Structure-Function Studies on Small Heat Shock Protein Oligomeric Assembly and Interaction with Unfolded Polypeptides*

The small heat shock protein (smHSP) and α-crystallin genes encode a family of 12–43-kDa proteins which assemble into large multimeric structures, function as chaperones by preventing protein aggregation, and contain a conserved region termed the α-crystallin domain. Here we report on the structural and functional characterization of Caenorhabditis elegansHSP16-2, a 16-kDa smHSP produced only under stress conditions. A combination of sedimentation velocity, size exclusion chromatography, and cross-linking analyses on wild-type HSP16-2 and five derivatives demonstrate that the N-terminal domain but not most of the the C-terminal extension which follows the α-crystallin domain is essential for the oligomerization of the smHSP into high molecular weight complexes. The N terminus of HSP16-2 is found to be buried within complexes which can accommodate at least an additional 4-kDa of heterologous sequence per subunit. Studies on the interaction of HSP16-2 with fluorescently-labeled and radiolabeled actin and tubulin reveal that this smHSP possesses a high affinity for unfolded intermediates which form early on the aggregation pathway, but has no apparent substrate specificity. Furthermore, both wild-type and C-terminally-truncated HSP16-2 can function as molecular chaperones by suppressing the thermally-induced aggregation of citrate synthase. Taken together, our data on HSP16-2 and a unique 12.6-kDa smHSP we have recently characterized demonstrate that multimerization is a prerequisite for the interaction of smHSPs with unfolded protein as well as for chaperone activity.

The small heat shock protein (smHSP) and ␣-crystallin genes encode a family of 12-43-kDa proteins which assemble into large multimeric structures, function as chaperones by preventing protein aggregation, and contain a conserved region termed the ␣-crystallin domain. Here we report on the structural and functional characterization of Caenorhabditis elegans HSP16-2, a 16-kDa smHSP produced only under stress conditions. A combination of sedimentation velocity, size exclusion chromatography, and cross-linking analyses on wild-type HSP16-2 and five derivatives demonstrate that the Nterminal domain but not most of the the C-terminal extension which follows the ␣-crystallin domain is essential for the oligomerization of the smHSP into high molecular weight complexes. The N terminus of HSP16-2 is found to be buried within complexes which can accommodate at least an additional 4-kDa of heterologous sequence per subunit. Studies on the interaction of HSP16-2 with fluorescently-labeled and radiolabeled actin and tubulin reveal that this smHSP possesses a high affinity for unfolded intermediates which form early on the aggregation pathway, but has no apparent substrate specificity. Furthermore, both wild-type and C-terminally-truncated HSP16-2 can function as molecular chaperones by suppressing the thermally-induced aggregation of citrate synthase. Taken together, our data on HSP16-2 and a unique 12.6-kDa smHSP we have recently characterized demonstrate that multimerization is a prerequisite for the interaction of smHSPs with unfolded protein as well as for chaperone activity.
Molecular chaperones belong to a class of proteins whose function is to interact with and stabilize proteins that are partially or totally unfolded, as is the case when proteins are in the process of being synthesized, translocated across a membrane, or damaged by conditions of cellular stress. Many chaperones are expressed at higher levels during biological stresses, and are members of heat shock protein (HSP) 1 families (1)(2)(3)(4). Whereas some chaperones (HSP70, HSP40, and HSP60) are involved in protein folding under normal conditions in vivo (5)(6)(7), others such as HSP104 (8 -10), inducible HSP70s (11), and small HSPs (12)(13)(14)(15) are known to play important roles in protecting organisms from stress.
The small HSPs (smHSPs) form a structurally divergent protein family with members present in Archaea, Bacteria, and Eukarya (16,17). The presence of an evolutionarily conserved ␣-crystallin domain distinguishes all smHSPs and ␣-crystallins (18 -20). This domain is preceded by an N-terminal domain, which is highly variable in size and sequence, and is followed by a short, poorly conserved C-terminal extension. Some smHSP genes contain an intron which delineates the N-terminal and ␣-crystallin domains (21,22), and structural studies support a two-domain structure (20,23,24) consisting mostly of ␤-sheets (25,26) for smHSPs. The C-terminal extensions of smHSPs appear relatively unstructured (27,28) and are known to undergo numerous modifications, including truncations (reviewed in Ref. 29).
A common feature of smHSPs is their formation of large oligomeric complexes. The simplest and best characterized smHSP quaternary structure is the 150-kDa trimer of trimers formed by Mycobacterium tuberculosis HSP16.3 (30). Plant smHSPs assemble into complexes of 200 -300 kDa (31), and the typical oligomeric size of ␣-crystallins and smHSPs from yeast and mammals is between 400 and 800 kDa (29,32). Several contrasting models have been proposed to account for the arrangement of subunits within the smHSP aggregate. The review by Groenen et al. (29) addresses this issue in detail, and suggests that although there is as yet no single consistent model for smHSP quaternary structure, it is likely that the smHSP N-terminal domain is buried in the aggregate whereas the C-terminal domain is exposed at the surface, and that all subunits are in equivalent positions within a roughly spherical aggregate.
Numerous in vitro studies have documented the ability of smHSPs to prevent the aggregation of various test proteins at elevated temperatures (26,(33)(34)(35). In the cell, cytoskeletal elements, which are particularly sensitive to cellular stresses (36), appear to be substrates for smHSPs. Specific binding of smHSPs to actin and the intermediate filaments desmin, vimentin, and glial fibrillary acidic protein have been demonstrated (37)(38)(39)(40). HSP27 has been implicated in regulating the dynamics of actin filaments (41,42), and ␣-crystallin participates in intermediate filament assembly (40). Furthermore, HSP27 and ␣-crystallin can enhance the survival of cells subjected to heat shock or oxidative stresses by conferring increased stability to actin fibers (13,(43)(44)(45).
