Amino Acid Proximities in Two Sup35 Prion Strains Revealed by Chemical Cross-linking*

Background: Different but currently uncharacterized chain folding patterns of the Sup35 protein determine the prion strain type. Results: Strain-specific chemical cross-linking is obtained. Conclusion: Chain topology constrained by cross-linking indicates that N-terminal residues of Sup35 are differently arranged in two prion strains. Significance: Different Sup35 folding patterns are revealed by general methods readily applicable to other prions and pathological amyloids. Strains of the yeast prion [PSI] are different folding patterns of the same Sup35 protein, which stacks up periodically to form a prion fiber. Chemical cross-linking is employed here to probe different fiber structures assembled with a mutant Sup35 fragment. The photo-reactive cross-linker, p-benzoyl-l-phenylalanine (pBpa), was biosynthetically incorporated into bacterially prepared recombinant Sup(1–61)-GFP, containing the first 61 residues of Sup35, followed by the green fluorescent protein. Four methionine substitutions and two alanine substitutions were introduced at fixed positions in Sup(1–61) to allow cyanogen bromide cleavage to facilitate subsequent mass spectrometry analysis. Amyloid fibers of pBpa and Met/Ala-substituted Sup(1–61)-GFP were nucleated from purified yeast prion particles of two different strains, namely VK and VL, and shown to faithfully transmit specific strain characteristics to yeast expressing the wild type Sup35 protein. Intra- and intermolecular cross-linking were distinguished by tandem mass spectrometry analysis on fibers seeded from solutions containing equal amounts of 14N- and 15N-labeled protein. Fibers propagating the VL strain type exhibited intra- and intermolecular cross-linking between amino acid residues 3 and 28, as well as intra- and intermolecular linking between 32 and 55. Inter- and intramolecular cross-linking between residues 32 and 55 were detected in fibers propagating the VK strain type. Adjacencies of amino acid residues in space revealed by cross-linking were used to constrain possible chain folds of different [PSI] strains.

translation termination factor recognizes stop codons to release elongating polypeptides from the ribosome. In the prion state, Sup35 adopts a strain-specific aggregation structure, which self-perpetuates by capturing and templating the conversion of soluble Sup35 molecules into its like. As a result, the functional cellular Sup35 concentration is reduced, causing readthrough of stop codons. Different [PSI] strains can titrate out different amount of soluble Sup35, resulting in varied degrees in translation readthrough.
[PSI] aggregates can be isolated from yeast cells to seed prion fibers from bacterially prepared recombinant Sup35 protein or its N-terminal prion-forming fragments (2,3). When introduced into yeast without the prion ([psi Ϫ ]), the fibers can give rise to the same [PSI] strain as the original seeds, indicating that the strain-specific conformation has been faithfully propagated in the test tube. Taking advantage of the ability to propagate specific [PSI] strain structures in vitro, we previously prepared infectious Sup35 fibers of four different prion strains, namely VH, VK, VL, and W8, which were identified earlier in the laboratory using high resolution strain-typing methods (3)(4)(5)(6). The VH and W8 strains caused strong translation readthrough (thus they were often called "strong strains"), and the VK and VL strains exhibited weaker effects (weak strains). The minimally required length of a Sup35 fragment to propagate prion infectivity is strain-dependent. Forty amino acid residues suffice for the VK strain, but for the VH, VL, and W8 strains, the first 53 residues of the Sup35 protein are required (5,7).
Electron diffraction studies revealed that fibers of the four strains all exhibited a 4.7 Å periodicity, characteristic of ␤-sheet-rich amyloid. Mass per length measurements on the fibers further established that there was approximately one Sup35 molecule for every 4.7 Å repeat length along the fiber axis (4,5). Therefore, Sup35 monomers must adopt a distinct folding pattern for each strain, which then stacks up every 4.7 Å to form a fiber (Fig. 1). Recent solid state NMR measurements on fibers nucleated by a spontaneously formed Sup35 amyloid sample, which mostly induced strong translation readthrough in yeast, further suggested that the same amino acid residue in each Sup35 molecule was aligned in register (8). The electron diffraction and mass per length measurements were fully compatible with the parallel in-register structural arrangement (8,9), but clearly contradicted another often cited Sup35 fiber model (10), which had each Sup35 monomer spanning more than 4.7 Å distance along the fiber axis, thus stipulating an unobserved mass per length measure that was significantly smaller than one molecule per 4.7 Å. The model also specified a very distinct type of unit cell arrangements, consisting of two Sup35 monomers arranged in a dyadic head to head (or tail-totail) fashion. Diffraction patterns predicted from such arrangements were not observed by x-ray fiber diffraction experiments (11).
