A Genetic Code Alteration Is a Phenotype Diversity Generator in the Human Pathogen Candida albicans

Background The discovery of genetic code alterations and expansions in both prokaryotes and eukaryotes abolished the hypothesis of a frozen and universal genetic code and exposed unanticipated flexibility in codon and amino acid assignments. It is now clear that codon identity alterations involve sense and non-sense codons and can occur in organisms with complex genomes and proteomes. However, the biological functions, the molecular mechanisms of evolution and the diversity of genetic code alterations remain largely unknown. In various species of the genus Candida, the leucine CUG codon is decoded as serine by a unique serine tRNA that contains a leucine 5′-CAG-3′anticodon (tRNACAG Ser). We are using this codon identity redefinition as a model system to elucidate the evolution of genetic code alterations. Methodology/Principal Findings We have reconstructed the early stages of the Candida genetic code alteration by engineering tRNAs that partially reverted the identity of serine CUG codons back to their standard leucine meaning. Such genetic code manipulation had profound cellular consequences as it exposed important morphological variation, altered gene expression, re-arranged the karyotype, increased cell-cell adhesion and secretion of hydrolytic enzymes. Conclusion/Significance Our study provides the first experimental evidence for an important role of genetic code alterations as generators of phenotypic diversity of high selective potential and supports the hypothesis that they speed up evolution of new phenotypes.


INTRODUCTION
A number of exceptions to the standard genetic code have been discovered in prokaryotic and eukaryotic organisms, involving nonsense-to-sense and sense-to-sense codon identity changes [1,2]. Twenty five of such alterations have been recorded in mitochondrial genetic codes of metazoan, fungi, red algae, green plants, alveolates, stramenopiles, haptophtes and euglenozoans [3]. The most remarkable alterations involve metazoan arginine AGA and AGG (AGR) codons, which changed their identity to serine at the base of the phylogenetic tree, and later on to translation-stop in vertebrates and to glycine in urochordates [4,5].
In bacteria and eukaryotic cytoplasmic systems, 18 genetic code alterations have also been recorded, but unlike in mitochondria, they involve, with one exception, nonsense-to-sense codon identity changes or codon unassignments (codons that vanished from genomes). For example, in Micrococcus spp., Mycoplasma spp. and Pseudomicrothorax dubius the AGA/AUA, CGG and UGA codons are unassigned, respectively. In Bacillus subtilis UGA codons are used to terminate mRNA translation (stop codons) and to insert tryptophan, creating readthrough proteins [6]. In various species of ciliates, 1 or 2 stop codons changed their identity to either glutamine (UAA and UAG), glutamate (UAA) or cysteine (UGA). Interestingly, these genetic code alterations apparently minimize nonsense errors arising from re-assembly of the ciliates fragmented genome [7][8][9][10].
Those genetic code alterations show that certain codons, namely codons that start with uridine (UNN) or adenosine (ANN), in particular stop (UAA, UAG or UGA), arginine (AGR), serine (AGY) and isoleucine (AUA) codons are more prone to identity changes than others. The only exception to this first codon base rule occurs in yeast mitochondria where cytosine starting codons (CNN), namely leucine CUN codons, changed their identity to threonine. Also, the leucine CUG codon altered its identity to serine in the cytoplasm of various Candida and Debaryomyces species [11][12][13]. These findings suggest that the strength of the interaction of the first codon-anticodon base pair plays an important role in the evolution of genetic code alterations. Indeed, the change of identity of UAA and UAG stop codons to glutamine in various ciliates involved first base misreading by glutamine tRNAs, which decode glutamine CAA and CAG codons [8,14]. But, it also indicates that other forces beyond codon-anticodon interaction play important roles in codon identity redefinition.
The unexpected flexibility of the genetic code described above is further highlighted by insertion of selenocystein (21 st amino acid) in the active sites of prokaryotic and eukaryotic selenoproteins and insertion of pyrrolysin (22 sd amino acid) in the active site of monomethylamine methyltransferase of Methanosarcina barkeri [15,16]. These expansions involving reprogramming of UGA and UAG stop codons, respectively, highlight the potential of genetic code alterations/expansions to generate functional innovation. This hypothesis is supported by recent artificial expansion of the genetic code through synthetic biology methodologies. Indeed, 49 non-natural amino acids have already been incorporated into E. coli, yeast and mammalian cells [17][18][19][20], to produce novel proteins of biotechnological and biomedical interest. This dramatic demonstration of genetic code flexibility also unveiled an extraordinary capacity of complex organisms to tolerate partial codon identity redefinition [21].
We are using C. albicans as a model system to elucidate the evolution of genetic code alterations. In this case, a unique serine tRNA (tRNA CAG Ser ) decodes leucine CUG codons as serine [22][23][24]. However, the tRNA CAG Ser is recognized by both leucyl-and seryl-tRNA synthetases (LeuRS and SerRS) and it is aminoacylated in vitro with both serine (97%) and leucine (3%) [25]. This unusual dual aminoacylation of the tRNA CAG Ser has been preserved to the present day [26,27], raising the intriguing possibility that it may play a role in C. albicans biology. It also provides strong support for the hypothesis that codon identity redefinition is driven through codon decoding ambiguity [28,29]. In here, we have reconstructed the early stages of the C. albicans genetic code alteration to shed new light into those questions. This partial reversion of CUG identity from serine back to leucine triggered morphogenesis, phenotypic switching, and up-regulated expression of genes involved in cell adhesion and hyphal development and increased secretion of proteinases and phospholipases. Important karyotype alterations were also observed. The overall data suggest that C. albicans CUG ambiguity is an important phenotypic diversity generator and highlight important and yet overlooked functional roles for genetic code alterations.