Although many smHSPs are present during development under physiological conditions, smHSPs are among the most highly inducible HSPs during heat shock or other stresses (15). Four of over 15 smHSP genes present in the nematode Caenorhabditis elegans (hsp16-1, hsp16-2, hsp16 -41, and hsp16 - 48) have been shown to be expressed only under stress conditions, in a tissue general manner (46 -49). More significantly, their expression correlates specifically with the presence of agents which induce protein damage (50).
In an effort to understand the oligomeric assembly and chaperone activity of smHSPs in general, studies on the structure and function of the 16-kDa C. elegans HSP16-2 wild-type protein and various derivatives were undertaken. We show that binding to unfolded protein and chaperone activity is strictly correlated with the ability of the smHSP to form multimers. The nonconserved N-terminal region is buried within the smHSP complex and is required for subunit assembly, whereas most of the C-terminal extension is not required for multimerization or chaperone activity, but is possibly involved in maintaining the solubility of the complex. Investigations on the binding of wild-type HSP16-2 and derivatives to the cytoskeletal elements actin and tubulin show that smHSPs exhibit little or no substrate specificity and have a high affinity for unfolded, but not for aggregated or native polypeptides. The conclusions from our in vitro mutagenesis studies on HSP16-2 are supported by data from an unusually diminutive (12.6 kDa), naturally-occurring smHSP from C. elegans (51).
Expression and Purification of the SmHSPs-The expression constructs were transformed into the BL21(DE3) Escherichia coli strain, grown to an OD 600 of 0.8 -1.0, induced with 1 mM isopropyl-1-thio-␤-Dgalactopyranoside, and harvested 3 h later (52). All HSP16-2 proteins formed insoluble inclusion bodies which were solubilized in TEND buffer (50 mM Tris-HCl, 0.1 mM EDTA, 50 mM NaCl, 1 mM DTT, pH 7.5) supplemented with 8 M urea. HSP16-2 was first purified by size-exclusion chromatography over Sephacryl S-100 in TEND, 4 M urea. The peak fractions were then dialyzed against TEND buffer, and the native complex was purified by separation on Sephacryl S300-HR in TEND buffer. The polyhistidine-containing proteins were purified by Ni 2ϩchelate affinity chromatography in 8 M urea as described by the manufacturer (Qiagen), and then dialyzed in TEND buffer. HSP12.6 was produced in E. coli as a soluble protein. Cells resuspended in TEND buffer supplemented with protease inhibitors (2 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, 10 g/ml pepstatin A, 10 g/ml leupeptin, 2 mM EDTA) were broken in a French press, and the clarified supernatant was chromatographed on Sephacryl S200-HR in TEND buffer. Peak fractions containing HSP12.6 were further purified by chromatography on a hydroxylapatite support. Purification steps were carried out on ice or at 4°C. All smHSPs were stored at Ϫ70°C after purification (and dialysis) except for H 6 ⌬130 -145 HSP16-2 which was only dialyzed before use because of its insolubility following freeze-thawing.
Size Exclusion Chromatography (SEC) of the SmHSPs-To estimate the size of the native smHSP complexes, approximately 5 mg of purified protein was subjected to SEC over Sephacryl S-300HR. The column was first calibrated with the following molecular mass standards: thyroglobulin (669 kDa), ferritin (440 kDa), catalase (232 kDa), bovine serum albumin (67 kDa), ovalbumin (43 kDa), and RNase A (14 kDa). The proteins were chromatographed on the column in TEND buffer at room temperature. Samples were separated on SDS gels and the elution volume of each was used to estimate the molecular weight.
Sedimentation Velocity Measurements-Sedimentation velocity experiments were carried out with a Beckman Optima XL-A analytical ultracentrifuge equipped with an AN 60Ti four-hole rotor and cells with two-channel 12-mm path length centerpieces. Sample volumes of 400 l were centrifuged at 60,000 rpm. Radial scans of absorbance at 278 nm were taken at 10-min intervals. Data were analyzed to provide the apparent distribution of sedimentation coefficients using the programs DCDT (53) and SVEDBERG (54). The partial specific volumes, v, at 20°C, were calculated from the amino acid compositions and the solvent density was 1.00 g/cm 3 . The degrees of hydration of the totally unfolded proteins was estimated based on the amino acid compositions by the method of Kuntz (55) according to Laue et al. (56). The degrees of hydration used for all calculations, 0.316, 0.314, and 0.337 g of H 2 O per g of protein, for wild-type HSP16-2, H 6 ⌬130 -145 HSP16-2, and H 6 ⌬1-44 HSP16-2, respectively, were the result of correcting the calculated degrees of hydration by a factor 0.7 obtained by comparing degrees of hydration for several proteins in their folded state to that based on their amino acid composition (57).
Cross-linking of SmHSPs-Cross-linking reactions containing 1.5 M smHSP monomers, 15 M BSA, and 2 mM freshly prepared BS 3 were carried out in cross-linking buffer (25 mM MES, pH 7.5, 25 mM NaCl, 0.5 mM DTT) for 30 min at room temperature and analyzed by Western blotting using an antibody against HSP16-2 (50).
Assay for H 6 HSP16-2 Binding to Ni 2ϩ -Chelate Affinity Resin-H 6 HSP16-2 was purified and dialyzed against NP buffer (50 mM NaCl, 50 mM NaPO 4 , pH 8.0). Two samples containing 175 g of H 6 HSP16-2 were prepared, one in NP buffer (sample N for "native") and the other in NP buffer containing 6 M urea (sample D for "denatured"). A small amount of Ni 2ϩ -chelate affinity resin (30 l) was added to samples N and D. The samples were mixed for 15 min, then washed twice with NP buffer (without urea for sample N and with 6 M urea for sample D, both at pH 6.3), and any bound protein was eluted with 100 mM EDTA. Aliquots of the pH 6.3 and EDTA washes of sample N and the EDTA washes of sample D were analyzed on a 12% SDS gel.
