mammalian expression vector (30).
Antibodies-Anti-human HSPB3 rabbit polyclonal antiserum was generated using a glutathione S-transferase-HSPB fusion protein as an antigen. Anti-human MKBP (c-2) and anti-rat p20 polyclonal antibodies have been reported previously (23,26). The antibodies were affinity purified prior to use. The anti-␣B-crystallin polyclonal antibody was purchased from Chemicon International Inc. (Temecula, CA), while the anti-HSP25 polyclonal antibody and anti-rat HSC70 polyclonal antibody were from StressGen Biotechnologies Corp. (British Columbia, Canada). The anti-human HSP27 monoclonal antibody and the antimouse desmin monoclonal antibody were purchased from Affinity Bioreagents Inc. (Colden, CO) and Zymed Laboratories Inc. (San Francisco, CA), respectively.
RNA Analysis-The tissue distribution of each sHSP mRNA was analyzed using a set of Human Multiple-tissue Northern blot membranes (CLONTECH). Total RNA from cultured cells was prepared using a Quick Prep Total RNA Extraction kit (Amersham Pharmacia Biotech). cDNA fragments corresponding to the whole ORF of each sHSP were radiolabeled with [␣-32 P]dCTP (1 ϫ 10 6 cpm/ml) and used to probe membranes to which the RNAs were blotted as described previously (23). The probes for myogenin and muscle creatine kinase were prepared as described previously (31). Hybridization was performed using ExpressHyb Hybridization Solution (CLONTECH) according to the manufacturer's instructions. The membranes were washed two times at 50°C for 20 min except for MKBP/HSPB2, which was washed at 53°C. Autoradiography was carried out at Ϫ70°C on Kodak x-ray film. When reprobing the membranes, the remaining radioactivity was removed by incubating the membranes in 5 mM Tris-HCl, pH 7.5, 2 mM EDTA, 0.1% SDS at 100°C for 10 min; radioactivity was confirmed to be negligible before the next hybridization.
RT-PCR was performed using Ready-To-Go RT-PCR beads (Amersham Pharmacia Biotech) according to the manufacturer's instructions. The primer sets for the mouse sHSPs were as follows: ATGTCGGGC-CGCACAGTGCC and CCTTCTCCGAAGCGCTGC for MKBP/HSPB2, ACTATGGCAAAAATCATTTTGAGG and CTTCTCCGTGGAGGCTG-AGT for HSPB3 (designed based on the published mouse EST sequence), ATGGACATCGCCATCCACCAC and TCTGAGAGTCCGGTGTC for ␣B-crystallin, and ACTGGGCATGGCCTTCCGTGT and TTACTCCTT-GGAGGCCATGT for glyceraldehyde-3-phosphate dehydrogenase.
Yeast Two-hybrid Analysis-Bait and prey plasmids were constructed by subcloning the whole ORF of each sHSP into pGBT9 or pGAD10 (CLONTECH Laboratories, Inc.), respectively, and an appropriate set of plasmids was simultaneously transformed into the yeast indicator strain HF7C by standard methods (32). After 4 days of growth at 30°C on selective culture plates lacking tryptophan and leucine, double transformants were replated onto histidine-lacking plates containing 10 mM 3-aminotriazole to examine the interaction of the introduced proteins. The results were also confirmed by examining the ␤-galactosidase activity of each clone with a standard filter assay method.
Cell Cultures-A subclone of C2C12 mouse myoblast (clone C2/4) was used in this study (33). C3H10T1/2 (clone 8) was purchased from American Type Culture Collection and 10TflagMyoD cells were generated by transfecting the pME-flagMyoD expression vector together with pSV2neo (34). All cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, with 400 g/ml G418 added in the case of 10TflagMyoD. Differentiation was induced by culturing the cells on type I collagen-coated tissue culture plates (Iwaki, Tokyo), and then switching the cells to serum-free ITS medium consisting of Dulbecco's modified Eagle's medium supplemented with 10 g/ml insulin, 5 g/ml transferrin, 10 nmol of selenite (Life Technologies, Inc., Grand Island, NY). For heat treatment, the cells were exposed to transitory heat shock by floating them on a water at appropriate temperatures for 15-60 min; the cells were then recovered by culturing at 37°C for appropriate times before harvest.
Preparation of Tissue Extracts for SDS-PAGE-Ten-week-old SD rats were fasted overnight and killed under ether anesthesia or by cardioectomy. Organs and tissues were excised and washed with icecold PBS, frozen immediately in liquid nitrogen, and then crushed into a powder with a CRYO-PRESS (Diatron, Ltd., Tokyo, Japan) precooled in liquid nitrogen. The powders were suspended in 10 volumes (v/w) of SDS sample buffer (2% SDS, 1 mM EDTA, 70 mM Tris-HCl, pH 6.7, 10% glycerol, 125 mM 2-mercaptoethanol, 0.02% bromphenol blue), homogenized with a POLYTRON homogenizer (KINEMATICA AG, Littau/ Luzern, Switzerland), and sonicated. A human skeletal muscle extract was prepared by similarly processing a small block of frozen limb muscle (biceps bracii) obtained for diagnostic purposes with informed consent. The samples were heat-denatured at 100°C for 5 min. Total protein concentrations in the samples were determined by spotting aliquots of each sample onto nitrocellulose membranes (Hybond-C extra, Amersham Pharmacia Biotech) and staining with Coomassie Brilliant Blue. The stained spots were quantified densitometrically using bovine serum albumin as a standard. The amounts of each sHSP in a certain volume of human skeletal muscle solubilized as described above were estimated by Western blot analysis based on standard data obtained using purified recombinant human sHSPs.
