cGMP-mediated phosphorylation of heat shock protein 20 may cause smooth muscle relaxation without myosin light chain dephosphorylation in swine carotid artery

Abstract

1. Nitrovasodilators such as nitroglycerine, via production of nitric oxide and an increase in [cGMP], can induce arterial smooth muscle relaxation without proportional reduction in myosin light chain (MLC) phosphorylation or myoplasmic [Ca²⁺]. These findings suggest that regulatory systems, other than MLC phosphorylation and Ca²⁺, partially mediate nitroglycerine-induced relaxation.

2. In swine carotid artery, we found that a membrane-permeant cGMP analogue induced relaxation without MLC dephosphorylation, suggesting that cGMP mediated the relaxation.

3. Nitroglycerine-induced relaxation was associated with a reduction in O₂ consumption, suggesting that the interaction between phosphorylated myosin and the thin filament was inhibited.

4. Nitroglycerine-induced relaxation was associated with a 10-fold increase in the phosphorylation of a protein on Ser¹⁶. We identified this protein as heat shock protein 20 (HSP20), a member of a family of proteins known to bind to thin filaments.

5. When homogenates of nitroglycerine-relaxed tissues were centrifuged at 6000 g, phosphorylated HSP20 preferentially sedimented in the pellet, suggesting that phosphorylation of HSP20 may increase its affinity for the thin filament.

6. We noted that a domain of HSP20 is partially homologous to the ‘minimum inhibitory sequence’ of skeletal troponin I. The peptide HSP20_110-121, which contains this domain, bound to actin-containing filaments only in the presence of tropomyosin, a characteristic of troponin I. High concentrations of HSP20_110-121 abolished Ca²⁺-activated force in skinned swine carotid artery. HSP20_110-121 also partially decreased actin-activated myosin S1 ATPase activity.

7. These data suggest that cGMP-mediated phosphorylation of HSP20 on Ser¹⁶ may have a role in smooth muscle relaxation without MLC dephosphorylation. HSP20 contains an actin-binding sequence at amino acid residues 110–121 that inhibited force production in skinned carotid artery. We hypothesize that phosphorylation of HSP20 regulates force independent of MLC phosphorylation via binding of HSP20 to thin filaments and inhibition of cross-bridge cycling.

Phosphorylation of the 20 kDa regulatory light chain of myosin (MLC) on Ser¹⁹ is the primary determinant of myosin's ATPase activity and contraction in arterial smooth muscle (Hai & Murphy, 1989). Most, but not all, stimuli induce a single steady-state relation between MLC phosphorylation and force (Ratz et al. 1989). One exception to regulation of force by MLC phosphorylation is the sustained phase of nitrovasodilator-induced relaxation, i.e. relaxation induced by cyclic 3′:5′-guanosine monophosphate (cGMP). Initially, the nitrovasodilator nitroglycerine relaxed swine carotid artery by reducing both myoplasmic [Ca²⁺] ([Ca²⁺]_i) (McDaniel et al. 1992) and the [Ca²⁺]_i sensitivity of MLC phosphorylation (Van Riper et al. 1997). However, after 10 min of nitroglycerine application, [Ca²⁺]_i and MLC phosphorylation increased back to levels similar to those observed without addition of nitroglycerine (McDaniel et al. 1992; Bárány & Bárány, 1993). These data suggest that the sustained phase of nitroglycerine-induced relaxation was no longer regulated by [Ca²⁺]_i or MLC phosphorylation. In essence, nitroglycerine induces relaxation despite the presence of activated myosin, i.e. myosin with phosphorylated MLC.

We tested several hypotheses: (1) nitroglycerine-induced relaxation without proportional MLC dephosphorylation results from increases in [cGMP]; (2) nitroglycerine-induced relaxation without proportional MLC dephosphorylation results from inhibition of myosin attachment to thin filaments; and (3) nitroglycerine-induced relaxation without proportional MLC dephosphorylation results from phosphorylation of a specific protein that interacts with the thin filament.

In this study, we provide evidence in support of all three hypotheses. We also confirmed the finding of C. M. Brophy's laboratory (Beall et al. 1997) that nitroglycerine-induced relaxation was associated with phosphorylation of heat shock protein 20 (HSP20).

METHODS

Tissues, myosin phosphorylation assay and cGMP assay

Swine common carotid arteries were obtained from a slaughterhouse. Physiological saline solutions, dissection of medial strips, mounting and determination of the optimum length for stress development at 37°C were performed as described (Rembold & Murphy, 1988). The intimal surface was mechanically rubbed to remove the endothelium. Phosphorylation of the smooth muscle-specific 20 kDa myosin light chain was estimated in tissues mounted isometrically for force measurement. Tissues were frozen and homogenized as described (Rembold & Murphy, 1988) and the percentage MLC phosphorylation was determined by the acidic shift of the MLCs on two-dimensional electrophoretic separation (Rembold & Murphy, 1988; McDaniel et al. 1991). After Coomassie Blue staining and optical scanning, phosphorylation was estimated as moles inorganic phosphate (P_i) per mole total smooth muscle-specific MLC (mol P_i (mol MLC)⁻¹). cGMP and cyclic 3′:5′-adenosine monophosphate (cAMP) were assayed by radioimmunoassay after homogenization in 0.1 n HCl as described (McDaniel et al. 1991).

³²PO₄ loading

Isometrically mounted swine carotid medial tissues were loaded with ³²PO₄ for 5 h (Van Riper et al. 1995). After washout of ³²PO₄ for 1 h, the tissues were pharmacologically treated (e.g. histamine or histamine plus nitroglycerine). Then the tissues were frozen in an acetone-dry-ice slurry and homogenized in 0.5 ml of a solution comprising (mM): NaF, 100; sucrose, 80; EDTA, 10; Tes, 10; pH 7.4 (Rapoport et al. 1982, 1983; Draznin et al. 1986). The homogenate was centrifuged at 14000 g for 10 min and the proteins in the supernatant were separated by two-dimensional electrophoresis (isoelectric focusing (pI range, 5–7) followed by 12 % SDS polyacrylamide gel electrophoresis) (Driska et al. 1981; Aksoy et al. 1983). Proteins were transferred to nitrocellulose membranes. The nitrocellulose membranes were stained with colloidal gold, imaged with a Molecular Dynamics laser scanner for protein, and imaged with a Molecular Dynamics phosphoimager for ³²P activity. Controls performed after these experiments revealed that the homogenization buffer used did not totally inactivate MLC phosphatase, so that MLC phosphorylation values were artificially high. Therefore, we could not estimate MLC phosphorylation levels in the ³²P experiments.

