Multiple Sites in αB-Crystallin Modulate Its Interactions with Desmin Filaments Assembled In Vitro

The β3- and β8-strands and C-terminal residues 155–165 of αB-crystallin were identified by pin arrays as interaction sites for various client proteins including the intermediate filament protein desmin. Here we present data using 5 well-characterised αB-crystallin protein constructs with substituted β3- and β8-strands and with the C-terminal residues 155–165 deleted to demonstrate the importance of these sequences to the interaction of αB-crystallin with desmin filaments. We used electron microscopy of negatively stained samples to visualize increased interactions followed by sedimentation assays to quantify our observations. A low-speed sedimentation assay measured the ability of αB-crystallin to prevent the self-association of desmin filaments. A high-speed sedimentation assay measured αB-crystallin cosedimentation with desmin filaments. Swapping the β8-strand of αB-crystallin or deleting residues 155–165 increased the cosedimentation of αB-crystallin with desmin filaments, but this coincided with increased filament-filament interactions. In contrast, substitution of the β3-strand with the equivalent αA-crystallin sequences improved the ability of αB-crystallin to prevent desmin filament-filament interactions with no significant change in its cosedimentation properties. These data suggest that all three sequences (β3-strand, β8-strand and C-terminal residues 155–165) contribute to the interaction of αB-crystallin with desmin filaments. The data also suggest that the cosedimentation of αB-crystallin with desmin filaments does not necessarily correlate with preventing desmin filament-filament interactions. This important observation is relevant not only to the formation of the protein aggregates that contain both desmin and αB-crystallin and typify desmin related myopathies, but also to the interaction of αB-crystallin with other filamentous protein polymers.


Introduction
Human aB-crystallin is a small heat shock protein (sHSP) that interacts with a variety of important cellular proteins with the capacity to polymerise into either filaments [1], or tubules [2] or fibrils [3]. The fact that aB-crystallin is often part of the histopathological signature used to characterise a variety of human diseases [4] highlights the potentially important role that aBcrystallin plays in their etiology. This became apparent when it was discovered that mutations in both aB-crystallin (R120G; [5]) and desmin [6] can cause cardiomyopathy, typified by aggregates containing both proteins [4]. The R120G mutation in aBcrystallin also induced the aggregation of desmin filaments in transfected cells [7]. The dissociation constant was increased two fold for the R120G mutant compared to the wild-type aBcrystallin [7], which appeared to encourage the increased interaction of desmin filaments leading to their aggregation in transfected cells and in the muscles of affected individuals. Previously it had been established that aB-crystallin modulated the assembly of intermediate filaments [8] and reduced the extent of filament-filament interactions in vitro [9]. Over-expression of wild-type aB-crystallin is capable of reversing intermediate filament aggregation in transfected cells suggesting that aBcrystallin was involved in regulating the local associations of intermediate filaments [10]. The fact that mutations inaBcrystallin caused the aggregation of desmin filaments [5,7] also supports this view. It is therefore important to identify the sequences in aB-crystallin that are responsible for the effects on intermediate filaments and particularly desmin because mutations in either can be the genetic basis of myopathy [11,12,13].
Pin array studies identified sequences in aB-crystallin involved in the recognition of a variety of different client proteins including desmin and GFAP, two examples of intermediate filament proteins [1]. These sequences were not unique to the interaction of aBcrystallin with either desmin or GFAP [1], evidence of the ability of aB-crystallin to recognise a wide range of potential protein clients [14,15,16,17,18,19,20,21,22,23]. The five sequences in aBcrystallin with the strongest binding to desmin were spread throughout the primary sequence from the N-to the C-terminus and included some that were involved in aB-crystallin oligomerisation. The pin-array assays did not consider the assembly status of the desmin or GFAP, a potentially important factor in the mechanism of aB-crystallin activity. Indeed, the interaction between the aB-crystallin peptides and desmin was inversely correlated with temperature in the pin array studies [1]. In contrast, the fraction of aB-crystallin that pelleted with desmin filaments in the sedimentation assays increased with temperature [7]. Therefore it is important to verify that the sequences identified using the pin arrays are involved in the interaction of aB-crystallin with desmin filaments.