Electrophoresis and Western Blot Analysis-One-dimensional SDS-PAGE (12 or 15% polyacrylamide) was performed according to the method of Laemmli (35). The separated proteins were transferred onto polyvinylidene fluoride membranes (Millipore Corp., Bedford, MA) which were then soaked in 5% skimmed milk. Immunoblots were processed for immunoreaction as described previously (36), and visualized by a chemiluminescence ECL system (Amersham Pharmacia Biotech).
Immunohistochemistry-C2C12 cells were fixed directly on collagencoated dishes using 3% formaldehyde in PBS for 30 min at room temperature, and permeabilized with 0.1% Triton X-100 in PBS for 15 min. Detergent extraction was performed by treating the cells with extraction buffer (10 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 10 mM NaCl, 0.1 mM phenylmethylsulfonyl fluoride, 0.5% Triton X-100) for 10 min on ice prior to fixation. The dishes were soaked in PBS containing 10% calf serum, cracked into several pieces, and treated with primary antibodies for 45 min at 37°C in a moist chamber. The samples were washed three times with PBS containing 0.1% Tween 20, then similarly incubated with BODIPY-conjugated goat anti-rabbit IgG antibody (Molecular Probes, Inc., Eugene, OR). After washing, the cells were mounted with VECTASHIELD (Vector Laboratories, Inc., Barlingame, CA) and observed and photographed digitally under a fluorescence microscope (Olympus BX-50) equipped with a cooled CCD camera (Photometrics, East Britannia Drive, Tucson, AZ).

Identification of a Novel sHSP with a Molecular Mass of 17
kDa Corresponding to HSPB3 cDNA-Among the six members of the mammalian sHSP family so far reported, HSPL27 is the only one that has not been identified at the protein level (27). Several errors in the nucleotide sequence of the HSPL27 cDNA were recently corrected in the data base to predict an ORF with 150 amino acids corresponding to a protein with an estimated molecular mass of 16,965 kDa (GenBank accession number U15590). Because this is identical to the sequence of HSPB3 published by Boelens et al. (28) as a corrected version of the HSPL27 cDNA, we also use HSPB3 as a revised name for HSPL27, according to the suggestion of the HUGO Human Gene Nomenclature Committee. To confirm the expression of endogenous HSPB3, we raised an anti-HSPB3 rabbit polyclonal antibody using a recombinant protein as an antigen, and detected a band with the predicted 17 kDa mass in human skeletal muscle (Fig. 3a). The migration of this band was identical to that of the protein product of the HSPB3 cDNA ectopically expressed in COS1 cells (Fig. 3a, lane 6; see "Experimental Procedures"). Therefore, we conclude that the corrected amino acid sequence for HSPB3 represents a novel sHSP with a molecular mass of 17 kDa expressed in vivo. Northern blot analysis of the poly(A) ϩ RNA from human tissues (Fig. 2a) revealed a consistent 0.84-kb transcript for HSPB3 specifically in skeletal muscle and heart (see below), although longer ex-posure produced an extremely faint band of 3.7-kb in the same tissues.
The alignment of the amino acid sequence of human HSPB3 with those of the other human sHSPs so far identified revealed it to be the most diverged member of the family (28), showing ϳ40% amino acid identity in the ␣-crystallin domain (Fig. 1a), which is the most highly conserved region among family members (5). In addition, it shows no similarity in the N-terminal region (hatched bar in Fig. 1b) where another short conserved stretch is detected in other family members.