Reverting CUG identity from serine back to leucine
The CUG identity alteration from leucine to serine was initiated 272625My ago through a mutant serine tRNA (tRNA CAG Ser ), containing a 59-CAG-39 anticodon, which could decode CUG codons (see introduction) [26,22]. Initially, this unique tRNA-CAG Ser competed with endogenous tRNA Leu , which decoded CUG codons as leucine [30], creating a new situation where both leucines and serines were inserted at CUG positions on a genome wide scale ( Figure 1). For reasons not yet fully understood, the wild type tRNA Leu disappeared from the Candida ancestor genome leaving CUG decoding exclusively to the mutant tRNA CAG Ser . Disappearance of the tRNA Leu should have abolished the ambiguous status of CUG codons, however the tRNA CAG Ser is recognized by both LeuRS and SerRS (see above), creating a serine tRNA that exists in 2 distinct forms, namely ser-tRNA CAG Ser (charged with serine) and leu-tRNA CAG Ser (charged with leucine). This ambiguous tRNA still exists in C. albicans [25].
We have re-created in C. albicans (CAI-4 strain) the high ambiguity status of CUG codons that existed 272625 My ago in the Candida ancestor ( Figure 1). For this, we have expressed Saccharomyces cerevisiae wild type and mutant tRNAs, which decode CUG codons as leucine, in C. albicans (Figure 2A-D). These tRNAs Leu competed with the novel tRNA CAG Ser for CUG codons at the ribosome A-site, but were not lethal. We have hypothesized that such genetic manipulation would increase CUG ambiguity and could uncover phenotypes associated to the residual ambiguity (3%) [25] of CUG codons in C. albicans.
Three S. cerevisiae tRNA Leu genes, namely a wild type tRNA UA-G Leu and two mutant tRNA CAG Leu (containing U 33 or G 33 in the anticodon-loop), plus a control tRNA AGA Ser ( Figure 2B-D), were successfully expressed in C. albicans CAI-4 cells, as shown by Northern blot analysis of tRNAs fractionated by acidic-PAGE, which separates charged from uncharged tRNAs [31] ( Figure 3A). This is in line with previous experiments that showed that the identity elements of C. albicans and S. cerevisiae LeuRS and SerRS are identical [32,33]. Transformation efficiency of plasmids carrying the S. cerevisiae tRNA Leu genes was lower than that of control plasmids ( Figure 3B) containing no tRNA gene or containing a cognate serine decoder tRNA (pUA16) ( Figure 2D). In other words, the tRNA Leu were slightly toxic to C. albicans, which supported the hypothesis that they were fully functional and could incorporate leucine at CUG positions. Remarkably, clones that survived transformation showed no decrease in growth rate ( Figure 3C), suggesting that they adapted to increased CUG ambiguity.

