A solenoid design for assessing determinants of parallel β-sheet registration

Abstract

A novel protein construct is presented that combines a homotrimeric, triple-stranded β-helix as a guest to a homotrimeric foldon unit from bacteriophage T4 fibritin. The β-helical solenoid selected is short (46 residues) and is part of a subdomain of the T4 cell-puncturing device. The resultant design is trimeric and displays greatly enhanced stability over each sub-component alone. The intended goal is a design that will enable evaluation of sequence determinants that promote in-register versus out-of-register parallel β-sheet homotrimerization. Towards that end, the importance of a set of three buried salt-bridges was evaluated by converting them to residues otherwise consistently found throughout the natural solenoid at the same positions. The critical role of the charged residues in the salt-bridges was evident in that their elimination resulted in amyloid-like aggregation.

Introduction

Homopolymeric self-assembly of β-strands is an established mechanism and motif for both pathological and functional protein self-assembly (Chiti and Dobson, 2006). A notable pathological example is the 40–42 residue Aβ peptide. In Alzheimer's patients, this peptide forms amyloid fibers as a component of plaques in and around neurons. Amyloid fibers, by definition, are composed of β-sheets in which the strands run orthogonal (cross-β) to the long axis of the fiber (Tycko, 2006). A notable functional example is the protein curli from gram-negative bacteria. Formation of amyloid by this protein is carefully orchestrated and is essential for biofilm formation (Chapman et al., 2002).

The topological intra-sheet arrangement of strands within fibers is varied. There are two central considerations. Parallel versus antiparallel arrangement and stacking registration. Most pathological amyloids for which there is structural data, reveal a parallel, in-register β-sheet topology. Notable examples include tau and Aβ from Alzheimer's (Margittai and Langen, 2004; Tycko and Wickner, 2013), islet amyloid polypeptide from type II diabetes (Luca et al., 2007) and α-synuclein from Parkinson's (Der-Sarkissian et al., 2003). Numerous exceptions exist. For example, diffraction analysis of short peptide amyloid sub-domains show a mixture of parallel and antiparallel structures, although the parallel structures are always in-register (Sawaya et al., 2007). Fibers formed from the functional amyloid, curli, are parallel, but not in-register (Shewmaker et al., 2009).

It is likely that the amyloid state is a generic property of all proteins (Fändrich et al., 2001). This is most likely the result of facile parallel in-register stacking (Astbury et al., 1935). Many residues can promote these interactions by making favorable contacts when spaced 4.7 Å apart. These residues particularly include Q, N, I, F and Y. In sufficient number, such as in glutamine expansion diseases (Wetzel, 2012), this lends itself to oligomeric, amyloid assembly. An important insight is that natural selection likely selects against amyloid formation (Wright et al., 2005). Kinetics plays an important role in this as globular protein folding competes with amyloid self-assembly (Jahn and Radford, 2008). This selective pressure may additionally be relieved by simply reducing the use of aggregation prone amino acids. Finally, selective pressure may result in sequences that, as a result of steric or charge interactions, are incompatible with in-register self-assembly.

The lack of clarity reflects the paucity of discrete, quantitative tools. This is a pervasive issue in amyloid research as a result of the intrinsic heterogeneity of species formed. In this work, we report on our first efforts to establish a scaffold upon which we can quantitatively assess the role of sequence in determining parallel in-register versus antiparallel out-of-register sequences. The premise is to use a naturally occurring triple-stranded, homotrimeric β-helical solenoid as such a topology must have evolved to promote out-of-register stacking. This solenoid is constructed as a guest attached to well-characterized host protein. By using a host, we create a high effective local protein concentration to serve as pressure for the homotrimer to assemble into amyloid. The result is a construct that switches between in-register and out-of-register alignment depending on sequence effects on the relative stability of the two forms.

Materials and methods

Materials

Buffers and salts were purchased from J.T. Baker. UltraPure™ guanidine hyrdrochloride (GuHCl) and ethylenediaminetetraacetic acid (EDTA) were purchased from Invitrogen. Molecular biology reagents were purchased from New England Biolabs. All oligonucleotides were synthesized at the W.M. Keck Foundation Biotechnology Resource Laboratory (Yale University, New Haven, CT, USA).