Muscle Is the Only Tissue That Simultaneously Expresses all Five sHSPs-As a first step in investigating the physiological functions of mammalian sHSPs, we performed comparative studies of their tissue distributions by Northern blot as well as Western blot analyses. Although there have been several studies in which the tissue distributions of three sHSPs, HSP27, ␣B-crystallin, and p20, were examined individually (18,19,26,(37)(38)(39), this is the first study that examines on the same basis all known members other than lens-specific ␣A-crystallin. Fig.  2b shows a blot of 2-g samples of poly(A) ϩ RNA isolated from various human tissues probed successively with the cDNAs for each sHSP. To compare the relative tissue distribution of each sHSP, the exposure time for each autograph was normalized so that the signal for each sHSP in heart was almost equal. Overall, the mRNAs of all sHSPs examined show clear constitutive expression in heart and skeletal muscle. As for tissue distribution, the mRNA for HSP27 is observed most broadly with the richest expression in skeletal muscle and heart. The mRNAs of three sHSPs, ␣B-crystallin, MKBP/HSPB2, and p20, show similar distribution patterns, with significant accumulations in skeletal muscle and heart. Other than muscle, they are also seen in prostate, ovary, intestine, and colon. As shown previously, ␣B-crystallin mRNA is also expressed in kidney and brain (37). Compared with these sHSPs, the expression of HSPB3 mRNA is unusual in its exclusive expression in striated muscles; it is also unique in its preferential expression in heart. Fig. 3 shows the results of Western blotting using polyclonal antibodies specific to each sHSP. As described above, the newly generated antiserum against HSPB3 detected a band of the predicted 17 kDa mass in human skeletal muscle (Fig. 3a, lane  5). The size of the band is clearly distinct from those of other sHSPs, suggesting that the antiserum is specific for HSPB3 and does not cross-react with other sHSPs. Fig. 3b shows the tissue distribution of each sHSP protein in rat tissue extracts  (50). b, schematic diagram comparing the overall structures of mammalian sHSPs. Notice that HSPB3 not only shows a completely diverged N-terminal sequence, but also has the shortest C-terminal tail following the ␣-crystallin domain. examined using these antibodies. Here, we confirm the ubiquitous expression of HSC70, the constitutive form of the HSP70 family, on our parallel blot (39). The exposure times were again normalized to the signal intensity in heart. The results are consistent with those of Northern blot analysis, although the unique tissue distribution pattern of each sHSP was demonstrated more clearly here than by Northern blot. This is due to the relatively increased signal intensities for HSP27, ␣B-crystallin, and p20 in non-muscle tissues, especially lung, but not for MKBP/HSPB2 and HSPB3. Another discrepancy with the Northern blot data is the appearance of the prominent band in prostate (asterisk at the top of Fig. 3b) detected by the anti-HSPB3 antiserum. Although the antiserum shows a lower titer and specificity for rat HSPB3, and the band indicates a smaller molecular mass (15 kDa) than for HSPB3 in muscles, we cannot exclude the possibility this band might represent an HSPB3-related protein expressed in rat prostate.
It should be noted that, although detected broadly, HSP27 expression shows a clearly restricted tissue distribution pattern as shown previously (38,39). On the other hand, ␣Bcrystallin and p20 show more limited distributions, occasionally discordant in, for example, kidney, prostate, and colon (18,26,38). However, there is an apparent trend for HSP27 to be expressed abundantly in tissues where ␣B-crystallin or p20 accumulate, implying some correlation between their expressions. The other two sHSPs, MKBP/HSPB2 and HSPB3, show more specific expressions qualitatively distinct from those of HSP27, ␣B-crystallin, and p20. Their expressions are rather specific to heart and skeletal muscles, and consistent with the results of Northern blot, the expression of HSPB3 is more skewed to heart.
In summary, the results indicate that all five sHSPs are constitutively expressed in restricted tissues, and there is a unique hierarchy in their tissue distribution patterns. The  1-5). The 30-kDa band observed in lane 3 is a nonspecific band that reacted with the anti-␣B-crystallin polyclonal antibody. Note that the 17-kDa protein that reacts with the anti-HSPB3 antibody is clearly distinct from bands with any other sHSP. This protein shows the same migration as exogenously expressed HSPB3 in COS1 cells (lane 6). In lane 7, COS1 cell extracts without the expression vector were loaded as a negative control. b, tissue-specific expression of each sHSP protein. Equal amounts of proteins extracted from adult rat tissues (10 g for HSP27, ␣B-crystallin, and HSC70, and 20 g for MKBP/HSPB2 and HSPB3) were separated by 15% SDS-PAGE and, after membrane transfer, subjected to immunoreaction with the indicated antibodies. The asterisk indicates a prominent cross-reactive band for the anti-HSPB3 antiserum observed in rat prostate with a smaller molecular mass (15 kDa). Coomassie Brilliant Blue staining confirmed that weak bands at the edge of the top panel for HSPB3 observed in spleen, thymus, and prostate arise from nonspecific staining of condensed proteins (data not shown). most significant result of this hierarchy is that, as far as we have examined, heart and skeletal muscle are the only tissues in which the five sHSPs commonly show constitutive, abundant expression. Using purified recombinant proteins for each sHSP as standards, we estimated the protein concentrations of HSP27, MKBP/HSPB2, ␣B-crystallin, and HSPB3 in human skeletal muscle (biceps bracii, which contains an almost even population of three cell types, typeI, IIA, and IIB (40)) 3 as 3.4, 0.30, 0.57, and 0.09 g/mg of protein, respectively.
