Binding determinants of the small heat shock protein, αB-crystallin: recognition of the ‘IxI’ motif

Abstract

Small heat shock proteins (sHSPs) play a central role in protein homeostasis under conditions of stress by binding partly unfolded, aggregate-prone proteins and keeping them soluble. Like many sHSPs, the widely expressed human sHSP, αB-crystallin (‘αB’), forms large polydisperse multimeric assemblies. Molecular interactions involved in both sHSP function and oligomer formation remain to be delineated. A growing database of structural information reveals that a central conserved α-crystallin domain (ACD) forms dimeric building blocks, while flanking N- and C-termini direct the formation of larger sHSP oligomers. The most commonly observed inter-subunit interaction involves a highly conserved C-terminal ‘IxI/V’ motif and a groove in the ACD that is also implicated in client binding. To investigate the inherent properties of this interaction, peptides mimicking the IxI/V motif of αB and other human sHSPs were tested for binding to dimeric αB-ACD. IxI-mimicking peptides bind the isolated ACD at 22°C in a manner similar to interactions observed in the oligomer at low temperature, confirming these interactions are likely to exist in functional αB oligomers.

Introduction

Small heat shock proteins (sHSPs) help to maintain protein homeostasis by interacting with partly unfolded, aggregate-prone proteins to prevent cell damage (Haslbeck et al, 2005; Bukau et al, 2006). In humans, there are 10 members of the ATP-independent sHSP family (HSPB1, HSPB2, etc.) of which αB-crystallin (‘αB’ or HSPB5) is a paradigm example (Horwitz, 1992). Though originally discovered in eye lens as a protein essential for maintaining lens transparency, αB is additionally known to be expressed in muscle and brain. Specific biological roles for αB continue to emerge. In mice, knockout of αB reduces life span by half whereas its overexpression protects cardiac tissue from necrosis and myocardial infarction following ischaemic stress (Brady et al, 2001; Ray et al, 2001). Dysfunctions of αB in humans are associated with the occurrence of neurodegenerative diseases such as Alzheimer’s disease and Alexander’s disease (Kato et al, 2001; Goldstein et al, 2003). Inherited missense mutations in αB are associated with desmin-related cardiomyopathy, congenital cataracts, and myopathy (Vicart et al, 1998; Selcen and Engel, 2003; Liu et al, 2006; Rajasekaran et al, 2007). Furthermore, high expression of αB in cancer cells confers resistance to chemotherapy and is correlated with poor clinical outcomes (Vargas-Roig et al, 1998; Moyano et al, 2006). Yet despite its critical roles in maintaining healthy cells, understanding of αB structure and function remains rudimentary.

As is common for mammalian sHSPs, αB exists as large polydisperse assemblies containing variable numbers of subunits (Aquilina et al, 2003). The inherent dynamics of the assemblies is a defining feature and is likely linked to αB function. For decades, high-resolution structural studies on functional, full-length αB and mammalian sHSPs in general were confounded by both polydispersity and dynamics. Developments in solid-state NMR (ssNMR), solution-state NMR, and electron microscopy are finally providing experimental means to obtain atomic-level information on αB. Each approach has its strengths and weaknesses and each requires quite different experimental conditions under which a measurement can be made. Thus, a unifying model should be congruent with all observations to the extent possible.

As a class, sHSPs are defined by the presence of a conserved α-crystallin domain (ACD) that mediates formation of dimeric building blocks. The central ACD is flanked by highly variable N- and C-termini that mediate formation of the larger oligomeric assemblies (Bagnéris et al, 2009; Jehle et al, 2009; Langanowsky et al, 2010; Baldwin et al, 2011). sHSP oligomers may be composed of a single type of subunit or of a combination of different subunits (Zantema et al, 1992; Kato et al, 1994; Sun and Liang, 1998; Sugiyama et al, 2000; Bukach et al, 2009; Engelsman et al, 2009). Atomic-level structures of ACDs from a number of mammalian sHSPs including αB have been determined by X-ray crystallography and ssNMR (Bagnéris et al, 2009; Jehle et al, 2010; Langanowsky et al, 2010; Baranova et al, 2011; Clark et al, 2011). Atomic-level structures are available for non-mammalian sHSP oligomers (Kim et al, 1998; van Montfort et al, 2001; Stamler et al, 2005; Hilario et al, 2011). Pseudo-atomic models for αB have been generated using data from a combination of methods including ssNMR, SAXS, negative-stain EM, and cryo-EM (Braun et al, 2011; Jehle et al, 2011). The growing database of structural information provides a rich resource for studies aimed at understanding how sHSP structure is related to function.

