Refinement of 3D structure of bovine lens αA-crystallin

International Journal of Biological Macromolecules 22 (1998) 175 – 185

Refinement of 3D structure of bovine lens aA-crystallin P.N. Farnsworth a,b,*, H. Frauwirth a, B. Groth-Vasselli b, Kamalendra Singh c a

Department of Pharmacology and Physiology, UMD-New Jersey Medical School, 185 South Orange A6e., Newark, NJ 07103, USA b Department of Ophthalmology, UMD-New Jersey Medical School, Newark, NJ, USA c Department of Biochemistry and Molecular Biology, UMD-New Jersey Medical School, Newark, NJ, USA

Abstract In absence of 3D structures for a-crystallin subunits, aA and aB, we utilized a number of experimental and molecular modeling techniques to generate working 3D models of these polypeptides (Farnsworth et al., 1994. In Molecular Modeling: From Virtual Tools to Real Problems (Eds. Kumosinski, T.F. and Liebman, M.N.) ACS Symposium Series 576, Ch. 9:123 – 134, 1994, ACS Books, Washington DC). The refinement of the initial bovine aA model was achieved using a more accurate estimation of secondary structure by new/updated methods for analyzing the far UV-CD spectra and by neural network secondary structure predictions in combination with database searches. The spectroscopic study reveals that a-crystallin is not an all b-sheet protein but contains 17% a-helices,  33% b-structures and  50% turns and coils. The refinement of the aA structure results in an elongate, asymmetric amphipathic molecule. The hydrophobic N-terminal domain imparts the driving force for subunit aggregation while the more flexible, polar C-terminal domain imparts aggregate solubility. In our quaternary structure of the aggregate, the monomer is the minimal cooperative subunit. In bovine aA, the highly negatively charged C-terminal domain has three small positive areas which may participate in dimer or tetramer formation of independently expressed C-terminal domains. The electrostatic potential of positive areas is modulated and become more negative with phosphorylation and ATP binding. The refined bovine aA model was used to construct aA models for the human, chick and dogfish shark. A high degree of conservation of the three dimensional structure and the electrostatic potential was observed. Our proposed open micellar quaternary structure correlates well with experimental data accumulated over the past several decades. The structure is also predictive of the more recent data. © 1998 Elsevier Science B.V. All rights reserved. Keywords: Small heat shock proteins; a-Crystallin; Lens; 3D molecular model; Micellar structure

1. Introduction The major crystallins in all mammalian eye lenses can be divided, according to genetic studies, into two families, a-crystallin and bg superfamily. a-Crystallin, a major protein in all vertebrate lenses, has a central role in controlling fiber cell supramolecular order which under normal conditions supports transparency. Conversely, under pathological conditions it is a major participant in lens opacification. Interest in a-crystallin has been heightened by identification of both aA and aB * Corresponding author: Tel.: +1 973 9724489; fax: + 1 973 9727950; e-mail: [email protected] 0141-8130/98/$19.00 © 1998 Elsevier Science B.V. All rights reserved. PII S0141-8130(98)00015-4

subunits in many cell types under normal and pathological conditions ([1] references therein), sequence homology of its subunits with small heatshock proteins (sHsps) [2], identification of aB as a sHsp [3] and co-aggregation with other sHsps [4]. a-Crystallin, in its role as a chaperone-like protein, responds to stressful conditions by sequestering dysfunctional proteins thus preventing abnormal polypeptide aggregation [5,6]. The constitutive expression of aA and aB subunits [7] and related sHsps during normal growth and development suggests an important role in the stress of supramolecular reorganization associated with cell division and differentiation. Recent experimental evidence for the binding of ATP [8,9] and

176

P.N. Farnsworth et al. / International Journal of Biological Macromolecules 22 (1998) 175–185