Aggregation Assays-The thermally-induced aggregation of a 900-l citrate synthase solution (150 nM monomers), with smHSPs or control proteins, was measured by light scattering at 320 nm in a Cary 210 Varian spectrophotometer equipped with a thermostatted cell compartment preheated to 45°C. The effect of HSP16-2 on the chemicallyinduced aggregation of CS was measured in an SPF-500C stirred-cell spectrofluorometer with excitation and emission wavelengths set to 500 nm and a band pass of 2 nm. CS was denatured in 6 M guanidinium chloride for at least 45 min and diluted 100-fold in a total volume of 2 ml to give a final monomeric concentration of 200 nM. All aggregation experiments were carried out in 50 mM NaPO 4 buffer, pH 7.5, and normalized to control experiments in which aggregation with CS alone was defined as being 100% of total CS aggregation.
Native Gel Analyses of HSP16-2⅐actin and HSP16-2⅐tubulin Binary Complexes-The kinetics of HSP16-2⅐actin and HSP16-2⅐tubulin binary complex formation were measured by diluting 35  Appropriate dilutions of HSP16-2 were prepared using "refolding buffer" (25 mM MES, pH 7.5, 25 mM NaCl, and 0.5 mM DTT). At each time point, 20 l of the reaction mixture was mixed with 5 l of glycerol and frozen immediately in an ethanol/dry ice bath. Samples were then separated on 4.5% nondenaturing gels (58). To measure the binding of HSP16-2 actin and tubulin intermediates formed along the aggregation pathway, 35 S-labeled denatured actin or tubulin was first diluted into refolding buffer and incubated for various lengths of time. At each time point, an aliquot was added to a tube containing HSP16-2 such that the concentration of HSP16-2 oligomers was 1.6 M and that of actin/ tubulin approximately 225 nM. The binding was allowed to proceed for 20 min at room temperature, and the reactions were analyzed on nondenaturing polyacrylamide gels. The yield of binary complex was quantitated using a PhosphorImager.
Fluorescence Measurements-All measurements were made at 20°C in a Spex fluorolog 2 spectrofluorometer. The excitation monochromator was set at 345, 468, and 492 nm and the emission recorded at 386, 535, and 517 nm for pyrenyl-actin, NBD-actin, and DTAF-tubulin, respectively.
Miscellaneous-SmHSP protein concentrations were determined using the Bio-Rad protein assay kit and bovine IgG as standard. The concentration of CS was determined by absorbance at 280 nm using an extinction coefficient of 1.55 ϫ 10 Ϫ5 M Ϫ1 cm Ϫ1 (67). The concentration of HSP16-2 and H 6 ⌬130 -145 complexes were calculated according to the average of the two molecular masses obtained by sedimentation velocity measurements (317 and 421 kDa, respectively). All other smHSP concentrations were based on the calculated molecular weights. Actin and tubulin protein concentrations were determined by the Lowry method (68). Molecular weights of 42,000 and 100,000 were employed for actin (69) and tubulin (70), respectively.

Production of HSP16-2 Variants for Structure-Function Studies
Although the sizes of smHSPs vary considerably, a detailed sequence comparison of known members suggests that the minimal functional unit consists of a core region of about 85 amino acids (the ␣-crystallin domain) which is flanked by an N-terminal region of at least 39 residues (in E. coli IbpA), and a C-terminal extension of at least 12 residues (in Neurospora crassa HSP30). Accordingly, the most compact smHSP characterized to date is E. coli IbpA, which has a very short C terminus of 14 residues and totals 137 amino acids (15.8 kDa). A novel family of 12.2-12.6 kDa smHSPs from C. elegans represents the only known exception to the apparent minimum size restriction of smHSPs; interestingly, one of these proteins, HSP12.6, was shown to lack the typical quaternary structural and functional characteristics of smHSPs (51). C. elegans HSP16-2 (145 amino acids) is only slightly larger than IbpA, and is therefore a good model smHSP for delineating the structural requirements for smHSP oligomerization and chaperone activity in these proteins. Expression constructs harboring wild-type HSP16-2 and six derivatives were created (Fig. 1A). All derivatives have a 4-kDa N-terminal tag which contains a polyhistidine sequence (H 6 ). H 6 HSP16-2 represents wild-type HSP16-2 with an N-terminal H 6 fusion. H 6 ⌬1-15, H 6 ⌬1-32, and H 6 ⌬1-44 HSP16-2 derivatives have approximately onethird, two-thirds, and all of the N-terminal domain of HSP16-2 deleted, respectively. Lastly, H 6 ⌬130 -145 HSP16-2 is equivalent to H 6 HSP16-2 with the last 16 C-terminal residues removed. All smHSPs were produced in E. coli and purified to Ն90% homogeneity (Fig. 1B). An alignment of C. elegans HSP16-2 and HSP12.6, as well as other protein sequences from various organisms (Fig. 1C), illustrates the three distinct regions found in smHSPs, and depicts the regions of the HSP16-2 FIG. 1. Summary of smHSP constructs, purified proteins, and alignment. A, schematic of all smHSPs used in this study, drawn approximately to scale. The two-exon structure (exon 1, light gray, and exon 2, dark gray) applies to hsp16-2 and hsp12.6, as both have a single intron in the same relative position. The polyhistidine-containing pRSET tag is shown in white. B, purified smHSPs separated on a 15% SDS gel. Lanes 1-7 represent the constructs shown in B from top to bottom, respectively. C, amino acid sequence alignment of smHSPs (with Genbank accession numbers): C. elegans HSP16-2 (M14334), HSP12.6 (also identified as F38E11.2; Z68342 and U92044), HSP16 -48 (K03273), SEC-1 (Z35640); bovine ␣A-crystallin (M26142); murine HSP25 (L07577); Drosophila HSP27 (J01101); soybean HSP17.6 (M11317); M. tuberculosis HSP16.3 (M76712). The first block of aligned sequences corresponds to the N-terminal domain, the second block corresponds to the ␣-crystallin domain, and the third block shows the C-terminal extension. Identical amino acids present in at least 4 of the 9 sequences are highlighted in black, and structurally similar amino acids (at least 4 of 9) are shaded. Arrowheads indicate where N-and C-terminal truncations to HSP16-2 were made. derivatives which are truncated relative to wild-type HSP16-2, HSP12.6, and other smHSPs.