Structures of sHSP oligomers and of isolated ACDs reveal some common features. The ACD is a 6–8 stranded β-sandwich structure and two ACDs usually associate in an anti-parallel fashion to form homodimeric units that serve as the building block for higher order structures (see Supplementary Figure S1) (Bagnéris et al, 2009; Jehle et al, 2010; Langanowsky et al, 2010; Baranova et al, 2011; Clark et al, 2011; Jehle et al, 2011). In addition, sequences from outside the ACD have been observed to be in contact with the ACD both in crystalline samples and by ssNMR. The interactions involve three ACD regions: (1) the β3 strand, (2) a groove on the edge of the ACD structure known as the β4/β8 groove, and (3) the surface formed by the dimer interface (Kim et al, 1998; van Montfort et al, 2001; Bagnéris et al, 2009; Langanowsky et al, 2010; Jehle et al, 2011) (see Supplementary Figure S1 for strand nomenclature).

In what is by far the most frequently observed interaction, a small conserved sequence motif called the IxI/V motif found in the C-terminal region of many sHSPs is bound in the β4/β8 groove of an ACD. The motif is defined as two Ile (or Val) residues separated by one residue. The interaction has been observed both in the context of homo-oligomers such as αB-crystallin, MjHSP16.5, and wheat HSP16.9, and in the context of ACD crystals, where the interactions take place among neighbouring dimers in the crystal lattice (Kim et al, 1998; van Montfort et al, 2001; Langanowsky et al, 2010; Jehle et al, 2011). Heteromeric IxI/V-β4/β8 interactions have been detected in solution between human sHSPs, αB and αA-crystallin (αA) by FRET experiments (Pasta et al, 2004). That the ACD-IxI/V interaction is observed in so many contexts implies some shared structural and/or functional role. To varying degrees, the interaction has been implicated in oligomer formation (Studer et al, 2002; Pasta et al, 2004). Deletions spanning the C-terminal motif result in a loss of higher order oligomeric structure in multiple sHSPs and mutation of the conserved motif residues Ile/Val (IxI/V) to Gly (GxG), Ala (AxA), or to Phe (FxF) also appears to abrogate the interaction (Studer et al, 2002; Pasta et al, 2004; Hayes et al, 2008; Takeda et al, 2011). Altogether, the observations suggest that the ACD-IxI/V interaction has a role in oligomer structure. Solution-state NMR studies at 25°C designed to specifically detect highly flexible regions detect the IxI/V motifs in some sHSP oligomers but the same experiments fail to observe this region in αB, implying the region is either bound within oligomer or is exchanging among multiple states (Ghahghaei et al, 2009; Benesch et al, 2010; Treweek et al, 2010). In contrast, a ¹³C-methyl-TROSY solution-state NMR study on αB oligomers at temperatures ranging from 24 to 50°C contained signals assigned to the IxI/V motif congruent with the residues being highly flexible and in an unbound state (Baldwin et al, 2011). This observation led the authors of that study to propose that the IxI/V interactions observed in structural studies of sHSPs in the solid-state or crystalline form may not accurately reflect the oligomer in solution (Baldwin et al, 2011). Here, we assess the inherent binding properties of an ACD and an IxI/V motif outside an oligomeric or crystalline context. Towards this end, we have tested the ability of peptides that mimic sequences from human sHSPs to bind to αB-ACD and identify determinants for the interaction using solution-state NMR.

