(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



Identification of a non-canonical ciliate nuclear genetic code where UAA and UAG code for different amino acids [1]

['Jamie Mcgowan', 'Earlham Institute', 'Norwich Research Park', 'Norwich', 'United Kingdom', 'Estelle S. Kilias', 'Department Of Biology', 'University Of Oxford', 'Oxford', 'Elisabet Alacid']

Date: 2023-10

The genetic code is one of the most highly conserved features across life. Only a few lineages have deviated from the “universal” genetic code. Amongst the few variants of the genetic code reported to date, the codons UAA and UAG virtually always have the same translation, suggesting that their evolution is coupled. Here, we report the genome and transcriptome sequencing of a novel uncultured ciliate, belonging to the Oligohymenophorea class, where the translation of the UAA and UAG stop codons have changed to specify different amino acids. Genomic and transcriptomic analyses revealed that UAA has been reassigned to encode lysine, while UAG has been reassigned to encode glutamic acid. We identified multiple suppressor tRNA genes with anticodons complementary to the reassigned codons. We show that the retained UGA stop codon is enriched in the 3’UTR immediately downstream of the coding region of genes, suggesting that there is functional drive to maintain tandem stop codons. Using a phylogenomics approach, we reconstructed the ciliate phylogeny and mapped genetic code changes, highlighting the remarkable number of independent genetic code changes within the Ciliophora group of protists. According to our knowledge, this is the first report of a genetic code variant where UAA and UAG encode different amino acids.

The genetic code is almost universal across life. The vast majority of organisms use the canonical genetic code, which has three stop codons (UAA, UAG, and UGA) and 61 sense codons that code for amino acids. Here, we report the discovery of an unexpected genetic code variant in an uncultured ciliate species from the Oligohymenophorea class, where the canonical stop codons UAA and UAG have been reassigned to code for lysine and glutamic acid, respectively. This is a particularly unusual genetic code reassignment as UAA and UAG differ at the wobble position and their evolution is thought to be coupled. We also report that the remaining stop codon, UGA, is enriched immediately downstream of genes in the same reading frame, suggesting a possible role in minimising deleterious consequences in the event of translational readthrough. Our work documents, for the first time, a genetic code variant where the codons UAA and UAG specify two different amino acids and shows that there are still unexplored genetic code reassignments awaiting discovery.

Funding: This work was funded by Wellcome though the Darwin Tree of Life Discretionary Award (218328 to NH and TAR) and supported by the Biotechnology and Biological Sciences Research Council (BBSRC), part of UK Research and Innovation, through the Core Capability Grant (BB/CCG1720/1 to NH), the National Capability in Genomics and Single Cell Analysis (BBS/E/T/000PR9816 to DS) and the National Capability in e‐Infrastructure (BBS/E/T/000PR9814 to NH) at the Earlham Institute. TAR is supported by a Royal Society University Research Fellowship (URF/R/191005 to TAR). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Here, we report the discovery of a novel variant of the genetic code in a ciliate belonging to the Oligohymenophorea class, where the meaning of the UAA and UAG codons have changed to specify different amino acids. Using G&T-Seq [ 28 ], we performed parallel genome and transcriptome sequencing of small pools of ciliate cells. Combining genome and transcriptome sequencing data from multiple independently amplified samples enabled co-assembly of a highly complete macronuclear genome assembly and annotation. Genomic and transcriptomic analysis revealed that the UAA codon has been reassigned to specify lysine, while the meaning of the UAG codon has changed to specify glutamic acid. We identified multiple suppressor tRNA genes of both types in the genome, supporting the genetic code changes. We show that UGA codons are significantly enriched in the 3’-UTR of genes suggesting that there is selective pressure to maintain tandem stop codons, which may play a role in minimising erroneous protein elongation in the event of translational readthrough. To our knowledge, this is the first report of a genetic code variant where UAA and UAG specify different amino acids.

