Heterologous prion-forming proteins interact to cross-seed aggregation in Saccharomyces cerevisiae

The early stages of protein misfolding remain incompletely understood, as most mammalian proteinopathies are only detected after irreversible protein aggregates have formed. Cross-seeding, where one aggregated protein templates the misfolding of a heterologous protein, is one mechanism proposed to stimulate protein aggregation and facilitate disease pathogenesis. Here, we demonstrate the existence of cross-seeding as a crucial step in the formation of the yeast prion [PSI +], formed by the translation termination factor Sup35. We provide evidence for the genetic and physical interaction of the prion protein Rnq1 with Sup35 as a predominant mechanism leading to self-propagating Sup35 aggregation. We identify interacting sites within Rnq1 and Sup35 and determine the effects of breaking and restoring a crucial interaction. Altogether, our results demonstrate that single-residue disruption can drastically reduce the effects of cross-seeding, a finding that has important implications for human protein misfolding disorders.

In contrast to the titration model, the cross-seeding model suggests that Sup35 and Rnq1 physically interact during [PSI + ] formation (Fig. 1B). In this manner, the misfolded and aggregated Rnq1 in the [RNQ + ] state induces Sup35 to misfold via bending the monomer out of its native conformation or templating a spontaneously misfolded monomer into a structure that is conducive for prion propagation. There is some evidence for the seeding model. Fluorescence microscopy demonstrated that newly-forming Sup35-YFP aggregates co-localized with existing Rnq1-CFP aggregates 19 . Promoting the interaction between Rnq1 and Sup35 via a Sup35-Rnq1 fusion protein also enhanced [PSI + ] formation, whereas fusion of Sup35 to other aggregation-prone molecules did not 20 . In addition, preformed aggregates of the Rnq1 prion-forming domain (PFD) have been shown to slightly increase the rate of Sup35 polymerization 19,[21][22][23] , and Rnq1 weakly co-immunoprecipitates with Sup35 in vivo 22 .
To gain greater insight into the aggregation of heterologous proteins, we set out to investigate the nature of the Sup35-Rnq1 interaction. Using a variety of techniques in vivo and in vitro, we have demonstrated a binding interaction that supports the veracity of the seeding model. We probed the functional consequences of disrupting this binding, and for the first time in any system, identified sites of interaction that are responsible for cross-seeding and facilitating the spread of aggregation between two heterologous amyloidogenic proteins.

Methods
Yeast strains, plasmids, cultures, and transformations. Yeast strains were cultured using standard techniques 24 . Yeast transformation with plasmid DNA was performed by the PEG/LiOAC method 24 . All yeast strains used were 74-D694. Complete lists of strains and plasmids are available in Tables S1 and S2, respectively. Protein purification. Sup35 constructs and Rnq1 were tagged with a C-terminal polyhistidine tail and purified on nickel affinity resin. Briefly, BL21 (DE3) bacteria were transformed with the plasmid of interest and grown at 37 °C in LB+ Amp media. At OD 600 = 0.6, protein expression was induced with 1 mM IPTG and cultures were grown at 25 °C for 6 hours. Cells were pelleted, washed, and frozen at −80 °C until use. Cells were thawed at 4 °C for 2hr, shaken in lysis buffer (Sup35: 20 mM Tris, 0.5 M NaCl, 5 mM imidazole, 2 M urea, 0.1% IGEPAL; Rnq1: 20 mM Tris, 8 M Urea, pH 8) at 4 °C for 30 minutes, and then sonicated 5 × 10 seconds at high power. Lysates were cleared by centrifugation at 20,000 × g for 30 minutes, and the supernatant was incubated with a nickel resin for 30 minutes at 4 °C. Resin columns and bound proteins were washed with washing buffer (Sup35: 20 mM Tris-HCl, 0.5 M NaCl, 20 mM imidazole, 2 M urea; Rnq1: 20 mM Tris, 8 M Urea, pH 6.3-5.9) and eluted with wash buffer + 400 mM imidazole for Sup35 or wash buffer at pH 4.5 for Rnq1. Elution fractions were pooled, concentrated, and methanol-precipitated prior to use. Rnq1 purification was performed as described previously 25 .
