FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W4381612243> ?p ?o ?g. }

• W4381612243 abstract "Full text Figures and data Side by side Abstract Editor's evaluation Introduction Results Discussion Methods Data availability References Decision letter Author response Article and author information Metrics Abstract Facioscapulohumeral muscular dystrophy (FSHD) is an incurable myopathy linked to the over-expression of the myotoxic transcription factor DUX4. Targeting DUX4 is the leading therapeutic approach, however, it is only detectable in 0.1–3.8% of FSHD myonuclei. How rare DUX4 drives FSHD and the optimal anti-DUX4 strategy are unclear. We combine stochastic gene expression with compartment models of cell states, building a simulation of DUX4 expression and consequences in FSHD muscle fibers. Investigating iDUX4 myoblasts, scRNAseq, and snRNAseq of FSHD muscle we estimate parameters including DUX4 mRNA degradation, transcription and translation rates, and DUX4 target gene activation rates. Our model accurately recreates the distribution of DUX4 and targets gene-positive cells seen in scRNAseq of FSHD myocytes. Importantly, we show DUX4 drives significant cell death despite expression in only 0.8% of live cells. Comparing scRNAseq of unfused FSHD myocytes to snRNAseq of fused FSHD myonuclei, we find evidence of DUX4 protein syncytial diffusion and estimate its rate via genetic algorithms. We package our model into freely available tools, to rapidly investigate the consequences of anti-DUX4 therapy. Editor's evaluation To provide a logical answer to the over-expressed DUX4 in FSHD, the authors took a sophisticated mathematical modeling approach and applied it to empirical data. The approach successfully predicts behaviors of proteins and cells, thereby suggests a model for pathogenicity. The result poses a potential to be expanded to understand molecular dynamics of other mutation-mediated rare diseases. https://doi.org/10.7554/eLife.88345.sa0 Decision letter eLife's review process Introduction FSHD is a prevalent (~12/100,000 Deenen et al., 2014), incurable, inherited skeletal myopathy. The condition is characterized by progressive fatty replacement and fibrosis of specific muscle groups driving weakness and wasting (Banerji and Zammit, 2021), which is accelerated by inflammation (Dahlqvist et al., 2020). FSHD is highly heterogeneous, with both the rate of progression and order of muscle involvement varying dramatically from person to person, even between monozygotic twins (Tawil et al., 1993). Around 75% of patients exhibit a descending phenotype with weakness beginning in the facial muscles, before progressing to the shoulder girdle and latterly lower limb, while the remaining 25% exhibit a range of ‘atypical’ phenotypes, including facial sparing and lower limb predominant (Banerji et al., 2020a). Despite this clinical range, FSHD is associated with significant morbidity and socioeconomic costs (Schepelmann et al., 2010). Genetically FSHD comprises two distinct subtypes: FSHD1 (OMIM: 158900, 95% of cases) and FSHD2 (OMIM: 158901, 5% of cases). Both subtypes bear a unifying epigenetic feature: derepression of the D4Z4 macrosatellite at chromosome 4q35. In FSHD1 this is due to truncation of the D4Z4 macrosatellite from the typical >100 to 10–1 units (Lemmers et al., 2010). In FSHD2, derepression is due to mutation in a chromatin modifier, typically SMCHD1 (Lemmers et al., 2012) but rarely DNMT3B (van den Boogaard et al., 2016) or LRIF1 (Hamanaka et al., 2020). In addition to D4Z4 epigenetic derepression, FSHD patients also carry certain permissive 4qA haplotypes distal to the last D4Z4 repeat encoding a polyadenylation signal (Lemmers et al., 2010). Each 3.3 kb D4Z4 repeat encodes the transcription factor DUX4, which plays a role in zygotic genome activation, after which it is silenced in somatic tissues (De Iaco et al., 2017). In FSHD, however, epigenetic derepression of the D4Z4 region allows inappropriate transcription of DUX4 from the most distal D4Z4 unit, with transcripts stabilized by splicing to the polyadenylation signal in 4qA haplotypes, allowing translation. Mis-expression of DUX4 protein is thus believed to underlie FSHD pathogenesis and DUX4 inhibition is currently the dominant approach to FSHD therapy (Tawil, 2020; Le Gall et al., 2020). However, DUX4 is extremely difficult to detect in FSHD patient muscle, with the vast majority of transcript and protein level studies failing to detect DUX4 in FSHD muscle biospies (Banerji and Zammit, 2021). When