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



The transcription factor DUX4 orchestrates translational reprogramming by broadly suppressing translation efficiency and promoting expression of DUX4-induced mRNAs [1]

['Danielle C. Hamm', 'Human Biology Division', 'Fred Hutchinson Cancer Center', 'Seattle', 'Washington State', 'United States Of America', 'Ellen M. Paatela', 'Molecular', 'Cellular Biology Program', 'University Of Washington']

Date: 2023-10

Translational control is critical for cell fate transitions during development, lineage specification, and tumorigenesis. Here, we show that the transcription factor double homeobox protein 4 (DUX4), and its previously characterized transcriptional program, broadly regulates translation to change the cellular proteome. DUX4 is a key regulator of zygotic genome activation in human embryos, whereas misexpression of DUX4 causes facioscapulohumeral muscular dystrophy (FSHD) and is associated with MHC-I suppression and immune evasion in cancer. We report that translation initiation and elongation factors are disrupted downstream of DUX4 expression in human myoblasts. Genome-wide translation profiling identified mRNAs susceptible to DUX4-induced translation inhibition, including those encoding antigen presentation factors and muscle lineage proteins, while DUX4-induced mRNAs were robustly translated. Endogenous expression of DUX4 in human FSHD myotubes and cancer cell lines also correlated with reduced protein synthesis and MHC-I presentation. Our findings reveal that DUX4 orchestrates translational reprogramming by suppressing the cellular proteome while maintaining translation of DUX4-induced mRNAs to promote an early developmental program.

Competing interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: SJT and DCH are co-inventors on a patent application submitted by the Fred Hutchinson Cancer Center that covers research presented here. Other authors declare that they have no competing interests."

Funding: This research was supported by the Flow Cytometry Shared Re-source, RRID:SCR_022613, and the Genomics & Bioinformatics Shared Resource, RRID:SCR_022606, of the Fred Hutch/University of Washington/Seattle Children’s Cancer Consortium (P30CA015704). This work was supported by grants from the National Institutes of Health P50AR065139 (SJT), R01AR045203 (SJT), F32CA254805 (DCH), R37CA230617 (ACH), R01GM135362 (ACH), and the Friends of FSH Research https://www.fshfriends.org/ and the Chris Carrino Foundation for FSHD https://chriscarrinofoundation.org/ (SJT). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data Availability: All data needed to evaluate the conclusions in the paper are present in the paper, supporting information, or the Gene Expression Omnibus repository. The Ribo-seq, Poly-seq, and RNA-seq data generated in support of this publication have been deposited in the Gene Expression Omnibus (GSE206439). The processed datasets, including gene expression, ribosome footprint p-sites, metadata, shell scripts and R code for preprocessing and downstream analysis, are available on Zenodo ( https://zenodo.org/record/7822959 ; DOI: 10.5281/zenodo.7822959 ). A GitBook with detailed description of our analysis is also available ( https://fredhutch.github.io/DUX4-IFNg-ribosome-footprints ). Flow cytometry data generated in this study are available in the FlowRepository.org (FR-FCM-Z6XR, FR-FCM-Z6XT, FR-FCM-Z6XS).

Copyright: © 2023 Hamm et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

In this study, we have implicated the DUX4 transcriptional program as a driver of broad translational suppression that reprograms de novo protein synthesis. We initially focused on the DUX4 suppression of MHC-class I and related interferon-stimulated proteins to characterize the mechanisms of protein suppression. We found that a brief pulse of DUX4 expression disrupts several key regulators of translation initiation and elongation, which are sufficient to suppress protein synthesis of IFNγ-stimulated MHC-I and immunoproteasome (iProteasome) subunits. Moreover, high-throughput translational profiling with ribosome footprinting and polysome gradients showed significant changes in MHC-I translational efficiency (TE) and broad translational suppression of many cellular mRNAs, whereas DUX4-induced mRNAs were robustly translated. Taken together, DUX4 activation of its transcriptional program resulted in the replacement of the prior cellular proteome and lineage identity with the DUX4-induced proteome enriched for ZGA-associated proteins. We propose that coordinated regulation of transcription and translation is employed by DUX4 to reshape the cellular translatome in both development and disease.