Strain-specific chain folding patterns of Sup35 monomers can be further constrained by tertiary adjacencies of amino acid residues, derivable with chemical cross-linking experiments. Two issues are relevant in selecting a proper cross-linking reagent: (i) the sequence of Sup35 prion domain is of low complexity, composed mainly of glutamine, tyrosine, asparagine, and glycine residues. This limits the choice of a cross-linker that reacts with specific functional groups of amino acid side chains. (ii) Prion fibers, as amyloids, are tightly packed. An exogenously added cross-linker might not be able to penetrate the fiber to reveal amino acid contacts inside (12). Given the considerations, the photo-reactive amino acid analog p-benzoyl-L-phenylalanine (pBpa) 2 was selected in this study. The pBpa crosslinker was first applied in the study of peptide-protein interactions by Kauer et al. (13), and later adopted by others to study amyloid aggregates (14). Upon UV irradiation at a wavelength between 350 and 360 nm, the photo-generated triplet state of the benzophenone group abstracts a hydrogen atom from geometrically accessible carbon-hydrogen bond of the polypeptide backbone and side chain methylene groups and then combines the resulting radical to generate a covalent link between the ketone carbonyl carbon and the attacked methylene carbon. Chin et al. (15) extended the amino acid repertoire for cellular protein synthesis by engineered an orthogonal tRNA/aminoacyl-tRNA synthetase pair that transferred pBpa at a uniquely engineered amber codon. Employing this technology, pBpa was biosynthetically incorporated into recombinant proteins, which were then converted into strain-specific prion conformations by nucleation. The cross-linking induced by UV irradiation was subsequently analyzed by tandem mass spectrometry.
Protein Expression and Purification-The plasmid pSup-BpaRS-6TRN, expressing the orthogonal tRNA CUA /pBpa-tRNA synthetase pair (15), was co-transformed into E. coli strain OverExpress TM C41 (DE3) (Lucigen) together with mutant pSup(1-61)(ϩ4M2A)-GFP-Strep(II)-His 6 plasmids, which expressed pBpa-containing proteins. The transformants were grown in LB medium supplied with pBpa (270 mg/liter; Bachem AG), ampicillin (50 g/ml), and chloramphenicol (17 g/ml) until A 600 reached 0.5. Isopropyl ␤-D-thiogalactopyranoside was added to a final concentration of 1 mM to induce protein expression at 37°C for 3 h. The cells were then collected by centrifugation. Cell pellets were lysed by sonication in buffer A (20 mM Tris, 100 mM NaCl, 5 mM imidazole, pH 7.6) containing 1 mM PMSF and 1ϫ complete EDTA-free protease inhibitor mixture (Roche). After centrifugation (12,000 ϫ g, 30 min at 4°C) to remove debris, lysate was loaded onto nickel-nitrilotriacetic acid column (Qiagen) pre-equilibrated with buffer A. Recombinant proteins were eluted in buffer B (20 mM Tris, 100 mM NaCl, 500 mM imidazole, pH 7.6) after washing with 10 column volumes of buffer A. The UV monitor was turned off during elution to minimize exposure. Eluted protein was subjected to ultracentrifugation at 200,000 ϫ g on top of a 30% sucrose cushion (g/ml in buffer A) for 2 h at 4°C. Soluble protein was aliquoted, snap frozen in liquid nitrogen, and stored at Ϫ80°C. Immediately before seeding, protein was thawed on ice and loaded onto a StrepTactin column pre-equilibrated with buffer W (100 mM Tris, 1 mM EDTA, pH 8.0) at 4°C. The column was washed with 5 column volumes of buffer W, and the protein was eluted with buffer E (buffer W containing 2.5 mM desthiobiotin). Protein concentrations were determined by The assembly scheme is based on data from electron diffraction and mass per length measurements; chain folding patterns, which are yet to be determined, are drawn more freely. Black and blue chains represent randomly mixed 14 N-and 15 N-labeled Sup35 polypeptides.