Electrospray tandem mass spectroscopy

Coomassie Blue-stained spots containing the 20 kDa protein of interest were excised from five heavily loaded two-dimensional gels, minced, destained, alkylated and trypsinized as described (Henzel et al. 1993). Peptides were then run on a POROS 10 RC reverse-phase microcapillary HPLC and its output was directed into a Finnigan-MAT TSQ7000 electrospray tandem mass spectroscope (Henzel et al. 1993). Collisionally activated dissociation spectra were interpreted and the proposed peptide fragment sequences were then compared with published protein sequences (CAD spectral information on SEQUEST and partial peptide sequences on http://prospector.ucsf.edu/cgi-bin/msdigest.exe).

Measurement of oxygen consumption

Oxygen consumption and contractile stress were measured in swine carotid artery rings placed in a 1.2 ml airtight chamber containing a Clark-style oxygen electrode (Instech Labs) and connected to a force transducer (Grass FT03) as described (Wingard et al. 1994). Oxygen consumption was calculated from the rate of decline in P_O2 divided by the wet weight of the tissue. Data reported are means ± 1 s.e.m. for the values obtained between 26 and 30 min after solution changes.

Centrifugation assays

Swine carotid arteries were pharmacologically treated and frozen in an acetone-dry-ice slurry (Rembold & Murphy, 1988). After air drying, the tissues were homogenized in 20 mM Tris pH 7.4 with 5 μl ml⁻¹ proteinase inhibitory cocktail (Sigma) on ice, and then immediately centrifuged at 6000 g for 10 min at 2°C in a microfuge as described (Golenhofen et al. 1999). Samples of the whole homogenate, the pellet and supernatant were then separated by one-dimensional isoelectric focusing (isoelectric point (pI) range 5–8), blotted to nitrocellulose, immunostained with a rabbit commercially produced polyclonal anti-HSP20 antibody and detected with enhanced chemiluminescence.

Binding of HSP20 and peptides to thin filaments

Skeletal G-actin was extracted and purified from rabbit skeletal muscle acetone powder as described (Spudich & Watt, 1971), in accordance with national guidelines. The actin was polymerized by dialysis at 4°C against binding buffer consisting of 40 mM Tris, 100 mM KCl, 5 mM MgCl₂, 1 mM DTT, 0.01 % NaN₃, pH 7.8 (Van Eyk et al. 1997). Smooth muscle tropomyosin (TM) was prepared as described (Bretscher, 1984) and dialysed at 4°C against several changes of binding buffer. The purified synthetic peptides were dissolved in actin-binding buffer and the pH checked. Protein concentrations were determined by the Lowry assay.

The actin centrifugation studies were performed as previously described (Van Eyk et al. 1997; Tripet et al. 1997). Briefly, each binding point was determined by mixing 5 nmol actin, 1.4 nmol tropomyosin (ratio of 7.1 mol actin per 2 mol tropomyosin – equivalent to 1 mol dimeric tropomyosin) and increasing quantities of peptide (0–23 nmol) in a total volume of 200 μl. The actin-TM-peptide mixture was incubated at room temperature for 10 min prior to being spun for 30 min at 100000 g in a TL-100 ultracentrifuge. Under these conditions, greater that 95 % of the actin was pelleted. The pellet was washed with binding buffer and dissolved in 90 μl 0.05 % aqueous trifluoroacetic acid. This was injected onto a Zorbax SB 300 reversed-phase column (250 mm × 4.6 mm i.d., 6.5 μm particle size, 300 Å (30 nm) pore size; Rockland Technologies) on a Varian HPLC system consisting of a 9012 solvent delivery system, 9100 autosampler and 9065 Polychrom. The various proteins and peptides were eluted using a 2 % eluent B per minute linear gradient, where eluent A was 0.05 % aqueous trifluoroacetic acid and eluent B was 0.05 % trifluoroacetic acid in acetonitrile, at a flow rate of 1 ml min⁻¹. The peak areas were determined by integration at 210 nm for the peptide and TM and at 280 nm for actin. The peak areas were converted to nanomoles based on a standard curve obtained for each protein and peptide. The amount of protein and peptide pelleted in the absence of actin was subtracted from the amount pelleted in the presence of actin. Experiments were performed in triplicate. The K_d for HSP20 peptides was determined by fitting the binding data to the equilibrium binding equation with the best fit at n = 2 binding sites.

ATPase activity assay

Smooth muscle myosin S1 was prepared immediately prior to use (Greene et al. 1983). Reactions consisted of 25 μM F-actin, 8 μM tropomyosin, 0.6 μM myosin S1 and variable concentrations of synthetic HSP20 peptides. The buffer contained 40 mM Tris HCl, pH 7.5 at 37°C, 50 mM KCl and 1 mM DTT. Quantification of released P_i was determined as described (Pollard, 1982). Reactions were performed in a 96-well ELISA plate and were triggered by addition of 2 mM ATP. Reactions were terminated by addition of an equal volume (100 μl) of 6 % (w/v) SDS, 3 % ascorbic acid, 1 n HCl and 0.5 % ammonium molybdate. After 5 min incubation, the blue colour was stabilized with 2 % sodium citrate, 2 % sodium m-arsenite and 2 % acetic acid. The plates were read at 650 nm on an E-max precision microplate reader (Molecular Dynamics, Sunnyvale, CA, USA). Phosphate was determined by comparison with standards. Phosphate release was linear over the period of the experiment. Experiments were performed in triplicate.

Measurement of force in skinned smooth muscle

Swine carotid artery tissues were obtained as described above. Tissues were kept in a Hepes-buffered (pH 7.4), physiologically balanced salt solution prior to skinning. Tissues were demembranated (skinned) with 1 % Triton X-100 as described (Strauss et al. 1992). Small fibre bundles (5.0 mm in length and 100–200 mm in diameter) were dissected from the muscle strip and mounted for isometric force measurements as previously described (Strauss et al. 1992). Fibres mounted for mechanical measurements were immersed in a 250 μl bath of relaxing solution maintained at 20°C for a minimum of 10 min. Fibres generated ∼2.5 mN of force within 2–10 min of stimulation (corresponding to a stress of 1.5 × 10⁵ N m⁻²). The system had a usable sensitivity range down to 0.01 mN. All subsequently measured values were standardized to the values obtained in the control contraction cycle. We controlled for skinned fibre ‘rundown’ by performing parallel time experiments with both active and control peptides. ‘Relaxing’ solution was 10 mM MgCl₂, 1 mM NaN₃, 7.5 mM Na₂ATP, 4 mM EGTA, 0.1 μM calmodulin (activity ∼6 × 10⁵ U mg⁻¹; Boehringer-Mannheim), 20 mM imidazole (pH 6.7), 10 mM phosphocreatine (Na₂PCr) and 10 U ml⁻¹ creatine phosphokinase as an ATP regenerating system. Relaxing solution contained Ca²⁺ at < 10 nM. The ‘contracting’ solutions were identical to relaxing solution except that they also contained either 2.0 or 3.6 mM CaCl₂, resulting in a buffered free calcium concentration of ∼0.8 or 25 μM, respectively, 2.0 mM free Mg²⁺, 7.2 mM MgATP, and had an ionic strength of 110 mM, adjusted with potassium acetate.