both single and double stranded DNA [10,11] to a-crystallin suggests that this protein is engaged in other important functions that remain to be defined. The crystal structures of some members of the bg-superfamily have been elucidated. In the absence of 3D structures of a-crystallin subunits, we have used a number of experimental and molecular modeling techniques to generate working models of both aA and aB. The initial molecular modeling produced elongate amphipathic subunits [12]. However, these models were criticized for the substantial amount of a-helical content. From analyses of earlier far UV-CD studies, it was concluded that, similar to g-crystallin, a-crystallin contained little or no a-helix [13]. In addition, Wistow [14] predicted that a-crystallin subunits contain sequence repeats and have tertiary structures consisting of two domains with each containing a related pair of motifs. He also identified C-terminal extensions in both subunits. These concepts have prevailed for the last two decades. Recently, we performed an extensive far UV-CD study of a-crystallin under varying temperature and protein concentration [15]. Analyses of these spectra by three new/updated curve fitting programs, viz., SELCON [16], CONTIN [17] and K2D [18] revealed a consistency in secondary structure for a given set of conditions. At 37°C and 0.263 mg/ml, the estimated secondary structure was  17% a-helix, 33% b-sheet and  50% turns and coils [15]. This secondary structure distribution is not consistent with the definition of a b-sheet protein, i.e. E (extended structure) = 45%. A comparison of the primary structure of g- and a-crystallin shows that the amino acids necessary for the Greek key fold do not exist in a-crystallin [19]. The profile of positional flexibility of amino acid residues in a-crystallin subunits reveals that the two subunits are asymmetric in contrast to the very symmetrical g-crystallins [19]. a-Crystallin subunits have a relatively small hydrophobic Nterminal domain, a connecting segment and a larger more flexible, hydrophilic C-terminal domain terminating in an 8–10 amino acid unordered structure. This reflects the asymmetry of the genes for both subunits, i.e. three exons in bovine aA encode the sequences for the N-terminal domain (1 – 63), connecting segment (64–105) and C-terminal domain (106–173) ([20] references therein). Although exons are thought to encode functional units [21], both subunits, aA and aB,

are traditionally described as two domain structures. The absence of repeats in the positional flexibility profile of a-crystallin subunits is inconsistent with the existence of sequence repetitions [19]. Conversely, in the flexibility plots for bg-crystallins, such repetitions are well documented. Based on the secondary structure estimated from far UV-CD spectroscopy and predicted by the neural network method [22] in combination with database searches, the initial molecular model of bovine lens aA-crystallin was refined. The subunit tertiary structure and the proposed quaternary structure of its aggregate correlate well with the polydispersity of the molecular mass, the binding of a- to g-crystallin in 1:1 ratio during heat and oxidative stress [23–25], the separation of ˚ [26], the a-crystallin bound melittin by 25 A electrostatic parameters of a-crystallin obtained from light scattering experiments [27,28], the flexibility and exposure of the C-terminal domain of a-crystallin and 8–10 residue unordered structure at the C-terminus [19,29,30], the location of ordered structure and position of hydrophobic segments in a-crystallin subunits [31,32] and the exposure of amino acids subject to proteolysis and phosphorylation ([1], references therein; [33–37]. 2. Material and methods

2.1. Secondary structure prediction of bo6ine lens aA The secondary structure was predicted by the PHD method of Rost and Sander [22] by submitting the sequence to the ‘predictprotein server’ at the European Molecular Biology Laboratory, Heidelberg, Germany (http://www.embl-heidelberg.de/predictprotein/predictprotein.html). This method is a neural network method where input and output are connected by ‘neurons’. The ‘neurons’ constructed from the sequence are assigned the weights based on secondary structure of known proteins. The ‘neurons’ are weighted in such a way that the input results in correct output. The correctness of the output is defined by the error factor in the output for each residue of the protein. To identify the secondary structure for sections of the molecule left unassigned by the PHD method, a BLAST (basic local assignment search tool) search [38] against SCOP database (struc-

P.N. Farnsworth et al. / International Journal of Biological Macromolecules 22 (1998) 175–185

tural classification of protein) [39] (http:// scop.mrc-lmb.cam.ac.uk/scop) was performed. This search resulted in identification of 15 proteins having the segmental homology \ 50% with aA. A list of these proteins and their homologous segments with their secondary structure is presented in Table 1. aA predicted secondary structure and predicted solvent exposure of amino acid residues by PHD method are shown in Fig. 1. Based on this prediction aA contains  9% a-helix, 22% b-structure and  70% turns and coils. According to the classification of Rost and Sander [22], aA is predicted as a mixed class protein (see legend of Fig. 1 for protein classification). For residues I3–Q6, S62– R65, F71 – D76, T86–E91, F92–H96, S126– S133, G136 – T139 and P159, the predicted secondary structure is extended (b-sheet) conformation (Fig. 1). The residues in predicted a-helical conformation are D24–F26, S51–V56 and I110 – F114 (Fig. 1). Inspection of these distributions revealed that some of the prediction required refinement. For example, an amino acid, P159 cannot constitute a b-sheet and only three residues D24 – F26 are not sufficient to form an Table 1 Protein segments that share homology with aA identified BLAST search against SCOP database PDB id

Protein

2TMD

Trimethylamine dehydrogenase Major urinary protein

1MUP

7AAT 1MIN

2XIS 1SCU 1LMN 1TSS 1PLC

Residuea

685–709 (140–164) 23–55 (56–88) 81–84 (132–145) Aspartate amino- 232–279 transferase (77–124) Nitrogenase 268–280 protein (109–121) 6–26 (50–70) Xylose isomerase 110–137 (116–143) Succinyl-CoA syn4–21 thase (107–124) Lysozyme (rain4–29 bow trout) (125–150) Toxic shock syn- 120–153 drome toxin (76–109) Plastocyanin 39–52 (139–152)