The gp5C_441–486-foldon gene in the pJexpress 414 plasmid was ordered from DNA 2.0 (Menlo Park, CA, USA). The gp5C_441–486-foldon gene coded for residues 441–486 of gp5C from bacteriophage T4 with the foldon domain, residues 457–483, of bacteriophage T4 fibritin attached C-terminally with a three-residue Ser–Val–Glu linker. The gp5-(His)₆ gene was a gift from Shuji Kanamaru (Tokyo Institute of Technology). gp5C_473–518 from gp5 was subcloned into the pJexpress 414 plasmid with the foldon domain for the expression of HS. Mutants of HS were made by splicing by overlap extension. Expression and purification were performed using a modified protocol received with the gp5-(His)₆ gene. Ni-NTA resin was used for affinity chromatography and gel filtration was performed using Superdex-200 resin for the purification of all proteins. Fractions corresponding to the trimeric form of a protein were collected and run a second time on the Superdex-200 column to confirm isolation of trimeric protein. EDTA was present in buffers after the elution from the Ni-NTA resin to prevent metal chelation by the (His)₆-tag. All spectroscopy measurements were made in a standard buffer of 5 mM sodium phosphate, 150 mM NaCl, pH 7.4 unless otherwise specified.

Spectroscopy

Full wavelength spectra and thermal melts in 0 M GuHCl circular dichroism measurements were made at a concentration of 15 µM trimeric protein (45 µM chain concentration) in standard buffer on an Applied Photophysics Chirascan circular dichroism spectrometer using 1 mm pathlength cuvettes. Spectra shown have had the buffer spectrum subtracted. Spectra were collected at 1 nm intervals from 200 to 250 nm with 5 s averaging per step with enabled adaptive sampling. The temperature of the sample holder was held at 20°C for the full wavelength spectra collection. The temperature was increased from 4 to 94°C in 5°C steps with 1 min equilibration time.

Chemical melts for 15 µM trimeric gp5C_441–486-foldon and the foldon domain alone in the standard buffer brought to the appropriate GuHCl concentrations were performed using a two-channel fluorimeter (QuantaMaster C-61 fluorescence spectrometer, PTI, London, ON) to monitor intrinsic tryptophan fluorescence. Chemical melts with HS in the standard buffer brought to the appropriate GuHCl concentrations were performed using an Aviv circular dichroism spectrometer Model 215. Sample holders in both instruments were held at 20°C.

Thermal melts in 2.5 M GuHCl were measured by CD for 15 µM trimeric protein concentration in 5 mM sodium phosphate, 50 mM NaCl, 1 mM EDTA, pH 7.4 buffer brought to a final concentration of 2.5 M GuHCl using the Aviv 215 spectrometer. Spectra were collected from 200 to 250 nm with 2 nm steps with 5 s averaging per step. Temperature was increased in 2°C steps with 3 min temperature equilibration.

Thioflavin T (ThT) emission spectra were collected for 15 μM trimeric protein in 5 mM sodium phosphate, 50 mM NaCl, 1 mM EDTA, pH 7.4 with 10 µM ThT at 20°C using a two-channel fluorimeter (QuantaMaster C-61 fluorescence spectrometer, PTI, London, ON) with 4 nm slit widths. The excitation wavelength used was 440 nm and the emission spectra were collected from 460 to 510 nm.

Data analysis

The chemical melt data was fit to the following equation to determine the ΔG of unfolding.

$$f(\theta )=\frac{({\theta }_N+m_N\times \theta )+({\theta }_D+m_D\times \theta )\times \text{exp}((-\mathrm{\Delta }G-m\times \theta )/(R\times T))}{1+\text{exp}((-\mathrm{\Delta }G-m\times \theta )/(R\times T))},$$

where θ_N is the signal for folded protein, θ_D is the signal for unfolded protein and m_N and m_D are the slopes of the folded and unfolded baselines, respectively, θ is the measured signal, m is the m-value and ΔG is the free energy determined from the fit. R is the gas constant 1.987 cal/(mol K) and T is the temperature in Kelvin.