Three aB-crystallin peptide regions (b3-strand, residues 73-85; b8-strand, residues 131-138 and the C-terminal sequences 155-165) gave some of the strongest interactions with desmin using the pin array approach [1] were selected for our studies. Recent crystallisation [24] and solution structural [25,26] studies confirmed that all three regions are surface exposed on the aBcrystallin subunit and were potentially available to bind client proteins such as desmin (see Fig. 1). Substituting the b3and b8strands with the equivalent sequences from aA-crystallin and C. elegans HSP12.2 produced aB-crystallin protein constructs that have been well-characterised previously in terms of structural changes and client protein interactions [16,17]. Exchange of either the b3-strand from aA-crystallin or C. elegans HSP12.2 with the equivalent aB-crystallin sequence was shown to have minimal effect on the secondary, tertiary and quaternary structure of aBcrystallin [17]. Replacement of the b8-strand of aB-crystallin with those from aA-crystallin and C. elegans HSP12.2 also did not affect secondary structure, but oligomer size was increased [16]. Likewise deleting the C-terminal sequences 155-165 altered protein oligomerisation but without significant effects upon protein secondary structure. Interestingly chaperone activity was decreased for all but one of the client proteins tested (bL-crystallin, alcohol dehydrogenase and citrate synthase) [14,16,17] for these five different aB-crystallin protein constructs demonstrating that Figure 1. Location of IF interactive domains in b3-strand (red), b8-strand (yellow), and C-terminal 155-165 residues (blue) of human wild-type aB-crystallin. The primary sequences for human wild type aB-crystallin, human wild type aA-crystallin, and C. elegans wild type HSP12.2 were aligned using the residue numbers for human aB-crystallin in ClustalX. The boxes and colors in the aB-crystallin sequence correspond with the interactive sequences labeled on the surface of the 3D model. The amino acid substitutions in the aB-crystallin protein constructs are indicated by bold-italics. The b3-strand of aB crystallin, 73 DRFSVNLDVKHFS 85 , was replaced with the corresponding sequences from aA-crystallin, DKFVIFLDVKHFS (aAb3), or HSP12.2, EKFEVGLDVQFFT (CEb3). The b8-strand of aB-crystallin, 131 LTITSSLS 138 , was replaced with the corresponding sequences in aAcrystallin, SALSCSLS (aAb8), or HSP12.2, STVKSHLA (CEb8). The 155-165 residues were deleted in aB-crystallin to create the D155-165 protein construct. The CEb3, CEb8, and D155-165 aB-crystallin protein constructs were designed to target the desmin interaction sequences. doi:10.1371/journal.pone.0025859.g001 all three sequences are intimately involved in and optimised for client protein recognition in aB-crystallin. Recent studies, found that both the b8-strand and C-terminal sequences, but not the b3strand in aB-crystallin, were responsible for regulating microtubule dynamics and preventing tubulin polymerisation [2,27]. Perhaps therefore, there are differences in the availability of the b3-strand, b8-strand and the C-terminal sequences 155-165 in aB-crystallin when a protein polymer is the client rather than individual protein subunits. For these reasons, we have determined the effect of the selected b3and b8-strand substitutions as well as the C-terminal 155-165 deletion on the interaction of aBcrystallin with desmin filaments.
In the present study, the data demonstrate that the substituted b3-strand, b8-strand and C-terminal sequences in aB-crystallin can alter the interaction of aB-crystallin with desmin filaments and therefore all three sequences contribute to the interaction of aBcrystallin with desmin filaments. The sequence substitutions in aBcrystallin involving the C.elegans b3-strand and b8-strand as well as the C-terminal 155-165 deletion all caused increased desmin filament-filament interactions. In contrast, the aA-crystallin b3strand substitution in aB-crystallin prevented desmin filamentfilament interactions even more effectively than wild type aBcrystallin, but in a temperature specific manner. The data suggest that the interaction of aB-crystallin with desmin filaments does not always lead to the prevention of desmin filament aggregation, which we discuss with respect to desmin filament aggregation as a histopathological characteristic of desmin related myopathies.

Sequence alignment and molecular modeling
The amino acid sequence of human aB-crystallin was aligned with the sequences of human aA-crystallin and C. elegans HSP12.2 using ClustalX [28] and mapped to a 3D model ( Fig. 1) for human aB-crystallin, which is in good agreement will all reported X-ray and NMR structures [18,24,26,28].

Purification of proteins
Wild type and mutant human aB-crystallin were purified from bacterial lysates as previously described using ion exchange and size-exclusion chromatography [29]. The wild type and the other five aB-crystallin protein constructs were soluble. Human desmin was purified from bacterial lysates as previously described using ion exchange chromatography [30,31]. Wild-type aB-crystallin, the five aB-crystallin protein constructs (aAb3, CEb3, aAb8, CEb8, and D155-165), and desmin were purified to .97% purity as determined by SDS-PAGE.

Assembly of desmin filaments
Assembly of desmin was performed as previously described [9,30]. Purified Desmin at 0.2 g/l in 6 M urea, 20 mM Tris-HCl pH8, 1 mM DTT, 1 mM EDTA, 0.2 mM PMSF (in the presence or absence of aB-crystallin at 0.08 g/l) was dialysed out of urea in a stepwise fashion by reducing the urea concentration to 4 M, then 2 M, then 0 M over a period of 24 h at 22uC. Desmin assembly was then initiated by dialysis into 20 mM Tris-HCl pH7.4, 50 mM NaCl for 16 h at either 22uC, 37uC or 44uC to ensure that assembly equilibrium has been reached.