The Expressions of ␣B-crystallin, MKBP/HSPB2, and HSPB3, but Not HSP27 and p20, Are Induced during the Initial Phase of Skeletal Muscle Differentiation-The abundant and specific expressions of the five sHSPs in skeletal muscle suggest that their expressions may be controlled by myogenic factors such as MyoD. Therefore, we next examined the change in the amount of each sHSP mRNA during the differentiation of mouse C2C12 myoblast cells induced by serum starvation (Fig. 4a). HSP27 and ␣B-crystallin mRNAs were easily detected in myoblasts (time ϭ 0), whereas only very weak signals could be detected for MKBP/HSPB2 and p20 after very long exposures. HSPB3 mRNA was not detected at all. Under our conditions, the induction of myogenin mRNA, which is the earliest known event accompanying myogenic differentiation (41), is observed within 7 h after serum deprivation. After that, the mRNA for muscle creatine kinase, a muscle-specific protein, begins to up-regulate within 15 h. Morphologically, 36 h after serum starvation, the cells begin to fuse to form multinucleated myotubes. Northern blot and RT-PCR analysis of the total RNA from differentiating C2C12 cells at the indicated times reveal the following features of sHSP expression (Fig. 4, a and c): 1) the amount of HSP27 mRNA remains unchanged during differentiation; 2) the mRNA for ␣B-crystallin is upregulated in conjunction with the induction of myogenin mRNA; 3) the mRNAs for MKBP/HSPB2 and HSPB3 are induced at a later stage of differentiation; and 4) the amount of p20 mRNA decreases somewhat during differentiation. Fig. 4, b and c, establish that the increase in ␣B-crystallin, MKBP/HSPB2, and HSPB3 mRNA is the result of myogenic differentiation controlled by the myogenic regulatory factor MyoD, and not due to stress stimulation caused by serum starvation. 10TflagMyoD cells (10TMyoD), which are stably transformed to myoblastic cells by the ectopic expression of MyoD, also show the induction of ␣B-crystallin and MKBP/ HSPB2 mRNA upon serum depletion, while the parental Swiss C3H10T1/2 fibroblast cells (10T1/2) do not. Because of the low amount of the HSPB3 mRNA, its induction is difficult to detect by Northern blot analysis (Fig. 4b); however, RT-PCR analysis clearly demonstrates the induction of HSPB3 in parallel with 3 R. Akutsu, Y. Sugiyama and A. Suzuki, unpublished results. 4. MKBP/HSPB2, HSPB3, and ␣B-crystallin mRNA expressions are regulated under the control of a myogenic factor, MyoD, and induced along with the myogenic differentiation of C2C12 myoblasts. a, changes in the expression of each sHSP mRNA during the course of mouse C2C12 myoblast differentiation. Differentiation was induced by serum withdrawal from the culture medium at time 0. Ten micrograms of total RNA isolated from C2C12 cells at the indicated times was subjected to Northern blot analysis using the cDNA fragments indicated on the left. b, MKBP/HSPB2 and ␣B-crystallin mRNA levels are regulated by the myogenic factor, MyoD. C3H10T1/2 mouse fibroblast cells were stably transformed to myoblasts by the exogenous introduction of MyoD. The transformant (10TMyoD) and host (10T1/2) cells were cultured either in growth (G) or differentiation medium for 24 or 48 h. Ten micrograms of total RNA from each cell was subjected to Northern blot analysis. In lane C, RNA from differentiated C2C12 cells was loaded as a positive control. Note that the mRNAs for MKBP/HSPB2 and ␣B-crystallin are specifically induced in 10TMyoD cells by serum starvation, while a reduction in p20 mRNA is observed in both cells. The signal for HSPB3 mRNA was too weak to analyze, whereas HSP27 mRNA was not detected in these cells. c, the results of RT-PCR analysis showing the induction of MKBP/HSPB2 and HSPB3 mRNAs during myogenic differentiation. One hundred nanograms of total RNA extracted from the indicated mouse cells on the indicated day was processed by reverse transcription (42°C, 30 min) followed by PCR using specific sets of primers for each sHSP. PCR cycle numbers were 30 for MKBP/HSPB2 and HSPB3 and 24 for ␣B-crystallin and glyceraldehyde-3-phosphate dehydrogenase (GAPDH). The products were visualized by ethidium bromide staining. d, the induction of sHSP proteins during the course of C2C12 differentiation. Ten micrograms (for HSP27, p20, and desmin) or 20 g (for MKBP/HSPB2, ␣B-crystallin, HSPB3, and HSC70) of protein from cell cultures prepared on the indicated day was separated by SDS-PAGE and immunoblotted with the antibodies indicated on the left. MKBP/HSPB2 mRNA in 10TMyoD cells (Fig. 4c). It should also be noted that ␣B-crystallin is up-regulated earlier than the induction of MKBP/HSPB2 and HSPB3 in these cells, too. A reduction in p20 mRNA was observed even in the parental cells, suggesting that this reduction may be due to serum depletion (Fig. 4b).
Consistent results were obtained by Western blot analyses of C2C12 cell extracts during the course of differentiation (Fig.  4d). Again, HSP27 is the predominant sHSP expressed in myoblasts, although ␣B-crystallin and p20 are also present in lesser amounts. During the course of differentiation, ␣B-crystallin starts to up-regulate, while the amount of p20 gradually decreases. Subsequently, MKBP/HSPB2 and HSPB3 are induced during the late phases of differentiation.
