Probing the origin of prion protein misfolding via reconstruction of ancestral proteins

Abstract

Prion diseases are fatal neurodegenerative diseases caused by pathogenic misfolding of the prion protein, PrP. They are transmissible between hosts, and sometimes between different species, as with transmission of bovine spongiform encephalopathy to humans. Although PrP is found in a wide range of vertebrates, prion diseases are seen only in certain mammals, suggesting that infectious misfolding was a recent evolutionary development. To explore when PrP acquired the ability to misfold infectiously, we reconstructed the sequences of ancestral versions of PrP from the last common primate, primate‐rodent, artiodactyl, placental, bird, and amniote. Recombinant ancestral PrPs were then tested for their ability to form β‐sheet aggregates, either spontaneously or when seeded with infectious prion strains from human, cervid, or rodent species. The ability to aggregate developed after the oldest ancestor (last common amniote), and aggregation capabilities diverged along evolutionary pathways consistent with modern‐day susceptibilities. Ancestral bird PrP could not be seeded with modern‐day prions, just as modern‐day birds are resistant to prion disease. Computational modeling of structures suggested that differences in helix 2 could account for the resistance of ancestral bird PrP to seeding. Interestingly, ancestral primate PrP could be converted by all prion seeds, including both human and cervid prions, raising the possibility that species descended from an ancestral primate have retained the susceptibility to conversion by cervid prions. More generally, the results suggest that susceptibility to prion disease emerged prior to ~100 million years ago, with placental mammals possibly being generally susceptible to disease.

Abbreviations

AIC

Akaike information criterion

ancPrP

ancestral PrP

ASR

ancestral sequence reconstruction

CD

circular dichroism

CWD

chronic wasting disease

MD

molecular dynamics

RT‐QuIC

real‐time quaking‐induced conversion

ThT

thioflavin T

1. INTRODUCTION

Prion diseases are invariably fatal transmissible neurodegenerative diseases affecting a number of mammalian species. They are caused by conversion of the protein PrP from its native cellular form, PrP^C, which is largely α‐helical, into a misfolded form, PrP^Sc, which is pathogenic, infectious, and which aggregates into beta‐sheet rich structures. ¹ , ² PrP^Sc recruits PrP^C via a seeded polymerization process likely involving templated conversion. ² , ³ High‐resolution structures of PrP^C have been solved for several species, revealing PrP^C to consist of a structured C‐terminal domain typically containing three α‐helices and two short β‐strands ⁴ , ⁵ , ⁶ , ⁷ and a disordered N terminus. PrP^Sc structures have proven more difficult to define, but current models include a four‐rung beta‐solenoid conformation ⁸ and a parallel in‐register β‐sheet geometry. ⁹ , ¹⁰ , ¹¹ Three different structures with the latter geometry have been described, with variations in structural motifs occurring in PrP^Sc linked to different strains of prion disease. ¹¹ Such structural differences are likely key to understanding how strains can or cannot convert PrP^C molecules into infectious forms, a crucial question for understanding pathogenesis in prion diseases. The ability for a given PrP^Sc conformation (strain) to induce conversion of PrP^C depends largely on the sequence similarity between the two, but also on the permissible conformations that can be adopted by a given PrP^C. ¹² The conformational repertoire of a specific PrP^C is also thought to underlie the species barrier, whereby prion strains from one species are less efficient at transmitting disease to other species.

The mechanisms by which misfolding and conversion occur have been challenging to discern and remain poorly understood and controversial. ² , ³ , ⁸ This challenge has hampered the search for effective disease‐modifying therapeutic strategies, with prion diseases remaining incurable. New strategies for probing misfolding that tackle the problem from a different direction are thus urgently needed. Moreover, the discovery that some species previously thought immune to prion diseases may in fact not be so, such as rabbits, ¹³ raises questions as to how widespread susceptibility really is within the animal kingdom. Adding to the urgency of this question is the fact that a wide range of species is now being exposed to prions in the wild from sources like chronic wasting disease (CWD), ¹⁴ , ¹⁵ increasing the potential for the zoonotic emergence of novel prion diseases.