Analysis of desmin aB-crystallin interactions by electron microscopy
Desmin, aB-crystallin and mixtures of both were diluted into assembly buffer to 100 mg/ml. A carbon film that had been coated onto freshly cleaved mica was then floated onto the surface of the sample prior to being negatively stained with 1% (w/v) uranyl acetate (Agar Scientific, UK) and retrieved with 400 mesh copper grids (Agar Scientific, UK). Grids were examined in an Hitachi H-7600 transmission electron microscope (Hitachi High-Technologies Corporation, Japan), using an accelerating voltage of 100 kV. Images were acquired using a CCD camera (Advanced microscopy Technology, Danvers, MA) and assembled into montages using AdobeH Photoshop CS (Adobe System, San Jose, CA).
Evidence of the association between desmin filaments and aBcrystallin particles in the EM images was statistically examined using likelihood ratio tests (LRT). For each combination of desmin and aB-crystallin protein construct, two representative images were selected for our analysis. Using ImageJ, a grid square was overlaid randomly over the image with a grid cell size equivalent to 11400 nm 2 . Forty cells from the 125 total were then randomly selected and the number of filaments and particles within each cell counted. We proposed that the relation between the mean number of particles in a cell, m, and the number of filaments in a cell, x, could be well described by, The parameter b 0 describes the background density of particles, b 1 is the maximum additional number of particles associated with filaments in the cell, and a describes how quickly each additional filament contributes particles. We also proposed that the variation in aB-crystallin particle numbers was negative-binomial distributed (NBD) to correctly account for potential variation among cells caused by unknown sources. Richards (2008) [32] provides details on how to calculate the likelihood under the assumption of a NBD. The null, which states that there is no association, is obtained by setting a = b 1 = 0. The test-statistic is G = 2(LL 1 -LL 0 ), where LL 1 and LL 0 are the maximum logs-likelihood associated with the general model and the null model, respectively. Under the null hypothesis, G is chisquare distributed [33] with 2 degrees of freedom as the general model has two additional free parameters: b 1 and a.

Analysis of desmin, aB-crystallin interactions using centrifugation
To investigate the interactions between desmin and the various aB-crystallin protein constructs, two separate sedimentation assays ( Fig. 2) were used to separate desmin filaments and their associated aB-crystallin from un-associated aB-crystallin [9,30]. In the high speed sedimentation assay, a 200 ml sample was layered onto a 100 ml 0.85 M sucrose cushion containing 10 mM Tris-HCl pH7.0, 50 mM NaCl, 1 mM DTT, 200 mM PMSF. Samples were then centrifuged at 30,584 rpm (RCF max = 80,0006g) for 30 min at 4uC using a Beckman Coulter TLS-55 rotor (k factor = 50) to give pellet (Calculated size of pelletted particles $100 S) and supernatant fractions. The supernatant was carefully removed and samples prepared preserving volume equivalence so that a direct comparison could be made between pellet and supernatant fractions. The samples were then separated on 12% (w/v) polyacrylamide gels by SDS-PAGE and the separated proteins visualised by Coomassie Brilliant Blue staining. Destained gels were imaged using a Fujifilm LAS-100. Stained bands were quantified using Fujifilm Image Gauge V4.0 software. The protein content in the pellet (P) and supernatant (S) fractions at each temperature was measured based on Coomassie Brilliant Blue staining densities after SDS-PAGE and then plotted as bar charts to summarise the complete dataset and provide an overview. The total protein was the sum of the densities in the S and P fractions for each sample. The amount of desmin or aB-crystallin in each fraction was the density of the selected band in that fraction divided by the total protein. Low-speed centrifugation will pellet only desmin filaments that have become associated with each other by filament-filament interactions as well as any associated aB-crystallin (see (Fig. 2) and [9,30]). This assay was initially developed to study interactions between keratin filaments [34]. Individual desmin filaments and unassociated aB-crystallin will not be pelleted under these sedimentation conditions. Immediately following assembly, samples were centrifuged at 4,900 rpm (RCF max = 2,5006g) for 10 min at 20uC using an Eppendorf 5417R benchtop centrifuge and standard fixed angle rotor (F-45-30-11, k-factor = 377). The supernatant was carefully removed from the pellet (Calculated S-value of pelletted material $2258S) and as with the high-speed assay, both fractions prepared for SDS-PAGE in a way to preserve the relative protein levels in each fraction so as to allow direct comparisons to be made when viewing the stained SDS-PAGE gels. Band intensities were quantified as described above.