In virtually all genetic code changes reported to date, the codons UAA and UAG have the same meaning, i.e., they are either both used as canonical stop codons or are both reassigned to the same amino acid [ 25 ]. This suggests that evolutionary or mechanistic constraints couple the meaning of these two codons [ 26 ]. One such constraint is wobble binding of a suppressor tRNA gene with a UUA anticodon, where uracil in the first anticodon position can bind to either adenine or guanine in the third codon position of mRNA [ 27 ]. Thus, acquiring a suppressor tRNA gene with a UUA anticodon could potentially change the meaning of both the UAA and UAG codons. Wobble binding has been experimentally demonstrated in Tetrahymena thermophila, where tRNA-Sup(UUA) was shown to suppress both the UAA and UAG codons, whereas tRNA-Sup(CUA) suppressed only the UAG codon [ 16 ]. The first report of nuclear genetic code variants where UAA and UAG have different meanings were reported in transcriptomics analyses where a Rhizarian species (Rhizaria sp. exLh) was shown to use UAG to encode leucine and in a Fornicate (Iotanema spirale) where UAG has been reassigned to glutamine [ 26 ]. However, in both cases, the UAA codon was retained as a stop codon, thus avoiding the problem of genetic code ambiguity due to wobble binding.

Several models have been proposed to describe genetic code changes. Under the “codon capture” model, a codon that is rarely used (e.g., due to GC content) is gradually eliminated from the genome followed by loss of the corresponding unused tRNA [ 21 ]. Due to random genetic drift the codon could reappear and be captured by a noncognate tRNA charged with a different amino acid, thus changing the genetic code. Such a process would be essentially neutral, not resulting in mistranslated protein products as the codon is eliminated from genes before the change in meaning occurs [ 17 ]. Alternatively, under the “ambiguous intermediate” model [ 22 ], reassignment of a codon takes place via an intermediate stage, where a codon is ambiguously translated via competing tRNAs charged with different amino acids, or in the context of stop codon reassignment, a suppressor tRNA competing with a release factor. This process would be driven by selection and result in the elimination of the cognate tRNA if the new meaning is advantageous. The “genome streamlining" model is more relevant to small genomes (e.g., organellar genomes or parasites) where there is pressure to minimise translational machinery [ 23 ]. More recently the “tRNA loss driven codon reassignment” mechanism was proposed to describe codon reassignments whereby tRNA loss, or alteration of release factor specificity, results in an unassigned codon that can be captured by another tRNA gene [ 24 , 25 ].

Tandem stop codons are additional stop codons located in the 3’-UTR within a few positions downstream of a gene in the same reading frame [ 18 ]. They are thought to act as “back-up” stop codons in the event of readthrough, minimising the extent of erroneous protein elongation. For example, in yeast there is a statistical excess of stop codons in the third in-frame codon position downstream of genes with a UAA stop codon [ 18 ]. Tandem stop codons have been shown to be overrepresented in ciliates that only use UGA as a stop codon, compared to eukaryotes that use the canonical genetic code [ 19 ]. The level of overrepresentation is greater in highly expressed genes [ 20 ]. Tandem stop codons are thought to be particularly important in ciliates where, following stop codon reassignment, readthrough events might occur at a higher frequency due to mutations in eRF1 [ 20 ].

Changing the meaning of codons from stop to sense requires modifications to the translational apparatus. In eukaryotes, the eukaryotic release factor 1 (eRF1) protein recognises the three standard stop codons in mRNA and triggers translation termination. Studies have shown that mutations in the N-terminus of eRF1 can alter stop codon specificity [ 8 , 13 – 15 ]. eRF1 specificity to recognise only the UGA codon has evolved independently via different molecular mechanisms at least twice in ciliates with reassigned UAA and UAG codons [ 14 ]. Acquisition of tRNA genes with anticodons that recognise canonical stop codons (suppressor tRNAs), via mutations, base modifications or RNA editing enables translation of canonical stop codons into amino acids [ 16 , 17 ].