One powerful approach for exploring how many species may be susceptible to prion diseases is ancestral sequence reconstruction (ASR). In ASR, bioinformatic analysis of protein sequences from modern‐day organisms is combined with phylogenetic trees encoding the evolutionary relationship between species to deduce ancestral protein sequences (Figure 1a). ¹⁶ , ¹⁷ ASR is a useful tool for studying how protein properties change over evolutionary time, ¹⁸ revealing detailed information about the evolution of proteins and their functional mechanisms ¹⁹ , ²⁰ , ²¹ , ²² , ²³ , ²⁴ and allowing efficient identification of the mutations causing changes in phenotype. ¹⁹ It is increasingly being used to elucidate sequence‐function relationships ¹⁹ , ²⁰ and explore the origins of changes in critical features of folding mechanisms such as stability, pathways, barriers, and cooperativity. ²¹ , ²⁵ Here, we use it to examine changes in the susceptibility to conversion that occurred during the evolution of PrP. Although ASR has been applied successfully to a wide range of proteins, it has not yet been used to study proteins like PrP that cause disease via misfolding.

FIGURE 1.

Reconstruction of ancestral PrP sequences. (a) Simplified example of ancestral sequence reconstruction (ASR). Given an alignment of modern protein sequences (right) and an evolutionary tree, the sequences of ancestral proteins (shown on tree) and mutations linking evolutionary branches (e.g., E3T) can be inferred statistically. (b) Phylogeny used for PrP ASR calculation: tetrapod species tree with branch lengths optimized using the JTT + Γ₈ + F model. Gray wedges summarize clades of PrP proteins: height is proportional to the number of sequences, length reports the branch length of the most diverged sequence in the clade. Cyan nodes: ancestors under investigation. (c) Sequence differences between human and ancestral proteins for PrP residues 90–230 (human numbering). Color saturation proportional to divergence. (d) Alignment of human and ancestral PrP for the domain studied. Only changes from human PrP are noted. Secondary structure of human PrP (PDB ID 2LSB) shown above sequences. Red box highlights region in helix 2 known to be important for aggregation.

PrP likely evolved from the ZIP family of transmembrane proteins when vertebrates emerged roughly 500 million years ago. ²⁶ , ²⁷ ZIP proteins have a soluble ectodomain with high sequence similarity to PrP. ²⁸ PrP is highly conserved in vertebrates, but prion diseases have only been observed in certain mammals, ¹ suggesting that pathogenic misfolding may have developed only recently. By reconstructing ancestral forms of PrP, back to the last common amniote ancestor ~320 million years ago, and determining their aggregation potential when exposed to prion isolates from a variety of modern‐day species, we identified the point in the phylogenetic tree where pathogenic conversion likely emerged: by the evolution of the last common placental mammal. Our findings suggest that all lineages of the placental mammals may, in principle, be susceptible to prion infection.

2. RESULTS

2.1. Reconstruction of ancestral PrP

To reconstruct the sequences of ancestral versions of PrP, we assembled a database of 161 full‐length PrP sequences from a wide variety of modern organisms containing PrP including mammals, birds, reptiles, and amphibians (File S1). We then generated a sequence alignment and reconstructed the evolutionary history (see Section 4). ¹⁶ , ¹⁷ , ²⁹ , ³⁰ Figure 1b shows the phylogeny used for reconstruction. Ancestral nodes of particular interest (Figure 1b, cyan) were chosen based on points of divergence toward or away from prion‐susceptible modern‐day species. These nodes included PrP from the last common primate ancestor (ancPrimate, ~43 million years ago); the last common primate‐rodent ancestor (ancPrimRod, ~87 million years ago); the last common even‐toed ungulate ancestor (ancArtiodactyl, ~64 million years ago); the last common placental mammal ancestor (ancPlacental, ~99 million years ago); the last common bird ancestor (ancBird, ~98 million years ago); and the last common amniote ancestor (ancAmniote, ~319 million years ago). We reconstructed the sequences of these proteins using the marginal reconstruction method implemented in PAML, ³¹ then extracted the sites homologous to the structured, protease‐resistant core of modern prions (residues 90–231 in humans) for use in our downstream analysis. Ancestral sequences were reconstructed with high confidence: calculating the posterior probability that each site for a given ancestor was the amino acid indicated by the reconstruction rather than some other amino acid, we found average posterior probabilities of ≥0.99 for all ancestors except ancAmniote, with the value for the latter being 0.85 (File S2). These results imply very few reconstruction errors for all ancestors except ancAmniote. However, the phylogenetic tree and evolutionary substitution model assumed in the reconstruction are approximations, introducing additional uncertainty in the reconstruction. We therefore expect that the reconstructed sequences are highly similar to, but very likely not identical to, the actual ancestral sequences that existed millions of years ago.