Selective Mutual Interactions between sHSP: Diverged Hetero-oligomeric States of sHSPs in Muscle Cells-A characteristic feature of sHSPs that has been conserved throughout evolution is their tendency to form large aggregates in the cytosol through their homo-and hetero-oligomeric activities (19,26,42). Fig. 5 summarizes the results of two-hybrid assays monitoring all combinations of interactions between pairs of the five sHSPs. First, the panel shows that all sHSPs except HSPB3 show homophilic interaction, although the activity in the case of p20 is relatively low. Second, each sHSP shows selective heterogeneous interactions with other sHSP members: as shown previously, HSP27 and ␣B-crystallin interact with each other, while MKBP/HSPB2 interacts with neither HSP27 nor ␣B-crystallin (23). On the other hand, p20 unexpectedly interacts with MKBP/HSPB2 as well as with HSP27 and ␣B-crystallin, raising the possibility that p20 is involved not only in the complex formed by HSP27 and ␣B-crystallin, but also that formed by MKBP/HSPB2. Another intriguing result is that HSPB3, unlike other sHSPs, does not show a homophilic interaction, but interacts exclusively with MKBP/HSPB2. Therefore, it is speculated that HSPB3 represents a very intimate partner of MKBP/HSPB2 in vivo.
The selective mutual interactions of mammalian sHSPs expressed simultaneously in muscle cells were further confirmed by gel filtration analysis of muscle cytosol combined with immunoprecipitation. Fig. 6a shows the elution pattern of the five sHSPs by gel filtration of the rat heart soluble fraction. Similar results were also obtained in analyses of rat skeletal muscle extracts (data not shown). As demonstrated previously, HSP27 together with p20 forms two broad peaks with apparent molecular masses around Ͼ500 and 50 kDa, while ␣B-crystallin co-elutes only with the former peak (26). On the other hand, MKBP/HSPB2 elutes independently of these two oligomeric complexes, forming one broad peak with an apparent molecular mass of 150 kDa (23). We examined the elution profile of HSPB3 and detected it in the same fractions as MKBP/HSPB2. This is consistent with the results in Fig. 5, and suggests that MKBP/HSPB2 forms a 150-kDa complex with HSPB3. Subsequent immunoprecipitation analysis against the above fractions using the anti-MKBP/HSPB2 as well as anti-p20 antibodies finally established the complex formation of sHSPs in muscle cells (Fig. 6, b and c): as shown in Fig. 6b, HSPB3, but no other sHSP, including p20, co-precipitates with the anti-MKBP/HSPB2 antibody from fractions corresponding to the MKBP/HSPB2 peak, confirming the oligomerization of MKBP/ HSPB2 and HSPB3 in the 150-kDa complex. On the other hand, as shown in Fig. 6c, immunoprecipitation by the anti-p20 antibody further supports the absence of p20 from this complex. These results indicate that, based on their oligomerization properties, the five mammalian sHSPs can be categorized into two groups, one comprising HSP27, ␣B-crystallin, and p20, and the other MKBP/HSPB2 and HSPB3, and that the two groups form mutually exclusive oligomers in the soluble fraction of muscle cells.
Besides the dominant forms of sHSP oligomers described above, Fig. 6 also suggests the presence of heterogeneity among sHSP oligomers. Especially, because MKBP/HSPB2 was detected faintly in the anti-p20 antibody immunoprecipitate from the 50-kDa fraction (Fig. 6c), a very minor fraction of p20 may associate with MKBP/HSPB2 rather than HSP27. Although this association could not be confirmed by reciprocal immunoprecipitation with anti-MKBP/HSPB2 antibody, probably because of the difference in the antibody titers (Fig. 6b), the results may reflect the interaction between MKBP/HSPB2 and p20 suggested by the two-hybrid assay (Fig. 5). In fact, coprecipitation of very minor fractions of p20 and MKBP/HSPB2 was detected in the anti-MKBP/HSPB2 or anti-p20 antibody precipitates, respectively, when the total extract was used as a starting material (Fig. 6d). As shown in Fig. 6d, ␣B-crystallin, but not HSP27, was detected in the anti-MKBP/HSPB2 antibody precipitate. Because the two-hybrid assay did not detect an interaction between ␣B-crystallin and MKBP/HSPB2, this finding suggests the possible presence of a tertiary complex of ␣B-crystallin, p20, and MKBP/HSPB2, in which ␣B-crystallin interacts with MKBP/HSPB2 via p20.