Thioflavin T kinetics. Purified Sup35NM and Rnq1 were resuspended to 300 µM in 7 M guanidine hydrochloride. Rnq1 fibers were formed at 4 °C, 25 °C, and 37 °C and then pooled. Fiber formation and kinetic assays were performed under agitation with acid-washed glass beads (Sigma) as previously described 26  Boiled gels. Yeast strains were cultured overnight in selective media. Cells were lysed in buffer (25 mM Tris-HCl pH 7.5, 50 mM KCl, 10 mM MgCl 2 , 1 mM EDTA, 10% glycerol, protease inhibitors) by mechanical disruption with acid-washed glass beads (Sigma). Lysates were normalized by total protein and mixed with SDS-PAGE sample buffer (200 mM Tris-HCl pH 6.8, 4% SDS, 0.4% bromophenol blue, 40% glycerol). Unboiled samples were loaded on a 12% polyacrylamide gel and run under 140 V of current until the dye front migrated halfway through the resolving gel. The current was stopped and the gel and glass plates were sealed in plastic sheets prior to boiling upright for 15 minutes in a 95 °C water bath. The gels were removed from the plastic and reinserted into the PAGE apparatus, where voltage was re-applied (140 V) until the dye front migrated to the bottom of the gel. This procedure was followed by standard western blotting with Sup35 and Rnq1 antibodies (Supplementary Table S3).

Prion manipulation and protein biochemistry.
[PSI + ] induction and SDD-AGE experiments were performed as previously described 8,27 . [PSI + ] induction frequency was calculated by counting the number of white or white-sectored colonies on each plate and dividing by the total number of colonies on the plate. Cytoduction was performed as previously described 28, 29 . Screen for Sup35 mutations. The screen for suppressor mutations was performed using SP5, a chimeric Sup35 protein that contains 5 repeat domains from the prion protein PrP in place of the endogenous repeat domains of Sup35 30,31 . SP5 was utilized in the screen because it is a robust inducer of [PSI + ] 30 . To introduce mutations, six error-prone PCR reactions were performed in the presence of 25 µM MnCl 2 using Taq polymerase and 35 cycles to amplify the NM domains of SP5. PCR reactions were combined to create three separate libraries, followed by SpeI/BclI digestion and ligation into p415TEF-SP5. 74-D694 sup35Δ rnq1Δ yeast cells propagating m.d. high [RNQ + ] and harboring episomal copies of SUP35 and rnq1-Q298R were transformed with each library, with selection on SD-His-Leu. Transformation plates were replica plated onto SD-His-Leu + 5-FOA to select for cells having replacement of the URA3-marked SUP35 plasmid. These colonies were replica plated onto YPD and YPD + 3 mM GdnHCl to identify white [PSI + ] colonies that were also curable. More than ~3,000 colonies were assessed, of which 193 were pink (putative [PSI + ]) and 38 of those were curable (suggestiong the prion). After plasmid recovery and sequencing, we identified 12 mutant SP5 alleles.
Crosslinking. PCR was performed on pAED4-SCNM-his7 to introduce the G7C mutation, using forward primer 5′GCGCATATGATGTCGGATTCAAACCAATGTAACAATCAGCAAAAC (mutated bases underlined) and reverse primer 5′CATCGATGAATTCTTAATGGTGG to amplify the His-tagged NM domain and introduce the mutation (underlined bases). The PCR product and pAED4 vector were digested with NdeI and EcoRI, ligated, and transformed into BL21 (DE3) cells for protein expression and purification as described above. Following methanol precipitation, Sup35NM-G7C and Rnq1 were dissolved to 0.1 mM in conjugation buffer (PBS, 2 mM EDTA, pH 7.4). The heterobifunctional crosslinker SMPB (Thermo #22416) was attached to Rnq1 by incubating 1.3 mg protein with a final concentration of 1 mM SMPB for 30 minutes at room temperature. Disulfide bonds in Sup35NM-G7C were reduced with TCEP (Sigma). Excess SMPB and TCEP were removed by passing proteins through Zeba desalting columns (Thermo #89890). Equimolar amounts of Sup35-G7C and Rnq1-SMPB were combined and incubated at room temperature for 30 minutes, followed by SDD-AGE and western blotting. Data availability. All data generated or analyzed during this study are included in this article or are available from the corresponding author.