Comparing the sequence divergence of these ancestral PrPs (Figure 1c), we found that they differed from one another and human PrP by between 4 and 91 amino acids. A notable feature of the ancestors chosen for study is the relatively high evolutionary divergence between ancAmniote and ancPlacental. There were 53 substitutions between ancAmniote and ancPlacental, corresponding to 0.002 substitutions per site per million years. In contrast, there were only 17 substitutions between ancPlacental and human PrP, corresponding to 0.001 substitutions per site per million years. The average rate of PrP evolution in the ordered domain thus dropped by a factor of 2 on the placental mammal lineage. Looking at the specific sequence differences between the ancestral proteins and modern human PrP (Figure 1d) revealed that the changes were scattered throughout the protein sequence. The one region that appears strictly conserved across all ancestors corresponds to the enriched Gly/Ala stretch between residues 113–128. Intriguingly, this conserved region is partially disordered in the structure of human PrP (PDB 2LSB). ³²

2.2. Structural characterization of ancestral PrP

Samples of ancestral PrP were prepared recombinantly, inserting the sequences reconstructed for each node into a plasmid for expression in Escherichia coli. In each case, we expressed the section of the protein homologous to the structured, protease‐resistant core of modern prions, residues 90–231 in human PrP. All ancestral PrPs were successfully expressed and purified following approaches used previously for modern PrP. ³³

Modern PrP^C has a high content of α‐helices. ³⁴ To test if ancestral PrP has similar structural properties, we measured circular dichroism (CD) spectra for each of the six ancPrPs we produced. CD spectra (Figure 2) showed that five of the six ancestors (ancPrimate, ancPrimRod, ancArtiodactyl, ancPlacental, and ancAmniote) all had strongly helical character similar to that of modern PrP, with peaks at 208 and 222 nm characteristic of α‐helices. The CD spectrum of ancBird, however, suggested a more disordered conformation with lower helical content.

FIGURE 2.

Circular dichroism (CD) spectra of ancestral PrPs. Most ancPrP spectra showed peaks at 208 and 222 nm consistent with α‐helical structure and similar to the spectrum for modern PrP (mouse). The exception was ancBird, where the 222 nm peak was greatly reduced.

To visualize the differences in ancPrP conformations, we used structure‐prediction tools and molecular dynamics (MD) simulations to model the structures in atomic detail. We first predicted structures for each ancestor from its amino acid sequence using AlphaFold2, a structure‐prediction tool based on machine learning. ³⁵ We then equilibrated these predictions in all‐atom MD simulations, to ensure that they were stable. The results (Figure 3, Figure S1) showed a generally strong similarity between the ancestral proteins and modern PrP. In ancBird, however, helix 2 was partly disrupted, lowering the helical content (Figure 3b). To validate these results as robust and not a prediction artifact, we generated structural models for four of the ancestral PrPs (AncAmniote, AncBird, AncArtiodactyl, and ancPrimRod) using, as structural templates, known modern PrP^C structures from species descended from these ancestors. We introduced the point mutations needed to match the ancestral sequences and equilibrated these initial structures in all‐atom MD simulations. The results (Figure S2) were comparable to those found using AlphaFold2, including the shorter helix 2 in ancBird.

FIGURE 3.

Structural models for ancestral PrP from molecular dynamics (MD) equilibration of AlphaFold2 predictions. (a) Structures of all ancPrPs except ancBird (gray) aligned with structure of human PrP (red). Three helices denoted H1–H3. (b) Alignment of ancBird (gray) with human PrP (red) showing disruption of H2 at bottom right. For each anPrP, the frame shown is the one closest to the centroid of the most‐occupied cluster from clustering analysis of the simulations.

2.3. Aggregation kinetics of ancestral PrPs

To test the aggregation capacity of ancestral PrPs, we performed real‐time quaking‐induced conversion (RT‐QuIC) assays, which involve shaking the samples in mild denaturant (0.5 M guanidine hydrochloride [GdnHCl]). ³⁶ , ³⁷ The formation of β‐rich aggregates is monitored in RT‐QuIC using the amyloid‐sensitive dye thioflavin T (ThT). We first probed the potential for these proteins to undergo spontaneous aggregation, by measuring aggregation in the absence of any prion seed. ³⁸ We found that all the ancestral PrPs were inefficient at spontaneously forming ThT‐positive aggregates, requiring at least 200 hr of shaking before aggregation was observed (Figure 4, Table 1). The least efficient was ancAmniote, which did not generate any ThT‐positive aggregates during the 400‐hour experiment (Figure 4f), followed by ancPrimRod and ancPrimate, for which only one replicate started aggregating before 350 hr. The ability to aggregate was quantified via the lag phase, the time required for the ThT signal to enter its exponential growth phase. ³⁸ We note that unseeded aggregation led to some variability in the lag phases between replicates, as is typical in the absence of seeding. ³⁸

FIGURE 4.