Ambiguous CUG decoding generated important phenotypic diversity
Interestingly, expression of those S. cerevisiae tRNA Leu genes in C. albicans triggered morphogenesis in both solid and liquid media ( Figure 4). The pUA15 clones displayed extensive morphological variation ( Figure 4B-C) and produced highly heterogenous cell populations containing elongated-ovoid cells, pseudohypha and true hypha (not shown). Some pUA15 clones produced hypha that occupied sectors or entire colonies. Notably, morphological events that gave rise to these phenotypes happened spontaneously without external inducing factors. As expected, control pUA12 and pUA16 clones had homogeneous morphology and formed smooth-white colonies similar to those of untransformed C. albicans CAI4 ( Figure 4A). Similar results were obtained with clones pUA13 and pUA14 (data not shown). Apart from morphogenesis, CUG ambiguity also induced phenotypic switching, which is a C. albicans phenotype characterized by reversible induction of opaque or myceliated sectors in white smooth colonies [34]. High frequency of phenotypic switching (63-88%) was obtained for all clones expressing S. cerevisiae leucine tRNAs (pUA13-15), but not for control clones (pUA12 and pUA16) ( Figure 4D).
CUG ambiguity also increased cell adhesion in liquid and solid media ( Figure 5A), and once more, this phenotype was exacerbated in pUA15 clones, as they displayed strong flocculation in liquid media ( Figure 5A). Interestingly, more than 50% of the genes involved in adhesion contain CUG codons. For example, the ALS gene family which encodes cell-surface glycoproteins that mediate The ancestor of Candida decoded the CUG codon as leucine using a single leucine tRNA (tRNA Leu ). This situation changed dramatically with appearance 272625My of a mutant serine tRNA that acquired a 59-CAG-39anticodon (tRNA CAG Ser ). The latter competed with the tRNA Leu for decoding of CUG codons, inserting both leucine and serine, at high level, at CUG positions, on a proteome wide scale. Such ambiguity decreased over time due to disappearance of the tRNA Leu gene, however charging of the tRNA CAG Ser with leucine and serine prevented complete change of identity of the CUG codon from leucine to serine. In order to elucidate why CUG ambiguity was preserved in C. albicans and clarify whether CUG identity could be partially reverted from serine back to leucine, we have reconstructed the early stages of CUG identity change (high level of ambiguity) in C. albicans using S. cerevisiae tRNAs that decode CUG codons as leucine. doi:10.1371/journal.pone.0000996.g001 adhesion to host surfaces [35], contain various CUG codons (3CUGs-ALS2, ALS3, ALS8; 4CUGs-ALS4; 5CUGs-ALS1; 11CUGs-ALS6; 12CUGs-ALS9; 18CUGs-ALS7). It is not yet clear whether the change of serine (polar) for leucine (hydrophobic) at CUG positions in the Als proteins is responsible for the flocculation and exacerbated agar adhesion observed in pUA15 clones. But, the strong adhesion phenotypes resulting from expression of ALA1, EAP1 and members of the C. albicans ALS gene family, namely ALS1 and ALS5 in S. cerevisiae, support the hypothesis that the replacement of serines with leucines at CUG positions increases adhesion [36][37][38][39][40][41].
Finally, expression of S. cerevisiae tRNA Leu (pUA15) in C. albicans increased production of extracellular hydrolases, namely secreted aspartic proteinases (SAP) and phospholipases, as determined on agar plates supplemented with BSA and egg yolk, respectively ( Figure 5B,C). Hydrolysis of these substrates lead to formation of an halo of precipitated peptides around the colonies, indicative of SAP and phospholipase production [42]. Since adhesion, SAPs and phospholipases are important virulence attributes [43][44][45], those phenotypes are relevant to C. albicans pathogenesis and indicate, for the first time, that codon ambiguity and the Candida genetic code alteration may play a role in infection. It will be most interesting to put this hypothesis to experimental test as a positive result would clearly show that the negative impact of codon decoding ambiguity could be overcome by high selective potential of novel phenotypic diversity.

