Molecular Vision 2005; 11:641-647 <http://www.molvis.org/molvis/v11/a76/> Received 7 June 2005 | Accepted 22 August 2005 | Published 29 August 2005

Download Reprint

NH₂-terminal stabilization of small heat shock protein structure: a comparison of two NH₂-terminal deletion mutants of αA-crystallin

Chaoxing Yang, John C. Salerno, Jane F. Koretz
 
 

Biochemistry and Biophysics Program, Rensselaer Polytechnic Institute, Troy, NY

Correspondence to: Jane F. Koretz, PhD, Biochemistry and Biophysics Program, Rensselaer Polytechnic Institute, Science Center, 110 8th Street, Troy, NY, 12180-3590; Phone: (518) 276-6492; FAX: (518) 276-2344; email: koretj@rpi.edu

Abstract

Purpose: To assess the role of the NH₂-terminal in αA-crystallin folding and chaperone-like activity.

Methods: Two NH₂-terminal deletion mutants of αA-crystallin were generated by standard mutagenesis methods, one with and one without a leader sequence in place of the first 50 residues. Aggregate size of each before and after thermal stress was assessed by FPLC, and chaperone-like activity was assessed using DTT-induced insulin denaturation.

Results: Both mutants assemble primarily into tetramers, and both exhibit similar levels of chaperone-like activity, but are less protective than recombinant αA-crystallin. After a cycle of heat stress to 70 °C, tetramers of the mutant without the leader sequence dissociate into dimers and monomers and show severely reduced chaperone-like activity. In contrast, the mutant with the leader sequence retains its tetrameric form and its chaperone-like activity.

Conclusions: The NH₂-terminal region is an important determinant of α-crystallin aggregate size, but is not required for folding of the α-crystallin domain, since the aggregate size and chaperone-like activity of the two mutants at room temperature are essentially the same. The leader sequence appears to increase the thermal stability of the α-crystallin domain and/or to contribute to the reformation of the active form after cooling, suggesting that the native NH₂-terminal also plays a role in α-crystallin's resistance to environmental stress.

Introduction

Crystallins are the predominant proteins in the mammalian eye lens, constituting more than 90% of its total dry mass, and approximately one third of this is accounted for by α-crystallin [1-3]. The α-crystallins are expressed as two isoforms, αA-crystallin and αB-crystallin, which are found, respectively, in the crystalline lens in a 3:1 M ratio. They are separately represented in other organs and tissues, however, with αB-crystallin the more common nonlenticular isoform [4-8].

α-Crystallin belongs to the small heat shock protein (sHsp) superfamily, members of which span all five kingdoms [9,10]. As is indicative of all sHsps, the primary sequence of α-crystallin is characterized by the presence of a highly conserved "α-core" region flanked by a variable NH₂-terminal region and a variable COOH-terminal region [10-14]. sHsp subunits show a range of molecular masses from 12 to over 40 kDa, with that of α-crystallin subunits about 20 kDa. There is also considerable variation in the size and organization of quaternary assemblies of these subunits from species to species. Although many sHsps exhibit characteristic numbers of subunits and quaternary symmetries, α-crystallin aggregates exhibit heterogeneous particle sizes and subunit numbers, generally around 30-40 [15-18]. While usually constitutive, most sHsps are also overexpressed under stress conditions, suggesting that they may serve a protective function [19-28]. It has been suggested that the sHsps can bind to partially unfolded proteins to prevent their superaggregation and/or further denaturation, and can keep them in a refoldable conformation until environmental conditions return to normal. At that point, the substrate protein is released from the sHsps, and can then be refolded by other chaperone systems, such as the Hsp 70 system [29-31]. Horwitz [32] demonstrated that α-crystallins exhibit this form of chaperone-like activity in vitro. Since they can suppress aggregation of other eye lens proteins, such as β- and γ-crystallin, under high temperature, there is a possibility that α-crystallins might act as chaperones for these lens structural proteins in vivo. Prevention of superaggregation of the denatured proteins would contribute to long term maintenance of lens transparency.