Results
Rnq1 aggregates seed the polymerization of Sup35. We set out to specifically test the cross-seeding model, which posits a physical interaction between Rnq1 and Sup35. We began examining the interaction between Rnq1 and Sup35 by assessing the kinetics of in vitro aggregation using Thioflavin T (ThT) 32 . ThT binds specifically to amyloid fibers, and an increase in the fluorescent emission signal indicates formation of amyloid. When incubated with agitation at room temperature, the prion forming domain of Sup35, denoted Sup35NM, will form amyloid after a reproducible lag phase 25,31 . This lag phase can be reduced or eliminated if Sup35NM aggregation is "seeded" by preformed Sup35NM fibers that provide a template for amyloid formation 25 . Previous work has demonstrated that the Rnq1 prion-forming domain (PFD) can increase Sup35NM polymerization, albeit inefficiently 19,21,22 . As "self-seeding" (i.e. propagating [PSI + ]) in vivo is more efficient than cross-seeding (i.e. inducing [PSI + ] in [RNQ + ] cells), this is somewhat expected. Indeed, self-seeding of Sup35NM in vitro with pre-formed Sup35NM fibers is much more efficient than heterologous seeding 21 . However, we now know that regions outside of the Rnq1 PFD are important for maintenance of the [RNQ + ] prion, and that the distinct conformations of Rnq1 aggregates propagated by different [RNQ + ] prion strains, or variants, dramatically affect the formation of [PSI + ] in vivo 7, 21-23, 28, 33-36 . Thus, we utilized full-length Rnq1 and aimed to create a mixed population of seed conformers that could emulate the aggregation states that may be present in vivo.
We purified recombinant Sup35NM and full-length Rnq1, and formed Rnq1 fibers in vitro to serve as "seeds" for the polymerization reaction. As we previously demonstrated that temperature can affect the conformation of Rnq1 aggregates, such as changing the amount of fiber branching and bundling 26 , we formed Rnq1 fibers at a range of temperatures and combined them to create a broad pool of potential seed structures. Thermal stability experiments confirmed the existence of Rnq1 aggregates (not shown). We then performed ThT aggregation assays with monomeric Sup35NM in both unseeded and seeded conditions ( Fig. 2A). We found that the addition of monomeric Rnq1 did not have an effect on the lag phase of Sup35 polymerization relative to the unseeded reaction. However, the presence of pooled Rnq1 seeds at 4% (v/v) concentration reduced the lag phase of Sup35NM aggregation as predicted ( Fig. 2A). Thus, the initiation of Sup35 aggregation in vitro is enhanced by the presence of Rnq1 amyloid. As this experiment used pure, recombinant protein and Rnq1 seeds were sufficient to enhance Sup35 polymerization, these results support the cross-seeding model (Fig. 1).

Sup35 stably binds Rnq1 ex vivo.