Cross-linking and Cyanogen Bromide Digestion-Fibers were placed on top of a 30% sucrose cushion (g/ml in buffer W) to sediment at 200,000 ϫ g for 2 h at 4°C. The pellet was resuspended in 100 l of buffer W to a concentration of 50 -80 M (monomer equivalent) and transferred to a 1.5-ml ice-cold microcentrifuge tube to be irradiated with UV light (365 nm, 0.9 J/cm 2 intensity) in a Spectrolinker XL-1500 UV oven (Spectronics). One hundred microliters of 2ϫ SDS sample buffer (125 mM Tris-HCl, pH 6.8, 50% (v/v) glycerol, 4% (g/ml) SDS, 0.02% (g/ml) bromphenol blue, 10% (v/v) ␤-mercaptoethanol, 10 M urea) were added, and the sample was loaded onto a 12% Tris-glycine SDS-gel for electrophoresis. Gel bands corresponding to monomer (labeled as UV1) and dimer (labeled as UV2) species were excised, cut into small pieces, and destained thoroughly in 50 mM ammonium bicarbonate containing 40% (v/v) methanol. They were subsequently rinsed in wash buffer (50 mM ammonium bicarbonate, 50% (v/v) acetonitrile), shrunk in 100% acetonitrile, dried in a speed vacuum concentrator (Heto), and then rewetted in formic acid (Sigma-Aldrich, Ն98%) containing CNBr (Alfa Aesar) in 100-fold molar excess to the amount of starting material (in monomer equivalent) (20). Water was added to dilute formic acid to 70% (v/v) to a final reaction volume of ϳ600 l. After 24 h of incubation at room temperature, the supernatant was collected. The remaining gel pieces were submerged in ϳ400 l of 40% (v/v) acetonitrile in the microcentrifuge tube, which was placed in a Branson 3200 ultrasonic cleaner and agitated for 10 min to elute peptides. The procedure was repeated once more. Supernatants were pooled together and lyophilized.
Mass Spectrometry-NanoLC-nano electrospray ionization-MS/MS analysis was performed on a nanoAcquity UPLC system (Waters) connected to the Orbitrap Elite hybrid mass spec-trometer (Thermo Electron) equipped with a PicoView nanospray interface (New Objective). Peptide mixtures were loaded onto a 75-m-inner diameter, 25-cm-length C18 BEH column (Waters) packed with 1.7-m particles with a pore size of 130 Å and were separated using a segmented gradient in 90 min from 5% to 40% Solvent B (acetonitrile with 0.1% formic acid) at a flow rate of 300 nl/min and a column temperature of 35°C. Solvent A was 0.1% formic acid in water. The mass spectrometer was operated in the data-depandant mode. Briefly, survey full scan MS spectra were acquired in the Orbitrap (m/z 350 -1600) with the resolution set to 120,000 at m/z 400 and automatic gain control (AGC) target at 10 6 . The 20 most intense ions were sequentially isolated for collision-induced dissociation (CID) MS/MS fragmentation and detection in the linear ion trap (AGC target at 10,000) with previously selected ions dynamically excluded for 60 s. Ions with singly and unrecognized charge state were also excluded.
Identification of Cross-linking-MS/MS spectra were assigned to specific peptides using the automated database search program MassMatrix (version 2.4.2) (21), which searched the database with a mass accuracy-sensitive probability-based scoring algorithm. CNBr-cleaved peptide sequences (designated F1-F5) and all possible cross-linked peptides (e.g. F1-F1 (with pBpa in the first F1 fragment linked to all possible residues in the second F1), F1-F2, and so on) were included in the database, and pBpa was assigned as a specific user-defined amino acid. Peptide mass tolerance was set to 10 ppm, and fragment mass tolerance was set to 0.60 Da. Assignments were confirmed, and cross-linked residues were deduced by careful visual inspection of b-and y-ion series. Tandem mass spectra of cross-linked peptides of mixed isotope type (i.e. 14 N-15 N) were assigned manually, using as a guide computer-assigned, visually confirmed spectra of monoisotopic species, eluted in the same UPLC droplets. Spectra of full mass scans were obtained using OEM-supplied software (Thermo Xcalibur TM mass spectrometry data system). For each pBpa fiber species, at least two independent data sets were obtained and analyzed to ensure reproducibility.