Real‐time quaking‐induced conversion (RT‐QuIC) assays of spontaneous aggregation in ancestral PrPs. Thioflavin T (ThT) fluorescence measured in five replicates for (a) ancPrimate, (b) ancPrimRod, (c) ancArtiodactyl, (d) ancPlacental, (e) ancBird, and (f) ancAmniote.

TABLE 1.

Lag phases from RT‐QuIC assays. Average (in hr) from N = 5 unseeded reactions and N = 3 seeded reactions; error is SD. When not all replicates aggregated (*), the result is displayed as greater than the average from the reactions that did aggregate.

Seed	AncAmniote	AncBird	AncPlacental	AncArtiodactyl	AncPrimRod	AncPrimate
Unseeded	n/a	165 ± 22	238 ± 20	263 ± 18	>350*	>315*
Hyper	n/a	n/a	25 ± 4	135 ± 3	214 ± 3	200 ± 13
Wis‐1	n/a	n/a	39 ± 16	30 ± 3	>106*	26 ± 1
MM1	n/a	n/a	21 ± 1	13 ± 1	87 ± 16	19 ± 2

We next repeated the RT‐QuIC assays, but this time seeding the reactions with infectious prions from hamster (Hyper strain, derived from transmissible mink encephalopathy passaged through hamster), cervid (Wis‐1 strain, a CWD strain from white‐tailed deer), or human (MM1 strain of Creutzfeldt‐Jakob disease) brains, in order to test the ability of the ancestral PrPs to be converted and thereby propagate the misfolding (Figure 5). Just as it did not aggregate spontaneously, ancAmniote did not exhibit seeded conversion, regardless of the prion strain used for seeding. This time, it was joined by ancBird, which also did not undergo seeded conversion. We tested that the lack of conversion in ancBird was not simply a function of low seed concentration (Figure S3a); because the spontaneous aggregation observed previously in ancBird was inhibited in the presence of noninfectious normal brain homogenate (Figure S3b), we speculate that such an inhibitory effect may have influenced the lack of aggregation in seeded reactions, too. All other AncPrPs, however, could be seeded to varying extents by the different prion strains, with all of them showing shorter lag phases than in the corresponding unseeded reactions (Table 1). Surprisingly, of the four AncPrP that could be seeded, the most consistently seeded by all strains was ancPlacental, which was the oldest of these four on the evolutionary tree. In the more recent mammalian ancestors (ancPrimate, ancPrimRod, ancArtiodactyl), the seeding efficiency differed more depending on the prion strain used for seeding. One interesting aspect of the strain dependence is that Wis‐1 was relatively inefficient at seeding aggregation of ancPrimRod, which is at the node of divergence between rodents and primates, in contrast to its notably higher efficiency at seeding aggregation of ancPrimate, which is closer to primates.

FIGURE 5.

Aggregation of ancestral PrPs seeded by infectious prions. Thioflavin T (ThT) fluorescence monitoring aggregation seeded by (a) hyper, (b) Wis‐1, and (c) MM1 prion strains. Ancestral PrP substrates (from top to bottom): ancPrimate, ancPrimRod, ancArtiodactyl, ancPlacental, ancBird, and ancAmniote.

3. DISCUSSION

Here, we have probed the structural properties and aggregation propensities of ancestral forms of PrP, with the intent of learning clues about the evolution of prion pathogenesis. We have summarized our results, as well as the previously measured aggregation behavior of modern PrP proteins, in Figure 6. We found that the oldest ancestor, ancAmniote, has the lowest ability to aggregate. Indeed, it was the only ancestral PrP we studied that did not spontaneously form ThT‐positive aggregates under RT‐QuIC conditions, in the absence of infectious seeds. Although spontaneous aggregation under these conditions does not prove pathogenicity, modern‐day PrP, which can form pathogenic structures, does normally aggregate under such conditions, ³⁸ suggesting that ancAmniote is unlikely to form infectious seeds spontaneously. Moreover, when we tested ancAmniote under seeded aggregation conditions, it could not be seeded by any of the three infectious prion strains we tested (originating from a rodent, an ungulate, and a primate), indicating that it is not a good substrate for propagating misfolding seeded by modern prions. Taken together, these results suggest that the aggregation propensity of PrP developed at an evolutionary point after the last common amniote ancestor.

FIGURE 6.