Known genetic code changes in ciliates involve reassignment of one or more stop codons to specify for amino acids. Most reported ciliate genetic code changes involve reassignment of both the UAA and UAG codons to specify glutamine as in Tetrahymena, Paramecium, and Oxytricha [ 8 ], or glutamic acid in Campanella umbellaria and Carchesium polypinum or tyrosine in Mesodinium species [ 9 ]. Other known modifications include reassignment of the UGA stop codon to specify tryptophan in Blepharisma [ 8 ], or cysteine in Euplotes [ 10 ]. The most extreme example of genetic code remodelling is found in Condylostoma magnum where all three UAA, UAG, and UGA stop codons have been reassigned and can specify either an amino acid (glutamine for UAA and UAG, and tryptophan for UGA) or signal translation termination depending on their proximity to the mRNA 3’ end [ 9 , 11 ]. Not all ciliates use non-canonical genetic codes. For example, Fabrea salina, Litonotus pictus, and Stentor coeruleus use the canonical genetic code [ 9 , 12 ].

The genetic code is one of the most conserved features across life, emerging before the last universal common ancestor [ 1 ]. Virtually all organisms use the canonical genetic code which has three stop codons (UAA, UAG, and UGA) and 61 sense codons that specify one of 20 amino acids, including a translation start codon (AUG). Variants of the genetic code, while rare, have been reported in several lineages of bacteria, viruses, and eukaryotic organellar and nuclear genomes [ 2 , 3 ]. Ciliate nuclear genomes are a particular hotspot for genetic code variation. The phylum Ciliophora is a large group of single-celled eukaryotes (protists) that diverged from other Alveolates more than one billion years ago [ 4 ]. Ciliates are highly unusual in that they exhibit nuclear dimorphism whereby each cell has two types of nuclei, the germline micronucleus (MIC) and the somatic macronucleus (MAC), each of which contains its own distinct genome structure and function [ 5 ]. The MIC genome functions as the germline genome and is exchanged during sexual reproduction. MIC genomes are typically diploid and are transcriptionally inactive during vegetative growth. The MIC genome undergoes rearrangement and excision of micronucleus-limited sequences to serve as a template to generate the transcriptionally active MAC genome [ 6 ]. MAC genomes typically contain short, fragmented, gene-dense chromosomes that are present at high ploidy levels (up to tens of thousands of copies) [ 7 ].

Results & discussion

Genome assembly of an oligohymenophorean ciliate We isolated a novel ciliate species Oligohymenophorea sp. PL0344 from a freshwater pond at Oxford University Parks, Oxford, UK. Attempts to establish a stable long-term culture were unsuccessful so we applied low input single-cell based approaches to generate genomic and transcriptomic data. Small pools of cells (5–50 cells) were sorted into a microplate using fluorescence-activated cell sorting (FACS). Parallel genome and transcriptome sequencing was performed using G&T-Seq, which relies on whole genome amplification using multiple displacement amplification (MDA) and transcriptome analysis using a modified Smart-seq2 protocol [28]. A de novo genome assembly was generated by co-assembling reads from 10 samples (totalling approximately 6 Gb). Following manual curation and removal of contaminant sequences, the resulting macronuclear genome assembly was 69.7 Mb in length, contained in 3671 scaffolds with an N50 of 59.6 Kb (Table 1). Approximately 89% of the corresponding RNA-Seq reads mapped to the genome assembly, indicating high completeness. GC content of the genome is low at 30.6% (Table 1), which is similar to previously sequenced ciliate genomes [12]. The mitochondrial genome was also recovered which is a linear molecule 35,635 bp in length with GC content of 25.33% and capped with repeats. The mitochondrial genome contains the small subunit (SSU) and large subunit (LSU) ribosomal RNA (rRNA) genes, 5 tRNA genes, 19 known protein-coding genes, and 13 open reading frames. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Table 1. Genome Assembly and Annotation Statistics. https://doi.org/10.1371/journal.pgen.1010913.t001 The nuclear encoded SSU rRNA gene sequence is 99.81% identical to an environmental sequence (AY821923) of an unnamed ciliate in the GenBank database, isolated from Orsay, France [29]. Maximum-likelihood phylogenetic analysis of the SSU rRNA gene placed it within a clade containing four unnamed ciliate species (AY821923, HQ219368, LR025746, HQ219418) and Cinetochilum margaritaceum (MW405094) with 100% bootstrap support (S1 Fig). Thus, based on the SSU rRNA gene, C. margaritaceum is the closest related named species. The SSU rRNA gene of C. margaritaceum is 96.03% identical to that of Oligohymenophorea sp. PL0344. C. margaritaceum belongs to the Loxocephalida order (Class Oligohymenophorea; Subclass Scuticociliatia), which is considered a controversial order due to its non-monophyly [30,31]. Our phylogenetic analysis places C. margaritaceum as a divergent branch relative to other members of Loxocephalida (S1 Fig), which is congruent with previous analyses [30,31], suggesting taxonomic revision is required.