Disulfide-linked Protein-Using the QuikChange II sitedirected mutagenesis kit, the plasmid pSup(1-61)-GFP-Strep(II)-His 6 was doubly mutated to encode Sup(1-61)(Y32C,Y55C)-GFP. Nickel-nitrilotriacetic acid and Strep-Tactin affinity purification were performed with buffers containing 5 mM DTT, which was subsequently removed from protein solutions by PD MiniTrap G-25 desalting column (GE) equilibrated with buffer O (100 mM Tris, 1 mM EDTA, pH 7.0). Trans-4,5-dihydroxy-1,2-dithiane (oxidized DTT, Sigma) was added in 100-fold molar excess to catalyze disulfide bond formation at 22°C for 18 h and then removed from the sample with desalting column. The content of free thiol in the sample was assayed by reaction with 5,5Ј-dithiobis(2-nitrobenzoic acid) (Sigma) (22), and the reading was normalized against that of the same protein, equally diluted, but not treated with oxidized DTT. Two buried cysteine residues in GFP were tested to be unreactive. Intermolecular disulfide linking was negligible as judged by nonreducing SDS-PAGE. To form fibers, oxidized protein was nucleated overnight at room temperature with a catalytic amount of VK and VL particles purified from yeast. To rule out that fibers were exclusively assembled with the small amount of unoxidized protein left in the sample, we showed that much more than 10% of the protein monomers were incorporated into the fiber. Fibers were collected by centrifugation and dissolved in 8 M guanidine hydrochloride, and protein concentration was determined with a UV spectrophotometer (NanoDrop 2000C). Unseeded protein similarly processed and the prion seeds contributed negligibly to the sedimented material.
Plasmid Shuffle-Plasmid shuffle was performed with a derivative of the 74-D-694 background (MAT␣ ade1-14(UGA) his3-11 leu2-3 trp1 ura3-52) (23), whose SUP35 gene was deleted from the chromosome and was instead carried on a centromere-based plasmid YCp33-I-SupF, containing a URA3 marker. Another centromere-based plasmid, YCp111-Kan-MXs-I-SupF, expressing the full-length Sup35 (SupF) under a 1.2-kb native promoter, containing LEU2 and G418-selectable KanMX markers, was used in mutagenesis to obtain mutant plasmids expressing the Sup35(D3R) and Sup35(R28D) proteins, respectively. The mutant plasmids, as well as the wild type parent (used as control), were transformed into the 74-D-694 derivative harboring the VK and VL strain. Five transformants each were randomly selected on solid synthetic complete media lacking uracil and leucine (SCϪLeu,ϪUra), transferred to and grown on rich YPD plates, and replica-plated to solid synthetic media containing 5-fluoroorotic acid to counterselect cells that had lost the YCp33-I-SupF (URA3) plasmid. After growth at 30°C for 2 days, the yeast colonies, now expressing only mutant Sup35, were further replicated to YPD plates to observe changes in colony color (by comparison with the wild type control). The [PSI] strain type of each colony was determined as described above. Experiments for each [PSI] strain were performed with two independent 74-D-694⌬SUP35 [YCp33-I-SupF] isolates.

Results
Infectious Fibers-Experiments were performed with fibers derived from mutant Sup(1-61)-GFPs, containing the first 61 amino acid residues of Sup35 followed by the GFP, which enhanced solubility and facilitated experimental observation. Sup(1-61) was sufficient to propagate strain-specific infectivities of VH, VK, and VL in vitro (5,7). Its short length would simplify subsequent analysis. Two affinity purification tags-a Strep(II) tag followed by a six-histidine tag-were added to the C terminus of GFP to facilitate protein purification.