DUX4 is detected in FSHD patient muscle, it is at very low levels, requiring highly sensitive techniques such as nested RT-qPCR for transcripts (Jones et al., 2012) and proximity ligation assays for protein (Beermann et al., 2022). Investigation of FSHD patient-derived myoblasts has confirmed this very low level of DUX4 expression (Banerji and Zammit, 2021). Single-cell and single nuclear transcriptomic studies find only 0.5–3.8% of in vitro differentiated FSHD myonuclei express DUX4 transcript (Jiang et al., 2020; van den Heuvel et al., 2019). Immunolabelling studies only detect DUX4 protein in between 0.1–5% of FSHD myonuclei (Snider et al., 2010; Rickard et al., 2015). As DUX4 is a transcription factor, it has been proposed that DUX4 target genes may represent a key driver of FSHD pathology. However, multiple meta-analyses have found DUX4 target gene expression to be a poor biomarker of FSHD muscle, except in the context of significant inflammation (Banerji et al., 2017; Banerji and Zammit, 2019), where it may be confounded by immune cell gene expression (Banerji et al., 2020b). Importantly, a recent phase 2b clinical trial of the DUX4 inhibitor losmapimod failed to reach its primary endpoint of reduced DUX4 target gene expression in patient muscle, despite improvement in functional outcomes (Jagannathan et al., 2022). Given the challenge of detecting DUX4 in muscle biopsies (Banerji and Zammit, 2021) it is unsurprising that no data was published relating to DUX4 expression changes during the losmapimod clinical trial. An understanding as to why losmapimod did not alter the expression of DUX4 targets in patient muscle biopsies is also lacking, but hypotheses include heterogeneity in muscle sampling, low baseline levels of DUX4 targets and thus limited dynamic range, slow reversibility of DUX4-induced epigenetic changes on target gene promoters and losmapimod having limited impact on DUX4 transcriptional activity in vivo. How can such a rare expression of DUX4 drive a pathology as significant as FSHD? DUX4 expression in FSHD patient myoblasts follows a burst-like pattern, and cells expressing DUX4 have significantly shortened lifespans, suggesting a gradual attrition of cells over time (Rickard et al., 2015). A mouse model in which DUX4 expression is induced in a rare burst-like manner bears striking histological and transcriptomic similarities to FSHD patient muscle (Bosnakovski et al., 2017; Bosnakovski et al., 2020). DUX4 expression in FSHD patient-derived, multi-nucleated myotubes also displays a gradient across myonuclei, suggesting that DUX4 may ‘diffuse’ either actively or passively from an origin nucleus to neighboring nuclei, bypassing their need to wait for the rare DUX4 burst (Rickard et al., 2015; Tassin et al., 2013). A deeper understanding of DUX4 dynamics and how it drives FSHD pathology is essential to move toward anti-DUX4 therapy. Not only would this explain DUX4 target gene expression as a suboptimal monitoring tool, but would enable optimal therapeutic design, through in silico perturbation of the parameters underlying DUX4 expression and toxicity. Despite considerable discussion of ‘rare-bursts’ and ‘diffusion’ no attempt has been made to understand DUX4 expression through stochastic processes or differential equations; the natural setting to place these dynamics. Here, we combine ordinary differentiation equation models with stochastic gene expression models to construct a tuneable in silico simulation of DUX4 regulation in FSHD cell culture, both in unfused myocytes and syncytial multinucleated myotubes. Through analysis of iDUX4 myoblasts, scRNAseq and snRNAseq of FSHD differentiated myoblasts we derive experimental estimates for the parameters of our model, including DUX4 mRNA degradation, transcription and translation rates, and DUX4 target gene activation rates. Simulation of our model provides a striking fit to DUX4 and DUX4 target gene expressing cell proportions seen in scRNAseq of FSHD myocytes. Importantly, our model predicts that DUX4 drives significant cell death, despite expression being limited to <1% of live cells. By comparing scRNAseq of unfused FSHD myonuclei to snRNAseq of multinucleated FSHD myotubes, we find evidence of DUX4 protein syncytial diffusion. We extend our model to examine DUX4 spreading between adjacent myonuclei and project our simulation onto the surface of a muscle fiber, in a spatially relevant model. Employing genetic algorithms to fit our spatial model to snRNAseq of FSHD syncytial myotubes, we provide