We hypothesized that the conformation of Rnq1 in [RNQ + ] cell lysates may differ from that readily obtained in the in vitro amyloid formation reaction. Therefore, we sought to test the binding of Sup35 to a biologically relevant conformation of Rnq1 isolated from [RNQ + ] yeast cell lysates. To do so, we bound recombinant, His-tagged Sup35 containing either the NM prion-forming domain or the non-prion C-terminal domain (CTD) to nickel affinity resin. We then prepared cell lysates from [RNQ + ][psi − ] yeast and incubated the lysate with the nickel resin to allow for binding between Rnq1 and immobilized Sup35. We hypothesized that the bound Sup35NM would "trap" untagged Rnq1 that would ordinarily flow through the column. We washed the column with increasing amounts of imidazole to displace the bound proteins, and collected unbound, wash, and elution (bound) fractions (Fig. 2B). We found that the Sup35 CTD-7xHis experienced weak or non-specific binding to Rnq1, as the detectable Rnq1 appeared predominantly in the unbound and intermediate wash fractions. However, there was a strong interaction between Rnq1 and Sup35 when the Sup35NM-7xHis protein was used as bait, as a large proportion of Rnq1 appeared in the bound fractions. Rnq1 was still present in the unbound fraction, yet not all Rnq1 is expected to be aggregated or Sup35-binding-competent in [RNQ + ] cells. Importantly, when no Sup35NM-7xHis was present on the beads, all Rnq1 protein was present in the unbound fraction, indicating the lack of non-specific binding of Rnq1 to the column. Thus, the strong binding between ex vivo Rnq1 to the Sup35 prion-forming domain appears to be specific and stable.
The rnq1-Q298R mutation causes a [PSI + ] induction defect. Upon demonstrating the physical interaction between Sup35 and Rnq1, we sought to identify the specific residues in each protein that were important to binding. To do so, we utilized a mutation in Rnq1, Q298R, that had been previously identified by our lab and demonstrated to impart a [PSI + ] induction defect (Fig. 3A) 37 . This work was performed with the "multi-dot high" variant of [RNQ + ] that is a strong inducer of [PSI + ]. The mutant was previously shown to be capable of propagating [RNQ + ] 37 , and our additional tests of thermal stability (Fig. 3B, Supplementary Figure S1) and mitotic stability (Fig. 3C) showed that the Rnq1-Q298R mutant maintains Rnq1 aggregates that are biochemically indistinguishable from WT Rnq1.
Though they propagate [RNQ + ] efficiently, cells harboring rnq1-Q298R were shown to have a [PSI + ] induction defect; that is, they become [PSI + ] at a much lower frequency than RNQ1 cells 37 . To verify this defect, we utilized a colorimetric assay to visualize the frequency of [PSI + ] induction in vivo. The colorimetric readout relies upon a well-described phenotypic [PSI + ] reporter, using an ade1-14 genetic background, which impacts adenine production in a manner that depends on nonsense suppression. Since Sup35 is a translation termination factor, its aggregation to form [PSI + ] increases nonsense suppression. The ade1-14 allele contains a premature stop codon in the ADE1 gene 38 . Yeast containing ade1-14 cannot make adenine and colonies grown on media containing adenine appear red due to the buildup of a pigmented metabolic byproduct. The presence of [PSI + ] allows cells to read through the ade1-14 premature stop codon and synthesize adenine. As such, [PSI + ] cells can grow on synthetic media lacking adenine (SD-Ade) and appear white or pink on rich media plates (YPD), whereas [psi − ] colonies are red.
Utilizing this colorimetric test to monitor [PSI + ] induction, we overexpressed Sup35 in [RNQ + ][psi − ] yeast expressing either Rnq1 or Rnq1-Q298R. To build upon our previous work, we examined two [RNQ + ] prion variants -multi-dot high (m.d. high) and single-dot high (s.d. high) -in order to assess the specificity of the mutant, which was previously characterized in only m.d. high. Cells were spread on rich media plates to allow for development of red or white colonies. The presence of productive Rnq1-Sup35 interactions would induce the formation of white [PSI + ] colonies. In agreement with our previous results, strains with WT Rnq1 aggregates induced [PSI + ] at a high level in both variants, whereas yeast with Rnq1-Q298R aggregates exhibited a significant decrease in [PSI + ] formation in the m.d. high strain (Fig. 3D). In combination with our observation that there is no detectable effect of Rnq1-Q298R on aggregate structure, this suggests that Q298 is an important site of interaction between Rnq1 and Sup35, and that mutating the glutamine residue leads to decreased or unproductive interactions with Sup35. Interestingly, rnq1-Q298R had no effect upon [PSI + ] induction in conjunction with the s.d. high [RNQ + ] variant (Fig. 3D), suggesting that the residues involved in cross-seeding differ between different prion strains. As our lab has demonstrated that distinct regions of Rnq1 are important for the propagation of different [RNQ + ] variants 35 , it is not surprising that there are also differences in the regions important for Sup35 interaction.