Results

The isolated ACD of αB (residues 64–152; called αB-ACD) forms well-behaved dimers that are amenable to solution-state NMR (Jehle et al, 2009). NMR resonance assignments for the αB-ACD have been previously reported and provide the ability to obtain residue-specific information regarding potential peptide binding to an isolated ACD dimer (Jehle et al, 2009). Solution-state ¹H–¹⁵N HSQC NMR spectra provide both a peak position (¹⁵N and ¹H chemical shift) and peak intensity for ¹⁵N-labelled proteins. Comparison of spectra collected on samples that contain increasing amounts of a putative binding partner will reveal perturbations in the position, the intensity of a peak, or both. Perturbations are caused by a change in the local environment of the nuclei responsible for the peak. In this way, events such as peptide or protein binding can be detected, quantified, and mapped onto known structures of the protein under investigation.

NMR binding experiments were performed by titrating peptides with sequences containing the C-terminal IxI motif of αB or other human sHSPs into NMR samples of ¹⁵N-αB-ACD. The effect of peptide addition was monitored in the ¹H–¹⁵N HSQC NMR spectra of αB-ACD collected at 22°C. As illustrated in Figure 1, significant changes in specific peak positions are observed as a consequence of addition of C-terminal αB-IxI peptide (Ac-PERTIPITREEK-NH2). Thus, the spectra confirm that isolated ACD dimers in solution can interact with a C-terminal sequence outside an oligomeric context. Notably, only a subset of resonances is affected by the presence of peptide, indicating that binding is not accompanied by a significant conformational change in the ACD dimer. Inspection of individual peaks in spectra collected throughout a titration revealed that some peaks exhibit intermediate exchange behaviour: a phenomenon in which peaks broaden under conditions where the bound state is not saturated and reappear as observable resonances at saturating peptide concentrations (Supplementary Figure S2). In addition, a subset of affected peaks titrate to multiple resonance positions (up to 3), indicating that these nuclei exist in multiple environments as a result of peptide binding (Figure 2). The observation of multiple resonances rather than one averaged resonance indicates that distinct ACD-peptide complexes exist and that they do not interconvert rapidly. The observed exchange behaviour allows us to estimate the exchange rate, k_ex, from the magnitude of the chemical shift differences among species. The chemical shift perturbations (CSPs) between the free and bound states are all quite similar for the amide group resonance for residue L94, which has well-resolved multiple peaks. This indicates a similar rate of exchange for each observed species. Under conditions where distinct resonances are observed for different forms present, the exchange rate (k_ex=k_on+k_off) between two forms is given by k_ex<Δω, where Δω is chemical shift difference between the two forms. In the case of αB-IxI peptide bound to αB-ACD, we have determined the upper limit for k_ex to be 60 s⁻¹ for the multiple states observed.

Figure 1.

Overlay of ¹H–¹⁵N HSQC spectra of ¹⁵N αB-ACD collected at 22°C, in the absence (black) and at saturating concentrations (five-fold) of αB-IxI peptide (red).

Figure 2.

αB-ACD peaks titrate to multiple resonances at saturating concentrations of αB-IxI peptide. Three distinct resonances are observed for the amide of L94 in the presence of αB-IxI peptide (upper left arrows), while a single dominant peak is observed in the presence of the HSPB2-VxI peptide. The resonance for T132 also titrates to multiple positions in the presence of αB-IxI peptide (lower left), while a single dominant peak is observed in the presence of the HSPB2-VxI peptide. Spectra of the peptide-free and peptide-saturated ACD are shown in black and red, respectively.

As mentioned above, resonance assignments for peaks in the αB-ACD spectrum provide residue-specific information from the titration data. For each peak that could be followed through the titration series, the difference between its peptide-free chemical shift and its peptide-bound position was measured, as shown in the histogram in Figure 3. For peaks that titrate to multiple peptide-bound states, the position of the most intense peak in the peptide bound spectrum was used in this analysis. A histogram in which the CSPs are ranked according to their magnitude reveals a group of peaks with CSPs significantly larger than all others (>0.25 p.p.m.). For a subset of peaks, the final positions in the peptide-bound spectrum could not be readily assigned because their resonances could not be followed due to intermediate exchange behaviour or limited peak intensity. Nevertheless, the fact that they are in intermediate exchange indicates that their CSP values are even larger than those peaks whose positions shift gradually throughout the titration. The residues that experience the largest perturbations as a consequence of peptide binding reside in the β4/β8 groove and the loop between β4 and β5 (β4/5 loop) of the αB-ACD. Less affected peaks are clustered at the C-terminal end of the β3 strand (Figure 3). Peaks that are unaffected by the addition of peptide derive from residues outside these regions (Figure 3). The large CSPs observed within the β4/β8 groove are consistent with the peptide binding to a surface similar to that observed in ssNMR and X-ray crystallographic studies (Figure 4). Thus, there is an intrinsic affinity between the αB-ACD and the αB C-terminal sequence that promotes their binding even outside the context of solid-state or crystalline samples.

