1. Genomics and Evolutionary Biology
  2. Human Biology and Medicine
Download icon

A feedback loop between nonsense-mediated decay and the retrogene DUX4 in facioscapulohumeral muscular dystrophy

  1. Qing Feng
  2. Lauren Snider
  3. Sujatha Jagannathan
  4. Rabi Tawil
  5. Silvère M van der Maarel
  6. Stephen J Tapscott Is a corresponding author
  7. Robert K Bradley Is a corresponding author
  1. Fred Hutchinson Cancer Research Center, United States
  2. University of Rochester, United States
  3. Leiden University Medical Center, Netherlands
  4. University of Washington, United States
Short Report
Cited
14
Views
2,264
Comments
0
Cite as: eLife 2015;4:e04996 doi: 10.7554/eLife.04996

Abstract

Facioscapulohumeral muscular dystrophy (FSHD) is a muscular dystrophy caused by inefficient epigenetic repression of the D4Z4 macrosatellite array and somatic expression of the DUX4 retrogene. DUX4 is a double homeobox transcription factor that is normally expressed in the testis and causes apoptosis and FSHD when misexpressed in skeletal muscle. The mechanism(s) of DUX4 toxicity in muscle is incompletely understood. We report that DUX4-triggered proteolytic degradation of UPF1, a central component of the nonsense-mediated decay (NMD) machinery, is associated with profound NMD inhibition, resulting in global accumulation of RNAs normally degraded as NMD substrates. DUX4 mRNA is itself degraded by NMD, such that inhibition of NMD by DUX4 protein stabilizes DUX4 mRNA through a double-negative feedback loop in FSHD muscle cells. This feedback loop illustrates an unexpected mode of autoregulatory behavior of a transcription factor, is consistent with ‘bursts’ of DUX4 expression in FSHD muscle, and has implications for FSHD pathogenesis.

https://doi.org/10.7554/eLife.04996.001

eLife digest

Genes are sequences of DNA that contain instructions for the cell that must be carefully controlled because it is not always appropriate or safe for these instructions to be followed. When genes are active, copies of the DNA are made using molecules of ribonucleic acid (RNA) and these can then be used as templates to make proteins.

One way genes can be controlled is by adding small chemical tags that mark them out to be activated or deactivated, known as epigenetic control. The muscle disease facioscapulohumeral muscular dystrophy (FSHD) is caused by the loss of the chemical tags that normally keep certain genes switched off in many cell types. One of these genes is DUX4, which in healthy males is normally only active in the testes, but in FSHD patients it is also active in other parts of the body.

Another way to control genes is by nonsense-mediated decay, where incorrect or incomplete RNA molecules are destroyed before they can be used to make defective proteins. In this study, Feng et al. show that when DUX4 is activated following the failure of epigenetic control in FSHD patients, the effectiveness of nonsense-mediated decay is also reduced. This results in the build-up of incorrect RNA molecules inside muscle cells, which can harm the cell. In fact, 13% of the incorrect RNA molecules that are normally destroyed in cells were found at higher levels when DUX4 was active.

To investigate how DUX4 could work, Feng et al. focused on another gene called UPF1 because cells without the protein encoded by this gene have similar defects in nonsense-mediated decay as cells with active DUX4. No difference was found in how often the UPF1 gene is activated in FSHD cells and normal cells. However, the amount of the protein encoded by UPF1 was lower in cells with FSHD than in normal muscle cells. The experiments show that the protein encoded by UPF1 is broken down as a result of the activation of the DUX4 gene, leading to problems with nonsense-mediated decay, which may result in the worsening of FSHD symptoms.

The twist in the tale is that DUX4 itself is also controlled by nonsense-mediated decay under normal circumstances. Therefore, in diseased cells, a failure in epigenetic control allows DUX4 to prevent its own destruction by tampering with nonsense-mediated decay. These findings offer new insights into the role of the DUX4 gene in FSHD. The next step is to test whether these defects in nonsense-mediated decay can explain any of the symptoms of FSHD, such as muscle inflammation.

https://doi.org/10.7554/eLife.04996.002

Main text

Facioscapulohumeral muscular dystrophy (FSHD) is typically an adult-onset muscular dystrophy characterized by muscle weakness initially affecting the face (facio), shoulders (scapulo), and upper arms (humeral). FSHD is caused by decreased epigenetic repression of the D4Z4 macrosatellite array in the subtelomeric region of chromosome 4q, due to either D4Z4 repeat contractions (Lemmers et al., 2010) or mutations affecting trans-acting epigenetic regulators of the D4Z4 repeat such as SMCHD1 (Lemmers et al., 2012), which results in the misexpression of DUX4 mRNA in skeletal muscle and possibly other somatic tissues. DUX4 encodes a double homeobox transcription factor that activates germline genes and repetitive elements (Geng et al., 2012) and causes apoptosis and atrophic myotube formation when misexpressed in skeletal muscle (Kowaljow et al., 2007; Vanderplanck et al., 2011; Wallace et al., 2011; Mitsuhashi et al., 2012). DUX4 is expressed in only a small fraction of nuclei (Snider et al., 2010), likely due to occasional ‘bursts’ of DUX4 expression. However, the mechanism(s) regulating DUX4 expression and toxicity remain incompletely understood.

We previously ectopically expressed DUX4 in immortalized (54-1) and primary (MB135) myoblasts and used RNA-seq to identify coding genes, repetitive elements, and non-coding RNAs induced by DUX4 (Young et al., 2013). Further analysis of this data showed that DUX4 expression also resulted in the increased abundance of many coding RNA isoforms containing premature translation termination codons upstream of splice junctions. These isoforms, which are predicted substrates for degradation by nonsense-mediated decay (NMD), were present at very low levels in control myoblasts. Following DUX4 expression, however, many such predicted NMD substrates increased in abundance and in many cases became the predominant mRNA product of the parent gene. For example, an isoform of the SRSF3 gene containing a well-characterized NMD-inducing cassette exon (Lareau et al., 2007; Ni et al., 2007) was present at low levels prior to DUX4 expression but became the dominant isoform thereafter in both 54-1 and MB135 cells (Figure 1A–B).

Figure 1 with 1 supplement see all
DUX4 expression inhibits nonsense-mediated decay.