Gene expression alterations
In order to elucidate how CUG ambiguity generated phenotypic diversity we have carried out gene expression profiling of C. albicans cells expressing the S. cerevisiae tRNA Leu (pUA15 clones), using DNA microarrays. However, the diversity of phenotypes and permanent switching between phenotypes in culture ( Figure 4) created cell variability and culture instability that prevented us from obtaining meaningful mRNA expression profiles (data not shown). To overcome this limitation, we have tried to stabilize some of the phenotypes on solid media, but, once again, this turned out to be very difficult to achieve due to very high switching between morphological forms. As the mRNA profiles represented average values of a variety of phenotypes, reproducibility was low and most genes failed to pass our microarray statistical filters. Despite this, we were able to detect 4 genes whose expression was consistently altered in the microarray data sets and were relevant to the pUA15 phenotypes, namely the hypha-specific G1 cyclinrelated protein HGC1 (2.64 fold), the hypha specific gene HWP1 (41,76 fold), the white-opaque switch regulator WOR1 (7 fold) and the transcription factor MCM1 regulator of hyphal growth (-1.84 fold) ( Figure 6). Expression of these genes was confirmed by Real Time PCR, as described in materials and methods.
The WOR1 gene (7 fold up-regulated) is a master regulator of white-opaque switching and its expression induced the whiteopaque phase transition [46,47]. This provides a likely explanation Figure 2. Transfer RNAs used in this study. In order to reconstruct the early stages of the CUG identity alteration in C. albicans, S. cerevisiae leucine tRNAs containing the anticodons UAG or CAG were expressed in C. albicans. A) The respective tRNA genes were cloned into plasmid pUA12, which is based on the C. albicans pRM1 vector. B) A leucine tRNA gene containing the near-cognate anticodon (59-UAG-39) for the CUG codon was used as a low decoding efficiency tRNA (pUA13). C) Two tRNA CAG Leu genes, containing anticodons cognate for the CUG codon were used for higher CUG decoding efficiency, one contained G 33 (pUA14; medium decoding efficiency) and the other contained U 33 (pUA15; high decoding efficiency), in the anticodon-loop. D) The S. cerevisiae tRNA AGA Ser gene was used as negative control (pUA16). doi:10.1371/journal.pone.0000996.g002 for the strong induction of white-opaque switching in pUA15 clones, whose cells switch at high frequency. The HWP1 gene (41,7669,96 up-regulated) encodes a hyphal-specific cell wall mannoprotein, which is a substrate for mammalian transglutaminases and plays a crucial role in adhesion of C. albicans to epithelial cells [48]. Interestingly, HWP1 expression was correlated with MCM1 repression (-1.84 fold), confirming previous results that showed that expression of one of these genes represses expression of the other [49]. MCM1 plays an important role in cell morphology and its expression is auto-regulated by a feedback control mechanism. Since both low and high Mcm1p levels lead to hyphal formation, it may act as a recruiting regulatory factor for morphogenesis in C. albicans. Depletion of Mcm1p induced transcription of HWP1, however no Mcm1p binding site was found in the HWP1 promotor and it is not yet clear how the former activates transcription of the later. Finally, the HGC1 gene was also up-regulated (2.64 fold). This gene is crucial for hyphal formation by promoting apical bud elongation and it is strongly induced during morphogenesis. It is not required for expression of hypha-specific genes (HSGs), like HWP1, but is positively regulated by the cAMP/PKA pathways and repressed by Tup1 and Nrg1 morphogenesis repressors [50].

Increased CUG ambiguity increased C. albicans ploidy