In addition to a role in FSHD, recent analysis of nearly 10,000 cancer transcriptomes from 33 different cancer types revealed DUX4 to be one of the most commonly expressed cancer-associated genes [ 31 ]. Full-length DUX4 expression in diverse cancer types is strongly correlated with increased expression of high-confidence DUX4 targets activated in the embryo [ 31 ]. DUX4 expression in cancers was associated with decreased major histocompatibility complex class I (MHC-I) expression, resistance to checkpoint inhibitors, and decreased patient survival rates [ 31 ]. DUX4 expression in several cancer cell lines was sufficient to prevent the induction of MHC-I expression in response to interferon gamma (IFNγ); however, it was unclear whether this represented an activity restricted to MHC-I or a broader activity of DUX4 in regulating protein expression.

DUX4 is also the causative gene of facioscapulohumeral muscular dystrophy (FSHD), a complex genetic disorder that results in epigenetic derepression of the DUX4 locus in skeletal muscle and progressive muscle atrophy [ 17 – 20 ]. The aberrant expression of DUX4 in skeletal muscle activates expression of genes associated with germline and stem cell development [ 21 , 22 ], characteristic of the early embryonic ZGA program. Although DUX4 is sporadically expressed in approximately 0.1% of FSHD muscle cells in culture [ 3 , 23 ], DUX4 target gene activation results in a host of pathogenic features including impaired myogenesis [ 24 ], oxidative stress and DNA damage [ 25 , 26 ], compromised mRNA quality control [ 27 , 28 ], and inflammation [ 29 , 30 ].

Emerging evidence has shown that broad translational suppression is a hallmark of reprogramming in embryonic and somatic stem cells [ 9 – 11 ], where low rates of translation are thought to promote an undifferentiated state. This is further supported by the finding that a rare population of totipotent mouse embryonic stem cells (ESCs) known as 2-cell-like cells (2CLCs), thought to recapitulate the naïve state of the preimplantation embryo [ 12 ], exhibit global repression of nascent protein synthesis [ 13 , 14 ]. Mouse Dux and human DUX4 belong to the conserved DUXC family of proteins found in eutherian mammals [ 15 ]. Functionally, expression of mouse Dux reprograms these rare populations of 2CLCs to have expanded developmental potential [ 1 , 2 , 12 ], and human DUX4 has been reported to drive a similar totipotent program in human induced pluripotent stem cells (iPSCs) and ESCs [ 2 , 16 ].

The double homeobox protein 4 (DUX4) gene encodes a transcription factor that is expressed in immune-privileged niches such as the preimplantation embryo [ 1 , 2 ], testis [ 3 ], and, possibly, thymus [ 4 ]. DUX4 is briefly expressed in the 4-cell human embryo and serves as a key transcriptional activator of the zygotic genome, driving expression of hundreds of coding genes and repetitive retroelements [ 1 , 2 ]. In addition to zygotic genome activation (ZGA), regulation of mRNA degradation and translation is essential to rapidly diversify the proteome during early development [ 5 ] and has been associated with increased developmental potential of human preimplantation embryos [ 6 ]. It is becoming abundantly clear that translational control, both globally and at the level of individual transcripts, helps mediate cell fate transitions. This includes the shift from the maternal to the embryonic developmental program, the balance of stem cell self-renewal and differentiation, and the plasticity of cancer [ 7 , 8 ].