HSP27 and ␣B-crystallin, but Not MKBP/HSPB2, Are Localized on Well Developed Actin Bundles Specific to Myotubes-The exclusive properties of the two groups of sHSPs observed in muscle cytosol suggests the presence of two independent sHSP systems with divergent functions. Considering that sHSPs are assumed to work as molecular chaperones (6 -8), sHSPs in different systems may have distinct molecular targets. We previously demonstrated that MKBP/HSPB2 interacts with DMPK and thus activates and/or stabilizes its activity by way of its chaperone-like activity (23). Fig. 5 indicates that, among the five sHSPs, MKBP/HSPB2 is the only one that interacts with DMPK, suggesting that the effect on DMPK is specific for the system containing MKBP/HSPB2. On the other hand, recent studies have demonstrated that one of the specific targets of HSP27 and ␣B-crystallin may be the actin-based cytoskeleton (12,14,16). Therefore, we next examined the cellular localization of three sHSPs, HSP27, ␣B-crystallin, and MKBP/ HSPB2, in differentiated C2C12 myotubes in view of their relationship with the actin cytoskeleton (Fig. 7). Even under differentiation conditions, not all myoblast cells differentiate into myotubes; some remain mononuclear and can be discriminated morphologically as well as by their expression of marker proteins such as troponin T (33). Consistent with the results of FIG. 5. The five members of the mammalian sHSP family show highly selective interactions with each other. All possible interactions between two of the five sHSPs, as well as the interaction of each sHSP with DMPK, were examined using yeast two-hybrid assays. Growth on a -ULWH plate (right) lacking uracil, leucine, tryptophan, and histidine indicates interaction. 3-AT is 3-aminotriazole, which was added to suppress the background growth of yeast.
Northern and Western blot analyses, anti-HSP27 antibody stains myoblasts in growth medium (Fig. 7A) as well as multiand mononuclear cells under differentiation conditions (Fig.  7B), while anti-␣B-crystallin and anti-MKBP/HSPB2 antibody stain only differentiated multinuclear cells (C and D; data not shown). Without any pretreatment before fixation, these sHSPs all appear dispersed in the cytoplasm showing no localization to specific structures, although the staining of MKBP/ HSPB2 appears more particulate (B, C, and D). However, pretreatment of cells with 0.5% Triton X-100 before fixation  7. HSP27 and ␣B-crystallin, but not MKBP/HSPB2, are localized on myotube-specific actin bundles. C2C12 myoblasts were cultured in growth medium (A, E, and I) or differentiated into myotubes by culturing in differentiation medium for 4 days. The cells were fixed in 3% formaldehyde with (E-L) or without (A-D) pre-extraction with 0.5% Triton X-100. The cells were immunostained for HSP27 (A, B, E, and F), ␣B-crystallin (C and G), or MKBP/HSPB2 (D and H). For E-H, actin filaments were also doubly stained with rhodamine-phalloidin (I-L, respectively). Note that the Triton X-100 extraction of cells cultured at physiological temperature removes most cytoplasmic signals and clearly reveals the localization of HSP27 and ␣B-crystallin on actin bundles in myotubes (compare with F and J, and G and K).
removes most of the cytoplasmic staining and allows the localization of HSP27 and ␣B-crystallin to be clearly visualized on filaments running along the long axis of myotubes (F and G). Rhodamine-phalloidin staining revealed these to be actin bundles (J and K). Importantly, the localization of HSP27 on actin filaments is scarcely observed in undifferentiated myoblasts, although they express HSP27 abundantly and develop many stress fibers (E and I). This suggests that the actin localization of HSP27 and ␣B-crystallin is specific to myotubes, which contain well developed actin bundles as a prototype of myofibril structures. On the other hand, no filamentous localization of MKBP/HSPB2 could be detected even in Triton X-100 pretreated C2C12 myotubes (H and L). This provides another example of the functional divergence of sHSPs involved in different complexes.
In Contrast to HSP27 and ␣B-crystallin, the mRNAs for p20, MKBP/HSPB2, and HSPB3 Are Not Induced by Heat Shock-Finally, we examined the heat-shock response of the expression of each sHSP. Previously, we demonstrated that the amounts of HSP27 and ␣B-crystallin mRNAs dramatically increase in differentiated C2C12 myotubes after heat shock, whereas the amount of MKBP/HSPB2 mRNA does not (23). Fig. 8 expands this observation to the other sHSPs, p20 and HSPB3, as well as to undifferentiated myoblasts. At first, it was shown that even in C2C12 myoblasts (Fig. 8a), the heat-induced accumulation of HSP27 and ␣B-crystallin mRNAs occurs to an extent similar to that observed in myotubes (Fig. 8, b-d). In contrast, the mRNAs for MKBP/HSPB2, p20, and HSPB3 were show no significant heat inducibility in either cell type even under more severe conditions (Fig. 8, a-d). DISCUSSION The recent expansion of the mammalian sHSP family provides a new avenue to understanding the physiological significance of this poorly understood HSP family. In the present study, we for the first time examined several essential features of five mammalian sHSPs so far identified, all except the lensspecific ␣A-crystallin, on the same basis using collected cDNAs and specific antibodies for each sHSP. In the course of this study, we confirmed the expression of the protein product of the HSPB3 gene previously reported in human heart as a 17-kDa protein (27,28). The accumulated data reveal the unique features of this HSP family.