Results
Some aB-crystallin protein constructs appear to increase filament-filament interactions as seen by electron microscopy Negative staining with uranyl acetate followed by electron microscopy was used to visualise the interactions between aBcrystallin particles and desmin filaments in samples prior to the sedimentation assays (Fig. 3). The wild type aB-crystallin formed 15-20 nm particles in agreement with observations by ourselves [8,31] and others [35], which were seen at all three temperatures ( Fig. 3; WT aB), Similarly the desmin filaments ( Fig. 3; Des) had a consistent morphology at the three different temperatures typically forming 10 nm filaments many microns long. When mixed together ( Fig. 3; [Des + WT aB]), both individual desmin filaments and aB-crystallin particles were readily apparent for all the various desmin-aB-crystallin combinations (Fig. 3), but now some of the aB-crystallin particles were observed to be associated with the filaments at the three different temperatures eg ( Fig. 3; [Des + WT  [8,9,30]. Desmin was assembled at 22, 37, or 44uC and centrifuged at 80,000 g. The pellet (P) will contain desmin filaments and any aggregates formed as a result of filament-filament interactions. Any aB-crystallin that associates with these filaments or their aggregates will also be cosedimented. The supernatant (S) will contain soluble aB-crystallin and also any assembly intermediates or unassembled desmin. Therefore this assay measures filament assembly and aB-crystallin binding to assembled filaments. (RIGHT) Schematic of the 2,500 g (low speed) centrifugation assay. Individual desmin filaments will not be sedimented by these sedimentation conditions, neither will aB-crystallin. Only when the assembled desmin filaments self-associate into filament aggregates, will these sediment. Therefore this assay measures filament-filament interaction. If aB-crystallin binds to these aggregates, then it too will be cosedimented. Unlike the high-speed assay, it is the aggregate-associated aB-crystallin which will sediment into the pellet fraction (P) rather than the individual filaments and their associated aB-crystallin. The supernatant (S) will contain the free desmin filaments, their associated aB crystallin, desmin assembly intermediates and the unassociated aB-crystallin particles. doi:10.1371/journal.pone.0025859.g002 aB], arrowheads). The electron microscopic data suggest that wild type aB-crystallin particles interact with desmin filaments and this was tested for significance using LRT (Fig. 4). There was strong statistical evidence that aB-crystallin particles were positively associated with the desmin filaments (LRT; G 2 = 26.0; P = ,0.001%). These data are visual confirmation that wild type aB-crystallin interacts with desmin filaments.
The various aB-crystallin protein constructs all formed particles that could be easily detected in the negatively stained samples (Fig.  S1) and they did not appear to be substantially different to those formed by wild type aB-crystallin (cf Fig. 3; WT aB). When the various aB-crystallin protein constructs were included in the assembly of the desmin filaments, two observations were clear. Firstly, the presence of aB-crystallin did not appear to change noticeably the morphology of the desmin filaments (Figs. 3 and 4). As with wild type aB-crystallin, particles were clearly seen associated with desmin filaments. For the selected examples, strong evidence for the positive association of particles with desmin filaments was seen (Fig. 4). The logs-likelihood score for the interaction of the CEb8 aB-crystallin particles with the desmin filaments at 44uC was striking (G 2 = 64.2, P = ,0.001%), suggesting increased interaction when compared to wild type aBcrystallin. This highlights the second of our observations. In the presence of both CEb8 and D155 aB-crystallin ( Fig. 3 Des + CEb8 aB and Des + d155 aB respectively) not only was there a very obvious increase in the aB-crystallin particles interacting with the desmin filaments at these higher temperatures ( Fig. 3 Des + CEb8 aB and Des + d155 aB, arrowheads), but also there also appeared to be increased desmin filament-filament interactions. The effect of the various aB-crystallin protein constructs on the assembly and filament-filament interactions of desmin was then quantified by high-speed and low-speed sedimentation assays.
Temperature dependent increase in the association of wild type aB-crystallin with desmin by low-and highspeed sedimentation assay Desmin filament assembly was conducted at 22uC, 37uC or 44uC in the presence or absence of the various aB-crystallin protein constructs. Pellet and supernatant fractions were analysed by SDS-PAGE and a representative experimental series is shown (Fig. 5). These data were combined with two other data sets and the % in each pellet (Figs. 5 and 6) and corresponding supernatant (data not shown) fractions calculated along with the standard error of the mean. The % of protein in the pellet fractions from both the low-and high-speed sedimentation assays from each sample are presented on the same bar in the charts (Figs. 5 and 6), with the proportion corresponding to the low speed assay being represented by the lower portion of each bar. This method of data presentation facilitates the comparison of both (high-speed and low-speed) sedimentation assays for each combination of aB-crystallin protein construct and desmin, which is needed to assess the interaction of aB-crystallin with desmin.
Desmin alone assembled efficiently and 94-98% sedimented into the pellet (P) fractions ( Fig. 5A and 6). This was determined by the high-speed assay and there was no significant difference between the three different temperatures. In contrast, the lowspeed centrifugation assay revealed that there was a temperature dependent increase in the proportion of desmin pelleted corresponding to 24%, 45% and 61% at 22uC, 37uC and 44uC respectively (Fig. 6, Des). This assay measured filament-filament interactions and the data therefore suggest that these interactions were temperature dependent. We excluded the possibility that desmin failed to assemble equally efficiently at the three different temperatures, because the high-speed assay revealed a similar % of desmin in the pellet fractions at the three different temperatures (Fig. 6, Des).
In the presence of aB-crystallin, 93-100% desmin sedimented into the high-speed pellet at the three temperatures indicating that wild type aB-crystallin did not alter the extent of desmin assembly (Fig. 5, HIGH SPEED; cf Des and Des + WT aB. Fig. 6; cf Des and Des + WT aB). In contrast, the low-speed sedimentation assay showed that the presence of aB-crystallin prevented filamentfilament associations and only 12% of the desmin filaments sedimented at 22uC and 37uC (Fig. 5, 6. LOW SPEED; cf Des and Des + WT aB). At 44uC, the preventative effect was lost and there was no statistically significant difference in the proportion of pelletable desmin in the presence or absence of wild type aBcrystallin ( Fig. 6; cf Des and Des + WT aB).
