|Home | About | Journals | Submit | Contact Us | Français|
Protein fibrils termed amyloid-like are associated with numerous degenerative diseases as well as some normal cellular functions. Specific short segments of amyloid-forming proteins have been shown to form fibrils themselves. However, it has not been shown in general that these segments are capable of driving a protein from its native structure into the amyloid state. We applied the 3D profile method to identify fibril-forming segments within the amyloid-forming human proteins tau, alpha-synuclein, PrP prion and amyloid-beta. Ten segments, six to eight residues in length, were chosen and inserted into the C-terminal hinge loop of the highly constrained enzyme RNase A, and tested for fibril growth and Congo red birefringence. We find that all 10 unique inserts cause RNase A to form amyloid-like fibrils which display characteristic yellow to apple-green Congo red birefringence when observed with cross polarizers. These six to eight residue inserts can fibrillize RNase A and are sufficient for amyloid fibril spine formation.
Amyloid fibrils are unbranched, fibrous aggregates associated with several degenerative diseases (Westermark et al., 2007), as well as denatured globular proteins (Astbury and Dickinson, 1935; Dobson, 1999), bacterial inclusion bodies (Wang et al., 2008) and several normal cellular functions (Si et al., 2003; Sekijima et al., 2005; Wang et al., 2008). Fibrils appear to arise from the spontaneous opening of proteins, the exposure of fibril-forcing segments and subsequent self-assembly. These fibrils share an enriched β-structure reflected in a cross-β diffraction pattern and Congo red birefringence (Glenner et al., 1972; Klunk et al., 1989; Sunde et al., 1997; Serpell et al., 2000). The atomic structures of GNNQQNY and NNQQNY of the Saccharomyces cerevisiae Sup35 prion revealed the atomic nature of the cross-β spine architecture: a steric zipper consisting of two interdigitating β-sheets (8.5 Å apart in this case) mating at a dry, complementary interface. Each sheet is built by stacking of identical segments along the fibril direction, separated by 4.9 Å (Nelson et al., 2005). Hydrogen bonds hold each sheet together, and van der Waals interactions bind the two sheets into the zipper spine.
The 3D profile method was used to identify segments within a protein that have high propensity to form a steric zipper by self-complementation, and hence possibly drive fibrillation of the protein (Thompson et al., 2006). In this method, every six-residue segment of a target protein is threaded onto the steric zipper template of NNQQNY. By optimizing the interdigitation, and using the RosettaDesign energy (Rohl et al., 2004), we calculated binding energies for putative steric zippers. Those with binding energies below an empirical threshold were predicted to form fibrils, and the results were validated by experiments (Ivanova et al., 2006; Thompson et al., 2006). Atomic structures of fibril-like microcrystals of these segments show a common cross-β spine architecture with subtle geometrical differences (Sawaya et al., 2007).
Ribonuclease A is an exceptionally stable and well-characterized enzyme (Libonati and Gotte, 2004; Smith and Raines, 2006; Marshall et al., 2008), cross-linked by four disulfide bonds, particularly suitable for experiments on induced fibrillation. In our experience of over 15 years of work with RNase A (Liu et al., 1998, 2001), the wild-type enzyme never forms amyloid-fibrils, even in a wide variety of destabilizing and denaturing solvents. Because the wild-type sequence does not form fibrils, we used RNase A to test the sufficiency of a polyglutamine insert for inducing amyloid fibril formation (Sambashivan et al., 2005). This work involved domain-swapped dimers. It was long known that RNase A is capable of forming domain-swapped oligomers (Crestfield et al., 1962). A C-terminal domain swapped dimer can form when a β-strand with residues 116–124 is exchanged between two subunits (Fig. 1) (Liu et al., 1998, 2001). The swapped domain is connected to the core domain (residues 1–111) by the hinge loop of residues 112–115. Hinge loops from both subunits undergo a conformational change to align in space so that they bridge both subunits of the dimer. The dimer is held together by interactions at the interface between the swapped domain and the core domain. The enzymatic activity of both functional units of RNase A is preserved. When a 10-residue polyglutamine segment was inserted in the hinge loop and the modified sequence was concentrated in mild acid, RNase A formed amyloid-like fibrils via domain-swapping (Sambashivan et al., 2005). An atomic model was built showing that it is possible to form an amyloid-like fibril, with domain-swapped RNase A molecules maintaining their native structure around a steric-zipper, cross-β spine.