Size-exclusion chromatography

Recombinant calmodulin was expressed in Escherichia coli strain BL-21 DE3 (Novy et al. 1993) and purified by hydrophobic interaction chromatography (Gopalakrishna & Anderson, 1992). Calmodulin was dialysed against actin-binding buffer with 1 mM CaCl₂ at 4°C, with several changes. Mixtures of calmodulin and synthetic peptides at molar ratios of 1:4 were incubated at room temperature for 15 min. The peptide-protein mixtures were then run in actin-binding buffer plus 1 mM CaCl₂ at 0.5 ml min⁻¹ on a Superdex-75 column (Pharmacia) attached to a Varian HPLC system. The concentration of KCl in the buffer was adequate to eliminate non-specific ionic interactions between the protein, peptide and support. This maintains ideal size-exclusion behaviour. The effluent was monitored at 210 nm. Peaks were collected, lyophilized and identified by reversed-phase HPLC as described above for the actin-binding assay.

RESULTS

Nitroglycerine-induced relaxation without proportional MLC dephosphorylation

Nitroglycerine-induced relaxation without proportional MLC dephosphorylation (i.e. uncoupling of force from MLC phosphorylation) is shown as a rightward shift in a steady-state MLC phosphorylation-force plot (Fig. 1). Only the higher nitroglycerine concentrations (1 and 10 μM) induced substantial relaxation and a rightward shift in the MLC phosphorylation-force relation. This suggests that relaxation without MLC dephosphorylation occurred only when nitroglycerine induced a substantial relaxation.

Figure 1. Dose-response for nitroglycerine-induced relaxation without proportional MLC dephosphorylation.

Relation between MLC phosphorylation and stress (force per cross-sectional area) in swine carotid arterial tissues stimulated with 0.5, 1 or 3 μM histamine (Hist) alone for 60 min (open symbols) or 3 μM histamine for 10 min followed by 0.01, 0.1, 1 or 10 μM nitroglycerine (NTG) for 50 min (filled symbols). The filled diamond shows the response after 3 μM histamine for 10 min followed by 300 μM 8-pCPT-cGMP for 50 min. Stress was normalized to the value produced by exposure to 109 mM [K⁺]_o (K109) for 10 min. Data are presented as means ± 1 s.e.m. (n = 6–20). Where error bars are not visible, the s.e.m. is smaller than the size of the symbol.

One possible explanation for relaxation without proportional MLC dephosphorylation is that the myosin light chains could be phosphorylated on amino acid residues (e.g. Ser¹, Ser², Thr⁹ or Thr¹⁸) that do not activate myosin's ATPase. Importantly, we previously found that addition of nitroglycerine to swine carotid artery induced a relaxation that was associated with MLC phosphorylation entirely on Ser¹⁹ (McDaniel et al. 1992). This result indicates that MLC phosphorylation at other, inactive sites cannot explain nitroglycerine-induced relaxation without proportional MLC dephosphorylation. Other vasodilators such as forskolin, which increase [cAMP], did not alter the relation between MLC phosphorylation and force in the swine carotid artery (McDaniel et al. 1991, 1994). These data suggest that relaxation without MLC dephosphorylation may be specific to increased levels of cGMP. However, we cannot rule out the possibility that this mechanism may be activated by larger increases in cAMP in the swine carotid artery or moderate increases in cAMP in other smooth muscles. Relaxation without MLC dephosphorylation has also been observed with some pharmacological treatments: Ca²⁺ depletion/repletion (Gerthoffer, 1987; Aburto et al. 1993), high extracellular [Mg²⁺] (D'Angelo et al. 1992) and okadaic acid (Tansey et al. 1990).

The role of cGMP in nitroglycerine-dependent relaxation without proportional MLC dephosphorylation

Fifty minutes after addition of 10 μM nitroglycerine to 3 μM histamine-stimulated tissues, [cGMP] was 0.35 ± 0.032 pmol (mg wet weight)⁻¹. This was a significant (P < 0.001; Student's t test), 564 % increase in [cGMP] compared with the value observed in tissues only stimulated with histamine (0.052 ± 0.015 pmol (mg wet weight)⁻¹, n = 4 experiments). Nitroglycerine also significantly increased measured [cAMP] by 36 % compared with histamine stimulation alone (0.73 ± 0.028 pmol (mg wet weight)⁻¹ for histamine plus nitroglycerine vs. 0.54 ± 0.038 pmol (mg wet weight)⁻¹, n = 4). This may represent a true increase in [cAMP] or an artifact secondary to the large increase in cGMP cross-reacting in the cAMP assay.

Relaxation without proportional MLC dephosphorylation was also observed after addition of 300 μM 8-(4-chlorophenylthio)-guanosine 3′:5′-cyclic monophosphate (8-pCPT-cGMP), a membrane-permeant cGMP analogue (Fig. 1, ♦). These data support our first hypothesis that nitroglycerine-dependent relaxation without proportional MLC dephosphorylation was caused by increases in [cGMP] rather than by other effects of nitroglycerine.

Potential mechanisms: nitroglycerine reduces oxygen consumption

Nitroglycerine-dependent relaxation without proportional MLC dephosphorylation suggests the presence of another force regulatory system in smooth muscle. This second force regulatory system appears to be regulated by cGMP and important in relaxation rather than contraction. There are several candidate regulatory systems for force in smooth muscle (see reviews in Bárány, 1996). These include: (1) thin filament regulation by caldesmon or calponin, (2) thick filament regulation by Ca²⁺ binding to myosin or by heavy chain phosphorylation, (3) regulation via protein kinase C by an unknown cascade, and (4) non-cross-bridge-mediated stress in the form of postulated cytoskeletal domain cross-links. None of these appear to be specifically regulated by nitrovasodilators.