Evolutionary changes in PrP mapped to the phylogenetic tree. Phylogeny is tetrapod species tree, with length defined by median divergence time as reported by the TimeTree database. ⁶⁶ Circles denote the measured features of the ancestral proteins: complete helix 2, spontaneous aggregation, and susceptible to seeding by at least on prion. The modern species on the tree are colored by the susceptibility of the organism to prion disease: orange (susceptible), blue (not susceptible), and gray (unknown).

Following the evolutionary tree toward modern‐day birds, we found that ancBird was the one other ancestral PrP that did not aggregate in response to prion seeds. Looking more closely at the structural and sequence differences of ancBird and ancAmniote as compared to the ancPrPs that could be seeded, we noted changes in alpha helix 2 (Figure 1d, Figure 3, Figure S1, Figure S2). This region is of particular interest when considering possible mechanisms of conversion. Previous work analyzing helix 2 in modern PrP by NMR spectroscopy and MD simulations has reported that the C terminus of helix 2 is frustrated in its helical state. ³⁹ Experiments have also found that a peptide derived from the helix 2 segment can form β‐rich amyloid‐like fibrils, suggesting that the instability of this region in PrP^C may play a significant role in the α → β transition that leads to conversion into PrP^Sc. ⁴⁰ , ⁴¹ , ⁴² Within the C terminus of helix 2, the sequence TVTTTT (residues 188–193 in human PrP) has a high propensity for beta sheet formation region. ⁴⁰ , ⁴³ , ⁴⁴ This threonine‐rich sequence is present in most modern‐day PrPs, with the notable exception of modern‐day bird PrP. ⁴⁵ , ⁴⁶ As it happens, this motif is present in all the ancestral PrPs we reconstructed except ancBird and ancAmniote, the two AncPrPs whose aggregation could not be seeded by infectious prions. We therefore conclude that the evolutionary branch of PrP that descended from amniote through to birds is one that never developed a threonine‐rich helix 2, and so the descendants on this branch do not support seeded aggregation. This conclusion correlates with the lack of any known prion disease susceptibility in modern‐day birds.

Turning to the evolutionary branches that derived from ancPlacental, one might have expected that the more closely related an ancestral PrP was to PrP from a modern‐day animal that is susceptible to prions, the more prone it would be to undergo aggregation when seeded by infectious prions from that animal. All four ancestral PrPs on these branches that we studied (ancPlacental, ancArtiodactyl, ancPrimRod, and ancPrimate) were indeed able to aggregate in response to seeding with prion strains from rodent, cervid, and human brains. However, evolutionary distance from modern‐day animals did not always correlate with lag phase.

The earliest ancestral PrP that we found could be seeded with prions was ancPlacental. This ancestor was also the one that was most consistently converted by all the prion strains we tested, with rodent, cervid, and human prions all inducing aggregation in less than 100 hr. PrP thus appears to have experienced an important evolutionary change between the ancAmniote and ancPlacental nodes that enabled it to undergo seeded aggregation, and hence (in principle) prion conversion. Curiously, subsequent to this jump to an aggregation‐prone variant of PrP at the ancPlacental evolutionary node, ancestral PrP from subsequent nodes became less efficient at being seeded by rodent prions, whether or not they were on‐pathway to cervids, rodents, or primates. The efficiency of seeding by cervid or human prions followed a more surprising course. Following the evolutionary tree from ancPlacental to ancArtiodactyl—the branch from which cervids descended—we found that ancArtiodactyl was efficiently converted by both human and cervid prions. Conversely, ancPrimRod, which leads to descendants of primates and rodents, was less efficiently converted by all prion strains tested. Moving closer toward modern primates, however, ancPrimate was again efficiently seeded by human and cervid prions, showing no clear pattern in how the susceptibility to seeding by prions has evolved since the ability to aggregate first developed at the ancAmniote stage.

Our results suggest that susceptibility to prion conversion evolved at least 100 million years ago, sometime between the ancestor of amniotes and that of placental mammals. It was subsequently preserved on many descendant lineages, including humans. Why was this susceptibility maintained, despite the devastating consequences of prion diseases for susceptible animals? It is important to recognize that animals did not evolve in the context of prion infection or exposure. The ability to aggregate spontaneously had to evolve first in order to generate seeds. Acquired prion disease then requires both a host with a susceptible PrP and exposure to an already misfolded pathogenic conformation of PrP. Two key factors would be expected to reduce the likelihood of selection against acquired disease. First, prions were unlikely ever to be sufficiently abundant in the environment to impose selective pressures. Second, even though spontaneous misfolding of an ancestral prion protein should in principle be selected against if associated with disease, the fact that incubation times characteristic of modern spontaneous prion diseases are long compared to host reproductive lifespans (e.g., prion diseases in humans peak in the seventh decade, ⁴⁷ long after reproduction has typically occurred) would diminish any selective pressure that might otherwise exist. Hence, instead of undergoing selection against misfolding, prion proteins may have retained features that predispose them to it (such as the conserved threonine‐rich motif in helix 2) owing to other evolutionarily beneficial aspects.

