Extra N-Terminal Residues Have a Profound Effect on the Aggregation Properties of the Potential Yeast Prion Protein Mca1

The metacaspase Mca1 from Saccharomyces cerevisiae displays a Q/N-rich region at its N-terminus reminiscent of yeast prion proteins. In this study, we show that the ability of Mca1 to form insoluble aggregates is modulated by a peptide stretch preceding its putative prion-forming domain. Based on its genomic locus, three potential translational start sites of Mca1 can give rise to two slightly different long Mca1 proteins or a short version, Mca1451/453 and Mca1432, respectively, although under normal physiological conditions Mca1432 is the predominant form expressed. All Mca1 variants exhibit the Q/N-rich regions, while only the long variants Mca1451/453 share an extra stretch of 19 amino acids at their N-terminal end. Strikingly, only long versions of Mca1 but not Mca1432 revealed pronounced aggregation in vivo and displayed prion-like properties when fused to the C-terminal domain of Sup35 suggesting that the N-terminal peptide element promotes the conformational switch of Mca1 protein into an insoluble state. Transfer of the 19 N-terminal amino acid stretch of Mca1451 to the N-terminus of firefly luciferase resulted in increased aggregation of luciferase, suggesting a protein destabilizing function of the peptide element. We conclude that the aggregation propensity of the potential yeast prion protein Mca1 in vivo is strongly accelerated by a short peptide segment preceding its Q/N-rich region and we speculate that such a conformational switch might occur in vivo via the usage of alternative translational start sites.


Introduction
Several proteins that can undergo structural conversion from a soluble state into an insoluble heritable prion conformation have been characterized in the yeast Saccharomyces cerevisiae including Sup35 forming the prion [PSI + ] or Ure2p and its prion state [URE3] [1][2][3][4][5]. A common feature of these yeast prion proteins is the prion-forming domain (PFD), a glutamine and/or asparaginerich (Q/N-rich) region with repeats of oligopeptides, such as the imperfect PQGGYQQYN repeats found in Sup35.
Sup35 is a translation termination factor and conversion to the [PSI + ] prion state inactivates Sup35 molecules thereby increasing the levels of nonsense suppression [1,6,7]. The Sup35 protein has three domains, an N-terminal PFD which is essential for the conversion into the [PSI + ] prion form, a middle domain and a C-terminal domain, which is essential for its function in translation termination [5,8,9].
The in vivo assay for monitoring [PSI + ] usually involves readthrough of nonsense alleles in auxotrophic markers, e.g. ade1-14 (UAG) or ade2-1 (UAA). Cells lacking the [PSI + ] prion fail to synthesize adenine in the presence of a nonsense allele like ade1-14. Accordingly, [psi 2 ] cells cannot grow on minimal media lacking adenine and in addition accumulate a red pigment on rich media under adenine-limiting conditions. [PSI + ] cells, however, grow on minimal media lacking adenine (-ADE) and do not accumulate the red pigment on rich media [1,10].
Recent database analyses and genetic screens were performed in order to detect new yeast proteins with key prion features including Q/N-rich regions, similar to prion forming domains of other yeast prions [11][12][13]. Nemecek et al. [13] detected Mca1 as a potential new prion protein by a genetic screen. Mca1 was described earlier as a metacaspase that regulates apoptosis in Saccharomyces cerevisiae [14]. Mca1 harbors a Q/N-rich region in its N-terminal domain that is characteristic for yeast prion proteins and aggregation-prone proteins. In addition, this putative prionforming domain of Mca1 contains several imperfect repeats of QQYG that are reminiscent of the imperfect PQGGYQQYNrepeats found in the yeast prion protein Sup35 (Figure 1). Nemecek et al. [13] fused random yeast DNA fragments to the middle and C-terminal domain of Sup35 (Sup35MC) and selected for clones with increased read-through of the ade2-1 nonsense allele caused by elevated aggregation of the Sup35MC fusion protein. Three different clones contained parts of the Q/N-rich N-terminal domain of Mca1 and the Ade + phenotype of the Mca1-Sup35MC fusion protein dominantly segregated as a non-chromosomal genetic element, typical for prion proteins. Furthermore, the authors investigated several characteristic properties of yeast prions including curability and metastability. They found that the Ade + phenotype could be cured by overexpression of Hsp104 and showed the reversible curability by spontaneous appearance of the prion protein in a previously cured strain. De novo formation of the Mca1 prion protein occurred by overexpression of the Nterminal domain of Mca1 suggesting that the isolated Ade + clone, which contained a portion of Mca1, fulfills the criteria of a yeast prion [13]. In another recent study, Alberti et al. performed a bioinformatic proteome-wide survey for proteins with prion-like properties in S. cerevisiae [11]. The authors experimentally investigated 100 prion protein candidates and found 19 new prions. Also in this study, Mca1 had been tested for its aggregation properties, albeit the authors did not find compelling evidence for [MCA + ] being a prion. Mca1 did neither form SDS-resistant aggregates in semi-denaturing detergent-agarose gel electrophoresis, which would indicate prion-like structures, nor did a Mca1-Sup35C fusion protein display an Ade + phenotype.