Although a crystal structure for α-crystallin has not yet been obtained, the structures of other two other members of the sHsp superfamily have been solved. The X-ray structure of Mj Hsp 16.5, a small heat shock protein from Methanococcus jannaschii [33] shows that Mj Hsp 16.5 forms a hollow sphere with 24 subunits, with the subunits first binding specifically into well oriented dimers. The core region of each monomer contains nine β strands in two apposed sheets. Dimers are formed by the interaction between the β-2 strand of one monomer and the β-6 strand on the dimerization loop of another monomer. The β-2 strand also interacts with its antiparallel strand β-1 by hydrogen bonds. The β-1 strand is probably important for stabilizing the interaction between β-2 and β-6, and may also have a function in determination of the α-core orientation. A similar dimer conformation is seen in the crystallographic structure of Hsp 16.9 from wheat [34], although its oligomer consists of 12 subunits in two 6-subunit rings rather than a 24-subunit sphere. The NH₂-terminals of both 16.5 and 16.9 subunits are oriented toward the interior of the oligomers, and the COOH-terminals face the exterior. α-Crystallin and other animal sHsps exhibit subtle differences from plant and microbial varieties, as determined by a combination of theoretical and experimental approaches. Sequence alignments indicate that the superfamily α-core region dimerization loop containing the β-6 strand is too short in the equivalent animal sHsp region to serve the same function. Spin label mapping of the core region [35] of α-crystallin demonstrates the characteristic α-core topology, but appears to be missing the equivalent of the β-1 strand found in Hsps 16.5 and 16.9. In the absence of a dimer interaction with a β-6 type strand from the other monomer of the pair, it is possible that the β2-equivalent strand is stabilized by other interactions from the NH₂-terminal and/or prior core sequence. Small dimers of the αB-crystallin core region are formed when residues 1-56 (the NH₂-terminal up to the beginning of the β-1 equivalent sequence), and much of the COOH-terminal, are removed. There is also some indication that the orientation of the two monomers forming the dimer differs from that of Hsp 16.5. When a mutant with a similarly designed NH₂-terminal deletion (residues 1-50) was constructed and expressed for αA-crystallin, in contrast, the primary oligomer was tetrameric [10,36]. However, both deletion mutants exhibit chaperone-like activity, albeit at lower efficiencies than those of the parent sequences.

A major difference between the design of the αA- and αB-crystallin deletion mutants was the replacement of the NH₂-terminal sequence of the αA-crystallin with a 15-residue serine/glycine leader sequence. This hydrophilic leader sequence was developed and characterized by Robinson and Sauer [37], who showed that it could improve the solubility of mutated proteins. Its addition to the truncated αA-crystallin as a replacement for the NH₂-terminal could potentially account for the difference in oligomer size, but also might confer properties on the mutant that a strict deletion would not exhibit. To determine the possible contribution of the leader to the deletion mutant's structure and activity, deletion mutants were constructed with and without the leader in place of the first 50 residues using site-directed mutagenesis methods. The aggregate size of the mutants was analyzed with FPLC gel filtration chromatography, and the chaperone-like activity of the mutants was analyzed using the DTT induced insulin denaturation assay. The structural and functional properties of the two constructs were very similar, except for their response to thermal stress.

Methods

Mutagenesis

The leader (or linker) mutant (LM51) was constructed by using the PCR-based site-directed mutagenesis method. The forward primer was designed according to the desired NH₂-terminal deletion of residues 1-50, and the start codon, the NdeI restriction enzyme recognition site, and the leader sequences at the NH₂-terminus were incorporated. The reverse primer was designed to incorporate the stop codon, and the XhoI restriction enzyme recognition site at the C-terminus of the leader mutants.

The deletion mutant (DM51) was constructed by using the Ligation Independent Cloning site-directed mutagenesis method. The primers for the PCR reaction were specially designed according to the vector manual of Novagen (Madison, WI). The forward primer encodes the protease Factor Xa recognition site. The NH₂-terminal fusion sequences can be removed with the protease Factor Xa. The pET 30 Xa/LIC vector carries a His-tag at the NH₂-terminus of the mutated αA-crystallin DNA and a kanamycin selective marker.