The Mammalian sHSP Family May Have Diverged to Form Specific Chaperone Systems for Skeletal Muscle and Heart-Originally, HSPs were defined as proteins whose synthesis is induced by heat or other physiological stresses (43). However, subsequent work has revealed that most HSPs are also constitutively expressed in various tissues (1). Tanguay et al. (39) have suggested that HSP90 and HSC70, which are thought to play housekeeping roles for proper cell growth and differenti-ation, are expressed rather ubiquitously, whereas the levels of HSP70 and HSP27 are specifically high in organs that are subjected to more immediate environmental aggression. Here, we show that the tissue distribution of HSP27 is broadest among family members, and that the other four sHSPs display more limited tissue distributions showing a unique hierarchy. Considering the suggested cytoprotective roles of HSP27 and ␣B-crystallin against physiological stress (11)(12)(13), the constitutive expressions of these sHSPs in restricted tissues suggest that, in addition to other HSPs with higher molecular mass, the cells in these tissues need to be protected by sHSPs continuously or without time consuming protein synthesis (38). The observed hierarchy of sHSP expression further suggests that HSP27 is the most essential, while the other members provide additional responses to different kinds of stress or protect different molecular targets specific to certain tissues.
The most important implication of the present study is that all five sHSPs are abundantly and rather specifically expressed in skeletal muscle and heart. In other words, the tissues where the five sHSPs are commonly expressed are confined to these muscles. This is a very significant feature peculiar to this HSP family. Furthermore, among the five sHSPs, the muscle specificity of MKBP/HSPB2 and HSPB3 expression is significantly and qualitatively different from that of the others. In addition, they are sharply induced along with the myogenic differentiation controlled by MyoD, suggesting that they represent the first examples of muscle-specific HSPs. Interestingly, consistent with these expression properties, the five mammalian sHSPs can be categorized into two basic groups, one comprising HSP27, ␣B-crystallin, and p20, and the other MKBP/HSPB2 and HSPB3, based on their oligomerization properties. Each group forms mutually exclusive oligomers, suggesting the presence of two independent sHSP systems in muscle cells. Taken together, the present results indicate that muscle cells develop highly sophisticated, unique sHSPs in two ways. First, they express the three ubiquitous forms of sHSP, HSP27, ␣B-crystallin, and p20, simultaneously in large amounts and let them form co-oligomers (␣B-crystallin/HSP27/p20 and HSP27/p20) unique to muscle cells. Second, they also express another independent sHSP system comprising MKBP/HSPB2 and HSPB3, and this system is scarcely observed in other types of cells (Fig. 9). Because MKBP/HSPB2 expression is detected in intestine, colon, and uterus (Fig. 3) 4 and Boelens et al. (28) showed Northern blot data suggesting the expression of HSPB3 in smooth muscle, this latter system may also participate in stress response in smooth muscle cells.
The close relationship of sHSPs with cytoskeletal components has been well demonstrated: HSP27 and ␣B-crystallin 4 Y. Sugiyama and A. Suzuki, unpublished results.  a and b, 44°C 15 min; c, 44°C 1 h; and d, 46°C 15 min), then recovered at 37°C for the indicated times. Ten micrograms of total RNA extracted from each cell was subjected to Northern blot analysis using the indicated probes. In lanes 1 and 6 of each panel, the total RNAs from untreated cells harvested in parallel at 1 and 8 h after the heat shock, respectively, were loaded. Asterisks indicate the signal on 18 S ribosomal RNA detected nonspecifically with a MKBP/HSPB2 probe.
have been suggested to interact with actin and intermediate filaments in vitro (15)(16)(17). In addition, it has been suggested that their cytoprotective role in non-muscle cells can be attributed, at least in part, to their ability to stabilize cytoskeletal structures such as stress fibers (12,14). On the other hand, HSP27, ␣B-crystallin, and MKBP/HSPB2 have been shown to be localized on specific sarcomeric structures of the skeletal or cardiac muscle myofibril, such as Z-bands or I-bands (16, 20 -23, 44). Together with these results, the present data strongly suggest that one of the essential functions of the mammalian sHSP family is to maintain and/or protect cytoskeletal structures. In striated muscle cells, which are frequently subjected to severe conditions of heat and oxidative and mechanical stresses especially during exercise (2), the sHSP family may develop to maintain constitutively the extremely developed cytoskeletal structure, myofibrils, in which many proteins, including filamentous oligomers, are precisely assembled. In this respect, it is very interesting that, in contrast with other HSPs such as HSP70 or HSP90, the amino acid sequences of the mammalian sHSPs are highly diverged from those of sHSPs found in prokaryotes and yeast (5,45). No orthologs of the individual mammalian sHSPs are found even in Caenorhabditis elegans. Therefore, the gene duplications that resulted in the sophisticated sHSP systems observed in mammalian muscle cells might not have progressed rapidly until the early stages of vertebrate evolution when muscle cells developed to produce stronger and more sustained contractile forces. Although more sequence information about sHSPs from other species is needed, it is very interesting to clarify which of the two independent mammalian sHSP systems developed first.