(A) RNA-seq read coverage of the SRSF3 gene in control and DUX4-expressing myoblasts. 54-1, immortalized myoblasts; MB135, primary myoblasts. Purple shading, NMD-inducing cassette exon. Red stop sign, termination codon. (B) Inclusion of the premature termination codon-containing cassette exon of SRSF3 illustrated in (A). Error bars, 95% confidence intervals as estimated by MISO (Katz et al., 2010). (C) Relative levels of transcripts produced from NMD reporter plasmids encoding either premature termination codon-containing (top) or normal (bottom) β-globin (Gl). Bar plot illustrates the ratio NMD(+)/NMD(−) of transcripts from the NMD(+) and NMD(−) constructs. (D) Isoform ratios of predicted NMD substrates generated by cassette exon alternative splicing in control and DUX4-expressing myoblasts (54-1 cells). Red/blue, cassette exons exhibiting increases/decreases of ≥10% in isoform ratios for the isoforms that are predicted NMD substrates. (E) Isoform ratios of mis-spliced isoforms of annotated constitutive splice junctions generated by abnormal 5′ and 3′ splice site recognition (54-1 cells). Color as in (D). (F) Global increases and decreases in relative levels of predicted NMD substrates generated by differential splicing. Annotated alternative splicing events are illustrated in upper panel, and alternative splicing and intron retention of annotated constitutively spliced junctions are illustrated in lower panel. Up/down arrows, percentages of predicted NMD substrates generated by alternative splicing exhibiting increases/decreases of ≥10% in isoform ratios in DUX4-expressing vs control cells. Enrichment for increased vs decreased levels of NMD substrates indicated in columns three and six.

https://doi.org/10.7554/eLife.04996.003

To determine whether increased levels of such normally degraded mRNAs were associated with reduced NMD efficiency, we used an exogenous reporter system. We transfected plasmids encoding either the wild-type β-globin open reading frame or β-globin with a premature termination codon that induces degradation by NMD (Zhang et al., 1998). Relative levels of the β-globin NMD substrate were twofold higher in DUX4-expressing vs control myoblasts, indicating that NMD is indeed compromised by DUX4 (Figure 1C).

We then determined how reduced NMD efficiency affected global levels of predicted NMD substrates. Restricting to cassette exon splicing events where one isoform, but not both, was a predicted NMD substrate, we found that ∼13% of such predicted NMD substrates increased following DUX4 expression, while ∼1.6% decreased, in 54-1 cells (Figure 1D). Impaired NMD also caused accumulation of aberrant mRNAs resulting from mis-splicing or incomplete splicing, which are common byproducts of the stochastic nature of the splicing process (Weischenfeldt et al., 2012). We identified and quantified alternative splicing of annotated constitutive junctions, finding that ∼13% of such junctions exhibited increased aberrant splicing in DUX4-expressing vs control cells, while only ∼0.25% exhibited decreased aberrant splicing (Figure 1E). The vast majority of these novel products of annotated constitutive junctions were present at very low or undetectable levels in control 54-1 myoblasts.

We next extended this analysis to all classes of splicing events, including mis-splicing and intron retention of constitutive splice junctions. DUX4 expression caused increased levels of predicted NMD substrates for all classes of splicing events in both 54-1 and MB135 cells (Figure 1F). These increases were generally more extreme in 54-1 than in MB135 cells, likely due to the ∼15-fold higher DUX4 expression achieved in 54-1 vs MB135 cells as well as the longer time period allowed for infection (48 hr vs 24 hr).

High levels of NMD substrates in DUX4-expressing cells were not simply a side effect of DUX4-induced apoptosis. TP53 knock-down (KD) prevented apoptosis following DUX4 expression in normal myoblasts, confirming previous reports that DUX4 toxicity is p53-dependent (Wallace et al., 2011). However, TP53 KD did not prevent DUX4-induced NMD inhibition (Figure 1—figure supplement 1).

DUX4 could potentially inhibit NMD by transcriptionally repressing components of the NMD machinery. However, no UPF or SMG NMD factors exhibited decreased mRNA levels following DUX4 expression, and most were up-regulated by two- to fourfold (Figure 2A). This expression pattern was reminiscent of a recent report that mRNA levels of most NMD factors increase following the knock-down of UPF1, encoding a central component of the NMD machinery (Huang et al., 2011). Therefore, we hypothesized that UPF1 mRNA and protein levels might be decoupled in DUX4-expressing cells. We measured levels of UPF1, which was not transcriptionally up-regulated in DUX4-expressing cells, and UPF3B and SMG7, which were transcriptionally up-regulated in response to DUX4. UPF1 protein levels were markedly lower in DUX4-expressing myoblasts than in control myoblasts, as were SMG7 levels, although to a lesser extent. In contrast, UPF3B levels were unaffected by DUX4 expression (Figure 2B).

DUX4 destabilizes UPF1 via the proteasome.

(A) Relative mRNA levels of NMD factors in DUX4-expressing vs control myoblasts (54-1 cells). Red, up-regulation by ≥1.5-fold. (B) Immunoblot for NMD factors UPF1, SMG7, and UPF3B in DUX4-expressing and control myoblasts (54-1 cells) at 36 hr post-infection. H3, histone H3 (loading control). (C) Immunoblot of total protein from a 36-hr time course of DUX4-expressing and control myoblasts (54-1 cells). H3, histone H3. Loading Control, a nonspecific band that serves as an additional loading control. (D) Quantification of UPF1 protein level from the immunoblot presented in (C), normalized to the nonspecific band that serves as a loading control. (E) Relative levels of transcripts produced from the NMD(+) and NMD(−) β-globin reporter plasmids. (F) Isoform ratios of endogenously produced NMD-degraded isoforms of SRSF2 and SRSF3. Time course identical to (C). Error bars, standard deviation. (G) Immunoblot of total protein from DUX4-expressing and control myoblasts (54-1 cells) treated with the proteasome inhibitor MG132 (10 µM; 8 hr treatment initiated 16 hr after infection with lentiviral expression constructs). Loading control H3, histone 3, has a long half-life (Toyama et al., 2013). (H) Quantification of UPF1 protein levels from three independent replicates of the immunoblot presented in (G), normalized to the nonspecific band that serves as a loading control. Error bars, standard deviation.

https://doi.org/10.7554/eLife.04996.005

To determine whether decreased UPF1 temporally correlates with DUX4-induced inefficient NMD, we conducted a time course following DUX4 expression in myoblasts. DUX4 was robustly detectable 12–14 hr after lentiviral infection, coincident with the beginning of a sharp decrease in UPF1 levels (Figure 2C–D). SMG7 showed a more modest decrease through the time course, while UPF3B levels were relatively constant. NMD substrates produced from the β-globin reporter, as well as endogenously produced from the SRSF2 and SRSF3 genes, exhibited increased levels 12–14 hr after lentiviral expression (Figure 2E–F). The close temporal coupling between DUX4 protein production, decreased UPF1 levels, and increased levels of both endogenous and exogenous NMD substrates suggests that insufficient levels of UPF1—and perhaps additional NMD machinery components such as SMG7—may contribute to inefficient NMD in DUX4-expressing cells.

The rapid decrease in UPF1 levels that we observed suggested that DUX4 might trigger UPF1 degradation. To test this, we treated DUX4-expressing or control myoblasts with MG132 to inhibit the proteasome. MG132 treatment restored normal UPF1 levels in DUX4-expressing myoblasts, while UPF1 levels in control myoblasts were unaffected (Figure 2G–H). As proteasome inhibition inhibits normal translation (Cowan and Morley, 2004; Mazroui et al., 2007) and therefore NMD, we were unable to test whether the restoration of normal UPF1 levels by proteasomal inhibition rescued NMD. However, the close temporal relationship between the onset of decreased UPF1 levels and increased NMD substrates strongly suggests that UPF1 degradation contributes to NMD inhibition in DUX4-expressing cells.