Conformation a–a b–a –b–b b– b a–b –b–a a –b a –a a–b a–b a–a – a– –b –

The numbers in parenthesis represent the sequence for aA from bovine.

a

177

Fig. 1. Predicted secondary structure of bovine lens a-crystallin by PHD method. H, a-helix; E, extended (b-structure); e, exposed; b, buried. ‘  ’ represents the extension of the modified helical and ‘solid line’ modified b-structure structure by homology search using BLAST against SCOP database. The total secondary structural distribution of aA-crystallin in this figure is  17%, a-helix,  28%, b-structure and 57%, turns and coils. According to the classification of Rost and Sander [22], all alpha protein contains, a-helix \ 45% and b-structure B5%; all beta: a-helix B 5% and b-structure \ 45%; mixed: a-helix \5%, b-structure B45% and b-structure \5%, a-helix B 45%, aA-crystallin is a mixed class protein.

a-helix. Although it is estimated that the PHD method for secondary structure prediction is  71% accurate, the inconsistencies and the presence of some very long segments of the aA sequence that are unassigned suggest that further refinement is required. Therefore, the segments were subjected to BLAST search against the SCOP database. The assignment of the residues to a specific secondary structure after BLAST search was also guided by the prediction of solvent exposure by the PHD method. An inspection of the exposure pattern near F27 suggests that the amino acid residues are buried and exposed in an alternate fashion. This favors the helical structure. In an a-helix if residue ‘i’ is exposed then there is a great possibility that residues ‘i +1’ and ‘i+2’ will be totally (or partially) buried whereas residues ‘i +4’ will again be exposed. We applied similar intuitive methods in assigning the secondary structure to a particular residue. For example, the gap of only one or two residues in the predicted extended structure suggests that these extended segments most likely form antiparallel b-sheets. The final secondary structure distribution of aA is shown in Fig. 1. With the modifications from the BLAST search

178

P.N. Farnsworth et al. / International Journal of Biological Macromolecules 22 (1998) 175–185

Fig. 2. Stereopair of the modified structure of bovine lens aA-crystallin. The hydrophobic residues have been represented in ‘cyan’ whereas charged and polar residues in ‘red’ and ‘yellow’, respectively. This figure shows that the N-terminal domain of ˚. aA-crystallin is highly hydrophobic while C-terminal domain is hydrophilic. The total length of the subunit is  76 A Fig. 3. Sequence comparison of bovine and dogfish shark aA-crystallin. Overall homology between the two sequences is 66%. The N-terminal domain of aA from two species is has greater homology than C-terminal domain.

P.N. Farnsworth et al. / International Journal of Biological Macromolecules 22 (1998) 175–185

179

results, the final secondary structure distribution is 17% a-helix, 27% b-sheets and  57% turns and coils. This distribution approximates the estimated secondary structure from far UV-CD.

Table 2 Sequence homology between aA-crystallin from various species aA

Bovine

Human

Chick

Shark

2.2. Incorporation of the assigned secondary structure in the aA model

Bovine Human Chick Shark

100 95.4 87.3 77.5

95.4 100 86.7 76.9

87.3 86.7 100 75.7

77.5 76.9 75.7 100

The secondary structure thus obtained was incorporated into our original model of aA. For this purpose, the molecular modeling package SYBYL 6.2 (Tripos Associates, St. Louis, MO) was used. First, the conformation of the residues was reformed guided by Fig. 1. The gaps containing unordered structure were filled by ‘loop search’ algorithm of SYBYL 6.2. The ‘loop search’ scans the entire SYBYL database of protein structures for the most homologous segments with minimum root mean square deviation of Ca. This loop is then inserted into the desired position. The structure thus obtained was subjected to simulated annealing by raising the temperature of the system to 1000° K and then decreasing it to 300° K in steps of 1 ps. The simulated annealing was performed by assigning Kollman United charges [40] to the protein. The resultant minimum energy structure is shown in Fig. 2.

2.3. Modeling of human, chick and dogfish shark aA-crystallins The 3D molecular models of aA from human, chick and dogfish shark were constructed using the bovine aA 3D model as the template molecule. For this purpose, the sequence of the target molecule was aligned by GAP and/or PILEUP algorithm of GCG (Wisconsin Sequence Analysis Package). The default parameters of GCG for

alignment were selected. The substitutions were performed in the template molecule using SYBYL 6.2. The models thus obtained were subjected to the’simulated annealing’ protocol in molecular dynamics package of SYBYL 6.2. All calculations were performed by using Kollman United charges [40] on the atoms of the target molecules. The structures were subjected to simulated annealing as described previously. The structures obtained after simulated annealing were energy minimized until the energy gradient reached 0.05. 3. Results and discussion