Transformations and overexpression

Wild type bovine αA-crystallin DNA was cloned into vector pET 3d⁺ (Novagen), and transformed into E. coli expression strain BL21 (DE3). The circularized ligation product of LM51 was first transformed into highly competent E. coli HB 101 strain (Novagen) with ampicillin selection, then transformed into E. coli BL21(DE3) plysS strain (Novagen) with chloramphenicol selection. The annealed LIC vector of DM51 was initially transformed into competent E. coli RossetaBlue cells which have tetracycline and chloramphenicol resistances, then was transformed into E. coli expression strain RossetaBlue (DE3) pLys S using the same transformation procedure. Protein overexpression was induced by IPTG. The cells were treated with BugBuster protein extraction reagent (Novagen).

Protein purification procedures

LM51 and recombinant α-crystallin were first purified with an anion exchange column, then purified with a gel filtration column. Samples in a buffer of 20 mM Tris and 100 mM NaCl (pH 7.9) were eluted on an anion exchange column HiPrep 16/10 Q XL (GE healthcare, Piscataway, NJ) using a NaCl gradient (100 mM to 500 mM) at a speed of 5 ml per min. The fractions that contained the most pure mutant protein were pooled together and concentrated with a Centricon spin column to about 1-2 ml, then further purified by gel filtration using a HiPrep Sephacryl S-100 column (Pharmacia Biotech) equilibrated with Tris buffer (20 mM Tris, 100 mM NaCl, pH 7.9). Concentrated sample (1 ml) was loaded on the column at a speed of 0.5 ml/min, and fractions were collected every 5 min. LM51 was eluted after 50 min.

Since the DM51 mutant had a His-tag, it was purified with a His-bind resin chromatography column (Novagen) charged with NiSO₄. The protein was concentrated with a YM-3 Centrifugal filter device (Millipore, Milford, MA) and centrifuged, and its buffer was changed to a protease digestion buffer to remove vector-incorporated residues, including the His-tag. The protease was removed by agarose when the digestion reaction was complete, and the digestion products were then purified again with the His-bind column to remove peptides with the His-tag.

FPLC gel filtration column analysis

A Superose 12 HR 10/30 gel exclusion column (GE Healthcare) was washed with buffer (20 mM Tris, 100 mM NaCl, pH 7.9), then calibrated by individually running each of five protein standards β-amylase (200,000), alcohol dehydrogenase (150,000), bovine serum albumin (66,000), carbonic anhydrase (29,000), and cytochrome C (12,400) at an elution speed of 0.5 ml/min and a pressure of 150 Pa (1 Pa=1 N/m²). For each mutant protein, 200 μl of a 0.25 mg/ml solution (about 0.05 mg) in Tris buffer (20 mM Tris-HCl, 100 mM NaCl, pH 7.9) was loaded on the gel filtration analysis column. Molecular weights of the different peaks were calculated from the elution times according to the standard curve, and aggregate sizes determined using the calculated molecular weights of the subunits. A similar analysis was performed for each mutant protein after incubation at 70 °C in a water bath for 30 min followed by cooling on ice and equilibration to room temperature.

Chaperone-like activity analysis

The chaperone-like activity of each of the mutants was analyzed using the DTT induced insulin denaturation assay [38]. Bovine insulin (Sigma, St. Louis, MO) was denatured with DTT, and the chaperone-like activity of the sample indicated by the difference in turbidity of DTT induced insulin denaturation in the presence and absence of the mutant protein. The sample with insulin alone and the sample with insulin and DTT served as negative controls, while the sample with insulin and recombinant αA-crystallin served as the positive control. Both mutant protein and insulin were dialyzed in buffer (50 mM imidazole, 100 mM NaCl, pH 7.9), and concentrations were adjusted to an insulin concentration of 90 μM and a 1:1 subunit to insulin ratio. The final concentration of DTT was 20 mM, and was added to the appropriate samples to begin the experiment. A 96 well plate was used to perform the assay, and three sets of replicates were tested simultaneously. Turbidity changes were monitored at 360 nm at room temperature with a SPECTRA MAX 190 every 30 s for 30 min. A similar assay for chaperone-like activity was also performed using samples of mutant proteins that had been incubated at 70 °C in a water bath for 30 min, cooled on ice, and allowed to equilibrate to room temperature.