Both DUX4 isoforms encoding the full-length protein contain a constitutively spliced intron within their 3′ UTRs, rendering them likely NMD substrates (Figure 3A). To test this, we used cells isolated from FSHD1 (54-2, which are isogenic to normal 54-1 cells but carry a contracted D4Z4 array) and FSHD2 (MB200) skeletal muscle (Krom et al., 2012; Schoenberg and Maquat, 2012; Young et al., 2013). We knocked down UFP1 in 54-2 and MB200 myoblasts to 24.3% and 32.4% of normal protein levels, respectively, and differentiated these myoblasts to myotubes to stimulate DUX4 transcription. DUX4 mRNA was expressed at approximately fourfold higher levels in UPF1 KD vs control KD myotubes, as was ZSCAN4 mRNA, which is transcriptionally activated by DUX4 (Figure 3B–D).

DUX4 mRNA is an endogenous NMD substrate.

(A) Schematic of the DUX4 mRNA. Intron 2, constitutively spliced intron within the 3′ UTR. Black, coding sequence; purple, 5′ and 3′ untranslated regions (UTRs). Red stop sign, termination codon. (B) Immunoblot of total protein from FSHD1 (54-2) and FSHD2 (MB200) myoblasts following transfection with a siRNA against UPF1 or a control non-targeting siRNA. α-tubulin, loading control. (C) Levels of DUX4 mRNA following control or UPF1 knock-down, measured 2 days after the initiation of myogenesis. Error bars, standard deviation. (D) Levels of ZSCAN4 mRNA following control or UPF1 knock-down, measured 2 days after the initiation of myogenesis. Error bars, standard deviation. (E) Schematic of chimeric constructs encoding the β-globin opening reading frame (cyan) followed by the DUX4 3′ UTR (purple) containing (top) or lacking (bottom) the second intron of DUX4's 3′ UTR (Intron 2). (F) Relative levels of transcripts from the Gl-DUX4-Intron2(+) and Gl-DUX4-Intron2(−) constructs following control or UFP1 KD in normal myoblasts (54-1 cells). For each construct, data are normalized such that the siControl treatment is 1. Error bars, standard deviation. (G) Immunofluorescence with an antibody against DUX4 following control or UPF1 knock-down, measured 2 days after the initiation of myogenesis in FSHD1 cells (54-2), which was prior to significant fusion. Box plot, percentage of DUX4+ nuclei as estimated by ImageJ (Fiji); n = 8 fields. Whiskers, max and min over the fields.

https://doi.org/10.7554/eLife.04996.006

We next sought to determine whether the intron-containing 3′ UTR of DUX4 contributed to the degradation of DUX4 mRNA by NMD. We created chimeric constructs containing the β-globin open reading frame followed by either the complete DUX4 3′ UTR or the DUX4 3′ UTR with the second intron removed (Figure 3E). We focused on the constitutively spliced second intron within the 3′ UTR because it lies 100 nt downstream of the termination codon, and therefore it is predicted to trigger NMD. Transcripts from the chimeric construct containing the complete DUX4 3′ UTR increased twofold following UPF1 KD in normal myoblasts—a substantial but smaller increase than we observed for the endogenous DUX4 mRNA, perhaps due to the chimeric nature of the β-globin + DUX4 3′ UTR construct—while transcripts from the construct lacking the second intron of the DUX4 3′ UTR increased only 1.5-fold. We conclude that the second intron of the DUX4 3′ UTR is important for NMD-induced degradation of the DUX4 mRNA (Figure 3F).

DUX4 exhibits variegated expression in FSHD muscle cells, with only a few percent of nuclei detectable as DUX4+ (Snider et al., 2010). Therefore, augmented DUX4 expression following UPF1 KD in myotubes could be due to increases in DUX4 mRNA in nuclei that are already DUX4+ and/or increases in the fraction of DUX4+ nuclei. Immunostaining of FSHD myotubes revealed that the fraction of DUX4+ nuclei increased from 0.3% to 2.1% following UPF1 KD, a substantial order-of-magnitude increase (Figure 3G). Together, our data show that NMD is an endogenous suppressor of DUX4 mRNA levels that contributes to the very low and variegated expression of DUX4, a characteristic feature of FSHD muscle cells.

As DUX4 expression inhibits NMD and NMD degrades DUX4 mRNA, we hypothesized that DUX4 and the NMD pathway might participate in a double-negative feedback loop (Figure 4A). This feedback loop predicts that DUX4 will indirectly stabilize its own mRNA by inhibiting NMD. To test this, we ectopically expressed DUX4 in FSHD1 and FSHD2 myotubes and measured levels of endogenously transcribed DUX4 mRNA. Ectopic DUX4 expression led to an approximately fivefold increase in endogenously transcribed DUX4 mRNA levels (Figure 4B). We next tested whether DUX4's spliced 3′ UTR, which is important for NMD-mediated degradation of DUX4 mRNA, contributed to this increase. We transfected our chimeric β-globin + DUX4 3′ UTR reporters into normal myoblasts and ectopically expressed DUX4. Levels of the NMD-susceptible construct containing the complete DUX4 3′ UTR increased 1.43-fold following ectopic DUX4 expression, while levels of the construct without the second intron of the DUX4 3′ UTR exhibited a more modest increase of 1.08-fold (Figure 4C). As with the UPF1 KD experiments, the chimeric construct exhibited more modest effect sizes in these feedback loop experiments than we observed for the endogenous DUX4 mRNA itself.

DUX4 and NMD form a feedback loop.

(A) Schematic of potential double-negative feedback loop between DUX4 and NMD, in which DUX4 inhibits NMD and NMD degrades DUX4 mRNA. (B) Levels of endogenously transcribed DUX4 mRNA following control treatment or ectopic DUX4 expression, measured 2 days after the initiation of myogenesis in FSHD1 (54-2) and FSHD2 (MB200) cells. (C) Relative levels of transcripts from the Gl-DUX4-Intron2(+) and Gl-DUX4-Intron2(−) constructs following control treatment or ectopic DUX4 expression in normal myoblasts (54-1 cells). For each construct, data are normalized such that the siControl treatment is 1. Error bars, standard deviation. (D) Schematic of potential model of interactions between DUX4 and NMD in healthy (top) and FSHD (bottom) muscle cells. In healthy cells, DUX4 mRNA is efficiently degraded by NMD; in FSHD cells, DUX4 triggers proteolytic degradation of UPF1 and inhibits NMD, resulting in the accumulation of DUX4 mRNA and protein.

https://doi.org/10.7554/eLife.04996.007

Together, our data demonstrate that the DUX4 3′ UTR targets DUX4 mRNA for NMD and that DUX4-mediated inhibition of NMD results in increased perdurance of the DUX4 mRNA as a possible mechanism of positive autoregulation (Figure 4D). It is unclear whether NMD-mediated autoregulation is intrinsic to normal DUX4 function or instead an abnormal consequence of inappropriate DUX4 expression in skeletal muscle. However, it is interesting to consider that this mechanism might contribute to the spreading of DUX4 expression between adjacent nuclei in a muscle fiber. Because muscle fibers contain arrays of closely spaced nuclei, the expression of DUX4 mRNA from one nucleus will distribute protein to the surrounding nuclei and induce a region of NMD inhibition. If one of the surrounding nuclei subsequently expresses DUX4, then that mRNA would be unusually stable due to locally inefficient NMD, thereby facilitating the spread of DUX4 mRNA and protein throughout the fiber.