Results

DUX4 activity induces prolonged suppression of antigen presentation factors We recently reported that DUX4 blocks IFNγ-stimulated induction of MHC-I and surface antigen presentation [31]. To determine the mechanism of DUX4-induced MHC-I regulation, we used a well-characterized cellular model system of human myoblasts with a doxycycline (DOX)-inducible DUX4 transgene (MB135iDUX4) [32]. DUX4 expression occurs in transient bursts in rare populations of ESCs [2,16] and is sporadically misexpressed in FSHD muscle cells [3,23], making it difficult to characterize downstream mechanisms endogenously. We have previously demonstrated that a short “pulse” of DUX4 in MB135iDUX4 myoblasts induced a transcriptional program representative of FSHD muscle cells and the early cleavage-stage embryo [33]. Pulsed DUX4 expression in this cell culture system enabled reproducible and synchronized DUX4 induction, permitting the investigation of mechanisms downstream of DUX4 that may have otherwise been masked by heterogeneous populations of DUX4-expressing cells. Using our MB135iDUX4 cell culture system, we tested both the immediate effect of continuous DUX4 expression, as well as the prolonged consequences following a brief pulse of DUX4 on the expression levels of several IFNγ-induced factors involved in immunogenic antigen presentation, including canonical MHC-I subunits (HLA-A, HLA-B, HLA-C) and iProteasome subunits (PSMB8, PSMB9, PSMB10). MB135iDUX4 myoblasts were treated with DOX for 20 hours or for a 4-hour period followed by washout to induce a “continuous” versus a “pulse” of DUX4 expression, respectively. Cells were exposed to IFNγ for the final 16 hours prior to harvest, collecting terminal time points at 20 hours for the continuous treatment, and at 44 hours for the pulse (40 hours after the DOX washout) (Fig 1A). Cells stimulated with IFNγ showed elevated protein levels of MHC-I and iProteasome subunits, whereas IFNγ induction of MHC-I and the iProteasome was suppressed in myoblasts continually expressing DUX4 (Fig 1B, left). Remarkably, a pulse of DUX4 elicited the same degree of suppression despite having diminished levels of DUX4 protein (Fig 1B, right). PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 1. Brief expression of DUX4 results in long-term suppression of MHC-I and iProteasome subunits. (A) Schematic of experimental time course. (B) Immunoblot analysis following treatment with or without DOX and IFNγ as noted. Cells expressing DUX4 continuously (left) or a pulse of DUX4 (right). (C) Normalized RNA-seq read counts; data represent mean ± SD of biological replicates, n = 3. Source data available in S1 Data. (D) Immunoblot analysis of extended experimental time course outlined in (A). DOX-inducible MB135 myoblasts expressing active DUX4 (left) versus transcriptionally inactive DUX4 carrying F67A mutation in the DUX4 DNA binding domain (center), or mutation of the first (L)LxxL(L) motif and deletion of the second (L)LxxL(L) motif within the C-terminal activation domain (mL1dL2, right; see [34] for mutation sequences). DUX4 E14-3 antibody detects epitope in active DUX4 and mutant DUX4 proteins. DUX4 targets H3.X and H3.Y are expressed upon induction of active DUX4, but not DUX4(F67A) or DUX4(mL1dL2). GAPDH serves as loading control. DOX, doxycycline; DUX4, double homeobox protein 4; IFNγ, interferon gamma; MHC-I, major histocompatibility complex class I. https://doi.org/10.1371/journal.pbio.3002317.g001 In a recent study, we found that the DUX4 protein was sufficient to inhibit IFNγ induction of interferon-stimulated genes (ISGs) at the mRNA level by interacting with STAT1 and preventing RNA Pol-II recruitment to STAT1-regulated genes [34]. Our current findings that suppression of IFNγ-stimulated factors persisted following a transient pulse of DUX4 suggested an additional mechanism of regulation downstream of DUX4 protein expression. We compared mRNA levels of MHC-I and iProteasome subunits following continuous or pulsed DUX4 expression and found that transcriptional suppression of IFNγ signaling by continuous DUX4 mostly recovered at 44 hours following the pulse of DUX4 (Fig 1C). These data indicated that a pulse of DUX4 induced posttranscriptional suppression of MHC-I and iProteasome proteins through a method distinct from its interaction with STAT1 to suppress ISG mRNA induction. To determine the necessity of DUX4 transcriptional activity for long-term protein suppression of MHC-I and the iProteasome, we performed a time course with active DUX4 and transcriptionally inactive DUX4 mutants carrying either a mutation in the DUX4 DNA binding domain (F67A) or mutations within the (L)LxxL(L) motifs of the C-terminal activation domain (mL1dL2). A prolonged time course revealed that protein suppression persisted for several days following a pulse of active DUX4, whereas transcriptionally inactive DUX4 mutants were insufficient to suppress MHC-I and PSMB9 at later time points (Fig 1D). Suppression of MHC-I and PSMB9 induction at the 20-hour time point was observed with the DUX4 F67A mutant, but not the mL1dL2 mutant (Fig 1D). This is likely mediated by inhibition of interferon signaling and ISG transcription by the DUX4 protein that requires the C-terminal activation domain and (L)LxxL(L) motifs [34], whereas long-term suppression of MHC-I and PSMB9 that persists after the loss of DUX4 protein required the transcriptional activity of DUX4. Collectively, these data suggest that DUX4 acts as a repressor of antigen presentation factors through 2 distinct mechanisms. Here, we focus on the finding that transient DUX4 expression activates a transcriptional program required for prolonged protein suppression. Subcellular fractionation of mRNAs and proteins showed that a pulse of DUX4 did not disrupt HLA-A, HLA-B, HLA-C, PSMB8, PSMB9, or PSMB10 mRNA nuclear export or protein localization to the cytoplasm following IFNγ simulation (S1 Fig). Additionally, treatment with proteasome inhibitor MG132 or autophagy inhibitor Bafilomycin did not rescue suppression of IFNγ-stimulated MHC-I, PSMB8, PSMB9, or PSMB10 following a pulse of DUX4 (S2 Fig), eliminating protein degradation as a causal mechanism. iProteasome production of immunogenic antigens has also been linked to MHC-I protein stability; however, siRNA-mediated knockdown of iProteasome catalytic subunits PSMB8 and PSMB9 in parental MB135 myoblasts did not impact IFNγ-induced MHC-I levels (S3A Fig). Furthermore, treatment with ONX-0914, a selective inhibitor of the iProteasome, did not reduce IFNγ-induced MHC-I levels (S3B Fig). Thus, MHC-I stability does not require iProteasome-dependent proteolysis. Together, these data suggest that DUX4 posttranscriptionally suppresses MHC-I and the iProteasome independently through methods of translational inhibition.