These experiments also measured the proportion of aB-crystallin that co-sedimented with the desmin filaments (Fig. 7A). In the absence of desmin, wild type aB-crystallin remains almost entirely in the supernatant fractions of both the high-and low-speed centrifugation assays (Figs. 4 and 6B). In the presence of desmin, however, there is a temperature dependent increase in the proportion of wild type aB-crystallin in the pellet fractions from the high-speed sedimentation assay (Figs. 4 and 6A). This corresponded to 9%, 17%, and 23% of the wild type aB-crystallin at 22uC, 37uC and 44uC respectively. These data confirm observations made in a previous study [7].
These data form the baseline for assessing the effects of changing the b3-strand, b8-strand sequences and deleting residues 155-165 in wild type aB-crystallin on the association of aBcrystallin with the desmin filaments and the subsequent effects on filament-filament interactions. All the various aB-crystallin protein constructs were soluble as determined by both sedimentation assays (Fig. 7B) and formed mono-disperse particles as judged by electron microscopy (Fig. S1). Only the D155-165 aB-crystallin protein construct showed an increased tendency to pellet and then only at 44uC (Fig. 7B, d155 aB).
Substituting the b8 strand of aB-crystallin and deleting the C-terminal residues 155-165 can promote desmin filament-filament interactions rather than inhibiting them By comparison to the b3-strand substitutions, changing the b8strand produced aB-crystallin protein constructs that showed either no significant improvement (aA-crystallin b8 chimera aB-crystallin Figure 3. Analysis of negatively stained samples desmin and wild type aB-crystallin protein by electron microscopy. The morphology of the assembled desmin filaments and the coassembled wild type (WT aB) and various aB-crystallin protein constructs was analysed by electron microscopy. Samples were negatively stained with uranyl acetate and viewed at an 100 kV accelerating voltage. Wild-type aB-crystallin formed monodisperse particles at all three temperatures (WT aB). This was also true for all the aB-crystallin protein constructs (Fig. S1). Desmin, when assembled alone, formed long smooth 10 nm filaments at all three temperatures (Des). When desmin was assembled with wild type aB-crystallin (Des + WT aB), the filaments were not aggregated and some aB-crystallin particles were seen to associate with the filaments (arrowheads). Unassociated particles are indicated (arrows). Coassembly of desmin with either aAb3, or CEb3 or aAb8 aB-crystallin gave similar results to wild type aB-crystallin at all three temperatures. In contrast, both the CEb8 and D155-165 aB-crystallin protein constructs increased desmin filament-filament associations at higher temperatures leading to filament aggregation along with increased aB-crystallin particle association (arrowheads). Bar = 100 nm. doi:10.1371/journal.pone.0025859.g003   The low-and high-speed sedimentation properties of each individual protein was determined at 3 different temperatures. The pellet (P) and supernatant (S) fractions were analysed by SDS-PAGE and the proportion of each protein in each fraction determined. By high-speed sedimentation assay, which measures the efficiency of desmin assembly, virtually all the desmin had pelleted at 22u, 37u and 44uC. By low-speed sedimentation assay, there was a temperature dependent increase in the proportion of desmin sedimented. The aB-crystallin remained largely in the supernatant fractions of both sedimentation assays. (B) Analysis of desmin pelleted by high-and low-speed sedimentation assay in the presence of either wild-type or the various aB-crystallin protein constructs at three different temperatures. aAb3 aB-crystallin (aAb3 aB) reduced the proportion of desmin filaments sedimenting at low-speed at 44uC. Conversely, the CEb8 and the D155-165 aB-crystallin protein constructs induced the complete low-speed sedimentation of desmin at 44uC. For each sedimentation assay, the band intensities were quantified and then combined with two other data sets to determine statistical significance and summarized in Figs. 5 and 6. doi:10.1371/journal.pone.0025859.g005 due to the increased aggregation of this aB-crystallin at 44uC, but this is not the case for the 37uC sample, a temperature at which this protein construct did not obviously aggregate ( Fig. 7B; d155 aB). For the CEb8 protein construct, no temperature dependent selfaggregation was observed ( Fig. 7B; CEb8 aB) and therefore the increased cosedimentation of the CEb8 construct with the desmin filaments seen by low-and high-speed centrifugation assay is due to an increased association of this aB-crystallin with the desmin filaments at 44uC (Figs. 4B and 6A; [Des + CEb8 aB]). The consequence of these increased interactions for both the CEb8 and D155-165 aB-crystallin constructs with the desmin filaments manifested itself as very obvious increased degree of desmin filament-filament associations (Fig. 6, blue bars for [Des + CEb8 aB] and [Des + d155 aB] compared with [Des + WT aB]) and also seen by electron microscopy (Fig. 3; [Des + CEb8 aB] and [Des + d155 aB]). The aggregation of the desmin filaments observed with these two aB-crystallin constructs at 44uC is very apparent when compared to the other representative images in Fig. 3. We conclude from these data that the b8 region and the C-terminal 155-165 residues also play an important role in the interaction of aBcrystallin with desmin filaments. The b8-strand substitution with the C. elegans HSP12.2 sequences and the deletion of the C-terminal residues 155-165 in aB-crystallin have both caused an increase in the interaction of aB-crystallin with the desmin filaments.