There are three potential mechanisms whereby nitroglycerine could induce relaxation without MLC dephosphorylation. (1) Nitroglycerine-induced relaxation could result from fragmentation or disassembly of thin filaments. (2) Nitroglycerine could also accelerate release of myosin from thin filaments so less force is generated per cross-bridge cycle (i.e. diminished power stroke without a change in the cycling rate of myosin). If either of these were the case, ATP consumption should increase from the futile cycling of phosphorylated myosin and fragmented actin filaments. (3) Conversely, if nitroglycerine were to induce relaxation by inhibiting the binding of phosphorylated myosin to thin filaments, ATP consumption should decrease. Steady-state oxygen consumption is an index of ATP utilization. In swine carotid artery, resting oxygen consumption was 42.3 ± 2.8 nmol min⁻¹ g⁻¹. Histamine, at a concentration of 3 μM, induced a contraction and significantly increased oxygen consumption to 69.9 ± 3.3 nmol min⁻¹ g⁻¹. Addition of nitroglycerine induced a 70 % relaxation and significantly decreased oxygen consumption to 45.2 ± 4.7 nmol min⁻¹ g⁻¹ (Fig. 2). These data suggest that nitroglycerine-induced relaxation did not occur by accelerating myosin release from actin or by inducing fragmentation or disassembly of actin filaments (mechanisms 1 and 2). Our data support the third mechanism: nitroglycerine-induced relaxation without proportional MLC dephosphorylation resulted from inhibition of phosphorylated myosin attachment to thin filaments. This inhibition could result from a nitroglycerine-sensitive regulatory system that is associated with either the thin filament or thick filaments.

Figure 2. Nitroglycerine reduced steady-state oxygen consumption in histamine-stimulated swine carotid arteries.

Oxygen consumption and contractile force were measured in swine carotid tissues at rest and following sequential stimulation with 109 mM [K⁺]_o (K109), after washout (Wash), 3 μM histamine (Hist), histamine plus 10 μM nitroglycerine (H + NTG), nitroglycerine alone (NTG) and after washout. Data are means ± 1 s.e.m. (n = 4–6).

Identification of phosphorylated HSP20 in nitroglycerine-stimulated smooth muscle

Since okadaic acid is known to uncouple force from MLC phosphorylation (Tansey et al. 1990), we hypothesized that a phosphoprotein would be involved in nitroglycerine-induced relaxation. Draznin et al. (1986) reported that nitroglycerine treatment of intact rat aorta induced phosphorylation of several proteins. We repeated these studies in the swine carotid artery. Tissues were loaded with ³²PO₄ and cellular proteins separated by two-dimensional gel electrophoresis. In histamine-stimulated tissues, we found that addition of 10 μM nitroglycerine for 50 min primarily increased ³²P labelling of a single protein with a molecular mass of approximately 20 kDa (labelled HSP20 in Fig. 3). Nitroglycerine also increased protein staining of the 20 kDa protein, consistent with a phosphorylation-dependent acidic shift in the pI of this protein (Fig. 3; the dephosphorylated form of HSP20 appears to comigrate with another protein). Histamine stimulation alone did not significantly alter ³²PO₄ labelling or protein staining of the 20 kDa protein. The ³²P labelling and concentration of other proteins, such as MLC and HSP27, did not significantly change with addition of nitroglycerine (data not shown).

Figure 3. Nitroglycerine induced phosphorylation of a 20 kDa protein, later identified as HSP20.

Representative gels resulting from homogenates of ³²PO₄-loaded swine carotid tissues that were unstimulated (top; Control), stimulated with 3 μM histamine alone for 60 min (middle), or stimulated with 3 μM histamine for 10 min followed by 10 μM nitroglycerine (NTG) for 50 min (bottom). Cellular proteins were separated in two dimensions: isoelectric focusing in the horizontal direction (approximately 6.0–5.5 pI, left to right) and SDS electrophoresis in the vertical direction (approximately 30–15 kDa, top to bottom). The left panels show protein staining with colloidal gold and the right panels ³²P activity. The arrows indicate the 20 kDa protein (labelled HSP20) and a second phosphoprotein that was identified as heat shock protein 27 (HSP27) by electrospray tandem mass spectroscopy (data not shown).

Figure 4 shows a correlation between ³²P labelling of the 20 kDa protein and contractile force. Histamine stimulation (3 μM for 60 min) alone induced a prolonged contraction that was not associated with a significant increase in ³²P labelling of the 20 kDa protein (upper tracing with bar showing ³²P labelling). Addition of 10 μM nitroglycerine to histamine-stimulated tissues induced a rapid and prolonged relaxation that was associated with significantly increased ³²P labelling of the 20 kDa protein (lower tracing). The increase in ³²P labelling was 9.1 ± 2.6-fold 5 min after addition of nitroglycerine and 8.2 ± 2.4-fold at 50 min (bars on lower tracing, P < 0.05 for both compared with unstimulated tissues, n = 4–7).

Figure 4. ³²P labelling of HSP20 correlated with nitroglycerine-induced relaxation.

Representative force tracings from the ³²PO₄-loaded swine carotid tissues that were stimulated with 3 μM histamine alone for 60 min (top trace), or stimulated with 3 μM histamine for 10 min followed by 10 μM nitroglycerine for 50 min (bottom trace). The bar graphs show means ± 1 s.e.m. of the ³²P labelling (as a percentage of control) of the protein later identified as HSP20. *Significantly different from control (P < 0.05).

We identified the 20 kDa protein as HSP20 (Swiss prot B53814) by electrospray tandem mass spectroscopy (Table 1). Trypsin digestion of the phosphorylated form of the protein resulted in peptides that produced collisionally activated dissociation spectra identical to those predicted from the published sequence of HSP20 (three representative peptides are shown in Table 1) (Kato et al. 1994). These data support our third hypothesis that nitroglycerine-induced relaxation without proportional MLC dephosphorylation results from phosphorylation of a specific protein. One of these mass spectroscopy-identified tryptic peptides from the 20 kDa protein was phosphorylated on a Ser residue. This phosphorylation corresponded to Ser¹⁶ of the HSP20 amino acid sequence. The amino acid sequence N-terminal to Ser¹⁶ (WLRRAS) resembles the consensus sequence for both cGMP-dependent protein kinase and cAMP-dependent protein kinase (Glass, 1990). Since we found above that nitroglycerine increased [cGMP] at the time point at which relaxation occurred without MLC dephosphorylation (Fig. 1) and when HSP20 was phosphorylated (Figs 3 and 4), our results are consistent with the hypothesis that cGMP mediates HSP20 phosphorylation.

Table 1.