Interestingly, based on its genetic structure three potential translational starts of the MCA1 gene from yeast (YOR197W) exist which could give rise to different isoforms. Two potential translational start sites of MCA1 are in immediate vicinity giving rise to Mca1 proteins that differ only slightly by two amino acids in length (453 and 451 amino acids, respectively), while the third one is more distant giving rise to a significantly shorter Mca1 protein of 432 amino acids. Which of these potential translational start sites are used in vivo is still unclear due to the lack of supportive experimental data. In addition, the annotated translational start of MCA1 was changed over the past years. Based on automated genome sequence comparison of closely related Saccharomyces species, the translational start site of MCA1 was recently moved downstream to the third AUG, giving rise to the short version of the Mca1 protein (432 amino acids) [15,16].
Here we show that dependent on which isoform of Mca1 is expressed, the aggregation propensity of Mca1 is strongly affected. While the short version Mca1 432 is completely soluble, the long versions of Mca1 aggregate in vivo. The conversion of Mca1 from a soluble protein into an insoluble conformation with prion-like properties is triggered by the additional peptide stretch present at the N-terminus of the long Mca1 versions. However, under regular growth conditions only the short soluble isoform is detectable in vivo.

Mca1 432 is predominantly expressed in vivo
We were intrigued by the possibility of Mca1 from S. cerevisiae being a potential new yeast prion protein and set out to examine the aggregation and prion properties of this protein more closely. The domain architecture and the putative prion-forming domain (PFD) of Mca1 are displayed in Figure 1. First, we determined which potential initiation codon of the MCA1 allele is used for translation in vivo. The use of the recently annotated start would result in a Mca1 protein of 432 amino acids (designated hereafter Mca1 432 ), while the use of the AUG start codons 57 or 63 nucleotides upstream of the currently annotated AUG would result in a Mca1 protein of 451 or 453 amino acid residues, respectively ( Figure 1). Interestingly, homologous Mca1 proteins from other fungi display similar features in respect to their translational start site. For example, Kluyveromyces lactis Mca1 contains a Q/N-rich N-terminal region and a potential second translational start site, giving rise to a 20 amino acids Nterminally extended isoform (data not shown).