HSP16-2 Quaternary Structure and Subunit Orientation
We first estimated the size of recombinant wild-type HSP16-2 by SEC on a calibrated Sephacryl S-300HR column. The majority of HSP16-2 elutes as a single peak between the molecular weight markers thyroglobulin (669 kDa) and ferritin (440 kDa), and has an apparent molecular mass of 550 kDa ( Fig. 2A). The estimated size of the complex is essentially identical to that of HSP16-2 isolated from a heat-shocked nematode extract (71). Interestingly, the H 6 HSP16-2 protein also assembles into large oligomeric complexes of somewhat larger size (680 kDa), as judged by SEC (Fig. 2B).
Sedimentation velocity measurements confirm the oligomeric nature of HSP16-2. Raw sedimentation data were fitted by nonlinear least-squares procedures as described by Philo (54). The data fitted relatively well to a two-component system involving a 10.5 S and a 14.7 S species with proportions of about 58 and 42%, respectively, at an HSP16-2 concentration of 1.0 mg/ml. Using the relation (s 1 /s 2 ) 3 ϭ (M 1 /M 2 ) 2 and bovine serum albumin as a reference (57), we obtained apparent M r values of 239,470 and 394,920 for the two species, consistent with the behavior of wild-type HSP16-2 as two populations of oligomers consisting of approximately 14 and 24 subunits. The frictional ratio values (f/f 0 ) suggest that HSP16-2 is slightly asymmetrical. The above sedimentation velocity data are summarized in Table I.
Since the H 6 HSP16-2 derivative also forms an aggregate, we determined whether its polyhistidine-containing tag was exposed. Aliquots of native or urea-denatured H 6 HSP16-2 were incubated with nickel-chelate affinity resin, and after washing the resin to remove nonspecifically adsorbed proteins, specifically bound protein was eluted with 100 mM EDTA. No detectable binding of the native H 6 HSP16-2 to the nickel resin was seen, whereas the unfolded H 6 HSP16-2, with its exposed H 6 tag, bound efficiently (Fig. 3). This suggests that the N-termi-

TABLE I Hydrodynamic parameters of wild-type HSP16 -2 and two truncated forms
The conformational parameters were calculated as described under "Materials and Methods," using the molecular mass and the partial specific volume (v) values determined from the amino acid composition. Experimental points representing the distribution of each molecular form were decomposed into Gaussian curves. The surface of each Gaussian is proportional to the concentration of the corresponding oligomeric form. Numerical values for the n-mer of each species were obtained by dividing the experimental molecular mass of each species by the corresponding theoretical molecular mass. s 0 20,w is the sedimentation coefficient; f and f 0 are the frictional coefficients and R s the Stokes radius. nal region of wild-type HSP16-2 is sequestered within the interior of the native complex.

HSP16-2 Interacts with Unfolded Citrate Synthase and Prevents Its Aggregation
To confirm that HSP16-2 possesses the chaperone activity of typical smHSPs, its effect on the aggregation of CS was studied. The denaturation and renaturation of CS has been well studied (72), and CS has been used in chaperone studies with HSP90 and associated proteins (73)(74)(75), GroEL (76), and smH-SPs (30,34,35). When CS is incubated alone at 40 -45°C it aggregates irreversibly, as detected by light scattering at 320 nm, and loses activity. Incubation of CS with increasing amounts of HSP16-2 results in a concomitant decrease in CS aggregation (Fig. 4A). At molar ratios of HSP16-2 complex to CS monomer of 1:1 or higher, aggregation of CS is almost completely inhibited (summarized in Fig. 4B). The H 6 HSP16-2 variant is also equally effective in preventing CS aggregation (51). When HSP16-2 was added to a solution of CS which had aggregated for 15 min, further aggregation was prevented but existing CS aggregates were not re-solubilized (not shown). This behavior of HSP16-2 parallels that of bovine ␣-crystallin and M. tuberculosis HSP16.3 (30,77).
The aggregation of proteins at high temperatures may involve unfolded intermediates which are different from those generated when the proteins are diluted into buffer from chaotropes such as urea or guanidine hydrochloride. We therefore examined the effect of HSP16-2 on the aggregation of 6 M guanidine hydrochloride-denatured CS which occurs upon 100fold dilution of the chaotrope. It was found that HSP16-2 interacts with and stabilizes CS diluted from denaturant essentially as efficiently as it prevents the thermally-induced aggregation of CS (Fig. 4C).

HSP16-2 Forms Stable Binary Complexes with Denatured Actin and Tubulin
Although smHSPs have been shown to bind to and stabilize actin filaments (38,41,45), detailed studies on the nature of the interaction between smHSPs and actin have not been reported. To further explore this property of smHSPs, we first investigated the formation of binary complexes between HSP16-2 and unfolded ␤-actin using a gel mobility assay.  6). The presence of a 2-fold molar excess of BSA has no effect on binary complex formation between HSP16-2 and denatured actin (lane 7), indicating that HSP16-2 but not BSA binds to the unfolded polypeptide.
A time course of binary complex formation between HSP16-2 and unfolded actin (Fig. 5B) reveals that approximately 70% of the binary complexes are formed within the first minute of incubation. Interestingly, identical kinetic experiments performed with HSP16-2 and labeled, denatured ␤-tubulin gave very similar results (Fig. 5C). This demonstrates that HSP16-2 FIG. 3. Affinity of native and denatured H 6 HSP16-2 to nickel affinity resin. Separate samples of native HSP16-2 complex and 6 M urea denatured complex were incubated with Ni 2ϩ -chelate affinity resin equilibrated in buffer without or with 6 M urea, respectively. The resins were then washed with a pH 6.3 buffer without or with 6 M urea, respectively, and the first two washes without urea were combined and analyzed on a 12% SDS gel (lane 1). Following the washes, any protein bound to the resin was eluted with EDTA with 2 column volumes. The EDTA washes without urea are shown in lanes 2 and 3, and the EDTA washes with urea are shown in lanes 4 and 5. Note that only the denatured HSP16-2 bound to the resin and therefore was eluted in the EDTA washes. has similar affinities for unfolded actin and tubulin, despite the fact that these proteins share no sequence homology. Although the binding of HSP16-2 to denatured actin is initially very rapid, it then proceeds with observable kinetics until a plateau is reached after about 20 min. The slower phase of binary complex formation may indicate that the actin intermediates formed at later times are less well recognized by the smHSP or that the available binding sites become progressively saturated.