3.1. Structure of aA 3.1.1. Sequence comparison An evaluation of the most highly conserved sequences in the evolution of aA-crystallin provides an insight into the functionally important regions of the protein. A multiple sequence alignment of bovine with the sequences of human, chick and dogfish shark aA-crystallin reveals that there is 62% overall sequence homology (Table 2). A sequence comparison of the bovine with human, chick and spiny dogfish shark aA shows that there is a sequence homology of 94, 86 and

Table 3 Isoelectric points of a-crystallins form various species Crystallin

aA Truncated aA (1–165) Deamidated aA (at Asn101) aB Truncated aB (1–165) Deamidated aB aA(3)+aB(1) Tr. aA(3)+Tr. aB(1)

Isoelectric points Human

Bovine

Rat

Chick

Shark

6.16 5.89 5.89 7.38 6.61 7.04 6.46 6.12

6.16 5.89 5.90 7.38 6.61 7.04 6.46 6.13

6.16 5.89 — 7.38 6.61 — 6.46 6.19

6.16 5.89 5.90 6.81 6.26 — 6.31 5.98

5.39

6.89

180

P.N. Farnsworth et al. / International Journal of Biological Macromolecules 22 (1998) 175–185

Fig. 4. Stereopair of the ribbon backbone representation of aA from four species: cyan, bovine; yellow, chick; red, human and green, dogfish shark. Despite the lower sequence homology of dogfish shark aA with bovine aA, their structures superimpose very well. This suggests the structural conservation of a-crystallin during evolution.

65%, respectively. The conservation of charged and polar amino acids is illustrated by the consistency of the isoelectric point, 6.16 (Table 3) for the vertebrates including human, bovine, rat and the chick. The cartilaginous dogfish shark, a distant relative has a lower pI. It is interesting to note that post-translational alterations that are known to occur in aA tend to decrease its isoelectric point (Table 3). These alterations and phosphorylation may contribute to the lower experimentally determined isoelectric point of aA extracted from the lens. They may also account for changes in aA solubility and its interactive properties with dysfunctional proteins [41]. Sequences of the N-terminal domain of aA in human and spiny dogfish, a very primitive species, are 80% homologous. The most highly conserved region is the residues 17–67 (Fig. 3). The sequence homology in aA N-terminal domains of the four species is higher than C-terminal domain. The sequence differences in both domains may represent adaptations required for lens fiber cell terminal differentiation. This highly controlled process provides the required species specific refractive index gradient.

3.1.2. Tertiary structure Similar to g-crystallin [42], a-crystallin is under severe evolutionary constraints [20]. aA and sHsps share significant sequence homology mainly in the C-terminal domain [2]. This is referred to as the a-crystallin domain. There are two highly conserved regions designated as HCR1 and HCR2 [25]. HCR1, a highly charged sequence starts in the connecting segment at H96, and ends in the C-terminal domain at I110 while HCR2, a hydro-

phobic segment, includes C-terminal domain residues S130–F141. Both of these regions have been identified as participants in the binding site for g-crystallin [25,43]. Another conserved region in aA is between R157 to the C-terminus, S173. This includes an unstructured, flexible C-terminal extension (l66–173) which appears to have a role in chaperone activity [25,41]. For all species studied, the tertiary structure of aA contains two domains, a hydrophilic C-terminal domain, which imparts solubility, and a hydrophobic N-terminal domain that serves as the driving force for subunit aggregation. The major change in the overall configuration of the refined subunit is a shortend peptide link between the two domains (F74–H79) and the integration of the second exon into the N-terminal domain (V64– I73) and the C-terminal domain (F80–E105). The root mean square deviation between the Ca of bovine with human, chick, and dogfish shark aA ˚ , respectively (Fig. 4). are 0.98, 1.06 and 1.17 A These data suggest that the structure is more highly conserved than the sequence. A comparison of the electrostatic potential contours of bovine and spiny dogfish shark aA shows that the charge distribution is also conserved (Fig. 5). For all species, the secondary structure of the N-terminal domain is mainly unordered structure with two a-helices while the C-terminal domain is rich in b-structure. In addition to six b-strands in the C-terminal domain, there are two small a-helices which constitute a putative helix-turn-helix motif. This motif contains a conserved sequence identified in other DNA binding proteins [11]. A comparison of far-UV CD spectra of independently

P.N. Farnsworth et al. / International Journal of Biological Macromolecules 22 (1998) 175–185

181

Fig. 5. The electrostatic surface potential of +1 (blue) and −1 (red) kT/e for (a) bovine and (b) dogfish shark aA-crystallin. For both species the electrostatic potential corresponding to the negative field extends over by  2/3 length of the subunit whereas the potential representing the positive electric field distributed locally occupies only smaller portions of the surface area. The post-translational changes which increase the negative surface charge of the C-terminal domain will increase the intensity of the negative field and result in increased repulsion between subunits. Although the general characteristic of the surface potential of bovine and dogfish shark aA are similar, there exist some fine differences between the two species. For example the positive field around Ser122 in bovine aA is more pronounced than in the dogfish shark. This suggests that the binding of ATP and phosphorylation may occur more readily for bovine than dogfish shark.