The close temporal coupling between the onset of DUX4 expression, decreases in UPF1 protein, and increases in NMD substrates (Figure 2) strongly suggests that DUX4-mediated degradation of UPF1 contributes to DUX4-induced NMD inhibition. In the absence of a direct mechanistic link between UPF1 degradation and NMD inhibition, we were unable to determine whether insufficient UPF1 protein levels are primarily responsible for DUX4-induced NMD inhibition or instead merely one of the several contributing factors. Nonetheless, as we are unaware of other reports of physiological stimuli triggering rapid UPF1 protein degradation, our data suggest that UPF1 proteolysis constitutes a potential new regulator of cellular NMD efficiency. DUX4 may prove a useful system to gain insight into the biological relevance of this mechanism for altering NMD efficiency. We previously observed that many of the most up-regulated genes following DUX4 expression in normal myoblasts are involved in the ubiquitin–proteasome system, including numerous E3 ubiquitin ligases (Geng et al., 2012). It is therefore tempting to speculate that DUX4-induced dysregulation of the ubiquitin–proteasome system is responsible for triggering UPF1 protein degradation. However, the precise mechanism by which DUX4 induces UPF1 proteolysis, and whether that mechanism is specific to the FSHD disease state, remains to be elucidated.

DUX4-mediated inhibition of NMD may contribute to FSHD pathophysiology through both cell autonomous and non-cell autonomous mechanisms. The accumulation of abnormal RNAs may cause direct or indirect toxic effects in muscle cells due to intrinsic toxicity of abnormal RNAs or a stress response to the production of abnormal proteins. FSHD muscle is frequently characterized by a T-cell infiltrate (Arahata et al., 1995), and it is possible that stabilized NMD substrates encode novel peptides with antigenic potential, contributing to an immune response (Pastor et al., 2010). Production of antigenic peptides could potentially enable even a small fraction of DUX4+ nuclei to induce widespread pathology within a muscle fiber. Directly detecting or measuring DUX4-induced NMD inhibition in FSHD muscle biopsies or in bulk populations of cultured FSHD muscle cells is not feasible due to the low fraction of DUX4+ nuclei present at any given time in the absence of ectopic DUX4 expression. With DUX4+ nuclei constituting only 0.3% of the bulk population of cultured FSHD muscle cells (Figure 3G), changes in the ratios of NMD and non-NMD isoforms in these DUX4+ nuclei are swamped by the normal levels expressed by the vast majority of DUX4− nuclei. Single-cell assays of NMD efficiency are likely required to demonstrate DUX4-induced NMD inhibition in unperturbed patient cells. However, future efforts to identify the downstream antigenic products or toxic effects of stabilized NMD substrates may prove fruitful even in a bulk cell population.

Consistent with the idea that NMD inhibition may contribute to DUX4 toxicity in skeletal muscle, it is interesting to note that the degree of NMD inhibition induced by DUX4 is comparable to that observed in previous studies involving genetic ablation of components of the NMD machinery. For example, a recent study of mouse embryonic fibroblasts lacking Smg1, which encodes a kinase responsible for phosphorylating UPF1, found that 9% of predicted NMD substrates created by alternative splicing exhibited increased levels relative to wild-type cells (McIlwain et al., 2010). In comparison, we found that 13% of such substrates were up-regulated following DUX4 expression (Figure 1), suggesting that DUX4-induced NMD inhibition causes profoundly abnormal RNA metabolism. As RNA toxicity is the major pathophysiologic mechanism in myotonic dystrophy, it is interesting to consider that RNA-mediated disease mechanisms may also have important roles in FSHD.

Materials and methods

Accession codes

FASTQ files for the DUX4 expression experiments were downloaded from the NCBI GEO database under accession number GSE45883 (Young et al., 2013).

Genome annotations

The UCSC knownGene (Meyer et al., 2013) and Ensembl 71 (Flicek et al., 2013) genome annotations were merged to create a single genome annotation. Splicing event annotations from MISO v2.0 (Katz et al., 2010) were then added to this merged genome annotation. Constitutive splice junctions were defined as those for which neither the 5′ nor 3′ splice site was alternatively spliced in the UCSC knownGene annotation. Predicted NMD substrates were annotated by identifying isoforms containing premature termination codons >50 nt upstream of a splice junction. For the purposes of predicting NMD substrates, open reading frames were assigned based on UniProt annotations (UniProt Consortium, 2012) when available, and Ensembl predicted reading frames when UniProt annotations were not available. For the purposes of RNA-seq read mapping, an additional annotation file consisting of all splice junctions annotated in the UCSC, Ensembl 71, and MISO v2.0 annotations was created. This splice junction file was then with a list of all possible junctions between the annotated 5′ and 3′ splice sites of isoforms in the annotation (to detect novel alternative splicing).

RNA-seq read mapping

Reads were mapped to the UCSC hg19 (NCBI GRCh37) genome assembly. RSEM (Li and Dewey, 2011) was modified to call Bowtie (Langmead et al., 2009) v1.0.0 with the -v 2 argument. RSEM was then called with the arguments --bowtie-m 100 --bowtie-chunkmbs 500 --calc-ci --output-genome-bam on the genome annotation. Read alignments with mapq scores of 0 and/or a splice junction overhang of less than 6 bp were then filtered out. Remaining unaligned reads were then aligned TopHat (Trapnell et al., 2009) v2.0.8b with the arguments --bowtie1 --read-mismatches 2 --read-edit-dist 2 --no-mixed --no-discordant --min-anchor-length 6 --splice-mismatches 0 --min-intron-length 10 --max-intron-length 1000000 --min-isoform-fraction 0.0 --no-novel-juncs --no-novel-indels --raw-juncs on the splice junction file (--mate-inner-dist and --mate-std-dev were calculated by mapping to constitutive coding exons with MISO's exon_utils.py utility). Alignments produced by this call to TopHat were then filtered identically to the alignments produced by RSEM. Reads aligned by RSEM and TopHat were then merged to create BAM files of all aligned reads.