Transient DUX4 activity broadly suppresses nascent protein synthesis Uncoupling of the transcriptome and proteome downstream of DUX4 is not limited to antigen presentation factors. We previously reported a discordant relationship between RNA and protein levels for a multitude of mRNAs in DUX4-expressing cells [28]. This, in combination with the dysregulation of multiple key translational regulators (Fig 2A), suggested that DUX4 might broadly alter cellular translation. Indeed, metabolic labeling with 35S-methionine/cysteine in MB135iDUX4 myoblasts treated with IFNγ showed that de novo protein synthesis was transiently suppressed following a pulse of DUX4, with an approximate 50% reduction at the 68-hour time point (Fig 2D and 2E), whereas induction of transcriptionally inactive DUX4(F67A) did not alter nascent protein synthesis (Fig 2E, right). We confirmed DUX4 inhibition of protein synthesis by labeling cells with methionine analog L-homopropargylglycine (HPG) followed by fixation and Click-iT chemistry, wherein fluorescence microscopy and flow cytometry showed a dramatic reduction in HPG-labeled peptides after a pulse of DUX4 (Fig 2F). As seen with 35S-methionine/cysteine labeling, HPG signal was lowest 44 hours after a pulse of DUX4, comparable to the degree of translational suppression induced by cycloheximide (CHX) treatment alone, and nascent protein synthesis started to recover in a subset of cells by 68 to 92 hours. These findings establish long-lived, yet transitory, suppression of protein synthesis downstream of the DUX4 transcriptional program that encompassed IFNγ-induced MHC-I and PSMB9 expression.