Results

Figure 1 shows SDS-PAGE gels of both deletion mutants after purification. The LM and DM mutants both expressed at levels comparable to the wild type protein. This indicated that both constructs fold comparably well after synthesis in vivo, and that both were stable enough to resist degradation. However, the DM mutant delivered a much lower yield than either the LM mutant or wild type αA-crystallin during the purification process.

FPLC analyses

The results of analytic FPLC showed that, as expected, neither the LM or DM construct formed high order aggregates comparable to wild type α-crystallin. This was expected because previous results have shown that aggregation above the level of tetramers depends on the presence of the hydrophobic NH₂-terminal region, and was confirmed by the absence of a peak at short elution times. The DM51 peak eluted slightly later than the major LM51 component at 20 °C (Figure 2), and this was consistent with the fact that the subunit molecular weight of DM51 was smaller than that of LM51.

The major peaks in the elution profiles of both constructs were asymmetrical. In both cases, there appeared to be a small peak at the leading edge, overlapping the major peak and partially eluting with it. On the other side of the major peak, there was a clearly evident shoulder that might contain one or two other minor peaks. Based on the calibration curve, the major peak of both mutants consisted of tetramers. The small leading edge of the major peak was presumably composed of octamers, while the trailing shoulder(s) of the mutants were probably contributed by dimers and monomers. The lack of resolution may indicate dynamic equilibration between these aggregation states on the time scale of the experiment.

The aggregate size of each mutant was further analyzed after a cycle of heating to 70 °C and cooling on ice. The major FPLC peaks representing tetramers of heat stress treated and untreated LM51 overlap closely, except that the heat stress treated LM51 had lost the small peak at the leading edge corresponding to octameric aggregates (Figure 3A). The FPLC elution profile of heat-treated DM51 was very different from that of untreated DM51 (Figure 3B). The sample exposed to heat stress exhibited two resolved peaks in the FPLC elution profile. The first peak eluted at 32 min, consistent with a tetramer, and the second peak eluted at about 34.1 min, consistent with dimers. The second peak itself was asymmetric, however, and appeared to be the sum of two peaks. The estimated elution time of the minor peak was consistent with monomers. Thus, while the direct consequence of a single cycle of heating and cooling of LM51 was the production of a more homogeneous preparation than the originally isolated material, the DM51 preparation became far more heterogeneous than the originally isolated form of either construct.

Chaperone-like activity assay

The chaperone-like activity of each mutant was tested before and after a cycle of heat stress and cooling using the optimized insulin denaturation assay at room temperature as described in Methods. The heat-treated sample for each mutant was heated up to 70 °C in a water bath for 30 min, then cooled directly on ice and equilibrated to room temperature. Both mutant proteins demonstrated much less chaperone-like activity than recombinant αA-crystallin under the optimized experiment conditions (Figure 4), but their levels of chaperone-like activity were very similar to each other. However, there was a significant difference in the chaperone-like activity of the two mutants after heat treatment. The chaperone-like activity of the heat-treated LM51 decreased only slightly compared to the untreated sample. In contrast, the chaperone-like activity of the DM51 was decreased dramatically by the heat stress treatment. The protection provided to insulin by the heat-treated DM51 against DTT denaturation was so low as to be essentially nonexistent.