Up-regulation of the WOR1 gene and the high percentage of opaque cells (mating competent) observed in pUA15 clones prompted the question of whether CUG ambiguity would induce mating in C. albicans. Indeed, pUA15 opaque cells formed conjugation tubes and mating figures which were readily observed by optical microscopy ( Figure 7A). Furthermore, flow cytometry analysis of pUA15 clones identified sub-populations of 4N, 6N and 8N cells ( Figure 7B), supporting the hypothesis that ambiguous C. albicans mated at high frequency. Since C. albicans is a diploid organism with a sexual life cycle [51], and considering that its mating locus (MTL) is heterozygotic in mating incompetent white cells (MTLa/a) and homozygotic in mating competent opaque cells (MTLa/a or MTLa/a) [52], the latter were isolated from pUA15 clones and were analysed for MTL homozygozity. All clones analysed were MTLa/a, suggesting that expression of tRNA Leu (pUA15) induced biased MTLa/a homozygoty ( Figure 7C), creating an excess of MTLa/a over MTLa/a cells. These results were in good agreement with up-regulation of the WOR1 gene as its expression is controlled by the MTLa1-a2 heterodimer, which, in turn, controls white-opaque transition and mating competence [46,47]. That is, switching from white-to-opaque phase required conversion of heterozygotic MTL to homozygotic MTL inducing mating competence [52]. Therefore, it is likely that MTL homozygoty induced by tRNA Leu derepressed the WOR1 gene, which, in turn, promoted the white-to-opaque transition and mating.
The presence of cells with high ploidy (6N and 8N) in pUA15 opaque cultures ( Figure 7B), prompted us to monitor ploidy variability. Most clones showed increased ploidy ranging from 4N to 8N (data not shown), however very large cells with remarkably high ploidy (.64N) were also observed ( Figure 8A). In general, ploidy variability between clones was higher than previously described [53,54] and raised the question of whether those high chromosome numbers could return to 2N over time. Since wild For this, total tRNAs were extracted under acidic conditions from pUA13, pUA14, pUA15, and pUA16 clones and fractionated on an acidic polyacrylamide gel, as described in materials and methods. These gels separated deacylated (-AA) from aminoacylated tRNAs (+AA), which were detected using a tDNA Leu/Ser probe labeled with [ 32 P]. B) Transformation efficiencies of plasmids encoding tRNA Leu , which decoded the C. albicans serine CUG codons as leucine, was significantly lower that that of control plasmids (pUA12 and pUA16), indicating that the leucine tRNAs were slightly toxic. C) However, clones that survived the transformation procedure adapted readily to CUG ambiguity and showed growth rates similar to control clones (pUA12). doi:10.1371/journal.pone.0000996.g003 type C. albicans undergoes a process of chromosome loss after mating, which decreases its ploidy from 4N back to 2N [55][56][57], we hypothesized that high ploidy in pUA15 clones could also be reduced. To test this hypothesis, clones with very high ploidy (large cells) were successively re-plated on fresh agar plates and their ploidy was monitored over time by flow cytometry. Indeed, ploidy reverted to 2N or 4N after several passages on fresh agar, confirming the above hypothesis ( Figure 8B). The above results also prompted us to check whether transformation of C. albicans with the pUA15 plasmid destabilized its karyotype. The C. albicans karyotype is characterized by frequent chromosome rearrangements, in particular of the chromosome R, which contains rDNA cistrons [58,59]. We wondered whether the tRNA Leu would affect rRNA metabolism and protein synthesis. Indeed, various rearrangements of the R-chromosome were readily observed ( Figure 8C). In particular, the size of R-chromosomes increased in some clones, decreased in others and these rearrangements affected most cells ( Figure 8C). It will now be most interesting to investigate whether CUG ambiguity affects ribosome assembly and the rate of protein synthesis.