Identification of the 20 kDa protein as HSP20 by electrospray tandem mass spectroscopy

Tryptic peptide	Amino acid sequence	MW
Mass spec. 1	RAS(p)APXPGXSAPGR	1430·4
B53814 (14–27)	RAS(p)APLPGLSAPGR	1430·5
Mass spec. 2	HFSPEEXAVK	1157·8
B53814 (82–91)	HFSPEEIAVK	1157·3
Mass spec. 3	VVGEHVEVHAR	1231·8
B53814 (92–102)	VVGEHVEVHAR	1232·4

Three tryptic peptides were identified by mass spectroscopy that were consistent with the published sequence of HSP20 (B53814). The numbers after B53814 refer to the residue position of the peptide. MW refers to the molecular weight either measured by mass spectroscopy or calculated from the published sequence of B53814. Amino acids are given as the single letter code. S(p), phosphorylated Ser; X = either I or L (which cannot be distinguished by mass spectroscopy).

Phosphorylated HSP20 preferentially sediments in the pellet with centrifugation

HSP20 is very similar to HSP27 and αB-crystallin (Kato et al. 1994), proteins known to bind to and alter actin filament dynamics (see Discussion). We evaluated whether phosphorylated HSP20 could be a thin filament-binding protein with a centrifugation assay. We stimulated swine carotid artery with 10 μM histamine alone or 10 μM histamine plus 10 μM nitroglycerine. After freezing, the tissues were homogenized in a low salt buffer, separated on a one-dimensional isoelectric focusing gel, blotted to nitrocellulose, and immunostained for HSP20. Histamine stimulation alone induced a dark band of immunostaining and a lighter second band at a more acidic pI (Fig. 5). This lower band is consistent with phosphorylated HSP20. Histamine-induced HSP20 phosphorylation was calculated to be 0.04 ± 0.02 mol P_i (mol HSP20)⁻¹ (mean ± 1 s.e.m., n = 4). Nitroglycerine-induced relaxation was associated with more intense staining of the phosphorylated HSP20 band. Nitroglycerine-induced HSP20 phosphorylation was calculated to be 0.19 ± 0.03 mol P_i (mol HSP20)⁻¹ (mean ± 1 s.e.m., n = 7), a value significantly greater than that observed with histamine alone (P < 0.05).

Figure 5. Phosphorylated HSP20 cosediments with thin filaments.

Representative blot resulting from homogenates of swine carotid tissues that were stimulated with 10 μM histamine alone for 60 min (lanes 1–3) or stimulated with 10 μM histamine for 10 min followed by 10 μM nitroglycerine for 50 min (lanes 4–6). Cellular proteins were separated by isoelectric focusing in the vertical direction (approximately 6.0–5.5 pI), blotted to nitrocellulose, immunostained with an antibody to HSP20 and imaged with enhanced chemiluminescence. Total homogenates were loaded in lanes 1 and 4. After centrifugation at 6000 g for 10 min, the pellets were loaded in lanes 2 and 5, and the supernatants loaded in lanes 3 and 6. Phos HSP20, phosphorylated HSP20.

Immediately after homogenization, some of the homogenate was centrifuged at 6000 g. Phosphorylated HSP20 preferentially sedimented in the pellet compared with the supernatant (this is most apparent in the samples from nitroglycerine-stimulated tissues; Fig. 5). When normalized to the amount of phosphorylated HSP20 observed in the homogenate, phosphorylated HSP20 was found in significantly greater amounts in the pellet (1.10 ± 0.16 times homogenate) compared with the supernatant (0.64 ± 0.16 times homogenate, P < 0.05 by paired t test). These data suggest that phosphorylation of HSP20 increased its affinity for a component of the pellet. Since HSP27 and αB-crystallin are known to bind to thin filaments (Miron et al. 1991) and since Brophy et al. (1999) found that HSP20 cosediments with actin in vitro, the most likely candidate for HSP20 binding is the thin filament. While our in vitro centrifugation assay results were significant, the relative concentration of HSP20 in the pellet may not reflect that observed in vivo since dissociation may have occurred during homogenization and/or centrifugation.

Stoichiometry of HSP20 compared with actin

From protein staining of two-dimensional gels (Fig. 3), the ratio of the amount of phosphorylated HSP20 to total MLC was 0.57 ± 0.10 (n = 4). Murphy et al. (1974) reported that the swine carotid artery contained 9.7 mg g⁻¹ of myosin and 30 mg g⁻¹ of actin. Based on their estimate that 60 % of swine carotid artery is cellular, calculated concentrations of myosin and actin of 68 and 1100 μM, respectively, can be obtained. If we assume (1) that HSP20 and MLC bind Coomassie Blue stain equally and (2) that the homogenization buffer extracted HSP20 and MLC equally, we can estimate HSP20 stoichiometry. Given our HSP20/MLC ratio of 0.57 and the relative molecular masses, the estimated concentration of phosphorylated HSP20 is 46 ± 8 μM, corresponding to a phosphorylated HSP20/actin ratio of 1/24. If we assume that our estimate of HSP20 phosphorylation (0.19 ± 0.03 mol P_i (mol HSP20)⁻¹ from Fig. 5) is reasonable, then the estimated concentration of total HSP20 is 236 ± 42 μM, which corresponds to a total HSP20/actin ratio of 1/5. Given the assumptions, this analysis suggests that HSP20 is present in a stoichiometry similar to that observed with troponin I (TnI) in skeletal muscle (1 TnI to 7 actin). It is therefore possible that HSP20 could be a thin filament regulatory protein.

Binding of HSP20 peptides to actin-TM

Interestingly, we noted that the amino acid residues 110–121 of HSP20 have a high degree of sequence homology with a region of cardiac and skeletal TnI termed the TnI inhibitory region (skeletal TnI_104-115 and cardiac TnI_136-147). Specifically, there is a conserved glycine (G), a phenylalanine (F) and four arginine (R) residues (Table 2). Therefore, we hypothesized that HSP20 may share some, but not all, of the thin filament regulatory properties of TnI. An important difference is that HSP20 appears to be regulated by a covalent modification (phosphorylation), while TnI is regulated by an allosteric interaction with troponin C (TnC).

Table 2.