Oligohymenophorea sp. PL0344 uses a novel genetic code Preliminary analysis of the genome sequence revealed that many coding regions contained in-frame UAA and UAG codons. Consistent with codon reassignments in other ciliate species, this suggested that the UAA and UAG stop codons have been reassigned to code for amino acids. Surprisingly however, the meanings of these codons do not match any known genetic code. An example gene (tubulin gamma chain protein), showing six in-frame UAA codons and six in-frame UAG codons, translated and aligned to orthologous protein sequences with representatives from across Eukaryota is displayed in Fig 1. Five in-frame UAA codons correspond to highly conserved columns in the alignment where lysine is the consensus amino acid (Fig 1). Four in-frame UAG codons correspond to highly conserved columns in the alignment where glutamic acid is the consensus amino acid, and another corresponds to a column with a mix of glutamic acid and aspartic acid (Fig 1). PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 1. Genetic code change in Oligohymenophorea sp. PL0344. Example multiple sequence alignment of a tubulin gamma chain protein and orthologous sequences spanning Eukaryota. The alignment has been trimmed for visualisation purposes to remove poorly conserved regions and highlight internal UAA and UAG codons. https://doi.org/10.1371/journal.pgen.1010913.g001 We used two complementary tools to analyse the genetic code further. First, we used the “genetic_code_examiner” utility from the PhyloFisher package [32], which predicts the genetic code by comparing codon positions in query sequences to highly conserved (> 70% conservation) positions in amino acid alignments from a database of 240 orthologous protein sequences. PhyloFisher identified 58 genes with 87 in-frame UAA codons that correspond to highly conserved amino acid sites. Of these, 74 UAA codons (85%) correspond to highly conserved lysine residues (Fig 2A). The second most numerous match was to arginine, another positively charged amino acid, with 9 (10%) hits. PhyloFisher identified 46 genes with 63 in-frame UAG codons that correspond to highly conserved amino acid sites. Of these, 56 UAG codons (89%) correspond to highly conserved glutamic acid residues (Fig 2B). The second most numerous match was to aspartic acid, another negatively charged amino acid, with 4 (6%) of hits. Amongst the genes identified by PhyloFisher, 27 contained both in-frame UAA codons and an in-frame UAG codons. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 2. Genetic code prediction for Oligohymenophorea sp. PL0344. PhyloFisher genetic code prediction for the (A) UAA and (B) UAG codons using the PhyloFisher database of 240 orthologs. Only well conserved (>70%) amino acids are considered. Colours correspond to amino acid properties and match the multiple sequence alignment in Fig 1. (C) Codetta genetic code prediction. Log decoding probabilities for the UAA and UAG codons are shown for each of the 20 standard amino acids. https://doi.org/10.1371/journal.pgen.1010913.g002 We also analysed the genetic code using Codetta [33,34]. Codetta predicts the genetic code by aligning profile hidden Markov models (HMMs) from the Pfam database against a six-frame translation of the query genome assembly. The meaning of each codon is inferred based on emission probabilities of the aligned HMM columns. From the whole genome sequence, 14,633 UAA codons and 10,160 UAG codons had a Pfam position aligned. Based on these alignments, Codetta also predicted that the UAA codon is translated as lysine and UAG translated as glutamic acid, each with a log decoding probability of zero (Fig 2C). Thus, these results indicate that Oligohymenophorea sp. PL0344 uses a novel genetic code where UAA is translated as lysine and UAG is translated as glutamic acid. This is the first time this genetic code variant has been reported. Furthermore, according to our knowledge, this is the first report of a genetic code variant where UAA and UAG have been reassigned to specify different amino acids. Genetic code variants were previously reported where UAG was reassigned to specify an amino acid (either leucine or glutamine) but UAA was retained as a stop codon in both cases [32]. This is significant as it suggests that the genetic code variant reported herein has overcome mechanistic constraints linking the translation of these two codons.