DISCUSSION
The identity of CUG codons is variable in the genus Candida. Indeed, C. glabrata decodes CUGs as leucine, C. cylindracea changed their identity completely to serine and the other Candida species decode them ambiguously [13,25,60,61]. These differences in CUG decoding arose from differences in the structure of the tRNA CAG Ser , which is the only cognate tRNA for CUG codons in Candida. Indeed, the various tRNA CAG Ser have identity determinants for both SerRS and LeuRS [1]. For example, the C. albicans tRNA CAG Ser contains identity elements for the LeuRS, namely m 1 G 37 and the middle base of the anticodon (A 35 ), but the discriminator base at position 73 (G 73 ) is specific for SerRS and not for LeuRS (A 73 discriminator) [62]. This base is critical for the recognition of tRNAs by aminoacyl-tRNA synthetases (aaRSs) and one would expect that the LeuRS would not recognize tRNAs with G 73 . Therefore, charging the tRNA CAG Ser with leucine [25] suggests that the Candida LeuRS may have evolved a unique mode of recognition of its cognate tRNA Leu . Interestingly, the presence of a unique guanosine in the turn of the anticodon-loop (G-turn) of the tRNA CAG Ser (G 33 ), a conserved position occupied by U 33 (Uturn), reduced leucylation efficiency of the tRNA CAG Ser [25]. That is, recognition of the ancestral tRNA CAG Ser by the LeuRS was efficient and G 33 acted as a leucine identity anti-determinant. The reason for G 33 selection is not yet clear, but one possibility is that it may have decreased the toxicity of the mutant tRNA CAG Ser during the early stages of CUG identity alteration [27,63].
The dual recognition of the tRNA CAG Ser by the LeuRS and SerRS indicates that there are two forms of the tRNA CAG Ser in the cytoplasm of C. albicans, namely Ser-tRNA CAG Ser and Leu-tRNA CAG Ser , which are charged with serine and leucine, respectively. These 2 tRNAs generate ambiguity at CUG codons since they compete with each other for CUGs at the ribosome Asite. Interestingly, such CUG ambiguity was not constant over the 272625My of evolution of the genetic code alteration (see introduction) [30]. It was high during the early stages of CUG identity change (when the tRNA CAG Ser gene appeared), and decreased gradually due to elimination of the tRNA Leu gene from the genome of the Candida ancestor [30]. Reconstruction of the high level of CUG ambiguity, which existed during the early stages of CUG identity alteration, provided the first insight on how the genetic chaos created at the onset of CUG identity change may have generated phenotypic diversity of evolutionary and adaptive relevance. In extant C. albicans, morphological variation alters cell surface antigens and it is likely that this pathogen uses its sophisticated capacity to generate morphological variation as a strategy to escape the immune system. Furthermore, secreted proteinases and phospolipases are important C. albicans virulence attributes and their increased activity in pUA15 clones may also be relevant to virulence. It will now be interesting to engineer stable high level mistranslation in C. albicans and test the virulence of the recombinant strains in mice models. The high phenotypic diversity of pUA13, pUA14 and pUA15 clones and constant transition between phenotypes prevented us from carrying out a detailed study of the impact of CUG ambiguity on gene expression. However, the strong up-regulation of the hyphal specific gene (HWP1) and of the master regulator of the white-opaque transition (WOR1), may explain, at least in part, some of the phenotypes observed. As a hyphal specific gene (HSG), high HWP1 expression, induced by CUG ambiguity, is in agreement with spontaneous morphogenesis events that generate filamentous cell populations. Furthermore, HWP1 up-regulation may also increase adhesion since Hwp1p is a glycosylphosphatidylinositol cell wall adhesin (GPI-CWP) and mediates attachment of C. albicans cells to human endothelial and epithelial cells [44]. The observed adhesion phenotype ( Figure 5A) may have also resulted from the combined up-regulation of HWP1 and WOR1 genes, since the later, previously known as EAP2 (enhanced adhesion to polystyrene), mediates C. albicans and S. cerevisiae adhesion to polystyrene and epithelial cells [41].
Apart from its putative role in adhesion, up-regulation of WOR1 may also explain the white-opaque phenotype since Wor1p is a transcriptional regulator of white-opaque switching [46,47]. Indeed, Wor1p is present in very low amounts in white cells and accumulates in opaque cells. WOR1 is repressed by the heterodimer MTL a1/a2 and is activated by Wor1p itself when cells become homozygous MTL aa/aa or MTL aa/aa. Increased accumulation of Wor1p triggered white-opaque switching and repressed its own transcription by a feedback regulatory mechanism [46,47]. The up-regulation of the HGC1 gene ( Figure 6), which is a hypha-specific gene encoding a G1 cyclinrelated protein, that plays a role in hyphal morphogenesis, was also significant. Since it is transcriptionally regulated by hyphainducing rather than cell cycle signals [50], its up-regulation in ambiguous cells supports the hypothesis that it functions independently of other cell cycle cyclins.
CUG ambiguity also generated karyotype alterations and formation of polyploids and aneuploids. Ploidy-shift has been associated with chromosomal rearrangements [53,54,64] which also generates morphological variation [65][66][67], antifungal resistance [68], adaptation to alternative carbon sources [69,70], or even with homozigosity of chromosome-V [71]. This suggests that part of the phenotypic diversity observed in pUA13, pUA14 and pUA15 clones may have resulted from karyotype alterations, or from a combination of up-regulation of the genes described above Figure 5. Increased CUG ambiguity resulted in higher hydrolytic activity and increased adhesion. A) Highly ambiguous cells (pUA15) exhibited strong adhesion phenotypes both in solid and liquid media. Adhesion to the solid agar surface resulted from cell-cell and cell-agar adhesion. In liquid media, cells showed a strong flocculation phenotype and sedimented even when grown with agitation (30uC for 2 days). B-C) Cells transformed with pUA13-14 (data not shown) and with pUA15 plasmids, had higher SAP and phospholipase activity than control cells, as determined by hydrolysis of BSA and egg yolk, respectively. Hydrolytic activity was quantified by measuring precipitation zones formed around the colonies, corrected by the colony diameter, in order to obtain Pz values. doi:10.1371/journal.pone.0000996.g005 and karyotype destabilization. Since some clones did not have karyotype alterations ( Figure 8C), but still displayed phenotypic variability, it is likely that the former does not play a main role in expression of phenotypic diversity. However, one should not exclude the hypothesis that genome destabilization contributes to exposure of hidden phenotypes through ambiguous CUG decoding. Finally, in E. coli, genetic code ambiguity induced by misreading tRNAs triggered translational stress-induced mutagenesis (TSM), due to synthesis of error-prone DNA polymerases [72]. This general mistranslation resulted in increased global error rates during DNA replication leading to heritable genetic changes. Since these hypermutagenic phenotypes result in rapid adaptation, an unexpected consequence of genetic code ambiguity (and genetic code alterations) is acceleration of genome variability and fast evolution of new phenotypes. This may explain evolution of the high plasticity of Candida morphology and the very high heterozigosity of its genome [73].