We  (Fig. 3E). Thus, the [PSI + ] induction defect in rnq1-Q298R is a consequence of the genetic mutation as opposed to a templating defect that is transmitted to WT Rnq1.

Suppressor mutations rescue the [PSI + ]-induction defect. After confirming the effect of rnq1-Q298R
on [PSI + ] induction, we sought to identify compensatory mutations in Sup35 that might represent putative binding sites between Rnq1 and Sup35. We performed a second site suppressor screen to identify mutations within the Sup35 prion forming domain that could restore the interaction with Rnq1-Q298R and rescue the [PSI + ] induction defect. Briefly, we transformed strains containing rnq1-Q298R with a pool of mutagenized SP5, a chimeric SUP35 construct with high [PSI + ] induction rates that allowed us to resolve fine differences between the rescuing abilities of putative suppressor mutations 30 . We then assessed the ability of yeast cells to form [PSI + ] (see Methods for screen details). Our secondary tests verified that [PSI + ] was formed by showing that the nonsense suppression phenotypes were reversible on plates containing 3 mM GdnHCl, which cures yeast prions 39 (not shown). To assess [PSI + ] formation, we spotted the screen candidates onto SD-Ade to test their growth (Fig. 4A, Supplementary  Figure S2A). The most promising mutants were those that grew strongly on SD-Ade relative to WT SP5 in the rnq1-Q298R genetic background. In all, we isolated 12 sup35 mutants containing a total of 16 nonsynonymous mutations within the "NM" prion-forming domain (Fig. 4B).

Sup35-N5Y robustly rescues the [PSI + ]-induction defect.
After cloning the 16 individual candidate mutations from SP5 into sup35, we then tested their ability to suppress the [PSI + ] induction defect of rnq1-Q298R. We again utilized [PSI + ] induction experiments to evaluate the sup35 screen candidates (Fig. 5A and not shown). We overexpressed either WT or mutant Sup35 in [RNQ + ] cells expressing either Rnq1 or Rnq1-Q298R from integrated alleles covering endogenous RNQ1. Both strains were sup35Δ and expressed the C-terminal domain of Sup35 to allow for shuffling of the mutant constructs. Importantly, [PSI + ] formation was never observed in [rnq − ] cells, and a high degree of induction occurred in WT [RNQ + ] cells. Moreover, the [PSI + ] induction defect was evident in [RNQ + ] rnq1-Q298R SUP35 cells. However, three Sup35 mutants rescued the defect and induced [PSI + ] to near-WT levels: N5Y, Q6R, and G116V (Fig. 5A, Supplementary Figure S2B). The G87C mutation, which was isolated in combination with G116V (Fig. 4A), did not restore [PSI + ] induction (Fig. 5A). Interestingly, despite M domain mutations also being identified and tested, the three rescuing mutations all fall within the N-terminal region of Sup35, which is the more aggregation-prone domain thought to be the most crucial for prion formation 17,40 .