expressed C-terminal domains of aA and aB with their respective intact subunits have been reported [44]. It was concluded that the secondary structure

of the intact subunits and their C-terminal domains are essentially identical. However, the significant differences in the two domains in

182

P.N. Farnsworth et al. / International Journal of Biological Macromolecules 22 (1998) 175–185

sequence, flexibility profile, and predicted secondary structure (Fig. 1) suggest that the secondary structure of the two domains are very different. This would not be surprising since the functions ascribed to each domain also differ. The application of a hydrogen–deuterium (H– D) exchange technique was applied to detect discrete a-crystallin conformational changes of specific hydrophobic regions with temperature [32]. The technique is also a probe for detecting ordered secondary structure, i.e. a low H–D exchange within a segment denotes a more ordered structure. The H – D exchange data for aA32 – 37 and aA72 – 75 show a low exchange at 10°C with a dramatic increase with increased temperature (10–60°C). The data are in agreement with temperature-induced exposure of hydrophobic surfaces determined by fluoresence spectroscopy using the hydrophobic probe bis-ANS [45]. It also correlates well with our refined model since the N-terminal domain aA32 – 37 and the connecting segment aA72 – 75 of our refined model are within an a-helix and a b-sheet, respectively [31]. Also other segments aA54 – 63 and aA85 – 96 have relatively low H–D exchange which is consistent with their presence in ordered secondary structure (Fig. 1). Therefore, the H–D exchange in these sequences correlates well with the ordered secondary structure in our refined aA 3D model. In contrast, the hydrophobic sequence aA3 – 10 which is 80% deuterated at 10°C is relatively insensitive to temperature increase. This reflects unordered structures such as the proposed turns and coils in this region of our aA model [31]. a-crystallin undergoes a number of post-translational changes which have been thoroughly reviewed ([1], references therein). A reasonable test of our model is the location of residues that are prone to modification. In our model, the location of serines subject to phosphorlation and residues that require surface exposure for proteolysis are on the surface of the C-terminal domain and/or along the subunit surface within the clefts.

3.1.3. Electrostatic properties The charge distribution and hydropathy profile of a protein defines the mode of its interaction with proteins and small ligands. To define the charge distribution of the refined structure of aA, the electrostatic potential contours of 91 kT/e was calculated and is displayed in Fig. 5. It is obvious that the distribution of charged residues in the C-terminal domain generates a strong negative electrostatic field with three regions of positive field

(Fig. 5). Two very small regions contain an isolated lysine 145 and two arginines 157 and 163 near the C-terminus. The largest positive field is within the a-crystallin domain and includes arginines 116, 117, and 119 and lysine 88. This region contains a consensus sequence known to participate in DNA binding [11] in other proteins and is proximal to S122 which is readily phosphorylated by cAMP-dependent kinase and appears to be age regulated [33–35]. At least one-third of the a-crystallin extracted from the lens is phosphorylated. Three other phosphorylation sites have been identified between 122 and 173, a region rich in serines [35]. The serines phosphorylated at these additional sites have not been identified. In vitro experiments have shown that in the presence of Mg2 + and 32P-ATP auto-phosphorylation occurs on one or more serines within the peptide sequence between C131 and K145 [36,37]. In the absence of Mg2 + , the binding of ATP to a-crystallin [8,9] and the identification of a putative binding site [46] within the auto-phosphorylation site satisfies a major requirement for this type of phosphorylation. The overall consequence of the binding of ATP and phosphorylation of multiple serines is an intensified negative field in the C-terminal domain. This will decrease the tendency for interdomain attraction within the aggregate and increase the interactions with aB and other more basic lens proteins. There is much to learn concerning the functional consequences of these interactions on lens supramolecular order. According to NMR data [30], the quaternary structure of an aggregate of both aA and aB subunits forms a tighter aggregate than the isolated aA and aB. This suggests a difference in the interactive properties of these subunits within the aggregate. A possible interactive profile of aA and aB were suggested earlier [12,47]. The release of aB from sHsp 27 during stress [48] and dissociation of aA and aB during terminal differentiation of lens fiber cells [49] suggest that control of these interactions may be of fundamental importance for the functional role of a-crystallin subunits in lens supramolecular order. Although short range order of lens proteins is thought to be sufficient for transparency, the application of two 13C NMR techniques sensitive to protein rotational motion [50] established the presence of both mobile and solid-like protein phases in the inner regions of the lens. The shift of a-crystallin from the soluble to the insoluble phase may contribute to this change. The consistency of morphology, composition and size in age matched lenses from a given species suggests that this process is exquisitely controlled.