Point mutations were introduced as described. (i) Routine tandem mass analysis of protein requires the sample to be fragmented to specific pieces of smaller masses. This is commonly achieved by the application of proteases and peptidases. Unfortunately, the Sup35 fusion protein was prone to aggregation in buffers optimal for peptidase activities and thus became resistant to enzymatic digestion. Moreover, the simple amino acid composition of the Sup(1-61) prion domain offered few peptidase recognition sites. A chemical digestion approach was therefore adopted. The amino acid residues Tyr 13 , Tyr 29 , Tyr 35 , and Tyr 46 were replaced with methionine residues, which reacted with CNBr under acidic conditions to form homoserine lactone and freed the C-terminal portion of the polypeptide (24). Formic acid was used as the solvent, which readily dis-solved Sup35 aggregates. (ii) The sole methionine residue in the native Sup(1-61) sequence is the initiating methionine, which is followed by a serine residue. CNBr reacts with Met-Ser dipeptide to yield two products (25) whose different masses split signals and complicate tandem mass analysis. Ser 2 and Ser 4 were thus replaced with alanine residues. The resulting mutant protein containing four methionine and two alanine substitutions is designated as Sup(1-61)(ϩ4M2A)-GFP. (iii) Three derivatives of Sup(1-61)(ϩ4M2A)-GFP were initially prepared, each of which contained a single pBpa substitution for cross-linking. The substitution was at Asp 3 , Tyr 32 , and Tyr 52 .
We first tested whether the pBpa-substituted mutants could still propagate strain-specific conformations in vitro, as judged by their fibers' specific infectivity to yeast. Solutions of pBpacontaining proteins were nucleated with prion particles purified from yeast harboring VH, VK, and VL strains. Fibers appeared in all nucleated solutions but not unseeded controls. They were fragmented by sonic disruption and then introduced into [psi Ϫ ] yeast cells, which expressed the wild type Sup35, to assay for strain-specific infectivity. VK and VL infectivity was propagated by all pBpa substitutions ( Table 1); none of the fibers nucleated by VH particles were infectious. Subsequent cross-linking analysis was performed with fibers seeded by VK and VL.
Photo Cross-linking-To distinguish intra-and intermolecular linkage, 15 N-labeled proteins were prepared from E. coli grown in minimal medium containing [ 15 N]ammonium chloride as the sole nitrogen source. Labeled proteins were mixed with natural 14 N preparations in equal molar ratio and then nucleated with yeast prion particles. Fibers were irradiated with UV light to activate cross-linking reactions and then dissolved in a SDS-containing sample buffer for PAGE. Unirradiated fibers, dissolved in the sample buffer, formed a single band on the gel at a position expected for the protein monomer (ϳ36.5 kDa). UV-treated samples exhibited an extra up-shifted band with the mobility of a dimer. Additional weak bands corresponding to higher oligomeric species were also observed (Fig.  2, A and D). Gel slices with bands corresponding to protein monomer (designated band UV1) and dimer (band UV2) were treated with CNBr. Digested peptides were collected for mass spectrometry analysis.

TABLE 1 Infectious fibers of mutant Sup(1-61)(؉4M2A)-GFP pBpa substituted proteins are nucleated by prion particles of the VK and VL strain type
Fibers are co-transformed with the YCplac111 (LEU2) plasmid into the 5V-H19 yeast background without the prion to check for infectivity. 5  Digested peptides were analyzed by nano LC-nano electrospray ionization-MS/MS, which utilized CID to fragment peptides to yield sequences as well as sites of cross-linking (26). MS/MS peaks were assigned by MassMatrix (21), an automated search and match program, which scored the overall peak assignment and mass matches with a probability-based algorithm.
Clearly different mass spectra were obtained for inter-and intramolecularly linked peptides that were derived from fibers nucleated from equal molar mixture of 14 N-and 15 N-labeled protein. The mass spectrum of an intramolecular link had a characteristic doublet isotopic pattern, corresponding to 14 N and 15 N species (Fig. 2, B and E). Intermolecularly linked peptides manifested a quadruplet isotopic pattern, corresponding to pure 14 N species, pure 15 N species, peptide 1 composed of 14 N-linked with peptide 2 of 15 N, and vice versa (Fig. 2, C and F). Furthermore, as expected, only intramolecularly linked Sup35 fragments were detected in UV1 bands, and intermolecular linking was exclusively detected in UV2 samples. Extra intramolecular linking was detected in UV2 samples, but it was of nonspecific nature, appearing in UV-treated samples of both strain types as well as unseeded protein solutions (see below).