There were two adjacent single mutants isolated during our screen, sup35-N5Y and sup35-Q6R. Since sp5-N5Y had the strongest growth on SD-Ade ( Fig. 4A and Supplementary Figure S2B), we decided to focus on the potential interaction between Sup35-N5 and Rnq1-Q298. First, we examined Sup35 aggregation biochemically via "boiled gel" assays 41 . In these experiments, unboiled cell lysates are loaded on SDS-PAGE gels. Native (non-denatured) aggregated material is too large to enter the resolving gel and becomes trapped in the wells and the stacking gel. Upon boiling the gel midway through electrophoresis, aggregated material breaks down and enters the resolving gel. Following western blots, two bands are visible per lane: a lower band of soluble material, and a higher band of insoluble material that was delayed in entering the gel (see Methods). We performed boiled gel experiments with cell lysates generated from overnight cultures of [RNQ + ][psi − ] cells containing either WT RNQ1 or rnq1-Q298R and overexpressing WT or sup35-N5Y (Fig. 5B, Supplementary Figure S3). We also examined a [rnq − ] control with WT RNQ1 and SUP35. As expected, the [rnq − ] control showed no Sup35 aggregation, since [PSI + ] rarely forms in the absence of [RNQ + ]. In WT cells, nearly all of the Sup35 was visualized  These mutations do not increase [PSI + ] formation in a RNQ1 genetic background. Horizontal axis labels denote SUP35 genetic status. The [rnq − ] control cells express RNQ1 and SUP35. More than 13,000 colonies were assessed over five biological replicates. Error bars represent mean ± s.e.m. The "*" symbols represent a significant difference between [PSI + ] induction with WT Sup35 in rnq1-Q298R versus induction with the indicated sup35 mutants, p < 0.005. (B) Boiled gel assays allow separate migration of monomeric and aggregated materials, confirming that Sup35 aggregation is reduced in the SUP35 rnq1-Q298R background relative to SUP35 RNQ1 cells. Sup35 aggregation is restored in sup35-N5Y rnq1-Q298R cells. We attribute the slightly aberrant monomer bands as an artifact of the experimental procedure, as samples were unboiled (and likely non-denatured) as per the protocol, and Sup35 may shed unevenly from aggregates in addition to the soluble pool. Western blot is a representative image from three independent experiments. in the aggregated band, indicating prolific [PSI + ] formation. Conversely, little aggregated Sup35 was detected in cells expressing rnq1-Q298R in conjunction with SUP35, mirroring the [PSI + ] induction defect. Finally, in rnq1-Q298R cells harboring the sup35-N5Y mutation, a dramatic increase in Sup35 aggregation was observed. The lack of complete restoration of Sup35 aggregation may be due to incomplete rescue by sup35-N5Y, or the shorter yeast culture period for this protocol (16 hours) versus [PSI + ] induction experiments (4 days), potentially indicating that Rnq1-Q298R propagates a sufficient yet non-optimal template for seeding [PSI + ] formation relative to WT Rnq1. Collectively, our results demonstrate that the Sup35 N5Y mutation rescues Sup35 aggregation and [PSI + ] formation when [RNQ + ] is propagated by Rnq1-Q298R.
The Sup35 N-terminus binds to Rnq1 in vitro. After demonstrating that Sup35 and Rnq1 physically interact and that Rnq1-Q298 and Sup35-N5 may be an important site of contact, we sought to verify the interaction by crosslinking in vitro. We decided to use a sulfhydryl-reactive crosslinker to take advantage of the lack of cysteine residues in Sup35. We chose to introduce a cysteine mutation close to the putative binding site so as not to disturb the residue(s) identified as being important in the interaction (N5 and the adjacent screen candidate Q6). We mutated the Sup35 G7 residue to cysteine to direct crosslinking to the N-terminal region of the protein. The G7C mutation of Sup35 has previously been shown to form and propagate the [PSI + ] state in a manner identical to WT Sup35 42 . After treating recombinant Rnq1 with a heterobifunctional crosslinker, SMPB, the two proteins were combined and analyzed via semi-denaturing detergent agarose gel electrophoresis (SDD-AGE), a protocol that allows for the visualization of large SDS-resistant aggregated species (Fig. 6). We utilized SDD-AGE so that we could visualize the full spectrum of outcomes of the Sup35-Rnq1 interaction; that is, the induction of additional aggregation. We also confirmed the presence of crosslinked heterodimers via SDS-PAGE (Supplementary Figure S4).