The present study tests the ability of high propensity, fibril-forcing segments to drive fully cross-linked RNase A enzyme into the amyloid state. We show that segments of six to eight residues from human proteins tau, alpha-synuclein (α-syn), prion (PrP) and amyloid-beta (Aβ) are sufficient to induce RNase A into amyloid-like fibrils. Our observations support previous work that suggest short peptide segments are important structural components of the amyloid fibril core (Lansbury et al., 1995; von Bergen et al., 2000; Balbirnie et al., 2001; von Bergen et al., 2001; Luhrs et al., 2005; Sawaya et al., 2007; Vilar et al., 2008).
Sequences of human proteins tau, α-syn, PrP and Aβ, and bovine pancreatic RNase A were analyzed with the 3D profiling method (Thompson et al., 2006) for fibril-forcing segments that form a steric zipper. With NNQQNY of the S.cerevisiae Sup35 as the backbone template, the binding energies of six to eight residue segments from these proteins were calculated. Fibril-forcing segments chosen for this study have energetic scores equal to, or less than, −3.7 kcal/mol/residue, which is the threshold for fibril formation via self-complementation (Thompson et al., 2006).
Two inactive RNase A mutants in the pET-32b vector were previously engineered by site-directed mutagenesis to generate H12A and H119A variants (Sambashivan et al., 2005). These two mutants were provided as templates to GenScript Corporation for the insertion of selected human prion- and amyloid-related segments, between residues G112 and N113 of the RNase A C-terminal hinge loop region. The H119A variant was also used as template for the insertion of GSRSRT from tau by site-directed mutagenesis in the same location. All constructs were confirmed by DNA sequencing.
Each new RNase A construct in pET-32b was transformed into BL21-CodonPlus(DE3)-RIPL cells (Agilent Technologies). Cells were grown in Teriffic Broth with 0.2% glucose, 0.4% glycerol and 100 µg/ml Ampicillin to an OD600 nm of 0.8–1.2. Protein expression was induced with 200 µM IPTG at 18°C for 20 h.
Cells were harvested by centrifugation at 6000g for 15 min. Cells were resuspended in 500 mM NaCl, 20 mM Tris pH 8.0 with protease inhibitors and lysed by sonication at 4°C. Cell debris was pelleted by centrifugation at 30 000g for 30 min, after which the supernatant was filtered with a 0.2-µm filter.
RNase A constructs were purified as described in Sambashivan et al. (2005), with the following changes. Pre-filtered supernatant for each construct was loaded onto a Hi-Trap Ni-NTA column (GE Healthcare) and eluted with an increasing imidazole gradient of 0–250 mM, in 500 mM NaCl, 20 mM Tris pH 8.0. Fractions with eluted protein were pooled and dialyzed against enterokinase buffer. Each sample was digested overnight with enterokinase (Invitrogen) and reloaded onto a Ni-NTA column. The flow-through containing cleaved RNase A protein was collected and dialyzed overnight against 100 mM NaCl.
Purified RNase A constructs were concentrated to an apparent concentration of 5–12 mg/ml, as detected by protein absorbance at 280 nm with dilutions made in 6 M Gdn-HCl, 100 mM Tris pH 8. Samples were incubated in 50% (v/v) acetic acid for an hour at room temperature, and later, lyophilized. Freeze-dried samples were resuspended in water to an apparent final concentration of 35 mg/ml and stored at room temperature.
Samples were applied to hydrophilic 400-mesh, carbon-coated, formvar films mounted on copper grids (Ted Pella, Inc.). After 2.5 min, samples were rinsed twice with 0.2-µm filtered water and subsequently stained with 1–2.5% uranyl acetate. Samples were dried and examined with a Hitachi H-7000 transmission electron microscope (TEM) at 75 keV or with a Phillips CM120 TEM at 120 keV.