Gene expression and isoform ratio measurements

Gene expression was quantified using RSEM as described above. Isoform ratios were quantified using two distinct methods. First, MISO was used to quantify isoform ratios for alternative splicing events contained in MISO's v2.0 annotations. Second, novel alternative splicing or intron retention of annotated constitutive splice junctions was quantified using reads crossing the 5′ or 3′ splice sites as previously described (Hubert et al., 2013). Differentially spliced events were defined as those with at >20 identifying reads (identifying reads support one or more, but not all, isoforms of a splicing event), a change in isoform ratio ≥10%, and a Bayes factor ≥5 (computed with Wagenmakers's framework [Wagenmakers et al., 2010]).

Viral infection and cell culture

We used previously described lentiviral constructs expressing full-length DUX4 or GFP as a control (Geng et al., 2012). Lentiviral particles were generated by the FHCRC Core Center of Excellence in Hematology Vector Production Core. Viral particle number was estimated with the WPRE element within the viral vector. Myoblasts were transduced at a MOI of ∼15 in the presence of 8 µg/ml polybrene. At this MOI, >85% of myoblasts were DUX4+ or GFP+. Unless otherwise noted, cells were collected for analysis 24 hr post-infection. Proliferating myoblasts were cultured in F-10-based growth media (Gibco, Carlsbad, CA) with 20% fetal bovine serum (Gibco) and 1% penicillin/streptomycin (Gibco), supplemented with 10 ng/ml rhFGF (Promega, Madison, WI) and 1 µM dexamethasone (Sigma, St. Louis, MO). Growth media was changed every other day, and proliferating myoblasts were cultured at ≤50% confluence. To initiate myogenic differentiation, myoblasts were switched to an F-10-based differentiation media containing 1% horse serum (Gibco) and 1% penicillin/streptomycin, supplemented with 10 µg/ml insulin (Sigma) and 10 µg/ml transferrin (Sigma) at 99% confluence.

Plasmid and siRNA transfection

The β-globin NMD(−) and NMD(+) plasmids were previously published as pmCMV-Gl Norm and pmCMV-Gl 39Ter (Zhang et al., 1998). Plasmid reporters were transfected with Lipofectamine 2000 (Life Technologies, Carlsbad, CA), unless otherwise noted. To control for transfection efficiency, a control plasmid phCMV-MUP was co-transfected with the reporter as previously described (Zhang et al., 1998). 2 µg of reporter along with 500 ng of control plasmid was used for transfecting cells in a six-well format. To measure DUX4-induced changes in NMD efficiency, cells were infected with lentiviral DUX4 or GFP 24 hr after transfection of the NMD reporters. For the DUX4 time course experiments, the NMD(−) and NMD(+) reporters were transfected along with phCMV-MUP using the SuperFect reagent (Qiagen, Valencia, CA), and the lentiviral transduction was performed 12 hr post-transfection. siRNAs against UPF1 (Thermo Scientific, Waltham, MA, On-Target siRNA #J-011763-07) and TP53 (Ambion, Silencer Select siRNA #4390824), as well as the control siRNA (Thermo Scientific), were transfected with Lipofectamine RNAiMAX (Life Technologies).

RNA isolation, real-time qPCR, and endogenous DUX4 mRNA measurement

Cells were lysed with TRIzol (Invitrogen, Carlsbad, CA) and the RNA was extracted according to the manufacturer's instructions. RNA was subsequently cleaned up with Qiagen RNeasy columns, with on-column DNase digestion. 1 µg of RNA was used for cDNA synthesis with Life Technologies SuperScriptIII First-Strand System. 2% of this cDNA was used as a template for real-time qPCR with Life Technologies Power SYBR Green Master Mix. qPCR primer sequences are provided in Supplementary file 1. Note that levels of endogenous DUX4 mRNA following ectopic DUX4 expression were measured with primers specific to the DUX4 mRNA's 3′ UTR (the DUX4 lentiviral construct contained the coding sequence alone). To determine how UPF1 KD affected DUX4 expression, proliferating 54-2 or MB200 myoblasts were transfected with siUPF1 or siControl and switched to differentiation media 48 hr post-transfection.

Western blotting

Protein extracts from the UPF1 KD experiments were generated by lysing cell pellets in protease inhibitor cOmplete ULTRA (Roche, Switzerland) containing RIPA buffer (Cell Signaling Technology, Danvers, MA) along with sonication. For the DUX4 time course and MG132 treatment, protein was extracted in parallel with the RNA from cells lysed in the TRIzol reagent. Protein pellets were resuspended in a sample buffer containing 5% SDS and 0.5 M unbuffered Tris base to ensure efficient solubilization. Protein concentrations were determined using the Bradford or BCA protein assay. 5 µg of total protein was used for Western blotting. Antibodies used in this study are: anti-UPF1 (Bethyl Laboratories, Montgomery, TX), anti-α-tubulin (Sigma), anti-H3 (Abcam, England), anti-UPF3B (Bethyl Laboratories), anti-SMG7 (Santa Cruz, Dallas, TX). HRP-conjugated (Jackson ImmunoResearch, West Grove, PA) secondary antibodies were used for protein detection in all experiments except for the time course and proteasome inhibitor studies (Figure 2). For the experiments reported in Figure 2, immunoblotting was performed using the LICOR system with the Odyssey blocking buffer and IRDye-conjugated secondary antibodies (LICOR, Lincoln, NE). Quantification was performed using ImageQuant software (GE Healthcare, Cleveland, OH) using the nonspecific band as a normalizer to account for differences in protein loading. Histone 3 served as an additional loading control, though its very high signal intensity made it an inappropriate normalizer for quantitative analyses.

Proteasome inhibition

54-1 cells transduced with DUX4 or GFP lentivirus were treated 16 hr post-infection with 10 µM proteasome inhibitor MG132 (Sigma). Samples were collected 8 hr after MG132 treatment, and UPF1 levels were estimated by immunoblotting. Histone H3, which has a long half-life (Toyama et al., 2013), was used as a loading control, in addition to the nonspecific band.

Fluorescence microscopy and quantification

Cells were permeabilized with PBS containing 0.5% Triton X-100, rinsed in PBS, and blocked in 1% BSA. Primary antibody against DUX4 (Abcam, ab124699) was diluted in blocking buffer at 1:500, and secondary anti-Rabbit TRITC (Jackson ImmunoResearch, 711-025-152) was diluted in blocking buffer at 1:400. For assaying apoptosis in DUX4-cells, Image-iT LIVE Red Poly Caspases Detection Kit (Life Technologies, I35101) was used. For both experiments, fluorescently labeled cells were then viewed under the ZEISS Axiophot fluorescence microscope. For each sample, pictures from eight random fields were taken. ImageJ (Fiji) was used for image analysis and quantification.