Discussion

Alignment of the αA-crystallin sequence with the sequences of other small heat shock proteins, most notably examples for which the crystallographic structure has been determined, provides guidance in designing the mutants so that the α-crystallin core region is preserved [10,36]. The structure of the core region of Hsp 16.5 consists of two β sheets of four strands each, with an additional strand (labeled β-6 by Kim et al. [33]) from each of the subunits looping over to interact with one of the sheets from the other monomer of the dimer pair, thus helping to define the dimer orientation. A very similar topology, including the dimerization loop, is observed for Hsp 16.9, and the sequence alignments indicate that this loop is characteristic of the sHsp sequences from plants and a majority of the unicellular organisms. Animal sequences, in contrast, appear to lack an equivalent region long enough to initiate and maintain this interaction [10,36].

Investigation of the topology of the core region of α-crystallin using spin labeling [11] has shown that it resembles those of the two solved structures. The major difference is in the characterization of the NH₂-terminal start of the core region. In Hsp 16.5, the first β strand (termed β-1 by Kim et al. [33]) forms the edge of one of the β sheets in the core region, with the β-2 strand providing the start of the second β sheet. A very short initial β strand that is part of the first four-strand sheet is also seen in the structure of Hsp 16.9, but these initial β strands are not formed by corresponding sequence regions of Hsp 16.5 and 16.9. In Hsp 16.5, the β-1 and β-2 strands are separated only by a short turn. In Hsp 16.9, a very short end strand ("β-0") corresponds to sequence elements near the extreme NH₂-terminus, and is separated from the second strand ("β-1" in the nomenclature of van Montfort et al. [34]) by sequence elements which form two α-helices in at least some of the monomers in the crystal structure.

The equivalent of a β-1 strand (in the Hsp 16.5 nomenclature) does not appear to be present at the start of the α-core region of α-crystallin, but this portion of the sequence nevertheless appears to be important for folding and functionality [39]. The selection of residue 51 as the start of the mutant sequence of αA-crystallin was thus based on the observation that sequence similarities to other animal Hsps are first apparent at this point. A similar strategy, and an equivalent start point, was presumably used in the creation of a similar αB-crystallin mutant [35].

The αB-crystallin deletion mutant was reported to exist in solution as a dimer, in contrast to our earlier characterization of the αA-crystallin deletion mutant as a tetramer. We hypothesized that the presence of the hydrophilic leader sequence, added to ensure the solubility of the deletion mutant, might have changed the αA-crystallin deletion mutant's physical properties and functionality. However, the leader sequence did not appear to change the deletion mutant aggregate size or size distribution, as characterized by gel filtration FPLC. The predominant species by far for both DM51 and LM51 is the tetramer, with other species present in much smaller amounts. Since the leader sequence does not appear to have changed the aggregation properties of the deletion mutant, there must be another reason why the equivalent αB-crystallin deletion mutant is only a dimer. One possibility is that the two α-crystallin isoforms differ enough to lead to different aggregation patterns; this is unlikely, since the sequence homology between the two is quite high, particularly in the core region.

A more likely explanation is that truncation of a portion of the COOH-terminal sequence of the αB-crystallin deletion mutant outside the core region inhibits tetramer formation. The crystal structures of both Hsp 16.5 and 16.9 show a specific interaction between dimers that leads to tetramer formation through a short β strand (termed β-10 in Hsp 16.5 by Kim et al. [33]) on the COOH-terminal; the αB-crystallin mutant truncation of Feil et al. [35] removes this strand, while in our NH₂-terminal deletion mutant of αA-crystallin, the COOH-terminal is left untouched. This "IPV" motif is conserved in a large number of small heat shock protein superfamily members; it does not form part of a β sheet, but the β backbone configuration places the two hydrophobic side chains on the same side of the backbone, in position to interact with an exposed hydrophobic surface on one side of the β core structure.