Conclusions
Genetic code alterations pose important new biological questions whose answers remain elusive. It is now clear that a number of them evolved through codon decoding ambiguity, required significant structural change of protein synthesis machineries and reprogrammed codon usage [1,5,30]. However, codon decoding ambiguity is toxic, decreases fitness and may ultimately lead to cell death, as is the case in multicellular organisms [19,74]. For these reasons, evolution of genetic code alterations through codon ambiguity is most puzzling. This study unveiled possible ways of overcoming the negative impact of codon ambiguity by high selective potential generated through generation of phenotypic diversity. The molecular mechanism used to generate such phenotypic diversity remains to be elucidated. However, the adaptive potential of the unveiled phenotypes strongly suggests that CUG ambiguity may have been preserved in Candida spp. as a novel generator phenotypic diversity. We have previously shown that codon ambiguity in S. cerevisiae creates a competitive edge under stress by inducing the general stress response and preadapting cells to sudden environmental changes [75]. Therefore, the toxicity of codon ambiguity is not an impediment to codon identity redefinition, supporting the hypothesis that codon misreading plays a critical role in the evolution of genetic code alterations and genetic code expansion.

Plasmids
A multi-cloning site was inserted (NruI/EcoRV) in the low-copy C. albicans vector pRM1 [76]. The resulting vector was named pUA12. For heterologous expression of the S. cerevisiae tRNAs genes in CAI4, a genomic DNA fragment containing S. cerevisiae tRNA gene (700 bp), previously amplified by PCR, was cloned into the multi cloning site of pUA12 plasmid. S. cerevisiae Leu-tRNA UAG and a 300 bp flanking region, upstream and downstream of the gene, was amplified with the following set of primers: 59-CCGCTCGAGC-GGCGACTGTCCAGACTTAGTAAAG CT-39 and 59-GCTC-TAGAGCCCGCTGTCGCCAGCGTTAGC-39. Genomic DNA containing S. cerevisiae tRNA GAG Leu gene was used as a template for PCR amplification using the forward primers: 59-GCTATGGGC-CCGCCTCCGGGTAGTTGCAACGGTACTCTGG CCGAG-TGGTCTAAGGCG-39 and 59-GCTATGGGCCCTAGTTGCA-ACGG TATCTGGCCGAGTGGTCTAAGGCGTCAGGTTC-AGGTCC-39; and reverse primer 59-ATGCATAAAAACAAAA-TTTGTTGAAA-39. These primers introduced a mutation in the first position of the anticodon changing it from 59-GAG-39 to 59-CAG-39 and allowed insertion of G or T at position 33 of the tRNA anticodon-loop. Upstream of this gene, a 250 bp fragment with the same sequence of the 59 flanking C. albicans Ser-tRNA CAG gene, was also inserted. This fragment was amplified by PCR from a XhoI/ApaI genomic DNA fragment, with the following primers: 59-

Plasmid transformation
Transformation of E. coli was carried out as described by [77] CAI4 transformation was performed by the spheroplast method as described in the Manual for Preparation and Transformation of Pichia pastoris Spheroplasts (version A, Invitrogen).

Northern Blot analysis
Acidic Northern Blot analysis was performed as described by . For total tRNA extractions, 250 ml cultures grown overnight in YEPD or MM-uri medium were harvested at an OD 600 of 0.7-0.9 and the pellets were frozen at 270uC overnight. Cells were resuspended in 5 ml lysis buffer (0.3M sodium acetate, pH 4.5, 10 mM EDTA), 1 vol. phenol equilibrated with sodium citrate pH 4.5 and baked glass beads. Cell suspension was vigorously shaken .30 seconds and incubated on ice for periods longer than 30 seconds, this procedure was repeated 8 times [78]. The aqueous phase containing RNAs was separated from the phenolic phase by centrifugation at 32006g for 20 min at 4uC and then transferred to a new Falcon tube and reextracted with fresh phenol. Aqueous phases containing RNAs were harvested by centrifugation at 32006g for 20 min at 4uC and applied to a 20 ml DEAE-cellulose column equilibrated with 0.1 M sodium acetate pH 4.5. tRNAs were eluted with 0.1 M sodium acetate/1 M sodium chloride and precipitated with 2.5 vol. ethanol, resuspended in 10 mM sodium acetate pH 5.0/1 mM EDTA, and stored a 220uC. The deacylated tRNAs were obtained by incubation at 37uC for 1 h in 1 M Tris pH 8, 1 mM EDTA buffer [26].
tRNAs were fractionated at 4uC in 7.5% acrylamidde-8 M urea (30 cm long, 0.8 mm thick), buffered with 0.1 M sodium acetate pH 5.0. The 7.5% acidic gels were run at 300 V until bromophenol blue dye reached the bottom [31]. Fractionated tRNAs were transferred to nitrocellulose membranes (Hybond N, Amersham). Membranes were pre-hybridized for 6 h in a Hybridization oven at 50uC in 50% formamide, 56SSC, 1% SDS, 0.04% Ficoll, 0.04% polyvinylpyrrolidone and 250 mg/ml sheared salmon sperm DNA [79]. Hybridization was performed overnight in the above buffer using probes labelled with [a-32 P]dCTP by PCR, using standard protocols [80], except that the amount of dCTP was reduced from 100 to 50 mM and 5 nmol (30 mCi) 6000 Ci/ mmol [a-32 P]dCTP was added to the reaction mixture. In order to decrease the background level of free radioactivity, 50 PCR cycles were performed to decrease the amount of non-incorporated [a-32 P]dATP. Membranes were washed at low stringency in 16SSC, 0.5% SDS at 50uC or at high stringency in 0.16SSC, Figure 7. Ambiguous CUG decoding induced mating. A) In pUA15 transformed cells, the number of opaque cells was very high. This phenotype is most likely explained by up-regulation of the white-opaque master regulator WOR1 gene ( Figure 6). Since C. albicans white cells (most common form) are mating incompetent and opaque cells (rare cells) are mating competent, we have verified whether pUA15 opaque cells formed conjugation tubes and mating figures in liquid culture. Both were readily observed using optical microscopy (white arrows). B) In order to confirm that mating occurred, the DNA content of pUA15 cells was analyzed by flow cytometry. Since C. albicans is diploid, 4N cells were expected. Surprisingly, higher ploidies (6N, 8N) were also observed suggesting that the cultures had significant number of aneuploid and poliploid cells. C) White to opaque transition and mating induced by CUG ambiguity occurs due to MTL homozygosity. Since mating requires transition from the heterozygotic mating locus (MTL a/a) found in white cells, to the homozygotic configuration (MTLa/a or MTLa/a) found in opaque cells, detection of aa/aa cells supported the hypothesis that CUG ambiguity induced mating. doi:10.1371/journal.pone.0000996.g007 0.5% SDS at 65uC for 1 h. The membranes were exposed overnight with intensifying screens and developed using a Molecular Imager FX (Biorad).