Recent work showing correlations with specific muscular dystrophies supports the critical importance of these myoprotective systems composed by diverged sHSP family members. We identified MKBP/HSPB2 as a binding protein for DMPK, activating and/or stabilizing the kinase activity through its specific chaperone-like activity (23). Since this kinase is considered to be important in the maintenance of myofibril integrity (25), we proposed that one of the causes of the disease is a defect in the muscle-specific, stress-responsive system composed of DMPK and MKBP/HSPB2. More directly, a missense mutation in the ␣B-crystallin gene was recently identified as the cause of desmin-related myopathy (46). Since this disease is characterized by a delayed accumulation of desmin aggregates, an essential component of Z-bands, this result strongly supports the idea that the chaperone activity of sHSPs is crucially important for maintaining myofibril structures. Although the molecular pathogenesis of these diseases should be clarified further, the results suggest that a defect in sHSP systems leads to a gradual accumulation of damage, which finally results in the late onset muscle degeneration observed in myotonic dystrophy and desmin-related myopathy.
Functional Divergence of the Two sHSP Systems in Muscle Cells-Several studies have demonstrated that the expression of ␣B-crystallin is regulated during skeletal and cardiac muscle development, and that its chaperone-like functions are also coupled to the activation of genetic programs responsible for myogenic differentiation and cardiac morphogenesis (47,48). Here, using a mouse myoblast cell line, C2C12, as well as 10TflagMyoD cells, which stably transform to myoblasts, we demonstrated the up-regulation of ␣B-crystallin expression during the very early stages of myogenic differentiation. Furthermore, we found the localization of ␣B-crystallin as well as HSP27 on well developed actin bundles in myotubes (Fig. 7). Importantly, this HSP27 localization is not observed in myoblasts, suggesting the possible involvement of these sHSPs in the initial organization of myofibril assembly in myotubes. They may interact with some proteins on actin bundles whose expressions themselves are induced along with myogenic differentiation. Interestingly, MKBP/HSPB2, which belongs to the other sHSP system in muscle cells, does not show a similar localization on actin bundles, although it localizes on Z-bands similarly to ␣B-crystallin in mature muscle cells. This suggests that the roles of MKBP/HSPB2 during the early stages of muscle differentiation may differ from those of HSP27 and ␣B-crystallin. In fact, the expressions of MKBP/HSPB2 and HSPB3 are more tightly regulated by the myogenic program and induced later than the up-regulation of ␣B-crystallin. The functional divergence between the two independent sHSP systems may be based on their specific interactions with distinct molecular targets, one of which is illustrated in the interaction between DMPK and MKBP/HSPB2.
Finally, we also demonstrate that ␣B-crystallin and HSP27 are the only sHSPs whose mRNAs are induced by heat shock. This further indicates that the sHSP system comprising MKBP/HSPB2 and HSPB3 may work dominantly under normal conditions. Although categorized into the group with ␣B-FIG. 9. Schematic model for chaperone systems comprising diverse members of the mammalian sHSP family. In addition to the system consisting of HSP27, ␣B-crystallin, and p20, which works rather ubiquitously, muscle cells develop an additional system consisting of MKBP/HSPB2 and HSPB3 to maintain the sophisticated cytoskeletal structure, myofibrils (see the text for details). Notice that the stoichiometry of each sHSP oligomer is illustrated arbitrarily. crystallin and HSP27, p20 mRNA shows no heat inducibility. Together with its unique response to serum starvation (Fig. 4), this sHSP may represent a variation of this group.
Complexity of sHSPs Oligomerization-The present results obtained by two-hybrid analysis monitoring the interaction between pairs of sHSPs are essentially consistent with the biochemical data from gel filtration analysis. This means that the distinct oligomerization of the five sHSPs observed in vivo arises from their highly selective interactions with each other (Fig. 6). To determine which regions are responsible for these selective interactions will be the focus of future investigations. The present results suggest that the unique nature of HSPB3 can be attributed to its sequence, which lacks the N-terminal conserved region and the C-terminal tail.
It should also be noted that the present analyses do not allow a precise discussion of the stoichiometry of each oligomer, and that there may be several variations in the composition of sHSP oligomers. In addition, our oligomerization data were all obtained with the soluble fraction of muscle cells. As has been suggested for ␣B-crystallin, HSP27, and p20 (26), sHSPs show an immediate redistribution from the cytosol to the insoluble fraction in response to several kinds of stress. We have observed that MKBP/HSPB2 and HSPB3 show similar responses to heat shock (23). 4 Because this redistribution of sHSPs may be an emergent response of the cells to protect themselves, identifying the oligomerization state and partners of each sHSP in the insoluble fraction is extremely important for understanding sHSP function. Interestingly, this early response is accompanied by the phosphorylation on the certain residues on HSP27 (for example, Ser-15 and Ser-90 in Chinese hamster HSP27 (49)) and dissociation of the large complex made by ␣B-crystallin, HSP27, and p20, suggesting that stressful conditions induce modification of the interactions between sHSPs. This suggests the intriguing possibility that the interaction between p20 and MKBP/HSPB2 detected in vitro (in two-hybrid assay), but not very strongly in the soluble fraction of skeletal muscle and heart, might operate during this dynamic process under stress providing cross-talk between novel sHSP members. Thus this interaction may exist and contribute to the regulation of the system in muscle cells. Our results provide a foundation for further studies to understand the complexity and dynamics of the sHSP stress-response system.