The chaperone-like activities of DM51 and LM51 were very similar. This would be expected, since the putative regions involved in interaction with partially denatured proteins are mainly in the COOH-terminal and core regions. However, there is a large difference between the chaperone-like activity of either of these mutants compared to the parent αA-crystallin. It is possible that the ability of the deletion mutants to bind partially denatured proteins has been reduced from that of the parent molecule, but an alternative explanation is more likely. The ability of α-crystallin subunits to protect against superaggregation of partially denatured proteins is a function of the size of the partially denatured protein. Presumably this is because they are arrayed over the surface of the α-crystallin oligomer. Although both DM51 and LM51 are capable of interacting with substrate molecules, it is likely that they are unable to keep the partially denatured proteins separate from each other as well, As a result, some superaggregation could potentially occur even if initial binding to individual DM or LM tetramers were as strong as with the parent protein. It is possible that the size of the α-crystallin particle is important for separating substrate molecules. Our results indicate that tetramer is probably the smallest aggregation size for the chaperone-like activity, The heating and cooling experiments showed that the dimer has much less chaperone-like activity than the tetramer. The large difference in chaperone-like activity between the mutants and the wild type αA-crystallin might also be due to the difference in their aggregation sizes, since the mutants formed mainly tetramers. This indicates that the deletion has weakened the interaction between tetramers, and this interaction is probably necessary for formation of aggregates above tetramer.

Thus the major difference between the two deletion mutants is the effect of thermal stress on their aggregate structure and functionality. The presence of the leader sequence on the NH₂-terminal appears to stabilize the mutant sequence sufficiently to preserve tetrameric structure and chaperone-like activity after exposure to a cycle of heating to 70 °C and rapid cooling with little or no change. The modest reduction in population heterogeneity we observed is analogous to annealing of the system. Folding of the protein during expression produces a slightly heterogeneous system, perhaps because of variability in parameters such as monomer concentration. Heating surmounts an activation barrier, and cooling then leads to a more homogeneous aggregation state.

The DM construct, in contrast, shows increased heterogeneity and almost total loss of chaperone-like activity after an annealing cycle that produced a more uniform population of LM molecules. This indicates that folding and/or annealing of the DM construct is significantly hindered with respect to the LM construct during cooling from 70 °C. The addition of an NH₂-terminal polypeptide would be expected to stabilize the α-core region interactions associated with one edge of each of the two β sheets, shielding the hydrophobic interior from water and inhibiting the mobility of the edge strands. The mechanism by which this is accomplished is likely to be insensitive of the sequence of the leader itself, which replaces a longer and significantly more hydrophobic native sequence, as long as the exterior of the cap structure formed is relatively hydrophillic. Such stabilization is not necessary to preserve the structure under standard conditions, since the properties of the DM and LM mutants are essentially identical, but becomes critical at higher temperatures where most proteins denature.

Several potential misfolding and misaggregation pathways are available to the DM construct. The absence of any NH₂-terminal sequence exposes the hydrophobic side of the conserved core structure flanked by the first β strand backbones of both sheets and containing the hydrophobic side chains packed between them. This is a surface that can participate in interactions with other dimers, or engage in intradimer interactions with hydrophobic regions in the COOH-terminal tail. In other proteins with the same folding topology (e.g., P23), the C terminus provides a cover for this region analogous to those provided by the NH₂-terminal region in Hsp 16.5 and Hsp 16.9. This would remove the COOH-terminus as a mediator of interdimer interactions, contributing to the observed heterogeneity in the size distribution of DM aggregates. In addition, removal of the NH₂-terminal sequence increases the possibility of alternative folding topologies in which the first β strand (by sequence criteria) is internal in one of the sheets. This would be expected to completely alter the aggregation of the mutant protein.

Why are these effects only observed after a cycle of heating and cooling? It is possible that heating to 70 °C provides access to local regions of stability that are not accessible at 25 °C, and that cooling traps a significant fraction of the molecules in these local minima. A more likely explanation is provided by the presence of the NH₂-terminal His tag during in vivo synthesis. This provides a hydrophilic sequence, which is in essence an alternative leader, so that the DM construct is synthesized, folds, and initially aggregates as an alternative version of the LM construct. After heating to 70 °C, however, cooling to 25 °C produces a distribution of states which were not observed previously.

Acknowledgements

The support of NIH grant EY10011 for a portion of this work is gratefully acknowledged.