The additional amino acids extension at the N-terminus of longer Mca1 versions should result in a molecular weight difference of 2.4 or 2.7 kDa compared to Mca1 432 . However, PFD containing proteins migrate aberrantly in SDS-PAGE gels [17][18][19]. Accordingly, we could not accurately determine the molecular weight of Mca1 expressed in vivo by immunoblotting of wild type lysate with polyclonal antibodies raised against Mca1. Also, our attempts to purify C-terminally TAP-tagged Mca1 expressed under authentic chromosomal conditions from S. cerevisiae failed due to the very low abundance of Mca1 and massive contamination by unspecifically co-purified proteins which did not allow the determination of its molecular mass by mass spectrometry. Therefore, we chose an alternative strategy to elucidate the translational start site used in vivo. We cloned two MCA1 genes with different translational start sites that served as standards to distinguish the long Mca1 versions from the short Mca1 protein. The short MCA1 gene contained only the third AUG (bp +1 to bp +1299 respective to the currently annotated MCA1 coding sequence) giving rise to Mca1 432 . The long version of MCA1 started at the first AUG (bp 263 to bp +1299 respective to the annotated MCA1 coding sequence). Due to cloning considerations the long MCA1 gene encoded an additional glycine residue after the initial methionine thereby resulting in a Mca1 protein of 454 amino acids (designated hereafter Mca1 454, Figure 1). Both MCA1 genes were cloned under control of a copper-inducible promoter into pRS313. To determine the authentic translational start of MCA1 in vivo, we additionally cloned the MCA1 containing DNA segment including its endogenous promoter and terminator regions (MCA1 endog = bp 2582 to bp +1799 respective to the annotated MCA1 coding sequence) into pRS313 yeast vector lacking the copper-inducible promoter. We transformed the plasmids encoding MCA1 454 (V414) and MCA1 432 (V413), as well as MCA1 under endogenous control (MCA1 endog = V415) into mca1D yeast cells lacking the chromosomal MCA1 gene. After copper induction of the plasmid encoded MCA1 454 and MCA1 432 genes, total cell lysates were prepared and all Mca1 variants were visualized by immunodetection using polyclonal Mca1 antibodies. As shown in Figure 2A  analyzed the aggregation properties of the different Mca1 isoforms by preparing lysates from mca1D cells expressing copper-inducible MCA1 454 or MCA1 432 , as well as MCA1 under endogenous control. The lysates were applied to sequential centrifugation analysis to differentiate between large aggregates that sediment by low-speed centrifugation, and smaller ones or aggregates with specific sedimentation properties such as Rnq1 aggregates that could be monitored by high-speed centrifugation. As it is evident in Figure 2B, endogenous Mca1 and the short version Mca1 432 controlled by the copper-inducible promoter were exclusively found in the supernatant but not in the pellet fraction implying that these Mca1 variants do not aggregate under the tested conditions. In contrast, the Mca1 454 variant revealed insoluble material by both, low-speed and high-speed centrifugation with 18,000 g and 100,000 g, respectively. Based on the fact that the same promoter drives both Mca1 variants, Mca1 432 and Mca1 454 , and that we detected similar total protein levels, the pronounced difference in the solubility can be attributed to the additional Nterminal amino acid stretch present in Mca1 454 .
To further analyze the aggregation properties of Mca1 in vivo, we fused the different Mca1 constructs to green-fluorescent protein (GFP) under the control of a copper-inducible promoter ( Figure 3A) and expressed the fusion proteins in cells lacking the chromosomal MCA1 gene. As discussed above, three translational In summary, we conclude that the short GFP-fusion version of Mca1 432 has only a latent aggregation tendency. However, the addition of 19 amino acids to the N-terminus of Mca1 432 converts the Mca1 protein into a strong aggregation-prone variant in vivo.
Next, we analyzed the N-terminal 19 amino acids of Mca1 for their biochemical properties and found an unusual hydrophobichydrophilic charge distribution ( Figure 1). We wondered whether the strong hydrophobic character within the first 10 amino acids is responsible for the strongly enhanced aggregation of Mca1 451 . Therefore, we divided the 19 amino acids into two parts and fused either the hydrophobic (Mca1 mut1 ) or the hydrophilic part (Mca1 mut2 ) of the 19 amino acids stretch to Mca1 432 -GFP and analyzed the different truncation mutants for their ability to form GFP aggregates. As displayed in Figure 3D, the frequency of aggregate formation was not altered by the different truncations, indicating that the presence of either hydrophilic or hydrophobic stretch at the N-terminus of the prion-forming domain of Mca1 is sufficient to destabilize the protein conformation.