Substitution of the b3 strand can increase the ability of aB-crystallin to inhibit desmin filament-filament interactions
In contrast to the C. elegans HSP12.2 b8-strand substitution and deletion of the C-terminal 155-165 residues, substituting the b3-strand in the wild type aB-crystallin with that from aA-crystallin caused a dramatic decrease in the proportion of sedimentable desmin at 44uC in the low-speed assay (Fig. 6 [Des + aAb3 aB] dark blue bar cf [Des + WT aB], light blue bar). The effect was very evident by the gel analysis of the 44uC samples (Fig. 5B, LOW SPEED cf Des + aAb3 aB and Des + WT aB). Quantification of these gel data (Fig. 6, Des + aAb3 aB) showed that there was a 5 fold difference in the sedimentable desmin by low-speed sedimentation assay (Fig. 6 [Des + aAb3 aB] dark blue bar cf [Des + WT aB], light blue bar). The results indicated a significant reduction in the extent of the filament-filament interactions in the presence of the aAb3-aB-crystallin protein construct. The high-speed sedimentation assays showed no change in the level of assembled desmin at 44uC when coassembled with the aAb3 aB-crystallin protein construct (Fig. 5B and 6. HIGH SPEED, Des + aAb3 aB) compared to desmin assembled in the presence of wild type aB-crystallin (Fig. 5B and 6. HIGH SPEED, Des + WT aB) in support of this conclusion.
Although the effect of substituting the b3-strand from C.elegans HSP12.2 (CEb3) was not as dramatic compared to that with the equivalent sequences from aA-crystallin (aAb3), there was still a significant reduction (,2 fold) in the level of desmin sedimented in the low-speed centrifugation assay at 37uC (Fig. 5B and 6; LOW SPEED [Des + CEb3 aB] cf LOW SPEED [Des + WT aB]). At both 22uC and 44uC, no additional effects for the CEb3 strand substitution into aB-crystallin were apparent (Figs. 5B and 6, LOW SPEED; cf [Des + CEb3 aB] and [Des +WT aB]). Once again this was not because of any effects of the aB-crystallin protein constructs upon desmin assembly per se as there was no significant difference in the pelletable desmin by high-speed sedimentation assays (Figs. 5B and 6 HIGH SPEED). Electron Figure 6. Desmin sedimentation characteristics in the presence of wild type and various aB-crystallin protein constructs. Bar chart of the low-speed (light and dark blue; lower portion of each bar) and high-speed (red) sedimentation assay data for desmin coassembled with either wild type (WT aB) or the various aB-crystallin protein constructs. The percentage of desmin in each pellet fraction at 22, 37 and 44uC was determined after both low-and high-speed sedimentation assay. The mean % from three independent experiments with its corresponding standard error was calculated for each and then plotted as a composite bar chart. The assembled desmin is pelleted by high-speed sedimentation assay. At low-speed, only the assembled filaments that have formed filament-filament interactions are pelleted. Neither temperature nor the presence of the various aBcrystallin protein constructs changed significantly the proportion of desmin pelleted in the high-speed assay. The significant differences are seen in the low-speed sedimentation assay. At 44uC, aAb3 aB-crystallin (Des + aAb3 aB) produced a significant reduction in desmin pellet fraction (dark blue bar). Conversely, both the CEb8 (Des + CEb8 aB) and D155-165 (Des + d155 aB) aB-crystallin protein constructs caused significant increases in the proportion of desmin pelleted at 44uC (dark blue bars). This was also true at 37uC for the D155-165 (Des + d155 aB) aB-crystallin protein construct. doi:10.1371/journal.pone.0025859.g006 Figure 7. Cosedimentation of aB-crystallin with and without desmin filaments. A. Cosedimentation of aB-crystallin with desmin filaments. Summary of the low-speed (light blue and dark blue) and high-speed (light red and dark red) sedimentation data for various aB-crystallin protein constructs coassembled with desmin, as quantified by gel densitometry. The percentage of aB-crystallin in the pellet fractions at 22u, 37u and 44uC was determined after both low-and high-speed sedimentation to quantify the association of aB-crystallin with the sedimented desmin filaments. The most striking observation is that both the CEb8 (CEb8 aB + Des) and D155-165 (d155 aB + Des) aB-crystallin protein constructs showed significant increases in desmin binding at 44uC as shown by the high speed assay (44uC, dark red bars). Conversely, aAb3 aB-crystallin (aAb3 aB + Des) showed significantly decreased association at 44uC at high speed. The CEb3 (CEb3 aB) and aAb8 (aAb8 aB) protein constructs showed similar sedimentation properties to wild type aB-crystallin. B. Aggregation of wild type and mutant aB-crystallins as measured by low-and highspeed sedimentation. Summary of the low-speed (blue) and high-speed (red) sedimentation data for the various aB-crystallin protein constructs as quantified by gel densitometry. The percentage of aB-crystallin in the pellet fractions at 22u, 37u and 44uC was determined after both low-and highspeed sedimentation assay to quantify the aggregation of the aB-crystallins. All the other protein constructs showed similar sedimentation properties to the wild type (WT aB) aB-crystallin, except D155-165 (d155 aB) aB-crystallin at 44uC, which showed increased aggregation by both low-(darker blue) and high-speed (darker red) sedimentation assay. doi:10.1371/journal.pone.0025859.g007 microscopy confirmed that the morphology of the desmin filaments and the aB-crystallin particles were similar to those seen in samples containing the wild type aB-crystallin and desmin (Fig. 3 cf [Des +WT] with {Des + CEb3 aB]).