Sequence homology of a region of HSP20 (NCBI number: B53814) with other actin-binding proteins

Protein	Database identification	Location	Amino acid sequence
Human cardiac TnI	Swiss P19429	126–148	D-L-T-Q-K-I-F-P-L-R-G-K-F-K-R-P-T-L-R-R-V-R-I
Human skeletal TnI	Swiss P48788	94–116	D-M-N-Q-K-I-F-D-L-R-G-K-F-K-R-P-P-L-R-R-V-R-M
			| | / | | | |
Human HSP20 (p20)	Pir B53814	100–122	A-R-H-E-E-R-P-D-E-H-G-F-V-A-R-E-F-H-R-R-Y-R-L
			**| | | | | | | | |**| | | | |*| |*
Human αB-crystallin	Pir S42754	101–123	G-K-H-E-E-R-Q-D-E-H-G-F-I-S-R-E-F-H-R-K-Y-R-I
Human HSP27	Pir A24017	121–143	G-K-H-E-E-R-Q-D-E-H-G-Y-I-S-R-C-F-T-R-K-Y-T-L
Chick HSP25	Swiss Q00649	119–141	G-K-H-E-E-K-Q-D-E-H-G-F-I-S-R-C-F-T-R-K-Y-T-L
Short peptide HSP20	—	110–121	Ac-G-F-V-A-R-E-F-H-R-R-Y-R-CONH2
Long peptide HSP20	—	100–121	Ac-A-R-H-E-E-R-P-D-E-H-G-F-V-A-R-E-F-H-R-R-Y-R-CONH2
Control peptide HSP20(R118G, R121G)	—	110–121	Ac-G-F-V-A-R-E-F-H-G-R-Y-G-CONH2

Sequences of cardiac TnI, skeletal TnI (underlined letters correspond to the TnI inhibitory peptide), HSP20, αB-crystallin, HSP27, HSP25 and the three peptides studied. Amino acids are shown as the single letter code. | identical amino acid; *, similar amino acid. Emboldened amino acid residues are those conserved between TnI and HSP20. The underlined glycine residues in the control peptide HSP20 sequence indicate those substituted for alanine. The sequence alignment was performed by ClustalW. (http://dot.imgen.bcm.tmc.edu:9331/cgi-bin/multi-align/multi-align.pl).

We characterized HSP20 amino acid residues 110–121 in a manner similar to that described for skeletal TnI_104-115 (see Discussion). Two HSP20 synthetic peptides were synthesized: a short peptide equivalent to skeletal TnI_104-115 (HSP20 residues 110–121, abbreviated HSP20_110-121) and a longer peptide equivalent to skeletal TnI_94-116 (HSP20 residues 100–121, abbreviated HSP20_100-121; Table 2). A control HSP20 peptide was also synthesized in which two arginine residues were replaced with glycine ((R118G)(R121G) HSP20_110-121; Table 2). These arginine residues were chosen based on the key role of their equivalent conserved residues in the TnI inhibitory region (TnI residues R115 and R112, Table 2; Van Eyk & Hodges, 1988).

The short peptide HSP20_110-121 bound to actin-TM (Fig. 6A, •). Binding increased further at high concentrations of HSP20_110-121 (Fig. 6A, inset). The calculated K_d was 155 ± 82 μM if it is assumed that two peptide molecules bind per mole of actin (it would be higher if 3 bind per mole of actin). HSP20_110-121 had a 2-fold lower affinity for actin-TM filaments than the published K_d for native TnI inhibitory peptide (TnI_104-115) and a slightly higher affinity for actin-TM than the TnI synthetic peptide (F106G, R115G) TnI_96-115 (Tripet et al. 1997).

Figure 6. The short peptide HSP20_110-121 specifically and stoichiometrically binds to actin-tropomyosin filaments only in the presence of tropomyosin and also reduced actin-tropomyosin-activated myosin S1 ATPase activity.

A, the quantity of synthetic HSP20 peptide bound to actin-TM at various concentrations was measured, and the data plotted as molar ratio of peptide bound to actin-TM filaments (actin:TM, 7.1:2). The synthetic peptide HSP20_110-121 (•) bound stoichiometrically to actin-TM filaments while control peptide (R118G)(R121G) HSP20_110-121 (□) and the long peptide HSP20_100-121 (▴) had weaker affinity for actin-TM. Inset, binding of higher concentrations of the short peptide HSP20_110-121 to actin-tropomyosin. B, actin-tropomyosin-activated myosin ATPase activity in the presence of the short peptide HSP20_110-121 (•) and the control peptide (R118G)(R121G) HSP20_110-121 (^). Only the short peptide HSP20_110-121 decreased ATPase activity. Inset, binding of the short peptide HSP20_110-121 to actin in the presence (•) and absence (^) of tropomyosin. HSP20_110-121 displayed enhanced binding in the presence of tropomyosin. Assays were done in triplicate. Data are means ± 1 s.e.m. Where error bars are not visible, the s.e.m. is smaller than the size of the symbol.

Stoichiometric binding of HSP20_110-121 to actin-TM filaments only occurred when the filaments contained tropomyosin (Fig. 6B, inset, •). Actin binding only in the presence of tropomyosin is a hallmark of TnI (Talbot & Hodges, 1979). This result suggests that HSP20 may share some of TnI's thin filament binding characteristics. The control peptide ((R118G)(R121G) HSP20_110-121) bound actin 20 times more weakly than HSP20_110-121 (Fig. 6A, □). This demonstrates that the Arg¹¹⁸ and Arg¹²¹ amino acid residues are important in actin binding. These data support our third hypothesis that nitroglycerine-induced relaxation without proportional MLC dephosphorylation results from phosphorylation of a specific protein that interacts with the thin filament. HSP20 appears to contain an actin-binding sequence.

Compared with HSP20_110-121, the longer synthetic peptide, HSP20_100-121, bound to actin-TM with reduced affinity (Fig. 6A, ▴). This is not uncommon since peptides with ‘extra’ amino acid residues (i.e. those not involved in binding) can reduce the affinity or biological activity. For example, the synthetic TnI peptide 96–131 is a poorer inhibitor when compared with TnI_96-116, the peptide containing only the actin-TM-binding site (Tripet et al. 1997). Since there is only a single amino acid residue that is semi-conserved between TnI_94-103 and HSP20_100-109 it is not surprising that the longer peptide binds more poorly to actin. Interestingly, HSP20 residues 100–109 contain four negatively charged and two positively charged amino acids and two His residues that are 100 % conserved or semiconserved between HSP25, 27 and αB-crystallin (Table 2). This suggests another role for this region of HSP20, a role shared by the other heat shock proteins. This region may be important in phosphorylation-dependent regulation of HSP20 and other heat shock proteins.

Effect of HSP20 peptides on actomyosin ATPase activity

If HSP20 functions as a TnI-like protein (i.e. preventing myosin attachment), then binding of HSP20 to actin should alter actin-activated myosin ATPase activity. The short peptide HSP20_110-121 decreased actin-activated myosin ATPase activity, but only at high concentrations (Fig. 6B, •). The control peptide ((R118G)(R121G) HSP20_110-121) did not alter actin-activated myosin ATPase activity (Fig. 6B, ^). This demonstrates that the Arg¹¹⁸ and Arg¹²¹ residues are important in the inhibition of ATPase by HSP20_110-121. ATPase measurements were not performed with the long peptide, HSP20_100-121.