Next, we investigated whether the destabilizing effect of these 19 amino acids is specific for Mca1 or perhaps portable to other proteins as well. To this end, we constructed a fusion protein containing the N-terminal 19 amino acids stretch of Mca1 451 fused N-terminally to firefly luciferase together with a GFP moiety at the C-terminus (FFL19 aa -GFP). Luciferase has no Q/N-rich region and displayed only minor protein aggregation on its own. However, the frequency of GFP aggregates was 3-fold increased for FFL19 aa -GFP although the aggregation was clearly less pronounced compared to Mca1 451 ( Figure 3D). We conclude that the increased ability to form aggregates of Mca1 451 and FFL19 aa is due to the presence of the additional N-terminal 19 amino acids stretch that presumably destabilizes protein conformations in general but shows a more dramatic effect when combined with the Q/N-rich region of Mca1.  . This indicates that the addition of either a hydrophobic, hydrophilic, or hydrophobic-hydrophilic stretch to the N-terminus of the Q/N-rich region of Mca1 is sufficient for destabilizing the protein conformation. The addition of the hydrophobic-hydrophilic 19 amino acids stretch to the N-terminus of firefly luciferase also increased the frequency of aggregate formation. Firefly luciferase lacks an aggregation-prone Q/N-rich region, suggesting that the 19 amino acids N-terminal extension has a general destabilizing effect on protein conformations. Such destabilizing effects of N-terminal extensions had been observed in previous studies investigating polyQ-proteins, which are not related to prions but also capable to convert their structures into amyloid-like fibers. A FLAG-tag was found to unmask the latent polyQ lengthdependent toxicity in polyQ-expanded exon I of Huntingtin protein (Htt) [21]. Moreover, a very recent analysis of the Frydman lab showed that polyQ aggregation kinetics is not solely a function of polyQ repeat length, but rather includes the critical contribution of the N-terminal 17 amino acid residues forming an amphipathic helix and promoting rapid Htt aggregation by direct modulation of Htt conformation [22]. Thus far, it is unclear why and how the N-terminal extension of Mca1 has such a dramatic impact on the aggregation properties of Mca1 and further analyses are required to resolve that finding on a mechanistic basis. However, we speculate that the addition of the hydrophobichydrophilic 19 amino acids stretch to the N-terminus of Mca1 or luciferase presumably destabilizes the protein conformation, thereby increasing the probability of aggregate formation, which in case of Mca1 allows for conversion of the protein into an aggregation-prone state potentially displaying prion-like properties.

The extended version of Mca1 fused to Sup35C displays a nonsense-suppressor phenotype
Intrigued by the possibility that the sole addition of N-terminal amino acids might trigger the aggregation properties of Mca1, we furthermore characterized the potential of the aggregation-prone version of Mca1 (Mca1 454 ) to induce an Ade + phenotype if fused to the C-terminal domain of Sup35. We found that Mca1 454 -Sup35C fusion constructs can complement a sup35D deletion strain and indeed displayed an Ade + phenotype dependent on the putative prion-forming domain of Mca1 454 . Only constructs containing the Q/N-rich domain of Mca1 454 fused to Sup35C were able to suppress the adenine deficient phenotype of strains harboring the ade1-14 mutation, as analyzed by growth on adenine minimal media as well as white color on rich media. Thus, the long Mca1 isoform, Mca1 454 , reveals an aggregation behavior reminiscent of yeast prion proteins like Sup35.
The question remains why Nemecek et al. [13] detected Mca1 as a yeast prion protein in their genetic screen, whereas Alberti et al. [11] did not detect Mca1 in their respective systematic screen for yeast prions. Based on the results presented in this study, we can speculate about this issue. Nemecek and coworkers [13] investigated a MCA1 fragment starting 161 base pairs upstream of the MCA1 coding region including all potential translation initiation sites fused to Sup35C on a plasmid called p20MCA. Thus, in addition to the shorter Mca1 432 version, theoretically also the longer Mca1 version could be expressed by their construct, perhaps at a level that could not be detected under their assay conditions but sufficient to trigger aggregation of Mca1. In contrast, Alberti et al. [11] cloned MCA1 for their study according to the currently annotated start site that is lacking the destabilizing N-terminal amino acids. This would resolve, together with our findings showing that the extra N-terminal amino acids modulate the aggregation properties of Mca1, the discrepancy in the conclusions regarding the prion properties of Mca1 that are present in the two studies [11,13]. Alternatively, and not mutually exclusive, it is also possible that differences in the experimental systems used by Alberti  Many intrinsic and extrinsic factors are known that can contribute to the conversion of soluble proteins into an aggregation-prone or prion-like state, including chaperones and various stress conditions. To our knowledge, this is the first report about a potential yeast prion protein that strongly varies in its aggregation and potential prion-like properties based on the translational start codon employed by nature. We provide evidence that the delicate balance of Mca1 conformation depends on its N-terminal start. Although there is no proof so far that such a translational switch of the start codon occurs in vivo for Mca1 or any other prion-like protein, we consider such a hypothesis as very attractive. It is tempting to speculate that perhaps yeast cells may also use the alternative translation initiation codons of MCA1 under special environmental conditions. It is known that alternative usage of ATG codons can be triggered, for example, in a hormone-dependent manner in the testis or by stress conditions [23].