Cloning of chimeric β-globin + DUX4 3′ UTR constructs

The genomic locus of the DUX4 3′ UTR (containing both introns) was amplified from a genomic fragment harboring 2.5 D4Z4 repeats (Gabriëls et al., 1999) (L42 clone; GenBank ID FJ439133.1). The DUX4 3′ UTR lacking the second intron was amplified from cDNA isolated from differentiated 54-2 cells (the first intron is frequently retained in DUX4 cDNA). The β-globin NMD(−) reporter backbone was linearized by forward and reverse PCR primers sitting downstream and upstream of the β-globin 3′ UTR (primers listed in Supplementary file 1). Amplicons of the DUX4 3′ UTR containing or lacking the second intron were flanked with sequences overlapping the linearized β-globin NMD(−) backbone lacking the β-globin 3′ UTR. The NEB Gibson Assembly Cloning Kit was used to insert the DUX4 3′ UTR fragments into the linearized β-globin NMD(−) backbone (New England Biolabs, Ipswich, MA).

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32

Decision letter

  1. Rachel Green
    Reviewing Editor; HHMI, Johns Hopkins University School of Medicine, United States

eLife posts the editorial decision letter and author response on a selection of the published articles (subject to the approval of the authors). An edited version of the letter sent to the authors after peer review is shown, indicating the substantive concerns or comments; minor concerns are not usually shown. Reviewers have the opportunity to discuss the decision before the letter is sent (see review process). Similarly, the author response typically shows only responses to the major concerns raised by the reviewers.

Thank you for sending your work entitled “A feedback loop between DUX4 and nonsense-mediated decay” for consideration at eLife. Your article has been favorably evaluated by James Manley (Senior editor), a Reviewing editor, and 2 reviewers.

The Reviewing editor and the reviewers discussed their comments before we reached this decision, and the Reviewing editor has assembled the following comments to help you prepare a revised submission.

In this short paper, the authors provide evidence for reduced NMD activity playing a role in Facioscapulohumeral muscular dystrophy (FSHD). Previous work had shown that mis-expression in skeletal muscle of the double homeobox transcription factor DUX4 causes FSHD. However, little was known about the pathways by which DUX4 expression in muscle cells leads to the muscle damage observed in FSHD. Here, the authors show that ectopic expression of DUX4 in immortalized and primary myoblast cell lines results in increased levels of transcripts that are known or predicted to be targets of NMD, suggesting an overall reduction of NMD activity. The authors speculate that this reduced NMD activity might lead to the expression of aberrant peptides in muscles that could either be toxic or activate an immune response against the muscle cells. The authors further show that concomitant with DUX4 expression the abundance of UPF1 protein is reduced (but not the mRNA). Since treatment of the cells with the proteasome inhibitor MG132 prevented the DUX4-induced UPF1 protein reduction, while MG132 had no effect on UPF1 abundance in cells that did not express DUX4, the authors concluded that DUX4 expression in muscle cells causes the proteasome-mediated degradation of UPF1.

In the second part of the paper the authors point out that the two DUX4 mRNA isoforms both contain introns in the 3' UTR, which is a well-known feature of NMD targets. They go on the show that the DUX4 mRNA is indeed targeted by NMD and that its 3' UTR is sufficient to induce NMD of a reporter gene. Together with the first part of the data, this results provide a possible explanation how occasional bursts of DUX4 transcription can trigger and enhance DUX4 mis-expression in these myoblast regions by an autoregulatory feedback loop that gradually inactivates NMD in the surrounding nuclei: DUX4-induced UPF1 degradation lowers NMD activity, which in turn leads to more DUX4 mRNA and hence protein and thus further downregulation of NMD. This is an attractive model, and the data for the feedback loop between DUX4 expression and NMD is compelling.

Overall, the manuscript is of considerable and broad interest and worthy of publication in eLife. However, there were several substantial concerns of the reviewers that should be addressed at least in the language of the manuscript:

1) The effects were overall modest, in particular in experiments with the 3' UTR intron and in the assessment of the fraction of nuclei expressing DUX4 protein in FHSD myotubes.

2) Lack of connection between observations in primary myoblast cell lines with overexpressed DUX4 and with authentic patient samples—perhaps too difficult to address, but perhaps it would be possible to ask whether authentic FSHD muscle cells contain increased proportions of predicted NMD substrates.

3) Lack of a potential molecular mechanism to explain how overexpression of a transcription factor leads to selective degradation of a set of factors involved in NMD. One possibility would be to more thoroughly evaluate the levels of other NMD factor (UPF2 etc.) or components of the proteasome itself. Without elucidation of some mechanistic link, this part of the story remains purely correlative and therefore less impactful.

Other more specific points follow:

1) Why is only a small fraction (13%) of predicted NMD targets affected by this mechanism?

2) Figure 2E: Can the authors explain the transient nature of the upregulation of the PTC-containing beta-globin NMD reporter? In the light of the proposed feedback loop, one would anticipate robust and enduring inhibition of NMD (as is observed for SRSF2 and SRSF3 transcripts; Figure 2F) and thus the beta-globin PTC+ levels to remain elevated after 20 hours post transduction with the DUX4 expressing lentivirus.

3) Figure 2E: The data points for the half-life measurement of DUX4 mRNA under siControl conditions scatter quite a lot; the authors should describe how they did the curve fitting. The conventional way would be to plot the RNA levels logarithmically and calculate the linear regression line through these data points. Furthermore, by limiting the time course to only 2 hours, the extrapolated half-life of 7.86 h in UPF1-depleted cells is most likely highly inaccurate. The time course should be expanded to 6 or 8 hours to get a more precise estimate of this half-life.

https://doi.org/10.7554/eLife.04996.009

Author response

1) The effects were overall modest, in particular in experiments with the 3' UTR intron and in the assessment of the fraction of nuclei expressing DUX4 protein in FHSD myotubes.

We agree that the experiments with the chimeric Gl-DUX4-Intron2 constructs generally yielded modest effect sizes, such as increases of 2- and 1.4-fold following UPF1 KD or ectopic DUX4 expression (Figure 3G, 4C). These effects were substantially smaller than the corresponding 4- and 5-fold increases in endogenous DUX4 mRNA levels that we observed following UPF1 KD or ectopic DUX4 expression. We are unsure why the chimeric construct exhibited more modest effect sizes than did the endogenous DUX4 mRNA, although it may be due to a combination of its plasmid-based nature and the absence of sequence features of the DUX4 mRNA other than its 3' UTR intron that contribute to its degradation by NMD. As the Gl-DUX4-Intron2 constructs are chimeric, we feel that the most realistic estimates of effect sizes stem from our measurements of the endogenous DUX4 mRNA rather than the Gl-DUX4-Intron2 constructs.

Outside of the chimeric construct, however, many of the other changes that we observed were quite dramatic. For example, the reviewers mention the fraction of DUX4+ nuclei in a FSHD myotube culture following control or UPF1 knock down. DUX4 exhibits a highly variegated expression pattern, with DUX4 detectable in only 0.1-1% of FSHD muscle cells (Snider et al., PLoS Genetics, 2010; Figure 3 from this manuscript). Following UPF1 knock down, we found that the fraction of DUX4+ nuclei increased from 0.3% to 2.1%, a change of an order of magnitude. As the extremely low fraction of DUX4+ cells in FSHD muscle is a defining characteristic of the disease, we feel that an order-of-magnitude increase in the fraction of DUX4+ cells is a substantial effect size, consistent with NMD playing an important role in the regulation of DUX4 levels. We have emphasized this effect size in the revised manuscript, and also updated Figure 3 to show additional microscopy fields.