Figure 3.

Histogram of calculated CSPs (black) for the αB-ACD in the presence of saturating concentrations of αB-IxI peptide. Positions where CSPs could not be determined due to exchange behaviour (and therefore have very large chemical shifts) are indicated in grey. Secondary structure elements are shown above.

Figure 4.

The chemical shift perturbations due to αB-IxI peptide binding map to the β4/β8 groove on the edge of the ACD. (Top) CSPs are mapped onto a secondary structure representation of the αB-ACD dimer (green and grey) (2klr). The most affected peaks (CSP>0.16) are shown in magenta on one monomer (green) and map to the β4/β8 groove and β4/5 loop. Residues where CSPs could not be quantitated due to exchange over large chemical shifts are shown in red. (Bottom) Surface representation of the αB-ACD dimer (2klr) rotated 45° with the same residue colouring as above. The C-terminal αB sequence PERTIPITREEK (white sticks) is shown as modelled previously with ssNMR restraints to illustrate the congruence between the chemical shifts observed for binding outside the oligomeric context and the structure observed in the context of αB oligomer.

The small CSPs (0.14–0.24 p.p.m.) observed for residues at the C-terminal end of the β3 strand may arise from two possible scenarios: (1) this region may comprise a second, lower affinity peptide binding site or (2) the observed CSPs may be due to small changes in the chemical environment of these residues caused by binding in the β4/β8 groove. To distinguish between the two possibilities, a point mutation in the β8 residue Ser135 was designed with the goal of disrupting the β4/β8-peptide interaction. In αB structures derived from ssNMR data and crystal structures, the side chain of Ser135 sits between two pockets into which the side chain of Ile159 and Ile161 of the IxI motif are bound (Langanowsky et al, 2010; Jehle et al, 2010). Substitution of Ser135 with a longer Gln side chain (S135Q) should occlude the binding pockets without significantly perturbing other elements of the structure. The ¹H–¹⁵N HSQC spectrum of S135Q-αB-ACD is highly similar to the wt-ACD spectrum, indicating that the mutation has not significantly altered the overall structure (Supplementary Figure S3). Spectra from a peptide titration carried out for ¹⁵N-S135Q-αB-ACD reveal that S135Q-ACD binds peptide with significantly lower affinity: even a 10-fold molar excess does not achieve saturation with the mutant ACD whereas for the wt-ACD, saturation was attained at a four-fold excess. Furthermore, the observed CSPs are much smaller in the mutant ACD and all affected peaks exhibit fast exchange rather than intermediate/slow exchange (Figure 5). Altogether, the observations indicate that the C-terminal peptide binds with lower affinity to the S135Q-ACD mutant and has a shorter lifetime in the bound state. Notably, CSPs at the C-terminal end of the β3 strand also fail to reach saturation at 10-fold excess peptide in the S135Q mutant. If these perturbations were due to a second, independent binding event, then they would not be expected to be disrupted by the S135Q mutation. The apparent coupling of the CSPs observed within the β4/β8 groove and the β3 region is strong evidence that they arise from the same binding event. Together with our previously published ssNMR data, the data presented here lead us to conclude that the IxI motif within the C-terminal peptide has an intrinsic affinity for the β4/β8 groove of αB-ACD and that alterations to the groove can modulate the affinity of this interaction.

Figure 5.