Thus far, we could not detect such a condition for Mca1 (data not shown). Interestingly, we found by screening yeast proteins harboring potential prion-forming domains for additional upstream start sites that Pgd1, a subunit of the RNA polymerase II mediator complex, also possesses an alternative, upstream start site. This finding indicates that other aggregation-prone proteins might also feature alternative start sites that may change their aggregation properties according to the isoform made in the cell. Additionally, Komar et al. [24] described an internal ribosome entry site (IRES) in the mRNA of URE2. Expression of Ure2 from that internal initiation site resulted in a truncated Ure2 protein that lacked the prion-forming domain. Importantly, the authors showed that this alternative, truncated Ure2 protein affected the [URE3] prion phenotype indicating that yeast cells can influence propagation of prion proteins by using different translation initiation sites depending on e.g. growth conditions or environmental signals.

Strains and plasmids used in this study
Strains and plasmids constructed and used in this study are listed in Table 1 and Table 2. Cloning strategies and primer sequences are listed in supplemental Table S1. The sup35D strain Y133 was generated by transforming strain Y119 [25] with PCRgenerated copies of the kanmx cassette amplified from plasmid pFA6a-KanMX6 [26] with primers containing regions homologous to the SUP35 locus: (CCATTGTACTGTAACAAAAAG-CGGTTTCTTCATGACTTGCTCGGcggatccccgggttaattaa and GCATTTACTTATGTTTGCAAGAAATTTACTCGGCgaattcgagctcgtttaaac, regions homologous to SUP35 locus indicated in capital letters).

Aggregation analysis of Mca1, Sup35 and Rnq1
Cultures for aggregation analysis were grown in appropriate media to mid-log phase and cell lysis was performed as described previously [5,27]. Protein aggregates were separated by low-and high-speed centrifugation (18,000 and 100,000 g, respectively) and subsequently analyzed by SDS-PAGE and Western blotting using standard techniques. Sup35, Rnq1 and Mca1 proteins were detected using polyclonal anti-Sup35 antibodies [7], polyclonal anti-Rnq1 antibodies [5] and polyclonal anti-Mca1 454 antibodies (this study), respectively. For preparation of crude cell extracts for Western Blot analysis NaOH lysis was performed.

Fluorescence microscopy
Cells containing the respective GFP vectors were grown overnight in appropriate selective media. Subsequently, the cultures were diluted into fresh selective media, expression was induced by addition of 150 mM CuSO 4 and the cells were grown for additional 24 hours. Fluorescence was observed using a Carl Zeiss fluorescence microscope at 1006 magnification using a standard FITC filter set.

Assays for Ade + formation and curing
Mca1-Sup35C fusions were assayed for their Ade + phenotype through the inability of aggregated Sup35 to terminate translation. Accordingly, readthrough of the ade1-14 (UGA) allele enables adenine biosynthesis and additionally prevents the accumulation of a red pigment.
Mca1-Sup35C fusion proteins were constitutively expressed in a strain lacking the chromosomal SUP35 gene and growth on media lacking adenine was analyzed after four and 18 days of incubation, respectively. The strains and plasmids used in this study are listed in Table 1 and Table 2.