In light of the above results, the ability of HSP16-2 to interact with actin and tubulin intermediates formed along the aggregation pathway was investigated. These intermediates are distinct from the molten-globule intermediates formed by non-cytoskeletal proteins in that they form relatively slowly (58). A constant amount of HSP16-2 was added to a reaction mixture containing actin which had been incubated alone in buffer for varying amounts of time. The binding reactions were allowed to proceed for 20 min, and then were analyzed on a nondenaturing gel (Fig. 5D). Under these conditions, the yield of binary complex formed decreases to almost zero over time. After a brief 2-min preincubation of the actin alone, a significant proportion (about 25%) of the actin fails to form a complex with HSP16-2, and after a 5-min preincubation, only approximately 50% of the actin is found in a binary complex, relative to the amount of actin bound to HSP16-2 in the absence of a preincubation. The same experiment with HSP16-2 and ␤tubulin gave almost identical results (Fig. 5E). It is noteworthy that the chaperonins HSP60 and CCT also have a decreasing ability to recognize actin intermediates which form during aggregation over the same time period (58). Overall, these studies suggest that HSP16-2 has a distinctly greater affinity for unfolded actin and tubulin intermediates which occur early on the aggregation pathway, but displays little protein specificity.
The affinity of HSP16-2 for denatured and native forms of actin and tubulin was further examined using fluorescentlylabeled derivatives. Addition of increasing amounts of HSP16-2 to denatured pyrenyl-actin and DTAF-tubulin resulted in a quenching of the fluorescence of these proteins (Fig. 6, A and  B). These data are well fitted on the simplifying assumption that HSP16-2 binds to labeled, denatured actin and tubulin with a single association constant of 2.8 ϫ 10 Ϫ6 M Ϫ1 . The stoichiometry of binding between the unfolded protein and the HSP16-2 complex is 1:1. In contrast, addition of up to a 5 M excess of HSP16-2 complex to the native pyrenyl-or NBDlabeled actin and DTAF-tubulin did not result in a quenching of the fluorescence of the labeled proteins. This further demonstrates that HSP16-2 has an equally high affinity for unfolded actin and tubulin, and also reveals that HSP16-2 has little or no affinity for the same native proteins. The native HSP16-2 oligomeric species and BSA are indicated by arrows, while binary complexes between HSP16-2 and unfolded actin are shown in the autoradiogram. The kinetics of binary complex formation between HSP16-2 and ␤-actin (B) or ␤-tubulin (C) were measured by diluting the labeled, unfolded target proteins into reactions containing a fixed amount of HSP16-2, incubating for the lengths of time shown, and analyzing the reactions on native gels as in A. The relative yields of binary complexes were quantified by PhosphorImager. The binding of HSP16-2 to ␤-actin (D) or ␤-tubulin (E) intermediates formed on the aggregation pathway was measured by diluting the labeled, unfolded target protein into buffer alone, incubating for the time periods shown, and then adding a fixed amount of HSP16-2. Binding reactions were allowed to proceed for 20 min before analysis on native gels as before.

In Vitro Mutagenesis of HSP16-2
Effects of C-terminal Truncation-The C-terminal extensions of smHSPs are variable in size and sequence, but often contain a region of limited similarity, having the motif (R/ K)X(I/V)X(I/V) (Fig. 1C). We therefore sought to determine whether the C-terminal extension of HSP16-2 (including the conserved motif RSIPI) plays a structural or functional role by studying H 6 ⌬130 -145 HSP16-2. Analysis of H 6 ⌬130 -145 HSP16-2 by SEC reveals that like H 6 HSP16-2, the N-terminally-tagged variant forms a somewhat larger aggregate than wild-type HSP16-2, with an estimated M r of 675,000 (Fig. 2B). This was confirmed by sedimentation velocity measurements (Table I), which as is the case for the wild-type protein, show that H 6 ⌬130 -145 HSP16-2 exists as two high M r species. The M r values are 322,395 and 520,385, consistent with oligomeric complexes of approximately 17 and 28 subunits.
We then tested the ability of H 6 ⌬130 -145 HSP16-2 to protect CS from aggregating at an elevated temperature (Fig. 7). Surprisingly, removal of the C-terminal extension has little or no influence on the chaperone function of HSP16-2: H 6 ⌬130 -145 HSP16-2 is as effective as the wild-type smHSP in suppressing CS aggregation (compare with Fig. 4B). Confirmation that this derivative interacts with unfolded proteins was obtained from fluorescence quenching experiments on pyrenyllabeled actin (Fig. 6C) and NBD-actin and DTAF-tubulin (not shown).
Interestingly, over 90% of H 6 ⌬130 -145 HSP16-2 precipitates from solution after freeze-thawing, while wild-type HSP16-2 and H6HSP16-2 remain completely soluble following the same treatment (not shown). This suggests that the solubility or perhaps stability of the HSP16-2 complex is compromised in the C-terminal truncation mutant.