In all crosslinked reactions, a very high molecular-weight product was visible in the wells of the gel. This product reacted with both Sup35 and Rnq1 antibodies ( Fig. 6 and Supplementary Figure S4), and was not present in control lanes containing Rnq1 alone, Sup35NM-G7C alone, or the two proteins combined without SMPB. Based on the spacer arm length of SMPB (11.5 Å), this indicates that Rnq1 and Sup35 physically bind within the first 15 amino acids of Sup35, and agrees with the genetic interaction found in our second site suppressor screen.

Discussion
The early events of protein misfolding and aggregation in human diseases have been difficult to elucidate using mammalian systems. A longstanding question has been whether interactions between heterologous proteins might contribute to the onset and progression of disease. Taking advantage of the tractable yeast prion model, we have utilized biochemical and genetic techniques to demonstrate the physical interaction of heterologous prion-forming proteins in vitro and in vivo. Utilizing a combination of genetic techniques and crosslinking, we identified specific amino acid residues that are important for the binding of Sup35 and Rnq1; namely, the N5 Figure 6. The N terminal region of Sup35 crosslinks to Rnq1. Site-directed crosslinking of Sup35-G7C to Rnq1 created high-molecular weight aggregates as visualized by SDD-AGE. Rnq1 alone, Sup35-G7C alone, and both proteins together without SMPB did not create large aggregates. The " + *" notation in the fifth lane indicates Sup35 monomer following treatment with TCEP. Western blot is a representative image from three independent experiments. In the lower blot, Rnq1 loading was confirmed by SDS-PAGE. Vertical white bars separate non-adjacent lanes of the same blot. The dashed line separates adjacent lanes of the same blot under differing film exposures for image clarity (full blots in Supplementary Figure S4). There is excess Rnq1 in lanes 1 and 3, samples prepared without crosslinker, to adjust for protein that is lost during sample desalting following attachment of the crosslinker (see Methods). Three trials of crosslinking were included in each experiment and the crosslinking experiment was repeated with different batches of purified proteins three times. residue of Sup35 and the Q298 residue of Rnq1. This indicates that the prion-forming domains of these proteins interact to transmit the prion state between distinct proteins. Many disease-associated proteins also have aggregation-prone regions that could facilitate promiscuous interactions with other proteins to facilitate their own aggregation 43,44 .
Importantly, although our work strongly supports the "seeding" model of Rnq1-Sup35 interaction, it does not disprove the inhibitor titration model (Fig. 1). It may be that a combination of both models will best explain the de novo formation of [PSI + ] in vivo. For instance, an inhibitor molecule might be titrated away from Sup35 to allow for the physical interaction between Rnq1 aggregates and Sup35 that our data show is necessary for [PSI + ] formation. The inhibitor itself, having affinity for both Sup35 and Rnq1, may even facilitate the Sup35-Rnq1 binding while it is being titrated by Rnq1. Molecular chaperones have been suggested as potential inhibitor molecules 18 . In support of this concept, previous research by our lab and others has demonstrated that several classes of chaperones can have opposing effects upon prion formation and propagation, sometimes in a conformation-dependent manner [45][46][47] . It is possible that the Rnq1 monomer itself may have an inhibitory effect upon Sup35 aggregation. Future research may implicate one or more of these proteins as an inhibitor of [PSI + ] formation. However, our results suggest that cross-seeding is sufficient for Sup35 polymerization in vitro and is clearly necessary for [PSI + ] formation in vivo.