The Congo red assay was carried out as described in Ivanova et al. (2006), with modifications. RNase A constructs that form fibrils were centrifuged at 20 000g. A pellet was observed in all samples post-centrifugation. Each pellet was washed twice by resuspension in 0.1-µm filtered water, centrifugation at 20 000g and removal of the supernatant. Each pellet was then resuspended in 100 µl of 0.1 mg/ml Congo red in 150 mM NaCl, 10 mM Tris pH 8.0 and incubated at room temperature for 30 min. After dye incubation, each sample was centrifuged for 5 min at 20 000g. The supernatant was discarded and the pellet was washed with 0.1-µm filtered water until the wash is no longer tainted red. Each pellet is resuspended in 5–20 µl of 0.1-µm filtered water. Sample drops were pipetted onto dust-free silanized coverslips and allowed to dry overnight in a dust-free petri dish. Dried samples were examined with an optical microscope in bright field and with cross polarizers.
Sequences for human tau, α-syn, PrP and Aβ were analyzed for fibril-forcing segments that form self-complementary steric zippers. Ten peptide segments from these proteins were chosen from a list of segments found to score equal to, or less than, −3.7 kcal/mol/residue (Table I). Two of these segments are eight residues long, while each of the remaining eight segments has six residues. All segments were also evaluated for area buried and shape complementarity at the predicted steric zipper interface (data not shown). These values were comparable to those calculated for NNQQNY of the S.cerevisiae Sup35 (Nelson et al., 2005).
RNase A gains the ability to form fibrils when the predicted fibril-forming segments are inserted in its C-terminal hinge loop (between G112 and N113) (Fig. 2). Each inserted fibril-forming segment, shown in Table I, expands the hinge loop by six to eight residues, distancing 12 residues at the C-terminus of RNase A, shown in green in Fig. 2, from the rest of the molecule, shown in pink. Fibrils were formed by all 10 constructs in identical conditions, as described in the Materials and Methods section, but with varying incubation times. Wild-type RNase A and the inactive H119A RNase A that lacks a fibril-forming insertion do not form fibrils under the same condition (data not shown). RNase A with an insertion of GSRSRT from tau, predicted to be unfavorable in forming a steric zipper, does not form fibrils (Fig. 2; Table I). Similarly, RNase A with a nine-glycine insertion was previously found not to form fibrils (Sambashivan et al., 2005).
The RNase A fibrils are unbranched with widths in the range of 70–120 Å. However, they vary in persistence length, the tendency to form either aligned bundles or less-ordered tangles, and in their resemblance either to flat ribbons or twisted threads in a cord. Electron micrographs show fibril morphologies that are representative for each RNase A construct (Fig. 2).
RNase A fibrils induced by fibril-forcing segments share structural features with amyloid proteins; the fibrils display a yellow to apple-green birefringence when dyed with Congo red and examined between crossed polarizers (Fig. 3). This specific feature of amyloid (Glenner et al., 1972; Klunk et al., 1989) is exhibited by these RNase A fibrils with varying degrees of intensity (Table I). Inactive H119A RNase A without a fibril-forming insertion binds to Congo red, but lacks birefringence when observed with crossed polarizers (Fig. 3).
Our study focuses on 10 short (six to eight residues) protein segments, predicted by the 3D profile method to form steric zippers. Crystallographic structures have shown to date that 6 of these 10 segments do in fact form steric zippers (Sawaya et al., 2007). The atomic structures of the remaining four segments are as yet unknown. When each of these 10 segments is inserted into the C-terminal hinge loop of RNase A, the enzyme is converted to the amyloid state.