To place these results in context, it is useful to compare our observations with previous studies where NMD was directly inhibited. For example, a recent study (Hurt et al., Genome Research, 2013), inhibited NMD with shRNAs against UPF1 or cycloheximide treatment, which prevents protein translation and thereby NMD. Hurt et al. then used RNA-seq to quantify global mRNA levels. Hurt et al. identified only 8, 14, or 40 predicted NMD substrates out of >20,000 queried mRNAs that were up-regulated by at least 4-fold in cells treated with one of two distinct shRNAs against UPF1 (8 and 14 mRNAs identified) or cycloheximide (40 mRNAs identified). In the context of these results, we feel that the 4- and 5-fold increases in endogenous DUX4 mRNA levels that we observe following UPF1 knock down or ectopic DUX4 expression are quite substantial.

As discussed further below in specific point #1, we also note that the degree of NMD inhibition that is caused by DUX4 expression is quite remarkable, matching the degree of NMD inhibition caused by genetic knock out of essential NMD factors.

2) Lack of connection between observations in primary myoblast cell lines with overexpressed DUX4 and with authentic patient samples—perhaps too difficult to address, but perhaps it would be possible to ask whether authentic FSHD muscle cells contain increased proportions of predicted NMD substrates.

We agree that our conclusions would be even further strengthened by a direct demonstration that DUX4 causes NMD inhibition in primary patient materials, in the absence of ectopic DUX4 expression. However, DUX4’s variegated expression pattern—with DUX4 detectable in only 0.1-1% of FSHD muscle cells (Snider et al., PLoS Genetics, 2010; Figure 3 from this manuscript)—typically renders such experiments infeasible. For this reason, most studies of DUX4’s molecular function have relied upon ectopic expression to ensure that the majority of assayed cells are DUX4+.

The challenge of detecting DUX4-induced NMD inhibition given DUX4’s variegated expression pattern can be readily seen with a statistical power estimation. For example, we used a poison exon-containing isoform of SRSF3 as an endogenous marker of NMD. This isoform increased dramatically in abundance (from ∼10% to 65% of SRSF3’s mRNA population) following ectopic DUX4 expression. It is therefore reasonable to expect that DUX4+ cells in a FSHD muscle cell culture may similarly express abnormally high levels of this SRSF3 NMD substrate in the absence of ectopic DUX4 expression. However, the variegated nature of DUX4 expression prevents the direct detection of such signals of NMD inhibition simply due to statistical power limitations. DUX4 expression is detectable in 0.3% of FSHD nuclei by immunostaining (Figure 3H). Assume that the SRSF3 NMD substrate is present at isoform ratios of 10% and 65% in DUX4- and DUX4+ cells, respectively. Measurements of the SRSF3 NMD substrate in a bulk FSHD cell population would therefore yield an isoform ratio of (0.003 x 65%) + (0.997 x 10%) = 10.165%. The difference between an isoform ratio of 10% for DUX4- cells and 10.165% for a mixture of DUX4+ and DUX4- cells is not detectable with RNA-seq or quantitative PCR.

In the future, we may be able to overcome the challenges presented by DUX4’s variegated expression pattern by studying single cells with assays such as single-cell RNA-seq or live cell imaging approaches. However, we feel that such assays are beyond the scope of this manuscript. We have added additional text to the discussion in which we briefly describe the challenges with detecting NMD inhibition in a bulk population of FSHD cells.

3) Lack of a potential molecular mechanism to explain how overexpression of a transcription factor leads to selective degradation of a set of factors involved in NMD. One possibility would be to more thoroughly evaluate the levels of other NMD factor (UPF2 etc.) or components of the proteasome itself. Without elucidation of some mechanistic link, this part of the story remains purely correlative and therefore less impactful.

We agree that this aspect of the manuscript is correlative. Given the close temporal connection between decreased UPF1 protein levels and increased levels of NMD substrates, we do feel that it is reasonable to speculate that UPF1 degradation is a likely contributor to NMD inhibition. In the revised manuscript, we are careful to state clearly that the data is suggestive but purely correlative. Nonetheless, as rapid UPF1 protein degradation has not been previously described to our knowledge, we feel that this section of the manuscript is a significant addition to the field, as it suggests a potential new mechanism for the cell to alter RNA surveillance efficiency.

In a previous study (Geng et al., Developmental Cell, 2012), we noticed that many genes that were differentially expressed following DUX4 expression were involved in protein ubiquitination, including numerous E3 ubiquitin ligases. It is possible, for example, that UPF1 is a substrate of one of the many E3 ligases that are up-regulated in DUX4-expressing muscle cells. While we do hope to eventually determine how DUX4 expression triggers UPF1 degradation, we feel that such assays are beyond the scope of this manuscript given the many different ways in which a transcription factor could indirectly influence UFP1 proteolysis. We briefly discuss this in the revised manuscript.

Other more specific points follow:

1) Why is only a small fraction (13%) of predicted NMD targets affected by this mechanism?

We agree that it is interesting that only a minority of premature stop codon-containing mRNAs exhibit increased levels following DUX4 expression and concomitant NMD inhibition. This could be due to imperfect annotation of NMD substrates, differential sensitivity of different NMD substrates to reduced UPF1 levels, or other factors.

It is important to note that this 13% is quite high when compared to other studies where NMD has been inhibited genetically or with RNAi. For example, a recent study (McIlwain et al., PNAS, 2010) used a gene trap to generate mouse embryonic fibroblasts lacking Smg1, which is responsible for UPF1 phosphorylation, a required step in NMD. McIlwain et al. then used RNA-seq to characterize the transcriptomes of Smg1-deficient cells, finding that “in the absence of Smg1, approximately 9% of predicted PTC-containing AS events displayed changes of more than 10%, and approximately 2% showed changes of more than 20%” (PTC = premature termination codon; AS = alternative splicing). Other studies conducted in worms, flies, and mammalian cells found that similarly small fractions of predicted NMD substrates were affected by depletion of core NMD factors.

Therefore, we agree that it is interesting that only 13% of predicted NMD substrates are up-regulated in DUX4-expressing cells. However, as this degree of up-regulation is comparable to that observed in previous studies where NMD was directly inhibited genetically, we believe that this 13% is a remarkably high fraction. We comment on this in the revised manuscript.

2) Figure 2E: Can the authors explain the transient nature of the upregulation of the PTC-containing beta-globin NMD reporter? In the light of the proposed feedback loop, one would anticipate robust and enduring inhibition of NMD (as is observed for SRSF2 and SRSF3 transcripts; Figure 2F) and thus the beta-globin PTC+ levels to remain elevated after 20 hours post transduction with the DUX4 expressing lentivirus.