The thermal melt data was fit to a sigmoidal curve for the data sets displaying a single transition. The thermal melt data that showed two transitions were fit to a double sigmoid. The fits were used to extract the T_ms, however, we lacked confidence in the baselines for the thermal melts in 2.5 M GuHCl so the second derivative of the transition was used to calculate the T_ms. The T_ms obtained from the fits and the second derivative of the transitions varied by at most 1.4°C. Fits of the data were performed using Igor Pro 6 (Wavemetrics, Lake Oswego, OR, USA). Structural renderings in Fig. 1 were made using the PyMOL Molecular Graphics System, Schrödinger, LLC.

Fig. 1.

Primary and 3D structures relevant to the current study. Throughout, three alternating colors are used to distinguish between each of the strands of the homotrimer. (A) The cell-puncturing device from bacteriophage T4. The trimeric β-helical region is shown as ribbon, while the remainder is shown as a semi-transparent, solvent accessible surface. (B) The 5-rung structure corresponding to gp5C_473–518. (C) Schematic of the construct referred to as HS in the main text. Specifically, His-tag, followed by gp5C_473–518 followed by foldon. Shown is a primary sequence, however, upon trimerization, each helix is assembled from contributions from all three strands. This is indicated for one of the three faces by using the three-color scheme. (D) A chain of solvent-exposed salt-bridges that run most of the length of the gp5C β-helix. Shown is the subset of residues present in gp5C_473–518 (rightmost, vertical stripe of residues shown in (C)). (E) A symmetric set of salt-bridges formed by the same pair of residues from each of the three strands. These are shown in the plane orthogonal to the helical axis of gp5C_473–518. Each salt-bridge includes the E486 residue and K488 residue of different strands. All structural renderings produced using PDBID: 1K28 (Kanamaru et al., 2002).

Results

Triangular, β-helical solenoids can be regarded as parallel β-sheet amyloid half-sites. That is, the interior of the solenoid is not amyloid, but the exterior is a fair representation of half of the inter-sheet interactions present in a cross β-sheet. Moreover, within the interior contacts of the solenoid, those interactions that involve intra-strand stacking of side chains are also amyloid-like. Naturally occurring triple-stranded solenoids have therefore overcome a particularly strong evolutionary challenge: three identical strands, with sequences biased towards β-sheet formation, come together to form β-sheets that are not in-register. Rather, triangular, triple-stranded solenoid formation is dependent on a consistent shift in register from one strand to the next. If the solenoid is a subdomain of a larger protein, the larger protein may further be holding the three strands at extraordinarily high effective concentrations. This likely increases the challenge of forming β-strands with the correctly shifted registration.

The bacteriophage T4 cell-puncturing device is a large (321 kDa) trimeric complex of the gp27 and gp5 proteins (Kanamaru et al., 2002). A gp5 protein maturational cleavage occurs between S351 and A352 that generates a stable complex (Kanamaru et al., 1999). Infection of a host cell, or exposure to increased temperature in the lab results in dissociation of the complex into two fragments, gp5* and gp5C (Kumar Sarkar et al., 2006; Nishima et al., 2011). The latter remains trimeric and contains an unusually long and symmetric triple-stranded β-helical subdomain (Fig. 1A). Each face of the triangular solenoid is one sheet in thickness, 8 residues across with a total of 18 parallel winds. This is a particularly uniform solenoid with a twist of 3° per rung. The 18-rung solenoid with the additional lysozyme subdomain displays long-term stability and has been capitalized upon for the development of nanostructures (Ueno et al., 2006). Most of the interior of this solenoid is solvent exposed which further distinguishes it from many other solenoids (Fig. 1B). Four of the eight side chains on each rung are presented outward, towards solution; a feature we have recently capitalized upon to investigate the role of amyloid lateral surface in nucleated fiber assembly and formation of pre-amyloid toxic species (Rubio et al., 2014) (Fig. 1B).