Tandem mass analysis of cross-linked Sup(1-61)(ϩ4M2A) (D3pBpa)-GFP fibers is summarized in Fig. 3A. The pBpa3 residue was intra-and intermolecularly linked to residue Arg 28 in fibers imparting the VL strain type (called "VL fibers" hereinafter; similarly, Sup(1-61)(ϩ4M2A)-GFP fibers that impart the VK strain characteristics are called "VK fibers") (Fig. 3, B and  C). Cross-linking was consistently detected in both 14 N and 15 N species. Experiments were repeated with independent sample preparations, and the data were reproducible. Ubiquitous intramolecular cross-linking between pBpa3 and Gln14 was detected in UV1 and UV2 bands of all specimens, including the sole UV1 band of unseeded protein control (Fig. 3D). The crosslinking was apparently derived from a non-prion conformation, although the cause of it, and why it also appeared in UV2 bands (but as intramolecular linking) remained to be investigated.
The Y32pBpa mutant reproducibly exhibited intramolecular cross-linking to Tyr 55 in both VK and VL fibers (Fig. 4). Similar to the D3pBpa mutant, nonspecific intramolecular cross-links between Y32pBpa and Arg 28 and between Y32pBpa and Asn 36 were detected in all specimens, including the unseeded protein control (not shown). The pBpa substitution at amino acid residue 52 did not yield any specific cross-linking. However, again, nonspecific intramolecular cross-linking with Tyr 45 was observed in every sample and control.
To further support observed specific cross-linking in VK and VL fibers, two more pBpa substitutions, R28pBpa and Y55pBpa, were prepared. Their ability to transmit VK and VL characteristics was confirmed by infection of yeast expressing the wild type Sup35 (Table 1). Although D3pBpa was cross-linked to Arg 28 in the VL fiber, no specific cross-linking between R28pBpa and Asp 3 was detected. The R28pBpa residue instead had nonspecific intramolecular cross-linking to the Gln 30 residue. The Y55pBpa mutant had specific intra-and intermolecular cross-linking to Tyr 32 in the VL fiber and also intermolecular cross-linking to Tyr 32 in the VK fiber (Fig. 5). Again, additional ubiquitous intramolecular cross-linking was de-tected between Y55pBpa and Tyr 45 in all samples. All results were reproduced with independent preparations.
Constraining the Chain Folds-The cross-links identified above were used to constrain strain-specific chain folding patterns of Sup(1-61)(ϩ4M2A). Tandem mass analysis suggested that Tyr 32 and Tyr 55 residues were close to each other in VK and VL fibers. In the VL fiber, Y32pBpa was intramolecularly linked to Tyr 55 , and Y55pBpa was similarly linked to Tyr 32 . The intermolecular Y55pBpa-Tyr 32 cross-link would further put the (Tyr 32 and Tyr 55 ) pair in each adjacent Sup35 subunit in close proximity, supporting a general parallel in-register Sup35 alignment, most clearly demonstrated by recent solid state NMR measurements on samples consisting mainly of strong [PSI] strains (8). (VL and VK are weak strains.) An intramolecular Y32pBpa-Tyr 55 link and an intermolecular Y55pBpa-Tyr 32 link were detected in the VK fiber. In light of a parallel in-register Sup35 arrangement mentioned above, the combined results indicated close proximity of residues 32 and 55 in the VK fiber as well. The absence of Y55pBpa-Tyr 32 intra-molecular cross-linking in the VK fiber probably reflected local asymmetry-for efficient cross-linking to happen, the Y55pBpa residue and a reacting methylene group would have to be in a  favorable stereogeometrical configuration, which occurred in the VL fiber but apparently did not in the VK structure.
We further tested whether Sup(1-61)-GFP, with residues 32 and 55 cross-linked, could be incorporated into [PSI] fibers in vitro (27). Sup(1-61)(Y32C,Y55C)-GFP was treated with oxidized DTT to obtain disulfide-linked monomers, which was then nucleated with prion particles purified from yeast harboring the VK or the VL strain. After overnight incubation at room temperature, numerous fibers were observed in seeded samples of both strain types; unseeded controls remained clear (Fig. 6). We further determined that 90% of the free cysteines were oxidized in the protein solution used for nucleation, and ϳ46 and 65% of the total protein was incorporated into the fiber after seeding with VK and VL particles, respectively. It was thus concluded that disulfide-linked monomers could indeed join a growing fiber and continue.