Three of these segments from α-syn have been found in previous work to be essential in forming a stable fibril core. In α-syn fibrils, NVGGAVVT (residues 65–72), GAVVTG (residues 68–73) and VTAVAQKT (residues 74–81) lie in β-strands 3 and 4 of a five-layered, β-sandwich fold (Vilar et al., 2008). Because deletion of these β-strands renders α-syn incapable of forming fibrils, they were proposed to be positioned in the fibril core (Vilar et al., 2008). In the present study, all three segments induce RNase A fibrils, the shorter GAVVTG subsegment of NVGGAVVT being sufficient to force RNase A into the amyloid state. In the β-sandwich model, GVATVA (residues 50–55) lies in β-strand 2, adjacent to β-strand 3 (Vilar et al., 2008), and we have shown that it too is sufficient to force RNase A fibril formation. In addition, an A53T mutation in the GVATVA segment was found to increase the aggregation propensity of α-syn, and lead to early onset of familial Parkinson's disease (Conway et al., 1998; El-Agnaf et al., 1998; Narhi et al., 1999; Li et al., 2001). It has not been determined if a GVTTVA segment is more capable than GVATVA in forming RNase A fibrils. It is not known whether GVTTVA enhances steric zipper formation or participates in long-range interactions that stabilize the overall β-structure of α-syn fibrils. In any case, Congo red birefringence observed in all the α-syn-RNase A constructs suggests that these tested segments can force β-sheet formation when inserted in the context of a globular protein. Neurotoxicity of α-syn is correlated with an increase in β-sheet structure from its native, random coil conformation (Serpell et al., 2000; Kim et al., 2007; Sandal et al., 2008). Although the β-sandwich model proposes that interactions between these segments hold the fibril core together (Vilar et al., 2008), our results show that each segment is sufficient for the formation of an amyloid fibril core by self-complementation.
Similarly, for Aβ in Alzheimer's disease, previous studies have shown that the segment GGVVIA (residues 37–42) is important in its pathogenesis. This segment was determined to form the fibril core of Aβ by mutagenesis and solid-state nuclear magnetic resonance, while a G37D mutation in this segment decreases Aβ cell toxicity (Lansbury et al., 1995; Kanski et al., 2002; Luhrs et al., 2005). The power of this segment to drive fibrillation is emphasized by our observation that GGVVIA is sufficient to force RNase A to form amyloid fibrils.
All induced RNase A fibrils in this study are in the amyloid state, as shown by Congo red binding and birefringence. Inactive H119A RNase A that lacks a fibril-forming insertion binds to Congo red, but does not display birefringence. Negatively charged Congo red may bind to positively charged RNase A (pI = 9.3) via electrostatic interactions (Liu et al., 1998), or bind to β-sheets in the native structure of RNase A. The lack of birefringence when observed between cross polarizers presumably reflects the absence of a steric zipper spine.
Similarly, the non-fibril-forming segment GSRSRT when inserted into RNase A displays weak Congo red birefringence although no fibril formation was observed. The GSRSRT segment is predicted to be weak in forming a steric zipper spine. Insertion of GSRSRT may allow RNase A to open up its native structure, allowing RNase A to sample more conformations and form new intermolecular interactions. Weak Congo red birefringence suggests more order in the GSRSRT aggregates observed. The new interactions, however, are insufficient for fibril formation.
We have identified fibril-forcing segments in the fibrillizing proteins tau, α-syn, PrP and Aβ. These segments form fibrils when in isolation from the rest of the protein (Sawaya et al., 2007) and when inserted in the C-terminal hinge loop of wild-type RNase A. From these observations, we postulate that under the right conditions, many proteins expose fibril-forcing segments, which may be sufficient to cause aggregation and fibril formation, as observed for normally soluble proteins by Astbury and Dickinson (1935), Dobson (1999) and Wang et al. (2008), as well as others.
This work was supported by the National Science Foundation (MCB-0445429); the National Institute on Aging (1R01 AG029430) and the United States Department of Energy Office of Biological and Environmental Research.
We thank Magdalena Ivanova, Stuart Sievers, Marcin Apostol, Mari Gingery, Shilpa Sambashivan and Lukasz Goldschmidt for discussion and technical help.
Edited by Valerie Daggett