Switching frequencies and phenotypic characterization
C. albicans grown overnight at 30uC in MM-uri were serially diluted to 1000 cells per ml. Approximately 50 cells were plated onto fresh agar plates and then allowed to grow at 30uC for 7 days in a humidified incubator to prevent drying of the agar surface. Sectored colonies displaying atypical colonies were scored and the data was analyzed for statistical significance using ANOVA. Colonies were photographed using a Stemi 2000-C dissecting microscope equipped with AxioCam HRc camera and Axio Vision Software from Zeiss. Cells were photographed using a Zeiss MC80 Axioplan 2 light microscope.

Real Time RT-PCR
Total RNA was prepared from C. albicans using hot acid phenol [81]. First-strand cDNA synthesis was carried out using the Superscript II RT kit from Invitrogen and its quantification was carried out in an Applied Biosystems 7500 Real Time PCR system using the SYBR Green I dye quantification assay (Power SYBR Green PCR master Mix). Primer concentrations were tested (0,2 to 0,4 mM) to ensure the lowest threshold cycle (C T ) and the highest signal magnitude against the target template and to ensure non-specific product formation, resulting from primer dimmerization. After amplification, reactions were checked for presence of non-specific products through dissociation curve analysis. Each gene was quantified using 9 replicas of both control (pUA12) and ambiguous (pUA15) clones and a mean value was calculated. Outliers were rejected using critical values of Dixon's ''Q'' parameter at 95% confidence level [82].
Determination of extra cellular hydrolytic activity C. albicans strains were screened for the production of extra cellular phospholipase and secreted aspartic proteinase activity by growing cells on MM-uri agar supplemented with 10% egg yolk (Merck) and 10% bovine albumin (Sigma), respectively. A 362 ml suspension of 10 7 cells/ml in PBS was plated on the surface of the agar medium and left to dry at room temperature. The culture Figure 8. Ambiguous CUG decoding induced karyotype rearrangements and ploidy-shift. A) In ambiguous cell lines (pUA15) polyploidy was predominant and very high ploidy was often detected (.32N). Aneuploidy was also observed (6N). B) However, after plating cells several consecutive times on fresh agar chromosome numbers were reduced indicating that most cells returned to low ploidy (2N or 4N). Ploidy reduction after mating normally occurs by chromosome loss in C. albicans and it is likely that such mechanism also played a role in ploidy reduction in pUA15 transformed cells. C) CUG ambiguity also promoted extensive rearrangements of the R-chromosome (highlighted in white circles). Chromosomes were separated by PFGE on 0.6% agarose gels under the following conditions: 120-300 s for 24h at 80 V, then 420-900 s for 48 h at 80 V. The numbers 1-7 and R identify C. albicans chromosomes. doi:10.1371/journal.pone.0000996.g008 was then incubated at 30uC for 3 days, after which the diameter of the colony and the precipitation zone around the colony was determined. Three different clones of pUA12, pUA13, pUA14 and pUA15 transformed cells were tested twice. The experiment was carried out on two different occasions. The extra cellular hydrolase activity was calculated using the formula [(1/Pz)-1], where Pz value represents the hydrolyse zone, i. e., the cloudy-zone-around-plus-colony diameter divided by the colony diameter [83]. Data obtained was submitted to an ANOVA statistical test.

Genome analysis
DNA content of C. albicans cells was determined using FACS analysis [84]. Karyotype analysis was performed using Pulsed-Field Gel Electrophoresis (PFGE) [85].

ACKNOWLEDGMENTS
We are most grateful to Jorge Rino for helping with the light microscopy studies, Alexander Johnson for providing oBP primers and Concha Gil for the pRM1 plasmid.