The combined data show that substituting the b3-strand of aBcrystallin can have beneficial effects for the ability of aB-crystallin to prevent desmin filament-filament interactions. The effects are temperature specific, but these data demonstrate that it is possible to improve this activity by changing the b3-strand, confirming that this domain plays an important role in the association of aBcrystallin with desmin filaments.
The role of aB-crystallin binding in the prevention of desmin filament-filament interactions The proportion of aAb3 that cosedimented with desmin filaments in the high-speed centrifugation assay at 44uC was significantly lower than that seen for the wild type aB-crystallin (Fig. 7A, 44uC [aAb3 aB + Des] and [WT aB + Des]). We interpret this to indicate reduced binding to the filaments. This was in addition to the ability of aAb3 to prevent desmin filamentfilament associations at 44uC, which was increased some 5 fold compared to wild type (Fig. 6, 44uC [Des + aAb3 aB] and [Des + WT aB]). Conversely, the proportion of CEb8 that cosedimented with desmin filaments in the high-speed centrifugation assay at 44uC was significantly higher than that seen for the wild type aBcrystallin (Fig. 7A, 44uC [CEb8 aB + Des] and [WT aB + Des]), indicating increased binding to the filaments. The ability, however, of CEb8 aB-crystallin to prevent desmin filamentfilament interactions at 44uC was decreased compared to wild type aB-crystallin (Fig. 6, 44uC blue bars [Des + Ceb8 aB] and [Des + WT aB]). Therefore the ability to bind to desmin filaments does not necessarily correlate with the prevention of desmin filamentfilament interactions. More specifically, increased aB-crystallin binding was not necessarily correlated with increased effectiveness in the prevention of filament-filament interactions.

Implications for aB-crystallin interactions with desmin filament
The data presented here are the first analysis of the regions in aBcrystallin previously identified as capable of interacting with desmin filaments. The strand-swapping approach used here has confirmed that the b3-strand is an important interaction site in aB-crystallin for desmin filaments. A 5 fold improvement at 44uC in preventing filament-filament interactions for the aAb3 protein construct is a direct measure that this sequence is indeed important. Equally dramatic results were also obtained for one of the b8 strandswapped protein constructs (CEb8) and for the C-terminal deletion of residues 155-165. For these two protein constructs (CEb8 and d155), a significant decrease in the ability of aB-crystallin to prevent desmin filament-filament interactions was observed (Fig. 6). The results emphasise the importance of all three regions to the interactions between aB-crystallin and self-assembling desmin filaments and subsequent filament-filament interactions (Fig. 8).
We also conclude from our data that it is reasonable to expect that changes to any interacting region can potentially produce positive as well as negative effects upon the observed activities of Figure 8. Summary of the influences of aB-crystallin on desmin filaments. The b3and b8-strands and the D155-165 sequences (C-terminal domain) in aB-crystallin were identified from peptide array studies as being desmin interaction sequences. In wild type aB-crystallin these sequences contribute to the interaction of the aB-crystallin oligomers with desmin filaments to prevent their self-association and the formation of filamentfilament aggregates. This activity can be increased by substituting the b3-strand from other small heat shock proteins (aA-crystallin and C. elegans HSP12.2). Substituting the b8-strand in aB-crystallin or removing the 155-165 residues appears to lead to the loss of this activity, but increases the binding of aB-crystallin to desmin filaments. This in turn will encourage increased filament-filament interactions, which in the case of the many point mutations in aB-crystallin linked to inherited myopathies, then leads to protein inclusion formation and the appearance of the histopathological feature of desmin-related myopathies -protein inclusions containing both desmin and aB-crystallin. doi:10.1371/journal.pone.0025859.g008 aB-crystallin toward desmin filaments (Fig. 8). Indeed summarizing the activity of the b3-strand, b8-strand and D155-165 aBcrystallin protein constructs toward other client proteins shows that both improvement and deterioration should be expected with such changes to client protein binding sequences (Table 1). Currently the data presented here do not distinguish between a direct or indirect interaction of these regions with the desmin filaments. Nevertheless the fact that aB-crystallin particles are seen to decorate desmin filaments (Fig. 3) and the various aB-crystallin protein constructs cosedimented to either greater or lesser extents (Figs. 4 and 5), we interpret to mean that aB-crystallin binds directly to the desmin filament and involves the b3-strand, b8strand and D155-165 sequences of aB-crystallin These results therefore confirmed the pin array studies that found multiple sites were responsible for the interaction of aBcrystallin with desmin [1]. Indeed the analysis of the effects of the cardiomyopathy causing mutant R120G aB-crystallin adds to this argument as this residue is outside of the b8-strand studied here, but is part of the b7-strand identified from the pin-arrays to be involved also in binding to desmin [1]. In patients, the R120G mutation leads to desmin filament aggregation and the formation of characteristic inclusions that were also enriched in aB-crystallin [5]. It was subsequently shown that the mutation also increased the binding affinity of aB-crystallin to desmin filaments by increasing the Kd by some two fold [7].