In addition to showing the importance of a physical interaction for [PSI + ] formation, our data demonstrate that the extreme N-terminus of Sup35 plays a critical role in facilitating this interaction. As the Sup35 N-terminal domain lacks a defined structure, it may be that this region of the protein is most freely available to interactions with other proteins. The charged M domain might then act as a stabilizer for whichever conformation is adopted by the disordered N domain. However, beyond the N-terminus, we uncovered other mutations that could partially rescue the [PSI + ] induction defect associated with Rnq1-Q298R (Fig. 4). This suggests that there are multiple residues where Sup35 and Rnq1 may interact en route to Sup35 aggregation. Though each of the three investigated Sup35 mutations was alone sufficient to restore interaction with Rnq1-Q298R, there may be multiple points of contact necessary before Sup35 assumes its prion conformation. By directing an interaction between Sup35 and Rnq1 at residues of interest (via a dimerizing tag or crosslinking agent), it may be possible to drive very high levels of interaction and subsequent aggregation in vivo. The total number of necessary interactions and their order of occurrence (if any) remain avenues of further investigation.
It is also likely that heterologous interactions of amyloidogenic proteins depends on the structure of the aggregate, or prion strain. Different prion strains adopt different tertiary protein conformations 48 . Various data suggest that these conformational differences result from beta sheets being formed either from distinct amino acid residues or from alternative packing of the same residues, thereby presenting different points of contact 48,49 . Our previous work indicated that the regions that form the Rnq1 aggregate β-sheets of the different [RNQ + ] variants may be distinct 35 . As we showed that these differences affected aggregate-chaperone interactions 47 , so might the intermolecular contacts with Sup35 be different. Indeed, our work with the multi-dot high variant of [RNQ + ] presented here identified Q298 as an important binding site with Sup35, but Q298R had no apparent effect on Sup35-Rnq1 interaction in the "single-dot high" prion strain (Fig. 3D). This suggests that the Rnq1 aggregates propagated in single-dot high [RNQ + ] cells might have a different set of binding residues with Sup35. While prion strains likely have distinct contact points, there may also be a cohort of interacting sites that remain consistent between prion strains. Such putative differences in contact of Rnq1 and Sup35 between the different [RNQ + ] variants might help explain their drastically different abilities to induce the formation of [PSI + ].
Furthermore, the [RNQ + ] strain present in a cell influences the distribution of [PSI + ] strains that form 22,27 . This phenomenon may also be explained by the existence of different interaction sites between the two prion-forming proteins. In our experiments, we observed our Sup35 mutants forming various strains of [PSI + ] under both WT and rnq1-Q298R genetic backgrounds (not shown). Thus, this suggests that the Sup35-N5 and Rnq1-Q298 residues comprise a key binding site that supports normal [PSI + ] formation with multi-dot high [RNQ + ]. Other sites of interaction may influence strain specificity. This would provide mechanistic detail for how distinct amyloid conformers can influence the types of aggregate structures formed by heterologous proteins, which might impact the progression of human diseases.
Indeed, many human pathologies involve the aggregation of misfolded, beta sheet-rich proteins 50 . Although these incurable diseases have profound effects upon patients and a multi-billion dollar impact on the healthcare system 51 , little is known about the factors that cause the de novo protein misfolding. It is possible that a seeding or templating mechanism contributes to the misfolding and subsequent aggregation of human proteins. In support of this concept, the Parkinson's disease-related protein α-synuclein has been shown to cross-seed the fibrillization of tau, which is implicated in Alzheimer's disease pathology 4 . Further, polymerization of amyloid-beta (Aβ) was recently shown to be promoted by the presence of misfolded islet amyloid polypeptide (IAPP), which aggregates in the islet cells of patients with type 2 diabetes 5 . This is intriguing in light of epidemiological studies that have identified type 2 diabetes as a risk factor for Alzheimer's disease 52 . While our work demonstrates that multiple binding sites may be involved in cross-seeding (Fig. 4), it also suggests that blocking one crucial site might be sufficient to alter the outcomes of the interaction. Further studies in yeast might provide a foundation for developing more focused therapeutic approaches to target these crucial interaction sites.