P.N. Farnsworth et al. / International Journal of Biological Macromolecules 22 (1998) 175–185

3.1.4. Quaternary structure of aA aggregates a-Crystallin extracted from the lens exists in solution as polydisperse aggregates ranging from 600 to 1200 kDa with average molecular mass of  800 kDa. Using our aA molecular model, an open micellar quaternary structure for an aA aggregate has been proposed (Fig. 6) [47]. In this model, the elongate a-crystallin subunits are arranged so that the hydrophilic C-terminal domain of each subunit occupies the surface of the aggregate and imparts aggregate solubility. The highly electronegative field surrounding this domain provides repulsive forces that obviate C-terminal interactions within or between aggregates. The subunits interact at the distal end of their hydrophobic N-terminal domains. This configuration enhances surface area exposure to the solvent by creating clefts between subunits where dysfunctional proteins bind [25,47]. The flexibility of C-terminal domains permit adjustments to meet the requirements for binding various proteins and small ligands. Further support for our model is provided by recent studies of other sHsps [51,52]. A sHsp12.6,

Fig. 6. Quaternary structure of bovine lens aA-crystallin. This figure represents the arrangement of the subunits through the diameter plane of a spherical aggregate consisting of 42 subunits (840 kDa). The subunits are arranged such that the hydrophobic N-terminal domain is inside the aggregate and provides the driving force for subunit interaction. The highly polar C-terminal domain on the aggregate surface imparts ˚ and the solubility. The length of each subunit is  76 A ˚ diameter of the central cavity is 50 A. Therefore, the radius ˚. of an spherical aggregate of 840 kDa is 100 A

183

found in Caenorhabditis elegans [51], contains the conserved a-crystallin domain. However, both the C- and N- termini are much reduced in size and, unlike other sHsps, do not form large oligomeric complexes. The recombinant sHsp12.6 is monomeric, synthesized in the soluble form and does not function as a chaperone. The fact that it has the shortest N- and C-terminal domains of any sHsp suggests that these deficits may be relevant to its monomeric state. Sequence comparisons of several sHsps suggest that the monomeric state of this sHsp is due to deletion of 16 amino acids from the N-terminus [51]. The deletion of 15 amino acids from sHsp16-2 was also sufficient to prevent aggregation [51]. This is in agreement with the observation that truncation of the C-terminus does not prevent aggregation of a-crystallin subunits [41]. More recent data on sHsp18.1 [52] are also in accord with our quaternary structure of a-crystallin aggregate. Further convincing evidence for the aggregative properties of the N-terminal domain of aA was demonstrated by a study of the aggregation of independently expressed N-(1–63) and C-terminal (64–173), (which in this case includes the connecting segment) domains [53]. The aA N-terminal domain is expressed in an insoluble form and exists as high molecular weight aggregates. In contrast, the C-terminal domain is expressed in a soluble form that exists as dimers and/or tetramers. This may be of functional importance since the formation of tetramers enhances auto-phosphorylation [37] and is concurrent with the binding of DNA [11]. Wistow [54] and Smulders et al. [55] have proposed a quaternary structure in which the tetramer is the minimal cooperative subunit. However, recent data from differential scanning microcalorimetry, spectroscopy and size exclusion chromatography led to the conclusion that the monomer is the minimal cooperative unit in a-crystallin aggregates [56]. This differs from both sHsp25 and sHsp27 in which the minimal cooperative unit is the dimer [57]. In contrast to aA, comparable aB C-terminal domains form a heterogeneous population of aggregates up to 18–20 subunits. When crosslinked, more than seven crosslinking products are observed [44]. This suggests random aggregation which is undoubtedly related to the heterogeneity of the electrostatic field of the aB C-terminal domain. There is considerable data to support the concept that the C-terminal domain occupies the surface of the aggregate. NMR studies have shown

184

P.N. Farnsworth et al. / International Journal of Biological Macromolecules 22 (1998) 175–185