A stable, trimeric protein, HS, was created by fusing gp5C_473–518 with a 27-residue foldon (Fig. 1C, Table I). The 473–518 residue sequence of gp5C was chosen because of several notable features including external salt-bridges (Fig. 1D), internal salt-bridges (Fig. 1E) and a recurring repetitive sequence mostly in keeping with the overall consensus sequence of the intact gp5C β-helical solenoid. The foldon, derived from residues 457–483 of the bacteriophage T4 fibritin, was selected as it is trimeric, small and has well-characterized refolding behavior (Güthe et al., 2004). This foldon has previously been used to create an independently expressible and crystallizable form of residues 490–575 of the gp5 β-helix, termed gp5_βf (Yokoi et al., 2010), as well other trimeric constructs such as collagen-like helices and adenovirus fiber shaft trimers (Frank et al., 2001; Papanikolopoulou et al., 2004). A trimeric fraction of HS, 33.4 kDa, was readily purified by size exclusion chromatography (Fig. 2). Moreover, reassessment of the purified trimeric fraction showed no evidence of re-equilibration to aggregate and/or monomer species. The foldon forms a dense cluster of three tryptophans, which gives a characteristic, positive contribution to the circular dichroism (CD) at 230 nm (Fig. 3A). The far-UV CD of HS shows both a β-sheet minimum at 216 nm, and the positive contribution of the foldon (Fig. 3A). This could similarly be observed for the previously studied gp5_βf. For both gp5_βf and HS, the contribution of foldon to the far-UV CD could be quantitatively subtracted indicating that all of the foldon in both proteins is folded. Importantly, the mean residue ellipticity of these difference spectra was directly compared (Fig. 3B). The near identity of the profiles suggests that the solenoid portion of HS is in a closely similar conformation to gp5_βf.

Table I.

The solenoid-foldon constructs discussed in this work

Construct name	Residues from gp5C	Mutations
HS	473–518	None
gp5C441–486-foldon	441–486	None
gp5βf	490–575	None
HSE486V	473–518	E486V
HSK488I	473–518	K488I
HSE486V, K488I	473–518	E486V, K488I
HSE486K, K488E	473–518	E486K, K488E
HSE486V, K488I, V502E, I504K	473–518	E486V, K488I, V502E, I504K

The name of the construct as used in the main text is given, the residues used from gp5C, and the mutations made to the solenoid, retaining the gp5C numbering.

Fig. 2.

Purification and stability of the HS complex. Representative size exclusion chromatography profile for the last step of HS trimer purification (solid). Trimer fractions collected and reapplied to the same column showed only a single peak (dotted). The expected molecular weights of trimeric and monomeric HS are 33.4 and 11.1 kDa, respectively. Column void volume was determined using dextrans. Size standards are indicated using dotted vertical lines, and included bovine serum albumin (66 kDa), carbonic anhydrase (29 kDa) and cytochrome C (12 kDa).

Fig. 3.

Structural assessment of HS by far-UV CD. (A) Far-UV CD of 15 µM of HS, foldon alone or gp5_βf in our standard buffer conditions. (B) Mean residue ellipticity computed after subtracting foldon spectrum from HS and gp5_βf, respectively. (C) CD spectra after thermal denaturation of foldon and HS in 5 mM sodium phosphate, 50 mM NaCl, 1 mM EDTA, pH 7.4 and 2.5 M GuHCl. The latter are collected after the first and second thermal transitions are complete, 78 and 102°C, respectively (Fig. 4).

The HS protein displays enhanced chemical stability compared with the isolated foldon. Monitoring both the CD at 230 nm and the intrinsic tryptophan fluorescence, the chemical (GuHCl) stability of the foldon alone was found to be 6.0 ± 0.1 kcal/mol (Fig. 4A), consistent with previous reports (Güthe et al., 2004). An HS-related construct, foldon linked instead to gp5C_441–486 (gp5C_441–486-foldon), showed a single unfolding transition that was unchanged from foldon alone. In contrast, HS showed a single unfolding transition that was shifted from the observed foldon unfolding transition. The unfolding of the foldon and HS was both reversible (identical results were obtained using either folded or unfolded starting states). The apparent stability of HS was 7.6 ± 0.1 kcal/mol. This is a ΔΔG of 1.6 kcal/mol, clearly indicating that the foldon is stabilized by the attached peptide.

Fig. 4.