Mutation of residue S135 in the β4/β8 groove disrupts IxI peptide binding. (A) Selected resonances from ¹H–¹⁵N HSQC of WT ACD (left) and the mutant S135Q ACD (right) in the absence and presence of αB-IxI peptide are compared. Spectra collected in the absence of peptide are shown in black. Spectra collected in the presence of 6-fold αB-IxI peptide (red) for the WT and S135Q ACD and 10-fold (green) peptide for S135Q are shown. In the S135Q mutant, residues K92 from the β4 strand and R120 (which is close to and behaves similarly to perturbed residues on the β3 strand) show a decrease in peptide-dependent chemical shift perturbation. The titration vector for the resonance of K92 in the WT spectra is shown as a grey arrow. (B) Peptide-induced CSPs observed for resonances of L94 and I133 are compared, with the same conditions and colouring as in (A).

Does the interaction between an isolated C-terminal peptide and an ACD dimer recapitulate that observed in the αB oligomer? To address this question, we compared the measured ¹⁵N chemical shifts for the isolated αB-ACD dimer in solution with those measured within the oligomer. The calculated difference is plotted for each residue in the histogram, shown as red bars in Figure 6. The largest differences occur at the termini of the excised ACD construct, the β3 strand, and at positions within the β4/β8 groove, suggesting that these regions differ in structure and/or environment between the ACD dimer alone in solution and in an oligomer. The calculation was repeated using ¹⁵N chemical shifts measured for the peptide-saturated state of the ACD in solution and the ssNMR oligomer, shown as black bars in Figure 6. Notably, in the presence of peptide all observable β4/β8 residues in the ACD in solution have shifted significantly towards their solid-state ¹⁵N chemical shifts (compare black and red bars). This remarkable concordance suggests the interaction between the αB-IxI peptide and the αB-ACD in the absence of oligomeric context recapitulates native intermolecular contacts within the αB oligomer. We note that a few peaks move away from their solid-state chemical shifts with the addition of peptide. The most significant of these deviations involves the perturbed positions at the C-terminal end of the β3 strand of the ACD. Thus, the secondary effects of peptide binding observed for this region in solution may not reflect a native effect and should be interpreted with caution.

Figure 6.

Comparison of ¹⁵N chemical shift between the αB-ACD in the context of the oligomer observed by ssNMR and isolated αB-ACD in solution. (¹⁵N_{solid state}−¹⁵N_{solution state}) differences were measured using the solution-state values observed both in the absence of peptide (red bars) and in the presence of saturating αB-IxI peptide (black bars). Note the greatly reduced values for (¹⁵N_{solid state}-¹⁵N_{solution state}) in the presence of peptide in the β4 strand and β4/5 loop (residues 89–98) and the β8 strand (residues 133–137). Positions where (¹⁵N_{solid state}−¹⁵N_{solution state}) could not be determined are assigned the value of 0.0 p.p.m. for clarity. For residues where multiple states were observed in the peptide saturated spectrum, the position of the most intense peak in the saturated spectrum was used.

To test the direct involvement of the IxI sequence in binding the β4/β8 groove, NMR experiments were performed using αB C-terminal peptides in which the IPI sequence was mutated either to GPG or APA. No detectable binding was observed for either mutant peptide even at high molar excess of peptide, consistent with the two Ile residues being the major determinants for the interaction (data not shown). This result raises the question as to whether other IxI/V sequences that exist (i.e., VxI or IxV) are capable of binding to the groove in αB. To address this question, NMR binding experiments were performed using peptides mimicking C-terminal sequences of HSPB1, HSPB2, and αA-crystallin (HSPB4), each of which contain an I/VxI or IxI/V motif (Table I). The NMR titrations confirm that each of these peptides binds to αB-ACD and produces CSPs that map to the β4/β8 groove. Differences in affinity and exchange behaviour are apparent in the heterologous titrations. For example, binding of the αA-IxV peptide and the HSPB2-VxI peptide exhibit fast exchange behaviour instead of the intermediate exchange observed with the αB-IxI peptide (Supplementary Figure S2). In a manner similar to the αB-IxI peptide titration, a subset of peaks in the αA-IxV peptide spectra titrates to multiple well-populated resonances throughout the titration. Throughout the HSPB2-VxI peptide titration, peaks titrate along a vector to a dominant saturated state. Additional multiple resonances are apparent at saturation, but are much more lowly populated than those observed in the αB-IxI or αA-IxV peptides (Figure 2). The observations permitted approximation of a K_{d app} value of ∼90 μM for HSPB2 peptide based on the dominant binding state CSPs (Supplementary Figure S4). This value should be viewed only as an upper limit estimate K_{d app} because alternative lowly populated bound states effectively serve as competitors to the dominant binding mode and therefore contribute to the observed binding isotherm. Further, our inability to detect additional binding modes for this peptide does not mean they do not exist, but rather that if they do exist they exchange rapidly. Altogether the data indicate that peptides that contain either an IxV or a VxI motif can bind the β4/β8 groove of the αB-ACD. In contrast, peptides that correspond to other regions of sHSPs that have been observed to make contact with an ACD in other structures display no detectable binding to αB-ACD under the conditions of our experiments (Table 1) (Bagnéris et al, 2009; Jehle et al, 2010). Thus, the binding observed in the solution-state NMR experiments is specific and cannot be attributed to a general promiscuity of αB-ACD for unstructured polypeptides.