Effect of HSP20 peptides on contractile force

If HSP20 functions as a TnI-like protein, then binding of HSP20 to actin should also relax smooth muscle. In skinned swine carotid artery, pretreatment with a high concentration of the short peptide HSP20_110-121 (300 μM) attenuated the initial phase and abolished the sustained phase of a pCa 5.4 Ca²⁺-induced contraction (Fig. 7, third contraction – HSP20_110-121 was present during the entire contraction). HSP20_110-121 treatment inhibited Ca²⁺-induced sustained force by 81 ± 7 % (n = 6, P < 0.05, force measured 20 min after increasing Ca²⁺). Pretreatment with 300 μM of the control peptide ((R118G)(R121G) HSP20_110-121) did not substantially alter the Ca²⁺-induced contraction (Fig. 7, second contraction).

Figure 7. The short peptide HSP20_110-121 attenuated contraction of skinned swine carotid artery.

Representative tracing of force generation by skinned swine carotid artery. The skinned tissue was contracted thrice by increasing pCa to 5.4. The first contraction was performed in the absence of any peptide. The second contraction was performed in the presence of 300 μM control peptide (R118G)(R121G) HSP20_110-121. The final contraction was performed in the presence of 300 μM HSP20_110-121. HSP20_110-121, but not the control peptide, significantly attenuated the initial phase of contraction and abolished the sustained phase of contraction. Peptides were added 5 min before [Ca²⁺] was increased.

Binding of HSP20 peptides to calmodulin

In addition to binding actin-TM, the TnI inhibitory peptide, TnI_104-115, tightly binds TnC and weakly binds calmodulin (Malencik & Anderson, 1984; Cachia et al. 1986; Tobacman, 1996; Solaro & Van Eyk, 1996). To determine whether HSP20 binds calmodulin, high-performance size-exclusion chromatography was performed in the presence of calcium. None of the HSP20 peptides interacted with calmodulin under these conditions (Fig. 8). However, it is possible that the peptide could form a very weak calmodulin complex with extremely fast on and off rates.

Figure 8. The short peptide HSP20_110-121 did not interact with calmodulin.

A, size-exclusion chromatography (SEC) elution profile of a 4:1 mixture of HSP20 and calmodulin in the presence of calcium. Fractions were collected (* and **) and analysed by reversed-phase HPLC. B, elution profile of the fraction collected at 21 min during the SEC run (*), which contained only calmodulin (which elutes off the reversed-phase column at 28 min). There was no peptide present in this fraction demonstrating that the peptide and calmodulin do not interact under these conditions. C, elution profile of the fraction collected at 35 min during the SEC run (**), which contained all of the peptide (which elutes off the reversed-phase column at 17.5 min).

DISCUSSION

Our results show that nitroglycerine-induced relaxation without MLC dephosphorylation was associated with increases in [cGMP]. A membrane-permeant cGMP analogue also induced relaxation without MLC dephosphorylation, suggesting that cGMP mediated the nitroglycerine response (Fig. 1). Nitroglycerine also decreased oxygen consumption, suggesting that nitroglycerine-induced relaxation is more likely to be caused by a mechanism that prevents myosin attachment to the thin filaments (Fig. 2). We found that nitroglycerine-induced relaxation was associated with phosphorylation of heat shock protein 20 on Ser¹⁶ (Figs 3 and 4). Finally, we found that a HSP20 peptide bound to thin filaments and inhibited force generation in skinned smooth muscle (Figs 5–7).

Characteristics of HSP20

HSP20 was originally discovered in extracts of human skeletal muscle because it coeluted with αB-crystallin and HSP27 (Kato et al. 1994). HSP20 is expressed in skeletal muscle, cardiac muscle, stomach, intestine and bladder (Kato et al. 1994). The laboratory of Colleen Brophy was the first to study the role of HSP20 in smooth muscle relaxation. They found that cyclic nucleotide-dependent relaxation was associated with HSP20 phosphorylation in bovine carotid smooth muscle (Beall et al. 1997). The human umbilical artery is an ‘experiment of nature’ since it does not exhibit cyclic nucleotide-induced relaxation. While the umbilical artery contains HSP20, Brophy's group reported that forskolin did not induce HSP20 phosphorylation, suggesting a linkage between HSP20 phosphorylation and relaxation (Brophy et al. 1997). Jerius et al. (1999) reported that flow-dependent vasodilatation, a nitric oxide-mediated process, was associated with HSP20 phosphorylation. Brophy et al. (1999) also reported that HSP20 cosediments with actin in vitro. These data suggest that phosphorylation of HSP20 may be integral in cyclic nucleotide-induced relaxation.

HSP20 is highly homologous to HSP27 and αB-crystallin (Kato et al. 1994), proteins known to affect actin filament dynamics. HSP25 (the rodent equivalent of HSP27) was first characterized as an inhibitor of actin polymerization (25 kDa actin inhibitory protein) (Miron et al. 1991). HSP25 is known to be present in vascular smooth muscle (Wilkinson & Pollard, 1993). Phosphorylation of HSP27 by p38 MAP (mitogen-activated protein) kinase increases the stability of actin filaments in cells exposed to cytochalasin D (Guay et al. 1997). Antibodies to HSP27 inhibited bombesin-induced contraction of rabbit anal smooth muscle (Bitar et al. 1991). αB-crystallin is a heat shock protein expressed in mammalian lens, skeletal muscle and heart, but not in smooth muscle (Klemenz et al. 1990). Dephosphorylated αB-crystallin, but not phosphorylated αB-crystallin, was found to stabilize actin filaments and prevented cytochalasin-induced actin depolymerization (Wang & Spector, 1996).

Most importantly, HSP20 is the only one of these small heat shock proteins containing a cGMP/cAMP-dependent protein kinase consensus phosphorylation site (RRAS) (Kato et al. 1994). This is a potential link between increases in cGMP, HSP20 phosphorylation and relaxation. Based on the sequence similarity of HSP20 with other actin-binding heat shock proteins, we hypothesized that nitroglycerine-dependent phosphorylation of HSP20 causes HSP20 to interfere with or inhibit the normal function of the actin-containing thin filaments despite activation of myosin by MLC phosphorylation. We therefore tested whether HSP20 bound to thin filaments. We found that phosphorylated HSP20 cosedimented with thin filaments better than dephosphorylated HSP20 (Fig. 5). This is consistent with but does not prove that phosphorylated HSP20 binds to thin filaments. Our result differs from that observed by Brophy et al. (1999). In cosedimentation assays with actin alone, they found that unphosphorylated HSP20 sedimented in the pellet while phosphorylated HSP20 remained in the supernatant. This result suggests that the binding of HSP20 to thin filaments may depend on the constituents of the thin filament.