Molecular ions revealed VL-specific intra-and intermolecular cross-linking between D3pBpa and Arg 28 , but no Asp 3 -R28pBpa cross-linking was detected. Similarly, this could be due to asymmetric stereogeometry in the two pBpa substitution mutants; in addition, if everything else is equal, the side chain of an interacting arginine residue was clearly a bit longer than that of an aspartate. Mutating the Asp 3 residue to arginine (i.e. Arg 3 and Arg 28 ) in the wild type Sup35 destabilized the VL strain in yeast, causing misnucleation to generate VK (10 [VK]/10); however, mutating residue 28 to a smaller aspartate (Asp 3 and Asp 28 ) in yeast did not cause VL to change (10 [VL]/10). Both mutants supported VK propagation (10 [VK]/10; 10 [VK]/10). The involvement of Sup35 N-terminal residues in the structure of VL, but less so in VK, was also consistent with our previous genetic finding that deleting residues 5-10 from the wild type Sup35 protein in vivo cured the VL strain but still allowed the cellular propagation of VK (7). Based on the analysis above, constrained chain folds of the VK and VL fiber are illustrated (Fig. 7).

Discussion
In this study, cross-linking was obtained from multiple Sup(1-61)(ϩ4M2A)-GFP derivatives, each of which had a single pBpa substitution at a different amino acid position. Different local structural perturbations might occur to result in crosslinking that did not correspond to amino acid contacts in the parental Sup(1-61)(ϩ4M2A)-GFP fiber and cause inconsistencies when linkages were taken together as constrains. The likelihood of such artifact was very low when cross-linking between a pair of amino acid residues was consistently detected, no matter which residue was replaced with pBpa, as in the case of residues 32 and 55. Otherwise, additional supporting evidence was sought, as in the case of residues 3 and 28 in the VL fiber.
The similarity between fibers of Sup(1-61)(ϩ4M2A)-GFP and of wild type Sup35 requires discussion. In the present work, all mutant proteins were nucleated by prion particles purified from yeast, and the resulting fibers were able to faithfully induce the same [PSI] strains in yeast expressing the wild type Sup35. This indicated that structural elements responsible for interacting with soluble Sup35 and converting it into strainspecific conformations were faithfully retained in all mutant fibers. Furthermore, no new element that caused mis-seeding was created, either by exposing a protected element or by de novo generation. The strain-specific folding topology of Sup(1-61)(ϩ4M2A)-GFP constrained by the cross-linkages thus should not conflict with that of the wild type Sup35 N-terminal domain.
The ability to distinguish strain-specific infectivity was arguably the most important aspect in the design of the present work. It is known that single amino acid changes in an amyloidogenic protein could cause the aggregates to adopt a completely different structure, and pBpa substitutions are presumably not exceptional. For example, the W8 strain of [PSI] was observed to transmute to VK when propagated by Sup35 with a single proline substitution at amino acid positions 5, 6, 7, or 43 (5).
With pBpa substitution at more amino acid positions, additional proximity information could be obtained to further constrain chain topologies of different [PSI] strains. We noticed that UV-activated pBpa cross-linking was inefficient such that less than 10% of irradiated polypeptides were actually crosslinked. Furthermore, with only a few CNBr-digested fragments from the proteins, we were operating far below the capacity of modern mass spectrometers and the analytical power of their software. It might be possible to introduce pBpa residues at several amino acid positions in a single protein and efficiently characterize the whole array of sparsely cross-linked products. This development would further mitigate the concern that each single pBpa mutant changes the fiber structure in a different way.
The experiments described here can be applied to other prions, such as Prp SC of mammals (28,29), as well as other pathological amyloids for which fine strain typing methods can be developed (30 -32). Tertiary contacts obtained by cross-linking experiments can be combined with additional secondary structural information attained by NMR (8,33) or hydrogen exchange mass spectrometry (34), to better define chain folding patterns. The knowledge would be useful in providing mechanistic insights in prion strain mutation and in validating future higher resolution fiber structures.