We note that the aggregation products of modern PrP RT‐QuIC reactions are not highly infectious, even though they are seeded with infectious prions. ⁴⁸ We therefore suspect that the ancestral PrP aggregates we generated are unlikely to be infectious, but proving this conjecture could be difficult: inoculating the closest living relatives with ancestral PrP aggregates might not provide a definitive answer owing to the high barriers for inter‐species prion infection. Nevertheless, our findings reveal the evolution of aggregation behavior that correlates with infectious misfolding, provide further support for the role of helix 2 in conversion, and offer some predictions for which species may be susceptible to prion diseases. Given the similarity in aggregation behavior between all ancestral nodes on the placental branch, we suggest that all mammals should be considered generally susceptible. The fact that some mammals appear resistant or less susceptible may be due to lineage‐specific amino acid changes that interfere with misfolding, such as D/E159 in dogs, S167 in horses, and S174 in pigs. ⁴⁹ , ⁵⁰ , ⁵¹ From an evolutionary perspective, the apparent resistance of these species is the special case, rather than the susceptibility of species known to suffer from prion diseases. Even in susceptible species, single‐point mutations can render a host immune to disease, including V127 in humans ⁵² ; whether these changes completely prevent conversion or simply reduce efficiency beyond the lifetime of the animal is difficult to prove.

Perhaps the greatest concern arising from our study is the efficiency with which cervid prions were able to convert ancPlacental, ancArtiodactyl, and ancPrimate. The first two are on‐pathway to cervid evolution, hence conversion of these constructs is not unexpected, but the lineage toward humans passes through primate‐rodent to primate. Cervid prions were less efficient at converting ancPrimRod, with lag phases of 106 hr and only two of three replicates aggregating, but did so quite efficiently for ancPrimate, with a lag phase of only 26 ± 1 hr. This result suggests that although the susceptibility to conversion by cervid prions may have diminished during the primate‐rodent stage of evolution, it redeveloped by the time primates diverged from rodents. Given the rise of cervid prions in the environment owing to endemic spread of CWD in wild cervid populations, ⁵³ the wide range of mammals that can be exposed to these environmental prions over time, and ongoing exposure of humans to CWD through hunting and consumption, the potential for prion transmission across more mammalian species remains concerning.

4. MATERIALS AND METHODS

4.1. Reconstruction of ancestral PrP sequences

We constructed a set of 161 PrP protein sequences sampled broadly across tetrapods. We identified sequences by BLAST against a variety of sequence databases and validated orthology by reciprocal BLAST against the human proteome. ⁵⁴ To find amphibian sequences, we used TBLASTN against a selection of amphibian transcriptomes. ⁵⁵ , ⁵⁶ We densely sampled taxa near evolutionary transitions of interest, including primates and rodents. The final sequence database had 7 amphibians, 23 sauropsids, and 131 mammals (File S1). We aligned the sequences using muscle 3.8.1551, ⁵⁷ followed by manual refinement in AliView 1.23. ⁵⁸ Our final alignment is available in File S1 (see alignment column). We generated a maximum likelihood phylogenetic tree using PhyML 3.3.2. ⁵⁹ We tested a collection of different models (including LG vs. JTT, ³⁰ , ⁶⁰ whether to fix a proportion of invariant sites, whether to allow rate variation, and whether to use alignment model frequencies). An Akaike information criterion (AIC) test strongly favored the JTT + Γ₈ + F model, yielding an AIC weight 4.8 × 10³⁰ times larger than the weight for the next‐best model. The resulting maximum‐likelihood tree successfully placed the amphibian, sauropsid, and mammal clades, with Shimodaira and Hasegawa approximate likelihood ratio test supports of 0.995, 0.986, and 1.000, respectively (File S3). ⁶¹ Because this gene is single‐copy and shows no evidence of lateral gene transfer, we used the species phylogeny for our ancestral reconstruction studies as previous work has shown a species‐aware reconstruction provides a better estimate of ancestral sequences. ⁶² We optimized branch lengths and rate parameters on this species tree using the JTT + Γ₈ + F model. Our final species tree, constructed using the OpenTree taxonomic database, is available as File S4.