References

1. Horwitz J. The function of alpha-crystallin in vision. Semin Cell Dev Biol 2000; 11:53-60.

2. Bron AJ, Vrensen GF, Koretz J, Maraini G, Harding JJ. The ageing lens. Ophthalmologica 2000; 214:86-104.

3. Horwitz J. Alpha-crystallin. Exp Eye Res 2003; 76:145-53.

4. Bhat SP, Horwitz J, Srinivasan A, Ding L. Alpha B-crystallin exists as an independent protein in the heart and in the lens. Eur J Biochem 1991; 202:775-81.

5. Benjamin IJ, Shelton J, Garry DJ, Richardson JA. Temporospatial expression of the small HSP/alpha B-crystallin in cardiac and skeletal muscle during mouse development. Dev Dyn 1997; 208:75-84.

6. Ishiguro Y, Kato K, Akatsuka H, Iwata H, Nagaya M. Chemotherapy-induced expression of alpha B-crystallin in neuroblastoma. Med Pediatr Oncol 1997; 29:11-5.

7. van Sechel AC, Bajramovic JJ, van Stipdonk MJ, Persoon-Deen C, Geutskens SB, van Noort JM. EBV-induced expression and HLA-DR-restricted presentation by human B cells of alpha B-crystallin, a candidate autoantigen in multiple sclerosis. J Immunol 1999; 162:129-35.

8. Tallot P, Grongnet JF, David JC. Dual perinatal and developmental expression of the small heat shock proteins [FC12]aB-crystallin and Hsp27 in different tissues of the developing piglet. Biol Neonate 2003; 83:281-8.

9. de Jong WW, Caspers GJ, Leunissen JA. Genealogy of the alpha-crystallin--small heat-shock protein superfamily. Int J Biol Macromol 1998; 22:151-62.

10. Salerno JC, Eifert CL, Salerno KM, Koretz JF. Structural diversity in the small heat shock protein superfamily: control of aggregation by the N-terminal region. Protein Eng 2003; 16:847-51.

11. Koteiche HA, Mchaourab HS. Folding pattern of the alpha-crystallin domain in alphaA-crystallin determined by site-directed spin labeling. J Mol Biol 1999; 294:561-77.

12. Muchowski PJ, Hays LG, Yates JR 3rd, Clark JI. ATP and the core "alpha-Crystallin" domain of the small heat-shock protein alphaB-crystallin. J Biol Chem 1999; 274:30190-5.

13. Guruprasad K, Kumari K. Three-dimensional models corresponding to the C-terminal domain of human alphaA- and alphaB-crystallins based on the crystal structure of the small heat-shock protein HSP16.9 from wheat. Int J Biol Macromol 2003; 33:107-12.

14. Caspers GJ, Leunissen JA, de Jong WW. The expanding small heat-shock protein family, and structure predictions of the conserved "alpha-crystallin domain". J Mol Evol 1995; 40:238-48.

15. Augusteyn RC, Koretz JF. A possible structure for alpha-crystallin. FEBS Lett 1987; 222:1-5.

16. Koretz JF, Augusteyn RC. Electron microscopy of native and reconstituted alpha crystallin aggregates. Curr Eye Res 1988; 7:25-30.

17. Radlick LW, Koretz JF. Biophysical characterization of alpha-crystallin aggregates: validation of the micelle hypothesis. Biochim Biophys Acta 1992; 1120:193-200.

18. Doss EW, Ward KA, Koretz JF. Preliminary studies on the aggregation process of alpha-crystallin. Exp Eye Res 1997; 65:255-66.

19. Wissing D, Jaattela M. HSP27 and HSP70 increase the survival of WEHI-S cells exposed to hyperthermia. Int J Hyperthermia 1996; 12:125-38.

20. Martin JL, Mestril R, Hilal-Dandan R, Brunton LL, Dillmann WH. Small heat shock proteins and protection against ischemic injury in cardiac myocytes. Circulation 1997; 96:4343-8.