Chemical and thermal denaturation of HS. (A) Representative chemical denaturation isotherms with fits for foldon, foldon linked to gp5C_441–486 and HS. The former two were monitored by changes to intrinsic fluorescence intensity at 280 nm. Changes in CD at 230 nm were also used for HS as orthogonal confirmation (not shown). (B) Thermal denaturation of 15 µM of foldon or HS in standard buffer monitored by CD at 230 nm. (C) Thermal unfolding of HS or foldon in 5 mM sodium phosphate, 50 mM NaCl, 1 mM EDTA, pH 7.4 and 2.5 M GuHCl. (D) Schematic of conjectured assignment of observed T_ms to conformational transitions in HS. The single T_m of foldon alone is likely a consequence of the first, rate limiting unfolding transition.

The HS solenoid construct displays exceptional thermal stability. At 15 µM, foldon alone undergoes a thermal transition with a midpoint, T_m, of 74.7°C (Fig. 4B). In marked contrast, HS begins to unfold around 84°C but little unfolding of HS was observed up to a temperature of 104°C. This suggests that the attached solenoid has a very small change in the entropy of unfolding. Given the water-filled, tube-like nature of gp5C, we conjecture this observation to derive from a very small hydrophobic effect. In addition, if the gp5C_473–518 component of HS is not fully unfolded, the foldon domain would be held together at high effective concentration. We surmise this to be the origin of the foldon displaying high thermal stability when coupled to gp5C_473–518.

HS is subject to thermal denaturation in the presence of GuHCl. Thermal denaturation of the foldon and HS was performed in 2.5 M GuHCl. At this concentration of GuHCl, the foldon unfolds with a T_m of 45.0 ± 0.5°C (Fig. 4C, Table II). In contrast, the HS protein unfolds with two transitions having T_ms of 69.3 ± 0.5 and 86.9 ± 1.0°C, respectively. This observable suggests the presence of two states that do not interconvert on the hours timescale. Unlike the chemical melt, however, heat does not result in full denaturation (Fig. 3C). Consistent with the observation is that the refolding of HS shows only partial reversibility, with the reversible fraction giving a single T_m of 56.3 ± 0.1°C. This T_m is similar to the lower T_m of the forward melt. A simple explanation of the unfolding is a sequential path, first of the host foldon, and second, the guest solenoid. This is unlikely as HS retains a contribution of positive ellipticity at 230 nm even after the first transition is complete (Fig. 3C). The reverse direction of the unfolding path, guest unfolds followed by foldon, would not result in two transitions given the much lower T_m of foldon alone (Fig. 4C). Thus, we believe that apparent thermal unfolding of HS to be a two-step process, first to a partially folded state.

Table II.

Thermal melt midpoints from thermal melts in 2.5 M GuHCl, with statistics from repeat measurements, for the indicated constructs

Construct	Tm1 (°C)	Tm2 (°C)
Foldon	45.0 ± 0.5a	NA
HS	69.3 ± 0.5	86.9 ± 1.0
HSE486V	72.8 ± 0.5	85.6 ± 1.4
HSK488I	NA	82.6 ± 0.4
HSE486K, K488E	72.6 ± 1.0	85.5 ± 2.5
HSE486V, K448I,V502E, I504K	54.1 ± 0.6	NA

NA refers to profiles for which only a single T_m was apparent.

^aMelt of foldon alone is not meaningfully assigned to T_m1 or T_m2.

An internal salt-bridge is critical to trimeric structure formation. The consensus sequence, determined by inspection, of the repeating sequence of the gp5_βf solenoid is (G N a T I X V ±), with solvent-exposed side chains in bold and ‘a’ corresponding to small aliphatics (G, A or V). Positions 1 and 7 are invariant and along with position 3, stabilize the β-helical turns. Position 8, ‘±’, corresponds to a chain of salt-bridges (Fig. 1D). In principal, such alternating charges should be sufficient for promoting staggered assembly of the β-strands. Parallel, in-register stacking of the same sequence would result in charge–charge repulsion. In addition, gp5C_473–518 possesses a set of three internal salt-bridges (Fig. 1E) that otherwise deviates from the consensus sequence. Mutation of these two residues to the consensus, HS_{E486V, K488I}, resulted in a system for which 100% of the protein aggregated under physiological solution conditions. Moreover, these aggregates were highly responsive to the amyloid indicator dye, ThT (Wolfe et al., 2010) (Fig. 5). This suggests that the exterior salt-bridge network is insufficient for the control of inter-strand registration. The aggregation prone nature of HS_{E486V, K488I} could be rescued by restoring either of the salt-bridge residues, HS_E486V and HS_K488I. We note that the latter appears to lose the lower T_m transition. Furthermore, the salt-bridges were tolerant to inversion, HS_{E486K, K488E}. Transfer of the set of salt-bridges to another location on the solenoid, HS_{E486V,K488I,V502E,I504K}, resulted in the formation of an apparently monodisperse, soluble oligomer, larger than a trimer and with a single T_m higher than that of the foldon alone (Table II). The formation of a uniformly sized oligomer suggests that the Glu486/Lys488 salt-bridge may at some point be regarded as a transferable motif once the additional residue requirements are determined. Regardless, it is clear that the folded structure of HS is sensitive to the presence and position of the charged residues in the internal salt-bridges.