We reconstructed ancestors using the marginal reconstruction method, ⁶³ as implemented in PAML 4, ³¹ assigning gaps by parsimony. The posterior probabilities for each site in each reconstructed ancestor are given in File S2. The average posterior probabilities for all ancestors except ancAmniote were 0.99 or higher. ancAmniote had an average posterior probability of 0.85. The reconstructed sequences, as well as summarized reconstruction quality statistics, are given in File S5. All files used for the reconstruction are available as a repository on github (https://github.com/harmslab/prp-reconstruction).

4.2. Protein expression and purification

All ancestral PrP variants were expressed and purified according to the protocol used for modern PrP, ³³ following approved biosafety guidelines in a containment level 2 lab.

4.3. Circular dichroism

CD was performed as previously described. ³⁸ A 200‐μl volume of ancPrP at a concentration of 0.5 mg/ml in 50 mM ammonium acetate (pH 4.0) was used in a 1‐mm cuvette with a Chirascan CD spectrometer instrument (Applied Photophysics), reading between 200 and 260 nm while sampling points every 1 nm. For each sample, 10 scans were averaged, and base‐line spectra were subtracted. Data were processed using Applied Photophysics Chirascan Viewer and Microsoft Excel.

4.4. Ancestral PrP structural modeling

Since there are no experimentally resolved structures of ancestral PrP, we modeled the structures of the ancestral proteins computationally using two complementary approaches. First, we used AlphaFold2 ³⁵ to make blind structure predictions based solely on sequence. Second, we generated initial estimates of the structures by identifying two reference PrP structures from modern species that have large sequence similarity with each of the ancestral PrPs, and then mutated the appropriate amino acids in the reference structures to match the sequence in the relevant ancestral PrP, using Molecular Operating Environment (Chemical Computing Group) to implement the mutations. The PDB IDs for the structures selected to perform the mutations were 1U5L and 1XYK (ancAmniote), 2LSB and 1B10 (ancPlacental), 1XYQ and 6FNV (ancArtiodactyl), 1B10 and 2LSB (ancPrimRod), and 1U5L and 1U3M (ancBird). We thus created three structures for each ancestor, one from AlphaFold and two others via mutation. Each structure was used to initiate a microsecond‐long all‐atom MD simulation with explicit water using Amber 18. ⁶⁴ The protein structures were parameterized using the ff14SB force field and were solvated in TIP3P water model with minimum margins of 12 Å using the tleap module of Amber. The simulations were performed with 0.15 M NaCl at 310 K. A clustering analysis was performed on the last 500 ns of each simulation to identify the most occupied clusters. After calculating the centroid of each cluster, the frame with the lowest root mean square deviation (RMSD) from the centroid of the most‐occupied cluster was selected as the representative equilibrated structure for each simulation.

The RMSD of the backbone alpha carbon atoms with respect to the starting structure for each simulation was computed to monitor convergence to a stable configuration. While most simulations were performed for 1 μs, the simulation of the ancBird structure starting from 1U3M was extended to 3 μs because it showed large fluctuations in the first microsecond.

4.5. Real‐time quaking induced conversion

RT‐QuIC was performed as described previously. ³⁷ All experiments were carried out within prion biocontainment biosafety level 2+ facilities, according to approved biosafety manual guidelines. Recombinant PrP protein in 6 M GdnHCl solution was diluted in RT‐QuIC buffer (20 mM sodium phosphate pH 7.4; 130 mM NaCl; 10 mM EDTA; 0.002% SDS) to a final protein concentration of 0.3 mg/ml (and residual 0.5 M GdnHCl). Reactions were left unseeded or were seeded with 0.00001% (weight/volume) homogenate from CJD‐infected, CWD‐infected, or hyper‐infected brains. The aggregation reactions were carried out in 96‐well plates (white plate, clear bottom; Costar 3610) sealed with thermal adhesive film (08‐408‐240; Fisherbrand). The samples were incubated in the presence of 10 μM ThT at 42°C with cycles of 1 min shaking (700 rpm double orbital) and 1 min rest. ThT fluorescence measurements (450+/210 nm excitation and 480+/210 nm emission; bottom read) were collected every 60 min. There were five technical replicates for unseeded reactions and three for seeded reactions. Lag phases were calculated as described previously. ³⁸

Brain homogenate with hyper prions was prepared from brain tissue taken from clinically affected Syrian golden hamsters (Envigo) at day 70 postinfection. Brain homogenate with MM1 prions was prepared from brain tissue obtained from the CJD Surveillance System brain bank. Hyper‐ and MM1‐infected brain tissue samples were homogenized in ultrapure water (sigma) with ceramic beads in an Omni Bead Ruptor to make 10% brain homogenates (wt/vol). Aliquots were stored at −80°C until further use. Wis‐1‐infected tissue was obtained from the brains of terminally ill, orally infected white‐tailed deer, and homogenate was prepared as described previously, ⁶⁵ with dounce homogenization and serial passage through different gauged syringes.