21. Trautinger F, Kokesch C, Herbacek I, Knobler RM, Kindas-Mugge I. Overexpression of the small heat shock protein, hsp27, confers resistance to hyperthermia, but not to oxidative stress and UV-induced cell death, in a stably transfected squamous cell carcinoma cell line. J Photochem Photobiol B 1997; 39:90-5.

22. Thomas JG, Baneyx F. Roles of the Escherichia coli small heat shock proteins IbpA and IbpB in thermal stress management: comparison with ClpA, ClpB, and HtpG In vivo. J Bacteriol 1998; 180:5165-72.

23. Woo J, Iyer S, Cornejo MC, Mori N, Gao L, Sipos I, Maines M, Buelow R. Stress protein-induced immunosuppression: inhibition of cellular immune effector functions following overexpression of haem oxygenase (HSP 32). Transpl Immunol 1998; 6:84-93.

24. Croute F, Beau B, Arrabit C, Gaubin Y, Delmas F, Murat JC, Soleilhavoup JP. Pattern of stress protein expression in human lung cell-line A549 after short- or long-term exposure to cadmium. Environ Health Perspect 2000; 108:55-60.

25. Lambot MA, Peny MO, Fayt I, Haot J, Noel JC. Overexpression of 27-kDa heat shock protein relates to poor histological differentiation in human oesophageal squamous cell carcinoma. Histopathology 2000; 36:326-30.

26. Sanz O, Acarin L, Gonzalez B, Castellano B. Expression of 27 kDa heat shock protein (Hsp27) in immature rat brain after a cortical aspiration lesion. Glia 2001; 36:259-70.

27. Aldrian S, Trautinger F, Frohlich I, Berger W, Micksche M, Kindas-Mugge I. Overexpression of Hsp27 affects the metastatic phenotype of human melanoma cells in vitro. Cell Stress Chaperones 2002; 7:177-85.

28. Yenari MA. Heat shock proteins and neuroprotection. Adv Exp Med Biol 2002; 513:281-99.

29. Wang K, Spector A. ATP causes small heat shock proteins to release denatured protein. Eur J Biochem 2001; 268:6335-45.

30. Haslbeck M. sHsps and their role in the chaperone network. Cell Mol Life Sci 2002; 59:1649-57.

31. Nath D, Rawat U, Anish R, Rao M. Alpha-crystallin and ATP facilitate the in vitro renaturation of xylanase: enhancement of refolding by metal ions. Protein Sci 2002; 11:2727-34.

32. Horwitz J. Alpha-crystallin can function as a molecular chaperone. Proc Natl Acad Sci U S A 1992; 89:10449-53.

33. Kim KK, Kim R, Kim SH. Crystal structure of a small heat-shock protein. Nature 1998; 394:595-9.

34. van Montfort RL, Basha E, Friedrich KL, Slingsby C, Vierling E. Crystal structure and assembly of a eukaryotic small heat shock protein. Nat Struct Biol 2001; 8:1025-30.

35. Feil IK, Malfois M, Hendle J, van Der Zandt H, Svergun DI. A novel quaternary structure of the dimeric alpha-crystallin domain with chaperone-like activity. J Biol Chem 2001; 276:12024-9.

36. Eifert C, Burgio MR, Bennett PM, Salerno JC, Koretz JF. N-terminal control of small heat shock protein oligomerization: changes in aggregate size and chaperone-like function. Biochim Biophys Acta 2005; 1748:146-56.

37. Robinson CR, Sauer RT. Optimizing the stability of single-chain proteins by linker length and composition mutagenesis. Proc Natl Acad Sci U S A 1998; 95:5929-34.

38. Farahbakhsh ZT, Huang QL, Ding LL, Altenbach C, Steinhoff HJ, Horwitz J, Hubbell WL. Interaction of alpha-crystallin with spin-labeled peptides. Biochemistry 1995; 34:509-16.

39. Yang C, Koretz JF. The structural and functional role of the region bridging the N-terminal and central core of alpha-crystallin, ARVO Annual Meeting; 2003 May 4-9; Fort Lauderdale (FL).