Effects of N-terminal Deletions-Perhaps the most conspicuous feature of smHSPs is the presence of a nonconserved Nterminal domain which is highly variable in size. Three Nterminal truncations of HSP16-2, ⌬1-15, ⌬1-32, and ⌬1-44 were used to probe the function of this domain. To examine the oligomeric nature of these N-terminal truncations as well as of other HSP16-2 derivatives, the purified proteins were subjected to chemical cross-linking using the homobifunctional cross-linker BS 3 (78). To minimize nonspecific intermolecular cross-linking of non-oligomerized smHSP subunits, cross-linking reactions were carried out in the presence of low protein (1.5 M monomer) and BS 3 (2 mM) concentrations, and a 10-fold molar excess of BSA (15 M). Reactions were analyzed by Western blotting with an anti-HSP16-2 polyclonal antibody (Fig. 8). All of the proteins yielded discernible monomeric, dimeric, and trimeric species. From the large number of products present in the high molecular mass range (above 100 kDa and at the origin of the separating gel), it can be concluded that the HSP16-2, H 6 HSP16-2, and H 6 ⌬130 -145 proteins also form higher-order oligomers which are too large to be well resolved on the 12% SDS gel. In contrast, the H 6 ⌬1-15, H 6 ⌬1-32, and H 6 ⌬1-44 HSP16-2 proteins yielded essentially no high M r cross-linked products but could be cross-linked to trimers (and possibly tetramers), suggesting that they exist as small aggregates. The presence of the N-terminal H 6 tag in each of these truncated proteins is unlikely to have resulted in disruption of the multimers since both H 6 HSP16-2 and H 6 ⌬130 -145 HSP16-2 are able to form large oligomeric complexes.
The sizes of the N-terminally-truncated smHSPs were estimated by SEC on a Sephacryl S-300HR column. These smHSP derivatives elute as proteins of 45-115 kDa (Fig. 2B), and therefore appear to form small aggregates of trimers-heptamers. These results resemble those obtained on the N-terminally-deleted ␣A-, ␣B-crystallin, and HSP25 (79,80). To obtain more accurate data on the subunit stoichiometry of one of the truncated smHSPs, sedimentation velocity measurements were performed on H 6 ⌬1-44 HSP16-2. These yielded molecular weights of 14,990 and 56,235, consistent with the behavior of H 6 ⌬1-44 HSP16-2 as a monomer (95%) with a small proportion (5%) of tetramers (Table I).
Since the N-terminally-truncated smHSPs did not form native-like oligomeric complexes, it was of interest to determine whether they still possessed molecular chaperone activity. The three derivatives were therefore tested for an ability to prevent thermally-or chemically-induced aggregation of CS. None of these smHSPs was effective in suppressing CS aggregation, even at a molar excess of 65 molecules of smHSP to 1 CS monomer (not shown). Furthermore, H 6 ⌬1-44 HSP16-2 did not quench the fluorescence of denatured forms of pyrenyl-actin (Fig. 6D) or NBD-actin and DTAF-tubulin, and had no effect on the fluorescence of the native labeled proteins (not shown).

The Properties of a 12.6-kDa SmHSP Support the HSP 16-2 Structure-Function Analyses
The C. elegans genome contains at least four highly conserved members of a novel class of smHSPs, of which one member (HSP12.6) is shown aligned to other smHSPs in Fig.  1C. HSP12.6 possesses a well conserved ␣-crystallin domain, but in comparison to other smHSPs, it has both the smallest N-terminal domain at 25 amino acids, and the shortest Cterminal extension, with only 2 amino acids. The present studies on wild-type HSP16-2 and derivatives would predict that  8. Chemical cross-linking of smHSPs. Samples of HSP16-2 and five derivatives were cross-linked with the cross-linking agent BS 3 , separated on a 12% SDS gel, transferred to a polyvinylidene difluoride membrane, and probed with an anti-HSP16-2 antibody. Monomers and cross-linked products corresponding to dimers, trimers, and tetramers are indicated. The separating gel origin and molecular weight markers are also shown.
HSP12.6 should be unable to multimerize due to its short N-terminal domain, and as a result should lack molecular chaperone activity. Indeed, our recent characterization of HSP12.6 revealed that unlike other smHSPs, HSP12.6 is monomeric and is unable to prevent the thermally-induced aggregation of citrate synthase (51).
We investigated whether HSP12.6 could interact with unfolded proteins even though it is incapable of preventing their aggregation. As with the mostly monomeric H 6 ⌬1-44 HSP16-2, HSP12.6 did not interact with unfolded, pyrenyl-labeled actin (Fig. 6D). Similarly, no quenching of fluorescence was obtained with native actin, or with native or unfolded tubulin (not shown). The structural and functional data on HSP12.6 therefore support the notion that the distal end of the smHSP N-terminal domain is required for multimerization, and that chaperone activity is only possible in the context of an oligomeric assembly. DISCUSSION The HSP16s are among the smallest members of the ␣crystallin family of HSPs, and we have shown here that they possess the multimeric structure and chaperone activities typical of these proteins. Very few mutational studies have been carried out on smHSPs, and these have concentrated mainly on mammalian ␣-crystallins (79 -83). To our knowledge, the present studies on HSP16-2 represent the most extensive structure-function analysis of a stress-inducible smHSP reported to date. The small size of the HSP16s is an advantage in such studies, and the combination of in vitro mutagenesis with functional chaperone assays, substrate binding measurements, and sedimentation studies reported here provide a number of important insights into the structure and function of smHSPs.
Structure of HSP16-2-Sedimentation velocity measurements reveal that recombinant wild-type HSP16-2 forms heterogeneous mixtures of two particles of 238 and 412 kDa, suggesting that the number of subunits per complex is approximately 14 and 24, respectively. In contrast, recombinant and natural HSP16-2 elutes as an ϳ550-kDa oligomer by SEC. The reason for the discrepancy between the two sizing techniques is unclear; it is possible that the HSP16-2 aggregates are loosely packed, or that they are non-globular as suggested by the frictional ratio values obtained (Table I). Examination of HSP16-2 by electron microscopy reveals globular but relatively heterogeneous particles with diameters ranging between 10 and 15 nm (not shown). Interestingly, a similar apparent dichotomy exists for mammalian HSP27, which by SEC behaves as an ϳ800-kDa oligomer, whereas by native gel electrophoresis migrates as a doublet of approximately 200 and 250 kDa (43). In sucrose gradients, HSP27 sediments as two distinct forms, one of 300 -400 kDa and another of ϳ65 kDa (84).