that the resonances of a-crystallin arose almost exclusively from C-terminal extensions [29,30] and that the C-terminal domain is the first to unfold in the presence of urea [58]. Further evidence for C-terminal surface exposure comes from the observation that its unstructured extension contains the major continuous epitope for aA [59] In addition, the susceptibility of the C-terminal domain to proteolytic cleavage provides evidence for substantial surface exposure ([1], references therein). The dimensions and the space available for the solvent within the aggregate correlates well with the experimental data [28,47]. The radius of a ˚ , includes 840 kDa spherical aggregate,  100 A ˚ the length of the subunit ( 76 A) and the radius ˚ ). Under various ionic of the inner cavity (24 A conditions [27], the hydrodynamic radius varies ˚ . The volume of the our between 93 and 98 A ˚ 3. Therefore, theoretical aggregate is  4.3× 106 A the volume occupied by the subunits,  1.2×106 ˚ 3, approximates only  30% of the total volume A of the aggregate. Light scattering measurements of a-crystallin solution [27] reveal that the hard sphere volume of the subunits within the aggregate is  27% of the total volume. Since the aggregate of aA contains  70% space available for solvent, aggregates of higher molecular mass can be constructed without significant change in the radius. The model provides substantial solvent space and exposed surface area of the subunits for binding dysfunctional proteins. The subunit exchange among a-crystallin aggregate [60] and the dynamic mobility of the subunits within the aggregates permit reorganization of available space to accommodate a variety of proteins. 4. Significance Several years ago, the original micelle hypothesis for the a-crystallin aggregate was proposed by Augusteyn and Koretz [61]; however, no representation of the molecular organization of subunits was presented. A study of the volume occupied by the subunits and space available for the solvent in the aggregate made it quite clear that the traditional micelle usually associated with relatively tightly packed amphipathic molecules like phospholipids was not tenable. We elected to base our concept of the quaternary structure on three dimensional working models of the subunits generated from experimental data and state of the art

computer assisted molecular modeling [12]. Our predicted working models of elongate, amphipathic aA-crystallin subunits and their open micellar quaternary structure [25,47] correlate well with earlier data on biochemical and biophysical properties as they relate to a-crystallin functions. Even more important is the fact that our model predicted the more recent findings on the quaternary structure. This clearly demonstrates the utility of our molecular models for correlating structure with function. The term micellar is used to denote the position of the highly charged C-terminal domain on the surface with the hydrophobic N-terminal domain directed inward. The use of ‘open’ implies that the C-terminal domains do not cover the surface. They are separated due to the interactions at the distal end of the N-termini, the shape of the subunits and the highly negative electrostatic field of the C-terminal domain. This creates clefts that renders increased surface area exposure and space for accommodating dysfunctional proteins. The importance of a-crystallin in modulating, maintaining, and altering lens fiber cell supramolecular order during lens terminal differentiation is well established. However, the details of the mechanisms involved in this exquisitely controlled process are yet to be determined. The working models of the 3-D structures are available for both formulating future experimental strategies and the interpretation of results. References [1] Groenen PJTA, Merck KB, de Jong WW, Bloemendal H. Eur J Biochem 1994;225:1. [2] Ingolia TD, Craig EA. Proc Natl Acad Sci USA 1982;79:2360. [3] Klemenz R, Frohli E, Steiger RH, Schafer R, Aoyama A. Proc Natl Acad Sci USA 1991;88:3642. [4] Merck KB, Groenen PJTA, Voorter CEM, Horwitz J, Bloemendal H, de Jong WW. J Biol Chem 1993;268:1046. [5] Horwitz J. Invest Ophthalmol Vis Sci 1993;34:10. [6] Jakob U, Gaestel M, Engel K, Buchner J. J Biol Chem 1993;268:1517. [7] de Jong WW, Lubsen NH, Kraft H. J Prog Retinal Eye Res 1994;13:391. [8] Palmisano DV, Groth-Vasselli B, Farnsworth PN, Reddy MC. Biochem Biophys Acta 1995;1246:91. [9] Reddy MC, Palmisano DV, Groth-Vasselli B, Farnsworth PN. Biochem Biophys Res Comm 1992;189:1578. [10] Pietrowski D, Durante MJ, Liebstein A, Schmitt-John T, Werner T, Graw J. Gene 1994;144:171.

P.N. Farnsworth et al. / International Journal of Biological Macromolecules 22 (1998) 175–185