AUTHOR CONTRIBUTIONS

Leonardo M. Cortez: Formal analysis (equal); investigation (equal); methodology (equal); visualization (equal); writing – original draft (supporting); writing – review and editing (supporting). Anneliese J. Morrison: Formal analysis (equal); investigation (equal); visualization (equal); writing – review and editing (supporting). Craig R. Garen: Methodology (equal); resources (lead); writing – review and editing (supporting). Sawyer Patterson: Formal analysis (supporting); investigation (supporting); methodology (equal); resources (supporting). Toshi Uyesugi: Formal analysis (equal); investigation (equal); writing – review and editing (supporting). Rafayel Petrosyan: Formal analysis (equal); investigation (equal); visualization (equal); writing – review and editing (supporting). Rohith Vedhthaanth Sekar: Formal analysis (equal); investigation (equal); visualization (equal); writing – review and editing (supporting). Michael J. Harms: Conceptualization (equal); data curation (lead); funding acquisition (equal); project administration (supporting); supervision (equal); visualization (equal); writing – original draft (supporting); writing – review and editing (equal). Michael T. Woodside: Conceptualization (equal); funding acquisition (lead); project administration (lead); supervision (equal); visualization (equal); writing – original draft (supporting); writing – review and editing (lead). Valerie L. Sim: Conceptualization (equal); funding acquisition (equal); project administration (supporting); supervision (equal); visualization (supporting); writing – original draft (lead); writing – review and editing (equal).

CONFLICT OF INTEREST

The authors declare that they have no conflict of interest with the contents of this publication.

Supporting information

FILE S1 Sequence alignment used for ancestral reconstruction

FILE S2 Sequence reconstruction probabilities

FILE S3 Maximum‐likelihood phylogenetic tree

FILE S4 Final phylogenetic tree

FILE S5 Reconstructed sequence

FIGURE S1 Structural models of ancPrP after MD equilibration of AlphaFold2 predictions. Representative structures from MD simulations of (a) ancPrimRod, (b) ancArtiodactyl, (c) ancPlacental, (d) ancBird, and (e) ancAmniote, showing the frame closest to the centroid of the most‐occupied cluster in the simulations. Insets: RMSD showing equilibration of structures. Helices are in red, β‐strands in yellow, loops in gray, and turns in blue.

FIGURE S2 Structural models of ancPrP after MD equilibration of mutated modern PrP. Representative structures from MD simulations of (a) ancPrimRod (top: based on mutation of PDB ID 1B10, bottom: based on mutation of PDB ID 2LSB), (b) ancArtiodactyl (top: based on mutation of PDB ID 1XYQ, bottom: based on mutation of PDB ID 6FNV), (c) ancPlacental (top: based on mutation of PDB ID 2LSB, bottom: based on mutation of PDB ID 1B10), (d) ancBird (top: based on mutation of PDB ID 1U5L, bottom: based on mutation of 1U3M), and (e) ancAmniote (top: based on mutation of PDB ID 1U5L, bottom: based on mutation of 1XYK), showing the frame closest to the centroid of the most‐occupied cluster in the simulations. Insets: RMSD showing equilibration of structures. Helices are in red, β‐strands in yellow, loops in gray, and turns in blue.

FIGURE S3 AncBird PrP aggregation kinetics under different concentrations. (a) No aggregation of ancBird PrP substrate (0.3 g/L) was seen when increasing the concentration of brain homogenate used for seeding as high as 0.1%, for any of the prion strains used. (b) No spontaneous aggregation of ancBird PrP was seen in the presence of normal (noninfectious) brain homogenate, in contrast to reactions without brain homogenate, suggesting that brain homogenate may have an inhibitory effect on aggregation.

Cortez LM, Morrison AJ, Garen CR, Patterson S, Uyesugi T, Petrosyan R, et al. Probing the origin of prion protein misfolding via reconstruction of ancestral proteins. Protein Science. 2022;31(12):e4477. 10.1002/pro.4477

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are either provided in the supplementary material or available from the corresponding authors upon reasonable request.