Table 1. Summary of C-terminal peptide sequences from different sHSPs binding to αB-ACD.

^aV/IxI/V motifs are shown in bold text.

^bWhere the peptide originates relative to the ACD within the sHSP is denoted N-terminal (NT) or C-terminal (CT).

Discussion

The αB-ACD binds αB C-terminal IxI motif outside the oligomeric context

Our results demonstrate that the αB-ACD has an inherent affinity for its C-terminal IxI motif and that the interaction can take place outside the context of either an oligomer or a crystal lattice. The major determinants for binding are the two Ile side chains, which interact directly with the β4/β8 groove of the ACD, yielding a bound state that is remarkably similar to that detected by ssNMR studies (Jehle et al, 2009). In addition to the β4/β8 groove, the β4/5 loop is strongly affected by peptide binding, while loops on the opposite end of the groove are not significantly perturbed. In the αB oligomer structure determined from ssNMR, the C-terminal region of a neighbouring dimer contacts the β4/5 loop but not the loops on the other end of the β4/β8 groove (Jehle et al, 2010). Similar binding modes are observed in crystal structures: the sequence N-terminal to the IxI motif is in close proximity to the β4/5 loop of a neighbouring subunit while sequences C-terminal to the motif exit the groove without making contact with β6+7/8 or β3/4 loops (Langanowsky et al, 2010; van Montfort et al, 2001; Kim et al, 1998; Jehle et al, 2010). Thus, the solution-state NMR data reported here are consistent with observations made for numerous sHSPs using multiple techniques. Solution-state NMR spectra that contain signals only from highly flexible and mobile residues failed to observe resonances from the IxI motif on αB-crystallin oligomers at 25°C. This is consistent with the sequence being less flexible than residues more C-terminal to it, whose signals were detected (Treweek et al, 2010). In contrast, the IxI motif was deemed highly flexible based on a solution-state NMR study on oligomeric αB in which Ile resonances were selectively detected by the methyl-TROSY approach (Baldwin et al, 2011). Altogether the available observations are most consistent with a model in which the IxI motif exists in an equilibrium between a free, flexible state, and a bound state. In the context of αB oligomers that contain between 24 and 32 subunits, free and bound IxI motifs, and empty and filled β4/β8 grooves, will exist simultaneously. Our results indicate a bound-state lifetime of 17 ms (k_ex≅60 s⁻¹) and a dissociation constant below 90 μM for the binding of αB’s IxI C-terminal sequence. These numbers are for a bimolecular binding reaction. In the context of an oligomer, the effective local concentrations of both binding components will be significantly higher than the conditions of our bimolecular, peptide-based NMR experiments. So, it is reasonable to conclude that the population of bound C-termini at any given time will be significant. As we have previously proposed, the equilibrium-free and bound states may be shifted towards one state or the other by a change in conditions such as pH and temperature (Jehle et al, 2010).