[11] Singh K., Groth-Vasselli, V., Kumosinski, T.F., Farnsworth, P.N. Biophys J 1996;70:A125 Int J Biol Macromol 1998;22:315–320. [12] Farnsworth PN, Kumosinski TF, King G, Grothvasselli B. Molecular Modeling: From Virtual Tools to Real Problems. In: Kumosinski TF, Liebman MN, editors. ACS Symposium Series, vol. 576. Washington: ACS Books, 1994:123. [13] Siezen RJ, Argos P. Biochim Biophys Acta 1983;748:56. [14] Wistow GJ. FEBS Lett 1985;181:1. [15] Farnsworth PN, Groth-Vasselli B, Greenfield N, Singh K. Int J Biol Macromol 1997;20:283. [16] Sreerama N, Woody RW. J Mol Biol 1994;242:497. [17] Provencher SN, Glockner J. Biochemistry 1981;20:33. [18] Andrade MA, Chacon P, Merolo JJ, Moran F. Prot Eng 1993;6:383. [19] Singh K, Zewgi D, Groth-Vasselli B, Farnsworth PN. Int J Biol Macromol 1996;19:227. [20] Wistow GJ, Piatigorsky J. Annu Rev Biochem 1988;57:479. [21] Gilbert W. Nature 1978;271:501. [22] Rost B, Sander C. Proteins 1994;20:216. [23] Wang K, Spector A. J Biol Chem 1994;269:13601. [24] Wang K, Spector A. Invest Ophthalmol Vision Sci 1995;36:311. [25] Singh K, Groth-Vasselli B, Kumosinski TF, Farnsworth PN. FEBS Lett 1995;372:283. [26] Farahbakhsh ZT, Huang Q-L, Ding L-L, Altenbach C, Steinhoff H, Horwitz J, Hubbel WL. Biochemistry 1995;34:509. [27] Xia J-Z, Aerts T, Donceel K, Clauwaert J. Biophys J 1994;66:861. [28] Singh K, Groth-Vasselli B, Farnsworth PN. Int J Biol Macromol 1995;18:205. [29] Carver JA, Aquilina JA, Cooper PG, Williams GA, Truscott RJW. Biochim Biophys Acta 1994;1204:195. [30] Carver JA, Aquilina JA, Truscott RJW, Ralston GB. FEBS Lett 1992;311:143. [31] Farnsworth PN, Groth-Vasselli B, Kumosinski TF, Singh K. Exp Eye Res 1997;64:853. [32] Smith JB, Liu Y, Smith DL. Exp Eye Res 1996;63:128. [33] Voorter CEM, Mulders JWM, Bloemendal H, de Jong WW. Eur J Biochem 1986;160:203. [34] Chiesa R, Gawinowicz-Kolks MA, Spector A. J Biol Chem 1987;262:1438. [35] Chiesa R, Gawinowicz-Kolks MA, Kleiman NJ, Spector A. Exp Eye Res 1988;46:199. [36] Kantorow M, Piatigorsky J. Proc Natl Acad Sci USA 1994;91:3112.

.

185

[37] Kantorow M, Horwitz J, van Boekel MA, de Jong WW, Piatigorsky J. J Biol Chem 1995;270:17215. [38] Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ. J Mol Biol 1990;215:403. [39] Murzin AG, Brenner SF, Hubbard T, Chothia C. J Mol Biol 1995;247:536. [40] Weiner SJ, Kollman PA, Case DA, Singh UC, Ghio C, Alagona G, Profeta S, Weiner PA. J Am Chem Soc 1984;106:765. [41] Takemoto L, Emmons T, Horwitz J. Biochem J 1993;294:435. [42] Slingsby C. Trends Biol Sci 1985;10:281. [43] Cooper PG, Aquilina JA, Truscott RJW, Carver JA. Exp Eye Res 1994;59:607. [44] Merck KB, Horwitz J, Kersten M, Overkamp P, Gaestel M, Bloemendal H, de Jong WW. Mol Biol Reports 1993;18:209. [45] Das KP, Surwicz WK. FEBS Lett 1995;369:321. [46] Groth-Vasselli B, Singh K, Farnsworth PN. Invest Ophthalmol Vision Sci 1995;36:885. [47] Groth-Vasselli B, Kumosinski TF, Farnsworth PN. Exp Eye Res 1995;61:249. [48] Zantema A, Verlaan-De Vries M, Maasdam D, Bol S, Van der Eb A. J Biol Chem 1992;267:12936. [49] Bhat SP, Horwitz J, Srinivasan A, Ding L. Eur J Biochem 1991;102:775. [50] Morgan CF, Schleich TW, Caines GH, Farnsworth PN. Biochemistry 1989;28:5065. [51] Leroux MR, Ma BJ, Batelier G, Melki R, Candido EP. J Biol Chem 1997;272:12847. [52] Leroux MR, Melki R, Gordon B, Batelier G, Candido EPM. J Biol Chem 1997;272:24646. [53] Merck KB, De Haard-Hoekman WA, Oudeessink BB, Bloemendal H, de Jong WW. Biochim Biophys Acta 1992;1130:267. [54] Wistow G. Exp Eye Res 1993;56:729. [55] Smulders RHPH, Martinus AM, van Boekel MAM, de Jong WW. Int J Biol Macromol 1998;22:187. [56] Gesierich U, Pfeil W. FEBS Lett 1996;393:151. [57] Dudich IV, Zav’ylov VP, Pfeil W, Gaestel M, Zav’yalova GA, Denesyuk AI, Korpela T. Biochim Biophys Acta 1995;1253:163. [58] Carver JA, Aquilina A, Truscott RJW. Biochim Biophys Acta 1993;1164:22. [59] Takemoto L, Emmons T. Invest Ophthalmol Vision Sci 1992;33:2024. [60] van den Oetelaar PJ, van Someren PF, Thomson JA, Siezen RJ, Hoenders HJ. Biochemistry 1990;29:3488. [61] Augusteyn RC, Koretz JF. FEBS Lett 1987;222:1.