Genome-wide polysome profiling identifies defects in translation initiation and elongation Because ribosome footprinting reports the distribution of ribosome protected fragments rather than the ribosome abundance per transcript, it can be less effective at identifying differences in translation that parallel changes in protein expression. Therefore, we turned to classical polysome profiling to directly measure changes in ribosome density per mRNA using sucrose gradient-based isolation from MB135iDUX4 myoblasts treated with a pulse of DUX4+IFNγ harvested at 68 hours versus IFNγ alone. We pooled RNA fractions representing sub-polysome (40S-60S-80S), low polysome (1 to 3 ribosomes), and high polysome (>3 ribosomes) populations and performed RNA-seq analysis (S8A Fig). We initially measured changes in steady-state mRNA translation, excluding DUX4-altered gene expression. RNA abundance in each polysome fraction was determined relative to total input mRNA reads normalized to an internal spike-in (S8B Fig and S4 Data). The high-to-sub polysome (high/sub) ratio identified 5,800 genes that had decreased polysome association following a pulse of DUX4, consistent with a broad suppression of translation initiation, and 323 genes with increased polysome abundance (|log2FC>1|; p-adj<0.01) (Fig 3F and S4 Data). GO analysis of the 323 genes enriched in the high polysome fraction showed an abundance of mRNAs characterized as ribosomal proteins and translation factors (Fig 3G). Many of these mRNAs contain 5-prime terminal oligopyrimidine (TOP) motifs [50] and are particularly sensitive to mTORC1 regulation of initiation factors [51,52] and eEF2K-eEF2 control of translation elongation [53]. Analysis of mRNAs with characterized TOP motifs [50] showed that TOP mRNAs remain associated with polysomes following a pulse of DUX4, while most other transcripts are depleted in the high polysome fraction (Fig 3H). This enrichment could reflect enhanced ribosome biogenesis used to poise cells for a rapid shift in translation rate upon recovery from DUX4 suppression. Conversely, mRNA enrichment in the high polysome fraction could result from stalled and accumulating ribosomes correlated with inhibited elongation and reduced protein expression. TOP mRNA-encoded ribosomal proteins RPL10A, RPL4, RPS6, and RPS15A were suppressed by DUX4 even though the mRNAs remain bound by polysomes (Fig 3I and 3J, and S8C Fig), consistent with inhibited translation elongation. To elucidate DUX4 posttranscriptional suppression of MHC-I and iProteasome subunits specifically, we similarly assessed their mRNA high-to-sub polysome ratios. HLA-A, HLA-B, and HLA-C mRNAs showed reduced polysome association indicative of impaired translation initiation; however, PSMB8, PSMB9, and PSMB10 mRNAs remained associated with polysomes, like TOP mRNAs (Fig 3I). Therefore, we propose that translation of MHC-I mRNAs is particularly sensitive to initiation defects, while a subset of mRNAs, including a subset of TOP mRNAs and iProteasome mRNAs, appear suppressed by DUX4 possibly through combinatorial inhibition of translation initiation and stalled elongation. Previous analysis of the DUX4-induced transcriptome in muscle cells showed that DUX4 can both activate and inhibit genes to prevent myogenic differentiation [24,54,55]. This led us to question whether a pulse of DUX4 also had a prolonged suppressive effect on myogenesis. Interestingly, several early myogenic markers, including ITGA7, PAX3, MEF2A, MEF2C, and MEF2D mRNAs, showed a loss of polysome abundance and decreased protein levels indicative of reduced translation (Fig 3I and 3J). While the master regulator of skeletal myogenesis, MYOD1, showed no change in polysome abundance, and mRNAs encoding the muscle-specific gene Desmin had an increase in polysomes (Fig 3I), both were suppressed at the protein level following a pulse of DUX4 (Fig 3J). Thus, consistent with our molecular and Ribo-seq analysis, polysome profiling supports DUX4-induced translational suppression of broad classes of mRNAs, including those involved in antigen presentation and lineage determination.