Fig. 5.

Fluorescent response of amyloid indicator dye to HS and related constructs. Representative emission spectra for 10 µM ThT in the absence and presence of 15 µM of each of the indicated protein constructs. The spectrum marked gp5_βf* is the gp5_βf spectrum scaled by 5/11 to facilitate a per β-strand comparison with the HS spectrum.

Discussion

In this work, we have excised a 40-residue β-helix motif from a much larger parent structure and shown that it can be stably assembled once attached to a trimeric foldon. We observe that the β-helix gives a net stabilization to the foldon, and has a CD spectrum compatible with the much larger parent structure from which it was excised. A disproportionate increase in thermal stability is observed and appears to map to the unfolding of the foldon with the solenoid remaining intact.

The HS construct can inform on the role of sequence in delineating in-register versus out-of-register assembly. Here, charged residues from internal salt-bridges were shown to be one of the required features for correct assembly of HS. Removal of the E486/K488 salt-bridge (HS_{E486V, K488I}) resulted in a wholly aggregation prone construct. The internal salt-bridge network can therefore be regarded as a component contributing to the prevention of aggregation. However, partial removal of the salt-bridge, HS_E486V and HS_K488I, rescues this aggregation. This suggests that it is not the salt-bridges, per se, that are essential to prevent aggregation. Rather, what is important are buried charges that are disfavored when organized as an in-register parallel stack. In contrast, note that HS_{E486V, K488I} retains salt-bridges running alongside the three solvent-exposed vertices of HS (Fig. 1D). These are plainly not sufficient to prevent aggregation. We believe that there are numerous motifs of potential relevance to stabilizing in-register versus out-of-register assembly (and vice versa). The intent and demonstration of this host–guest construct is that of providing a means of quantitatively exploring these motifs that affect in-register versus out-of-register assembly.

Mutations that rescued HS aggregation always resulted in thermal melting T_ms matched to that of the parent HS. As a tool, this coincidence is meaningful as it is an indication that only the foldon is subject to unfolding during thermal assessments. Consider the refolding pathway of the foldon in isolation (Güthe et al., 2004). The last two steps include these transitions: dimer + monomer ↔ trimeric intermediate ↔ native (Fig. 4D). Attachment of the foldon strands to a folded and thermally resistant solenoid, i.e. HS and variants, would show no change in apparent T_ms so long as the solenoid remains intact. In this model, the lower T_m of HS corresponds to the transition from native to trimeric intermediate. The higher T_m reflects the dissociation of the foldon strands, albeit still tethered to the structured solenoid. In any case, it is the presence of two thermal transitions that provides a facile measurement of the folded state of the solenoid.

The formation of parallel β-strand stacking from peptide sequences is documented in systems ranging from functional fungal prions (Ross et al., 2005) to pathogenic mammalian prions (Cobb et al., 2007) to short zipper-peptides from larger amyloidogenic systems (Sawaya et al., 2007; Ruschak and Miranker, 2009). It is a struggle to achieve thermodynamic and structural insights into the determinants of the self-assembly process as a result of the heterogeneous nature of the both intermediate and final oligomeric states. The construct developed here is a first step towards establishing a quantitative tool for generating insights into parallel amyloid assembly.

Funding

This work was supported by the National Science Foundation [grant number 0907671 to A.D.M.], and the National Institute of General Medical Sciences at the National Institutes of Health [grant numbers GM102815 and GM094693 to A.D.M. and T32GM007223 to E.M.W.].