Suppressor tRNA genes tRNA genes were annotated using tRNAscan [35], resulting in the annotation of 320 tRNA genes, including 15 that are predicted to be pseudogenes. Amongst the annotated tRNA genes are 23 putative suppressor tRNA genes. These are tRNA genes with anticodons complementary to canonical stop codons (UAA, UAG, or UGA). The annotated suppressor tRNA genes include 12 tRNA-Sup(UUA) genes and 10 tRNA-Sup(CUA) genes. tRNAscan also predicted a single tRNA-Sup(UCA) gene, however this was low scoring and was not predicted by ARAGORN [36], an alternative tool to identify tRNA genes. tRNAscan also predicts the function of tRNAs. Many of the tRNAscan isotype predictions were consistent with the predicted genetic code (i.e., UAA = lysine and UAG = glutamic acid), however several putative tRNA genes had low-scoring or inconsistent isotype predictions. To better characterise the suppressor tRNA genes, we compared their sequences to the non-suppressor tRNA genes. Eight of the twelve predicted tRNA-Sup(UUA) genes were most similar to tRNA-Lys genes with UUU or CUU anticodons (68.49% to 80.95% identical) (S1 Table), consistent with the genetic code prediction that UAA has been reassigned to specify lysine. An example tRNA-Sup(UUA) predicted to function as a lysine tRNA is shown in Fig 3A. All ten tRNA-Sup(CUA) genes were most similar to tRNA-Glu genes with CUC or UUC anticodons (69.44% to 93.06% identical) (S1 Table), consistent with the genetic code prediction that UAG has been reassigned to specify glutamic acid. An example tRNA-Sup(CUA) predicted to function as a glutamic acid tRNA is shown in Fig 3B. Similarly, analysis using phylogenetic networks clusters most of the suppressor tRNA genes with tRNA genes of their predicted function (S2 Fig). We also identified a tRNA gene for selenocysteine, tRNA-SeC(UCA) (Fig 3C), suggesting that the UGA codon is used both as a stop codon and to encode selenocysteine. Thus, all 64 codons can specify amino acids as has been reported in other ciliate genomes [37]. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 3. Example tRNA Genes. (A) Predicted secondary structure of an example tRNA-Sup(UUA) predicted to function as a lysine tRNA. The wobble position is highlighted. According to wobble-binding rules, uracil at this position can bind to either adenine or guanine in the third codon position of mRNA, allowing the suppressor tRNA to recognise both UAA and UAG stop codons. (B) Predicted secondary structure of an example tRNA-Sup(CUA) predicted to function as a glutamic acid tRNA. (C) Predicted secondary structure of the tRNA-SeC(UCA) for selenocysteine. https://doi.org/10.1371/journal.pgen.1010913.g003 UAA and UAG codons differ only in the wobble position. According to wobble-binding rules, uracil in the first tRNA anticodon position (“wobble position”) (Fig 3A) can bind to either adenine or guanine in the third codon position of mRNA [27], allowing tRNA with a UUA anticodon to recognise both UAA and UAG codons. It has been experimentally demonstrated that T. thermophila tRNA-Sup(UUA) can recognise both UAA and UAG codons [16]. It has been suggested that wobble binding is a possible explanation as to why UAA and UAG virtually always have the same meaning [9]. Considering that Oligohymenophorea sp. PL0344 has tRNA-Sup(UUA) genes for lysine and tRNA-Sup(CUA) genes for glutamic acid, this raises the question: are UAG codons ambiguously translated as both glutamic acid and lysine? If not, how has it overcome the mechanistic and evolutionary constraints that appear to couple the translation of these two codons? Presumably, if wobble binding allows tRNA-Sup(UUA) to recognise the UAG codon, it would be less efficient than tRNA-Sup(CUA) and outcompeted, possibly resulting in some degree of stochastically translated protein products with glutamic acid residues substituted by lysine at UAG codon positions. Attempts to establish a stable culture were unsuccessful, and while we can overcome this problem to generate a genome assembly using low-input sequencing methods designed for single-cell analysis, such low-input approaches are not available for proteomics. Without proteomics data, it is not possible to determine if UAG is ambiguously translated. Additionally, while the genomic and transcriptomic data provide strong evidence that lysine and glutamic acid are the major translation products of UAA and UAG codons, respectively, we cannot rule out the possibility that other amino acids are (mis)incorporated at these sites which could be detected using mass-spectrometry [38,39]. Furthermore, from suppressor tRNA gene sequences alone, it is not possible to determine if they incorporate modified nucleotides which could alter codon-anticodon binding specificity.