The H 6 HSP16-2 variant, which forms fully active oligomeric complexes that are slightly larger than those of wild-type HSP16-2, was used to determine the orientation of the HSP16-2 monomers within the native complex. It was found that the 4-kDa N-terminal polyhistidine-containing tag was inaccessible for binding to a Ni 2ϩ -chelate affinity matrix. Thus either the tag is collaterally positioned in the complex and masked by the HSP16-2 protein and neighboring subunits (which would very likely have a detrimental effect on the assembly of the complex), or, more likely, the tag is accommodated within a central cavity in the complex. Furthermore, the fact that HSP16-2 can accommodate an additional 4 kDa (36 residues) at its N terminus, making this domain larger than those of ␣-crystallins and Drosophila HSP27, and only 8 residues shorter than that of murine HSP25 (refer to the alignment in Fig. 1C), suggests that the arrangement of subunits within smHSP complexes provides considerable freedom for N-termi-nal regions of different lengths and amino acid sequence. Indeed, the alternatively-spliced ␣A ins -crystallin variant which contains a 23-residue insertion near the N-terminal/␣-crystallin domain boundary also assembles into oligomers (85). Interestingly, the increased length of the N-terminal region in the ␣A ins -crystallin leads to an increase in the size of the native complex, a situation analogous to that seen with H 6 HSP16-2.
The present work shows that deletion of only approximately one-third (15 residues) of the N-terminal domain of HSP16-2 results in a dramatic reduction in the size of the complex as judged by SEC and cross-linking analysis, suggesting that the region involved in multimerization may be distally located in the N-terminal domain. It is therefore conceivable that the HSP16-2 monomers have an extended conformation with the N-terminal regions oriented toward the interior of the complex and providing the necessary contacts for subunit aggregation. In support of this hypothesis, HSP12.6, which has 16 fewer residues than HSP16-2 in the N-terminal region, is monomeric and predicted to be asymmetrical by sedimentation velocity analysis (51).
The studies by Merck and co-workers (79,80) suggest the existence of at least two major sites of interactions between smHSP monomers: one within the N-terminal domain, which permits the assembly of the complex, and the other within the C-terminal domain, which appears to have an inherent ability to form smaller oligomers. Similarly, Chang et al. (30) found that mycobacterial HSP16.3 nonamers could be disassembled into trimers by treatment with 4 M urea or 1 M guanidine hydrochloride, suggesting that two different types of subunit interactions exist: one involved in trimer formation and another which participates in the oligomerization of trimers. While our SEC and cross-linking experiments on N-terminallytruncated HSP16-2 mutants also suggest that smHSP complexes may assemble from small oligomers, sedimentation velocity measurements of H 6 ⌬1-44 HSP16-2 and HSP12.6, which possess complete ␣-crystallin domains, show these proteins to be largely monomeric. While it is therefore possible that Nterminally-truncated smHSPs exist as monomers that are in slow equilibrium with higher-order oligomeric species, early cross-linking experiments using a series of cross-linkers with different spacer lengths suggest that ␣-crystallin aggregates are not likely to be built up of smaller clusters (86), and a recent study provides evidence that the minimal cooperative unit of ␣-crystallin is the monomer (87).
The removal of the last 16 residues of HSP16-2 has no effect on multimerization, demonstrating that this region does not participate in subunit assembly. Interestingly, the C-terminal truncation affects the solubility of the oligomer, perhaps by destabilizing the native protein structure. A role for the Cterminal extension in the solubilization of ␣-crystallin particles has recently been suggested (82). It has also been shown by 1 H NMR studies that the last 8 -10 and 18 C-terminal residues of ␣-crystallins and HSP25, respectively, form flexible extensions on the surface of the oligomers (27,28). Taken together, these data suggest that the C-terminal extensions of HSP16-2 and other smHSPs are on the surface (exposed to solvent), where they impart increased solubility and perhaps stability to the complexes.
Function of SmHSPs-Under stress conditions, heat shock proteins such as HSP60, HSP70, HSP90, and smHSPs likely prevent partially unfolded proteins from aggregating and becoming insoluble (33,34,74,88,89). Whereas most HSPs perform necessary functions in non-stressed cells and are also induced following cellular insults (1, 10, 15), the C. elegans HSP16-class smHSPs are produced exclusively when nematodes are exposed to a heat shock or to proteotoxic agents (48,50). Our studies on HSP16-2 confirm the ability of this specialized class of chaperone to interact with and stabilize proteins which become unfolded following heat or chemical treatment.
It is noteworthy that in eukaryotes, cytoskeletal proteins are one of the most sensitive targets of cellular stresses, and there is increasing evidence that smHSPs specifically interact with cytoskeletal elements, modulating their polymerization, and protecting them during stress conditions (13, 36 -38, 40, 41, 43). We have shown that the affinity of HSP16-2 for native actin and tubulin is negligible, whereas it is very high for the same thermally-or chemically-unfolded proteins. The interaction of cardiac ␣B-crystallin with actin and desmin filaments is enhanced during ischemic conditions, where a decrease in the pH of the cytosol is known to affect the stability of cytoskeletal elements (38). HSP16-2 has nearly identical affinities for actin and tubulin, and can also prevent the aggregation of citrate synthase, suggesting that this chaperone has a general affinity for unfolded proteins. Other smHSPs also display little protein specificity, so whether smHSP chaperones have specialized roles in protecting the cellular cytoskeleton network during stresses remains to be seen.
Recent data suggest that ␣-crystallin preferentially interacts with thermally-generated early unfolding intermediates which have exposed hydrophobic regions and are committed to the aggregation pathway (90 -92). The present study demonstrates that HSP16-2 is equally capable of preventing the aggregation of thermally-and chemically-unfolded citrate synthase. Furthermore, we show that HSP16-2 has a considerably greater affinity for actin and tubulin unfolded intermediates which occur early on the aggregation pathway, and has little or no affinity for native or aggregated proteins.