Relevance of desmin filament binding to histopathological aggregates of desmin and aB-crystallin
Another feature to emerge from the data presented here is that a significant increase in the binding of aB-crystallin to desmin filaments does not necessarily result in a similar increase in the ability of aB-crystallin to prevent the self-association of desmin filaments. This observation was first made with R120G aBcrystallin and GFAP filaments using a simplified viscometry assay [36] and was later confirmed for desmin filaments using the lowspeed sedimentation assay [7]. Both the CEb8 and D155-165 aBcrystallin protein constructs studied here showed increased cosedimentation with desmin filaments by high-speed sedimentation assay (Fig. 7A). Electron microscopy (Fig. 3) revealed this was direct binding to the desmin filaments, which coincided with an increase in desmin filament-filament associations as measured by the low-speed centrifugation assay (Fig. 6). In contrast, the aAb3 aB-crystallin construct significantly inhibited desmin filamentfilament associations, but the cosedimentation of this protein construct with desmin filaments was also significantly decreased at 44uC (Figs. [4][5]. Compare these data to the detailed analysis of the Q151X myopathy-causing mutation in aB-crystallin where increased desmin cosedimentation of Q151X aB-crystallin was accompanied by a very significant decrease in desmin filamentfilament associations [31]. The current study therefore confirms the importance of the b3and b8-strands in interactions with desmin filaments. It also adds to previous observations that increased binding of aB-crystallin to desmin filaments does not necessarily correlate with the prevention of desmin filamentfilament associations. This is reminiscent of the situation in desmin related myopathies where the characteristic histopathological feature of the disease is protein aggregates containing both aBcrystallin and desmin [11,12,13].
A role of aB-crystallin in modulating interactions between biopolymers?
Further quantitative studies are required to define the relationship(s) between binding of aB-crystallin and the polymerisation and subcellular distribution of important biopolymers such as intermediate filaments, microtubules and actin filaments. This poses an obvious question concerning the selection or hierarchy in the interaction of aB-crystallin with the different polymers and how this involves the different interaction sequences.
For the client protein T4 lysozyme, aB-crystallin has both high and low affinity binding sites [37,38]. Binding to the high affinity site appeared to induce structural changes in the client protein itself. For desmin, the measured dissociation constant [7] is equivalent to the low affinity site on T4 lysozyme, but how this might affect the subunit geometry within the filament has not yet been determined. The pin array studies show that the b3-strand, b8-strand and 155-165 regions are all involved in binding to all three cytoskeletal proteins [1]. Refinement of these studies has shown that peptides derived from the b8-strand and the Cterminal 15-165 region could inhibit tubulin assembly whereas the sequences in the b7-strand and including R120 actually promoted tubulin assembly [2]. Interestingly the b3-strand was not involved in the binding to microtubules, but the studies here have identified Numbers shown are normalised values of protein aggregation as determined by light scattering. Aggregation was calculated as (light scattering in the presence of chaperone)/(light scattering in the absence of chaperone) for the relevant client proteins. Where normalised values became greater than 1.00, this indicates increased protein aggregation. The results summarized here for the various aB-crystallin protein constructs are from previously published work (aAb3 and CEb3 see reference [17]; aAb8 and CEb8 see reference [16]; D155 see reference [14]). Median number of subunits was determined by size exclusion chromatography. Data from far-and near-UVCD spectroscopy were used to analyze secondary and tertiary structure of the aB-crystallin protein constructs. All aB-crystallin proteins consisted of large polydisperse oligomers and had far-and near-UVCD spectra unchanged (UC) from wild type (WT) aB-crystallin at 37uC and 50uC. doi:10.1371/journal.pone.0025859.t001 this region as a desmin filament interacting domain. This offers the possibility that the combination of interaction sites could be a key to polymer recognition. The role of aB-crystallin in the selfassembly of biopolymers and particularly the three main cytoskeletal elements [39], remains to be fully determined. Figure S1 Electron microscopy characterisation of the aBcrystallin proteins used in this study. The aB-crystallin samples were stained with 1% (w/v) uranyl acetate and then processed for electron microscopy. All proteins appeared as monodisperse particles. Scale bar represents 100 nm.