Could HSP20_110-121 binding to thin filaments reduce force generation?

Interestingly, we noted that the amino acid residues 110–121 of HSP20 have a high degree of sequence homology with a region of cardiac and skeletal TnI termed the TnI inhibitory region (skeletal TnI_104-115 and cardiac TnI_136-147). Specifically, there is a conserved glycine, a phenylalanine and four arginine residues (Table 2). TnI is part of the troponin complex that is responsible for Ca²⁺-dependent regulation of striated and cardiac muscle contraction. Binding of the inhibitory region of TnI to actin-tropomyosin depends on Ca²⁺ binding to TnC (Tobacman, 1996; Solaro & Van Eyk, 1996). The skeletal TnI peptide TnI_104-115 is the minimum TnI amino acid sequence required to inhibit the actin-TM-myosin (or myosin S1) ATPase activity (Talbot & Hodges, 1979). In TnI-extracted skinned cardiac muscle, addition of synthetic TnI_104-115 alone is sufficient for muscle relaxation (Van Eyk et al. 1993). The majority of the conserved or semi-conserved amino acid residues shared by HSP20 and TnI are known to be crucial for the binding of TnI to actin-TM. Grand et al. (1982) showed that the proton magnetic resonance signals of one or more of the five arginine residues in skeletal TnI_96-116 were perturbed by actin binding. Single glycine amino acid substitution throughout the synthetic skeletal TnI_104-115 indicated that Arg¹¹⁵, Arg¹¹², Lys¹⁰⁷, Leu¹¹¹ and Val¹¹⁴ are important for inhibition of the acto-TM-myosin S1 ATPase activity (Van Eyk & Hodges, 1988; Van Eyk et al. 1993; Van Eyk et al. 1997).

We therefore tested peptides from this region of HSP20 that is analogous to TnI. We found that the short peptide, HSP20_110-121, abolished force in Ca²⁺-contracted swine carotid artery (Fig. 7) and partially inhibited the acto-S1-TM ATPase activity (Fig. 6). HSP20_110-121 also displayed tropomyosin specificity, a hallmark of intact TnI and the peptide TnI_104-115 (Campbell et al. 1992; Tripet et al. 1997). These data suggest that HSP20_110-121 may regulate contractile force in a manner analogous to TnI_104-115. Our data are most consistent with the hypothesis that binding of HSP20_110-121 to the thin filament prevents attachment of phosphorylated cross-bridges to the thin filament (thereby reducing oxygen consumption).

However, HSP20_110-121 is not a complete analogue of TnI_104-115. High concentrations of HSP20_110-121 were required to abolish Ca²⁺-activated force in skinned carotid artery (Fig. 7). HSP20_110-121 also bound less tightly to actin-TM compared with the published values for TnI_104-115 (Van Eyk & Hodges, 1988; Van Eyk et al. 1997; Tripet et al. 1997). Additionally, unlike intact TnI and TnI_104-115 (Cachia et al. 1986), HSP_110-121 did not bind calmodulin (Fig. 8), suggesting that HSP20_110-121 cannot be regulated by a Ca²⁺-calmodulin-dependent mechanism (like TnI-TnC in striated muscle). These differences between HSP20 and TnI may represent differences in the regulation of these two proteins. The peptide HSP20_110-121 does not contain the phosphorylation site (which is at Ser¹⁶). Since relaxation without myosin dephosphorylation was associated with phosphorylated HSP20, not dephosphorylated HSP20, we hypothesize that phosphorylation of HSP20 regulates binding of HSP20 to thin filaments. Our centrifugation data (Fig. 5) are consistent with the hypothesis that phosphorylation of HSP20 increases the affinity of HSP20 for the thin filament. The relatively low affinity of HSP20_110-121 and its lack of calmodulin binding may be necessary to prevent dephosphorylated HSP20 from inhibiting thin filament function.

Intact HSP20 may have thin filament effects similar to those documented in other similar heat shock proteins (Miron et al. 1991; Wang & Spector, 1996; Guay et al. 1997). The actin-binding sequence in HSP20_110-121 is conserved in other small heat shock proteins such as αB-crystallin, HSP25 and HSP27 (Table 2). The fact that this region bound actin in a TM-specific manner but only bound weakly may reflect the initial function of these proteins. Early in evolution, weak interactions between heat shock proteins and actin-TM may have been important in the response to acute cellular stress, e.g. binding to protect the contractile proteins. This could explain the difference between these proteins and TnI. It is possible that the presence of a cGMP-dependent protein kinase phosphorylation site allowed HSP20 to evolve from a heat shock protein into a protein with additional functions.

Conclusion

Our working hypothesis is that cGMP-mediated phosphorylation of intact HSP20 increases the affinity of HSP20 for the thin filament. We hypothesize that this binding inhibits interaction of phosphorylated myosin with the thin filament, inducing a relaxation despite high levels of [Ca²⁺] and activated myosin. Data supporting these hypotheses are as follows. (1) Increases in [cGMP] and a membrane-permeant cGMP analogue induced relaxation without MLC dephosphorylation (Fig. 1). (2) cGMP-dependent relaxation was associated with phosphorylation of heat shock protein 20 on Ser¹⁶, a consensus site for cAMP/cGMP-dependent protein kinase (Figs 3 and 4). (3) Phosphorylated HSP20 cosedimented with thin filaments, consistent with phosphorylation-dependent regulation of the binding of HSP20 to thin filaments (Fig. 5). (4) A HSP20 peptide (HSP20_110-121) bound to thin filaments in a manner similar to TnI and inhibited force generation in skinned smooth muscle (Figs 6 and 7). (5) Increases in cGMP also decreased oxygen consumption, suggesting that a mechanism that prevents myosin attachment to the thin filaments was responsible for the reduction in force (Fig. 2). Evaluation of this hypothesis will require evaluation of intact phosphorylated HSP20 in actin binding, ATPase and skinned fibre experiments.

These data have implications beyond the pharmacology of nitrovasodilators. Nitric oxide is released both by the endothelium and by non-adrenergic non-cholinergic (NANC) inhibitory nerves (Furchgott & Vanhoutte, 1989; Li & Rand, 1989; Bult et al. 1990). It is possible that nitric oxide-mediated HSP20 phosphorylation may be the mechanism responsible for endothelial- and NANC-mediated smooth muscle relaxation. As such, HSP20 could have a role in diseases such as hypertension, gastrointestinal motility disorders and erectile dysfunction. It is interesting that amino acid residues 110–121 of HSP20 are highly conserved among the other heat shock proteins. Thus, other small heat shock proteins may work in a similar manner with their actin affinities altered by phosphorylation by their respective kinases.