Finally, we note the presence of multiple bounds states for αB IxI peptide at saturation in the NMR spectrum. Two IxI/V motif binding orientations in the β4/β8 groove have been observed in crystallographic studies (‘parallel’ and ‘antiparallel’ to the β8 strand; Langanowsky et al, 2010). Due to the presence of two β4/β8 grooves per ACD dimer, if the two peptide orientations are independent and of approximately similar affinity in solution, three bound species would be predicted: parallel/parallel, antiparallel/antiparallel, and parallel/antiparallel in a 1:1:2 ratio. It is notable that for the very limited number of examples where three resonances can be clearly and unambiguously observed, their ratios appear to be approximately as predicted. Thus, the simplest interpretation of the observed behaviour is that the αB-IxI peptide can bind in two orientations in solution. The slow exchange behaviour observed between binding modes indicates that a slow step, presumably peptide dissociation, governs the inter-conversion between binding modes. We cannot, however, rule out other possible explanations for the presence of multiple bound states. For example, if binding to the ACD is not restricted to a single peptide conformation, then the multiple bound states observed by NMR may originate from variability in the bound peptide conformation as opposed to its orientation. Whether the observed data result from one or a combination of these two possibilities cannot be unambiguously distinguished from the data at hand.

Binding to the β4/β8 groove requires an I/VxI/V motif

Of the peptides tested, only peptides with I/VxI/V motifs were observed to bind to the αB-ACD. Substitution of both Ile side chains with the smaller hydrophobic side chain Ala abolished all detectable binding. Whether all combinations of I/V in the two positions and whether Leu or other residues are also allowable in one or both positions will require a more thorough and systematic study. The structure of the β4/β8 groove is defined by the side chains that face into the groove from the two β strands. Substitution of the β8 residue, Ser135, with a larger side chain (Gln) significantly inhibited IxI binding. This result is consistent with the structure derived from ssNMR in which Ser135 resides between the two hydrophobic binding pockets and may provide a functional explanation to the strong conservation of small amino-acid residues observed at this position in sequence alignments (Kim et al, 1998; Taylor and Benjamin, 2005; Baldwin et al, 2011). Furthermore, the lowered affinity of the S135Q mutant for the IxI-peptide and the much shorter lifetime of the mutant bound state suggest that interpretation of studies in which this residue is mutated and/or modified should be approached with care (Baldwin et al, 2011).

Hetero-oligomers and C-terminal peptide binding

Hetero-oligomerization of αB with αA crystallin and HSPB1 has been reported and αB/HSP20 hetero-oligomers have been indirectly observed (Zantema et al, 1992; Kato et al, 1994; Sugiyama et al, 2000; Bukach et al, 2009; Engelsman et al, 2009). In contrast, attempts to observe αB/HSPB2 hetero-oligomers have failed to detect these species (Sugiyama et al, 2000; Engelsman et al, 2009). Such observations raise the question of what determines formation of sHSP hetero-oligomeric species. Given the high degree of conservation within the β4/β8 sequences of human sHSPs and the variability of sequences flanking the IxI motifs in the C-terminal regions, an appealing possibility is that β4/β8-to-IxI interactions help to define hetero-oligomers. If this were the case, then we would expect to detect interactions with the C-terminal peptides from αA and HSPB1, but not from HSPB2. As summarized in Table I, the prediction is not supported by our results. Nevertheless, interactions between the IxI motifs from αA-crystallin and HSPB1 and the αB-ACD suggest such an interaction is possible in hetero-oligomers. Whether and how the interaction contributes to hetero-oligomer formation or function must await further investigation.

Conclusion

It is possible to recapitulate interactions that normally occur within the context of an sHSP oligomer using a simplified system of isolated parts. The remarkable concordance of the NMR data acquired in solution on isolated ACD dimers and unstructured peptides with those data collected on oligomers in the solid-state indicate that important and relevant insights can be obtained using either approach. This provides a powerful strategy with which to study functionally relevant interactions involving sHSPs and their binding partners, be they other sHSPs or partly folded client proteins.