Translation of DUX4-induced mRNAs Our previous studies showed that mRNAs transcriptionally induced by DUX4 are translated into protein [28]. Comparing the high polysome fractions of DUX4 pulse+IFNγ samples to IFNγ-treated samples provided a measure of how DUX4 changes the overall translatome and showed that DUX4-induced mRNAs were indeed associated with polysomes (Fig 4A and S4 Data). Thousands of mRNAs were reduced in the DUX4 pulse+IFNγ high polysome fraction (7,765 genes; log2FC<−1, p-adj<0.01); whereas 256 polysome-bound mRNAs were significantly up-regulated (log2FC>1, p-adj<0.01), many of which are well-characterized DUX4-target genes [56]. Immunoblot analysis confirmed that several DUX4-induced mRNAs in the high polysome fraction correlated with translation of these mRNAs (Fig 4B). PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 4. DUX4 orchestrates translational reprogramming through broad inhibition of translation concurrent with de novo translation of DUX4 target mRNAs. (A) Differential RNA-seq analysis of high polysome fractions (high/high). Volcano plot showing log2 fold-change of DUX4 pulse+IFNγ harvested at 68 hours versus IFNγ (significance defined as basemean>50, |log2FC>1|, p-adj<0.01); see S4 Data. Polysome-bound mRNAs up-regulated by DUX4 highlighted in red (n = 256 genes). Direct DUX4 target genes highlighted in orange (n = 70 genes). (B) Immunoblot analysis of total protein lysate harvested for polysome profiling samples representing biological replicates, n = 3. (C) Schematic of mRNA with structured 5′ UTR that impedes translation. (D) Box plot showing 5′ UTR analysis of predicted MFE per 100 nt, including all annotated mRNAs, a subset of direct DUX4 targets, mRNAs up- or down-regulated following a pulse of DUX4 (defined as mRNAs differentially expressed in (A)). See S5 Data for 5′ UTR sequences and analysis. Statistical comparisons were conducted using Mann–Whitney U test, **** p < 1 × 10−8. (E) Immunofluorescence of HPG Click-iT labeled nascent proteins in differentiated FSHD myotubes and SuSa cells costained for DUX4-target genes H3.X/Y. MB2401 myotubes serve as a control cell line that does not express DUX4 (scale bars, 50 μm). (F) RT-qPCR analysis of unsynchronized SuSa cells in log growth phase (log) relative to a time course following release from synchronization with gapmer-mediated CTRL or DUX4 kd. Data represent mean ± SD of biological replicates, n = 3; see S1 Data. (G) FACS analysis of MHC-I surface levels on SuSa cells 72 hours after synchronization and treated with and without IFNγ. (H) RT-qPCR analysis of SuSa cells treated with or without IFNγ and sorted based on high versus low MHC-I surface levels highlighted in (G); see S1 Data. (I, J) Flow cytometry analysis of MHC-I surface levels of SuSa cells 72 hours after synchronization and gapmer-mediated CTRL or DUX4 kd with and without IFNγ treatment. (K) Model of ribosome abundance and translation efficiency resulting from DUX4-induced inhibition of translation initiation and elongation. CTRL, control; DUX4, double homeobox protein 4; FACS, fluorescence-activated cell sorting; FSHD, facioscapulohumeral muscular dystrophy; HPG, L-homopropargylglycine; IFNγ, interferon gamma; kd, knockdown; MFE, minimum free energy; MHC-I, major histocompatibility complex class I; nt, nucleotide; RNA-seq, RNA sequencing; RT-qPCR, quantitative reverse transcription PCR; UTR, untranslated region. https://doi.org/10.1371/journal.pbio.3002317.g004 The enrichment of DUX4-induced mRNAs in the high polysome fraction might reflect their increased abundance following DUX4 expression or a relative resistance to the DUX4-mediated translational inhibition, or both. Thermodynamic stability and RNA secondary structures within the 5′ UTR of an mRNA can influence translation initiation rates of distinct transcripts in cis (Fig 4C), with higher predicted minimum free energy (MFE) showing increased translation efficiency [57,58]. The annotated 5′ UTRs of the 256 genes induced by DUX4 had a significantly higher predicted MFE per 100 nt relative to the average of all annotated 5′ UTRs, whereas the average MFE of 5′ UTRs belonging to the 7,765 genes repressed by DUX4 was significantly lower (Fig 4D). Furthermore, selective usage of alternative transcription start sites (TSS) have been shown to alter 5′ UTR sequences to influence cell type–specific protein synthesis [59]. We have previously observed that some DUX4-bound repetitive elements are co-opted to form alternative promoters for DUX4 target genes [22,56]. To account for noncanonical TSS and splicing events, we annotated the functional 5′ UTRs of 70 direct DUX4 targets based on RNA-seq alignment and published DUX4 ChIP peaks (S5 Data and S8D Fig). Indeed, DUX4 targets were predicted to have less structured 5′ UTRs on average (Fig 4D). Therefore, DUX4-induced mRNAs are predicted to be less susceptible to inhibition of translation initiation, which correlates with their observed increase in protein expression and polysome association.

[END]
---
[1] Url: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3002317

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.