As the reviewers point out, levels of the NMD(+) β-globin reporter increase at the same time (∼12 h following infection) as the endogenous SRSF2 and SRSF3 markers, but the NMD(+) β-globin reporter up-regulation does not last for the entire time course. This is surprising because (1) SRSF2 and SRSF3 NMD substrates exhibit increased levels throughout the time course, (2) UPF1 levels decrease monotonically throughout the time course, and (3) our RNA-seq data from MB135 and 54-1 cells was obtained at 24 hours and 48 hours time points, respectively, and both samples exhibited obvious global increases in NMD substrate levels. Therefore, while we are confident that NMD inhibition occurs throughout the time course, we are unsure why this enduring behavior is not reflected in the NMD(+) β-globin reporter. We are unaware of a previous study that has used this reporter to study the dynamics of NMD efficiency along a time course. For unclear reasons, such a plasmid-based reporter may be a more useful tool for measuring NMD efficiency in a static, rather than dynamic, context.

3) Figure 2E: The data points for the half-life measurement of DUX4 mRNA under siControl conditions scatter quite a lot; the authors should describe how they did the curve fitting. The conventional way would be to plot the RNA levels logarithmically and calculate the linear regression line through these data points. Furthermore, by limiting the time course to only 2 hours, the extrapolated half-life of 7.86 h in UPF1-depleted cells is most likely highly inaccurate. The time course should be expanded to 6 or 8 hours to get a more precise estimate of this half-life.

To estimate the half-life of DUX4 mRNA, we computed the best-fit exponential model using untransformed data, which is mathematically identical to the reviewers’ suggestion of computing a best-fit linear model for log-transformed data. We apologize for not explaining this half-life estimation procedure more clearly in the initial submission.

We agree that the data points exhibit substantial variation, particularly for the siControl treatment. Unfortunately, this high level of variability is an unavoidable consequence of DUX4’s variegated expression. Since DUX4 is detectable in only 0.3% of nuclei by immunostaining for the siControl treatment (Figure 3), DUX4 mRNA levels are very low even in large cultures of FSHD cells. Estimates of DUX4 mRNA levels therefore are noisy and become increasingly so following transcription shutoff.

We agree with the reviewers that extrapolating a ∼8 hour half-life from a 2 hour time course is prone to substantial measurement error. We accordingly repeated the DUX4 mRNA half-life measurement multiple times with a longer 8 hour time course. However, due to the aforementioned difficulty of accurately measuring DUX4 mRNA following transcription shutoff even in a large culture of FSHD muscle cells, we were unfortunately unable to obtain reliable data from the longer time courses.

As the DUX4 mRNA half-life measurement was not an essential component of the manuscript—since we provide multiple other sources of evidence that DUX4 mRNA is a NMD substrate, due in part to its spliced 3' UTR—we have decided to remove the half-life data entirely from the revised manuscript. We agree with the reviewers that this data is noisy, and feel that it is a distraction from the otherwise clean data that we present here.

https://doi.org/10.7554/eLife.04996.010

Article and author information

Author details

  1. Qing Feng

    1. Computational Biology Program, Public Health Sciences Division, Fred Hutchinson Cancer Research Center, Seattle, United States
    2. Basic Sciences Division, Fred Hutchinson Cancer Research Center, Seattle, United States
    Contribution
    QF, Performed NMD inhibition experiments and computational analyses, and created Figures 1, 3, and 4
    Competing interests
    The authors declare that no competing interests exist.
  2. Lauren Snider

    1. Human Biology Division, Fred Hutchinson Cancer Research Center, Seattle, United States
    Contribution
    LS, Performed NMD inhibition experiments
    Competing interests
    The authors declare that no competing interests exist.
  3. Sujatha Jagannathan

    1. Computational Biology Program, Public Health Sciences Division, Fred Hutchinson Cancer Research Center, Seattle, United States
    2. Basic Sciences Division, Fred Hutchinson Cancer Research Center, Seattle, United States
    3. Human Biology Division, Fred Hutchinson Cancer Research Center, Seattle, United States
    Contribution
    SJ, Identified UPF1 proteolysis and created Figure 2
    Competing interests
    The authors declare that no competing interests exist.
  4. Rabi Tawil

    1. Department of Neurology, University of Rochester, Rochester, United States
    Contribution
    RT, Provided critical biological reagents and information
    Competing interests
    The authors declare that no competing interests exist.
  5. Silvère M van der Maarel

    1. Department of Human Genetics, Leiden University Medical Center, Leiden, Netherlands
    Contribution
    SMM, Provided critical biological reagents and information
    Competing interests
    The authors declare that no competing interests exist.
  6. Stephen J Tapscott

    1. Human Biology Division, Fred Hutchinson Cancer Research Center, Seattle, United States
    2. Department of Neurology, University of Washington, Seattle, United States
    Contribution
    SJT, Provided overall oversight of the project
    For correspondence
    1. stapscot@fhcrc.org
    Competing interests
    The authors declare that no competing interests exist.
  7. Robert K Bradley

    1. Computational Biology Program, Public Health Sciences Division, Fred Hutchinson Cancer Research Center, Seattle, United States
    2. Basic Sciences Division, Fred Hutchinson Cancer Research Center, Seattle, United States
    Contribution
    RKB, Performed computational analyses and provided overall oversight of the project
    For correspondence
    1. rbradley@fhcrc.org
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon 0000-0002-8046-1063

Funding

Damon Runyon Cancer Research Foundation (DFS 04-12)

  • Robert K Bradley

Ellison Medical Foundation (AG-NS-1030-13)

  • Robert K Bradley

Friends of FSH Research

  • Stephen J Tapscott

National Institutes of Health (R01AR045203)

  • Stephen J Tapscott

National Institutes of Health (P01NS069539)

  • Stephen J Tapscott

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Lynne Maquat for sharing the β-globin NMD reporter and MUP transfection control plasmids. This research was supported by the Damon Runyon Cancer Research Foundation DFS 04-12 (RKB), the Ellison Medical Foundation AG-NS-1030-13 (RKB), Fred Hutchinson Cancer Research Center institutional funds (RKB), Friends of FSH Research (SJT), NIH/NIAMS R01AR045203 (SJT), and NIH/NINDS P01NS069539 (SJT).

Reviewing Editor

  1. Rachel Green, Reviewing Editor, HHMI, Johns Hopkins University School of Medicine, United States

Publication history

  1. Received: October 1, 2014
  2. Accepted: January 7, 2015
  3. Accepted Manuscript published: January 7, 2015 (version 1)
  4. Accepted Manuscript updated: January 15, 2015 (version 2)
  5. Version of Record published: February 5, 2015 (version 3)

Copyright

© 2015, Feng et al

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 2,264
    Page views
  • 339
    Downloads
  • 14
    Citations

Article citation count generated by polling the highest count across the following sources: PubMed Central, Crossref, Scopus.

Comments

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)