Materials and methods

Expression and purification of ¹⁵N labelled αB-ACD

Expression of a TEV cleavable N-terminally His-tagged αB-ACD (64–152) was carried out using methods and conditions previously described for the αB-ACD (Jehle et al, 2009). Cultures were pelleted then resuspended in a buffer solution containing 20 mM Tris (pH 8) 100 mM NaCl, 1 mM PMSF and EDTA-Free Protease Inhibitor Cocktail (Sigma-Aldrich) and were lysed by French Press. Lysate was diluted five-fold in 6 M Urea, 100 mM NaCl and centrifuged at 43 000 g. The supernatant was applied to a Ni-NTA Agarose column (Invitrogen). Protein bound to the column was refolded by three column volumes (CV) of 25 mM Tris, 150 mM NaCl (Buffer A), 10 mM Imidazole, pH 7.6 and His-tagged αB-ACD was eluted with the addition of Buffer A containing 250 mM Imidazole, pH 7.6. The TEV cleavable His tag was removed by dialysing overnight into Buffer A containing 5 mM DTT in the presence TEV protease. Following TEV digestion, the protein was reapplied to a Ni-NTA column and eluted with the addition of Buffer A containing 50 mM Imidazole pH 7.6. αB-ACD was further purified by anion exchange using a MonoQ column (GE) and SEC as previously described (Jehle et al, 2009). Purified αB-ACD was dialysed into water and lyophilized for storage.

Peptides

Peptides were generated by Life Tein Custom Peptide Synthesis Services with acetylated N-termini and amidated C-termini. Additional αB-IxI peptide was generously provided by the Gestwicki Laboratory (University of Michigan). Peptides were initially solubilized in water and the pH was adjusted to neutral, before the addition of phosphate and salt. Peptide concentrations were calculated based on the dry weight of lyophilized peptide and final peptide stock solutions ranged from 30 to 65 mM peptide, in 25 mM sodium phosphate and 150 mM NaCl pH 7.5.

NMR data collection and analysis

NMR experiments were performed on a Bruker 500 MHz Avance III spectrometer equipped with a 5-mm z axis gradient, triple resonance probe. ¹⁵N-labelled, lyophilized αB-ACD was resuspended in 25 mM sodium phosphate (pH 7.5), 150 mM NaCl. Peptide titration spectra were collected on 600 μM αB ACD (300 μM A) samples in 25 mM sodium phosphate, 150 mM NaCl, 10% D₂O at 295K. Titration points at one-fold (600 μM), two-fold, and four-fold peptide were collected for all peptides that bound the ACD. Additional titration points were collected with the αB-IxI peptide and the αA-IxI peptide to confirm titration assignments and saturation points. Peptides that did not perturb the αB-ACD spectrum at four-fold excess were defined as peptides that do not bind the ACD and their presence in solution was confirmed by ¹H–1D NMR. Because of the inherent acidity of peptide solutions used, the pH of all NMR samples was confirmed prior to data collection.

All spectra were processed using NMRPipe and data analysis was performed with NMRViewJ (Johnson and Blevins, 1994; Delaglio et al, 1995). Assignments for amide resonances in peptide-bound αB-ACD were obtained by following the resonances during the course of the titration. Some resonances in the spectra of αB-ACD bound to αB-IxI peptide could not be assigned unambiguously due to intermediate exchange behaviour where the resonances disappear during the intermediate titration points and reappear at the excess peptide concentration. To resolve ambiguities in αB-IxI peptide bound αB-ACD assignments, peak trajectories were compared to trajectories in αA-IxI and HSPB2 peptide titrations where the resonances could be followed unambiguously. CSPs were calculated based on the formula {(Δδ_HN²+(Δδ_N/5)²)}^1/2 where Δδ_HN and Δδ_N are the chemical shift differences in amide proton and nitrogen resonances in p.p.m., respectively. Δδ_N is divided by 5 since the spectral width in ¹⁵N-dimension is five-fold greater than the spectral width in the ¹H dimension. CSPs were plotted onto a model of 2KLR using Pymol.

The K_{d app} for the HSPB2-VxI peptide binding to αB-ACD was determined using NMRViewJ (Johnson and Blevins, 1994). Six peaks from the HSPB2-VxI peptide titration were fit to a quadratic binding curve using a base 10 quadratic fit and 250 simulations and an average K_{d app} for all peaks fit was calculated.