Genome annotation and codon usage analysis Genome annotation incorporating RNA-Seq data and protein alignments from other ciliates resulted in the annotation of 22,048 transcripts from 20,141 gene models (Table 1). BUSCO analysis estimates that the genome annotation is highly complete with 94.7% of BUSCO genes recovered as complete, which compares favourably to other high quality ciliate genomes (S2 Table). The median intron size of 57 bp (Table 1) is similar to previously sequenced ciliate genomes, such as Tetrahymena thermophila and Oxytricha trifallax [7,37] but not as short as the extremely short introns (15–25 bp) found in Stentor coeruleus or Paramecium tetraurelia [12,40]. We defined a subset of genes as “highly expressed” based on the 10% of genes with the highest transcripts per million (TPM) values for comparison below. Codon usage is biased towards using codons with lower GC content. This bias is reduced in highly expressed genes which have higher GC content compared to all genes (38.51% versus 34.12%), similar to previous reports in Paramecium and Tetrahymena [37,41]. The reassigned codons are widely used across genes with 95.9% of genes containing both a UAA codon and a UAG codon. However, their usage is reduced in highly expressed genes (S3 Table). Reduced codon usage in highly expressed genes could indicate translational inefficiency, or that selective pressure to retain canonical lysine and glutamic acid codons is higher in highly expressed genes. Very little is known about translation termination efficiency in ciliates. This is particularly interesting for ciliates that use only UGA as a stop codon, as UGA is known to be the least robust stop codon and the most prone to translational readthrough [42]. The sequence composition surrounding a stop codon influences the rate of stop codon readthrough. The nucleotide immediately downstream of a stop codon (+4 position) is particularly important, with several studies demonstrating that presence of a cytosine following UGA substantially increases the rate of readthrough [43–45]. Interestingly, examining the sequence composition surrounding stop codons in Oligohymenophorea sp. PL0344, cytosine appears to be avoided following the stop codon (S3 Fig). This is particularly noticeable in highly expressed genes (S3 Fig) where the proportion of genes with a cytosine following UGA is significantly reduced (chi-squared test, p-value = 7.3e-10). This trend has also been observed in P. tetraurelia and T. thermophila [41]. Tandem stop codons potentially play an important role as “back-up” stop codons, minimising the extent of protein elongation in the event of readthrough [18]. Here, we analysed tandem stop codons by counting UGA codons in the first 20 in-frame codon positions downstream of genes. Our results show that UGA codons are significantly overrepresented (chi-squared test, p-value < 0.05) in the first four in-frame codons downstream of genes (Fig 4). 12.3% of genes have at least one UGA codon within the first six in-frame codon positions downstream of genes, similar to the proportion reported for T. thermophila (11.5%) where UAA and UAG have also been reassigned to encode amino acids [19]. For comparison, the reassigned UAA and UAG codons are not overrepresented in this region. The frequency of UGA codons at these positions is greater for highly expressed genes whereby 13.6% of highly expressed genes have at least one UGA codon within the first six in-frame codon positions downstream of genes (Fig 4). These data add support that there is selective pressure for ciliates with reassigned UAA and UAG codons to maintain tandem UGA stop codons at the beginning of the 3’-UTR. It is tempting to speculate that these additional UGA stop codons play a role in minimising deleterious consequences of readthrough events. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 4. Enrichment of tandem stop codons. The proportion of codon positions occupied by UGA in the 20 in-frame codon positions immediately downstream of all genes and highly expressed genes. Positions where UGA is significantly overrepresented (chi-squared test, p-value < 0.05) are indicated with an asterisk. https://doi.org/10.1371/journal.pgen.1010913.g004

[END]
---
[1] Url: https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1010913

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.