The mechanism by which smHSPs interact with and prevent unfolded polypeptides from aggregating is unknown. Although it is generally agreed that hydrophobic regions in smHSPs are important for polypeptide binding (83,93), hydrophilic interactions may be necessary as well (81,94). Interestingly, Lee et al. (95) recently found that a hydrophobic reporter molecule, bis-ANS, could be photoincorporated into specific regions of pea HSP18.1 which may be important for substrate binding. One site of incorporation was localized to the sequence ADLPGLK-KEEVKV (homologous regions found 13 residues into the ␣crystallin domains of other smHSPs; see Fig. 1C). This region lies within the first 30% or so of the ␣-crystallin domain, and is well conserved in HSP16-2 and other smHSPs (Fig. 1C). This region was not altered in our deletion studies. The fact that the inactive, low-M r HSP16-2 N-terminal deletion mutants and the naturally monomeric HSP12.6 both retain this sequence but lack chaperone activity lends support to the idea that multimerization is essential for substrate binding.
Since most of the C-terminal extension of HSP16-2 is dispensable for polypeptide binding and chaperone activity, and is likely to be on the surface of the oligomeric complex, the interaction with unfolded polypeptides presumably involves other regions of the smHSP, such as the clefts formed between subunits. Such clefts can only exist in the context of an oligomeric assembly, which we have shown is necessary for polypeptide binding. Moreover, studies with unstructured spin-labeled peptides suggested that these were not bound on the surface or clustered in an interior cavity of the ␣-crystallin complex, but rather were strongly immobilized at least 25 Å apart in polar environments (94). A detailed computer-generated model of the interaction between partially unfolded ␥-crystallin and ␣-crystallin also predicts the site of polypeptide interaction to be within clefts in the ␣-crystallin complex (96).
Structure-Function Model for SmHSPs-Molecular chaperones such as those belonging to the HSP60 and HSP104 fami-lies are active as homo-or heteroligomers, requiring multiple subunit contacts with the unfolded polypeptide for activity (5,97). SmHSPs also form large oligomeric assemblies of one or two types of subunits, but their quaternary structures are much more variable, ranging from toroids (98) to roughly globular particles of various sizes, shapes, subunit stoichiometries, and compositions (29,30,35). Several plausible models for smHSP quaternary structures have been proposed (reviewed in Ref. 29), but the lack of detailed structural data do not yet allow a definitive choice to be made among these. However, the micellar-like model first proposed by Augusteyn and Koretz (99) seems best able to accommodate the variability in subunit stoichiometry and particle morphology seen in many smHSPs, as well as the findings reported here, i.e. that a heterologous protein sequence, appended to a distal N-terminal region essential for multimerization, can be sequestered within the particles. Furthermore, our data support a model in which multimerization is a prerequisite both for the interaction of the chaperone with unfolded protein and for chaperone activity. A model consistent with our findings and summarizing some structural and functional characteristics of smHSPs is presented in Fig. 9.
Possible Function of HSP16-2 in Vivo-The function of HSP16-2 in vivo may be to interact with and stabilize a large variety of unfolding proteins, effectively buffering the cell against the deleterious effects caused by misfolded and aggregated proteins. Linder et al. (100) reported on a novel 18-kDa C. elegans protein most similar (32-38% overall amino acid sequence identity) to the C. elegans HSP16 smHSPs and named SEC-1 (for small embryonic Chaperone-1) (see Fig. 1C for align- FIG. 9. Model for smHSP oligomeric structure and interaction with unfolded polypeptides. This schematic depicts a hypothetical cross-section through a smHSP complex with an arbitrary number of subunits and quaternary structure. SmHSP monomers are shown as having a two-domain structure consisting of the N-terminal domain and C-terminal domain (which includes the ␣-crystallin domain and Cterminal extension). Double-headed arrows indicate possible sites of interaction between different smHSP monomers and between smHSP subunits and unfolded substrates. a, a proposed feature of smHSPs is their central cavity, which can accommodate N-terminal domains of varying lengths and sequence. b, N-terminal domain interactions are necessary for subunit assembly and are likely to modulate the arrangement of subunits within the smHSP quaternary structure. The multimerization domain may be restricted to a small section (e.g. the distal end) of the N-terminal region. c, the chaperone activity of smHSPs relies on their ability to interact with and stabilize unfolded polypeptides; this interaction is likely to be mediated through multiple contacts with ␣-crystallin domains in clefts formed by neighboring subunits. d, although the role of the C-terminal extension in the interaction of the smHSP with unfolded polypeptides is unclear, it is probable that this region is exposed to solvent and is largely responsible for maintaining the solubility of the smHSP complex. e, there is evidence that smHSPs assemble cooperatively from small oligomers, but the exact region of interaction between subunits, and whether this is a general feature of smHSPs, is unknown. ment for sequence). SEC-1 is detectable only during early embryogenesis, where it performs an essential but unknown function. Interestingly, SEC-1 confers thermotolerance to E. coli, even though its expression is not augmented by environmental stresses. Therefore, although the physiological functions of this smHSP are not redundant with those of the stress-inducible HSP16s, it is likely that both possess similar structural and functional characteristics as molecular chaperones.
Despite considerable effort, conditions were not found under which HSP16-2 could refold denatured test proteins in vitro. The fate(s) of the unfolded proteins bound to smHSPs is poorly understood. It is now established that many molecular chaperones cooperate in refolding denatured proteins (6,89,101,102), and are involved in targeting proteins for degradation (Ref. 103, also reviewed in Ref. 104). Recently, HSP25 and HSP70 were shown to act in concert in vitro in the refolding of citrate synthase (105), and Lee et al. (95) demonstrated that denatured firefly luciferase bound to HSP18.1 could be refolded in the presence of rabbit reticulocyte lysate (which contains the chaperones necessary for refolding denatured luciferase) and ATP. It is therefore likely that in vivo, the smHSP-bound unfolded proteins are folded by molecular chaperones such as those involved in de novo protein folding (HSP40/HSP70, HSP60; Ref. 5), or under certain conditions are degraded by the cellular proteolytic system.