an estimate for the syncytial diffusion rate of DUX4 protein. We package our model into three freely available user interfaces, presenting an in silico toolkit to assess the impact of specific anti-DUX4 therapies on FSHD cell culture in a rapid, cost-effective, and unbiased manner. Results Compartment and promoter-switching models of DUX4 expression Here, we consider two models of DUX4 expression in FSHD myocytes, a deterministic model of FSHD cell states we call the compartment model, and a stochastic model of gene expression we call the promoter switching model. We first describe the compartment model. FSHD single myocytes can express DUX4 and, therefore, DUX4 target genes (Rickard et al., 2015; Kowaljow et al., 2007; Heher et al., 2022). We have previously demonstrated, via a library of DUX4 expression constructs, that DUX4 target gene activation is also necessary for DUX4-induced cell death (Knopp et al., 2016). We thus propose that an FSHD single myocyte at a given time t, occupies one of the following five states/compartments, defined by its transcriptomic distribution: St – a susceptible state where the cell expresses no DUX4 mRNA and no DUX4 target mRNA (DUX4 -ve/Target gene -ve: DUX4 mRNA naive cell) Et – an exposed state where the cell expresses DUX4 mRNA but no DUX4 target mRNA (DUX4 +ve/Target gene -ve: DUX4 transcribed but not translated) It – an infected state where the cell expresses both DUX4 mRNA and DUX4 target mRNA (DUX4 +ve/Target gene +ve: DUX4 transcribed and translated) Rt – a resigned state where the cell expresses no DUX4 mRNA but does express DUX4 target mRNA (DUX4 -ve/Target gene +ve: i.e., a historically DUX4 mRNA expressing cell) Dt – a dead state We propose a model in which the single FSHD myocyte can transition through these five states according to five parameters: VD – the average transcription rate of DUX4 in a single cell d0 – the average degradation rate of DUX4 mRNA TD – the average translation rate of DUX4 from mRNA to active protein VT – the average transcription rate of DUX4 target genes in the presence of DUX4 Dr – the average death rate of DUX4 target positive cells We allow cells to transition from DUX4 mRNA negative states St and Rt to DUX4 mRNA positive states Et and It, respectively at the rate of transcription of DUX4, VD , with transition in the reverse direction occurring at the degradation rate of DUX4 mRNA, d0. Transition from state Et to state It requires DUX4 mRNA to be translated to active protein and DUX4 target mRNA to be expressed and thus occurs at the rate TDVT. Lastly, DUX4 target gene +ve states It and Rt transition to the dead cell state Dt at rate Dr. Our compartment model can be represented schematically (Figure 1A) or equivalently as a system of ordinary differential equations (ODEs) (Figure 1B). Figure 1 Download asset Open asset Overview of models. (A) Schematic of the compartment model describing the transition between five Facioscapulohumeral muscular dystrophy (FSHD) states according to five underlying parameters. (B) Ordinary differential equations describing the compartment model. (C) Schematic of the promoter-switching model of gene expression. There are important assumptions in our compartment model: Cells are assumed not to proliferate over the evolution of the model. We restrict applications to differentiating cells that have exited the cell cycle. The death rate of DUX4 target gene -ve cells is negligible in comparison to the death rate of target gene +ve cells. This assumption is justifiable given published data on the death rate of DUX4 target gene-positive cells (Rickard et al., 2015). The transition from DUX4 target gene -ve cell to DUX4 target gene +ve cell states is irreversible, i.e., we assume that the volume of target transcripts induced by DUX4 is sufficiently large, such that the rate of their degradation to zero is negligible in comparison to the death rate of target positive cells. This assumption is justified given that DUX4 is a potent transcriptional activator and pioneer factor (Knopp et al., 2016; Choi et al., 2016). In what follows, we derive experimental estimates for the 5 parameters underlying the compartment model. A further preliminary is the promoter-switching model. This model is precisely the two-stage telegraph process, which has a long history of study in stochastic gene expression (Shahrezaei and Swain, 2008; Vu et al., 2016; Kim and Marioni, 2013). Under this model, we assume that the promoter of a gene n can occupy one of two states: active and inactive, and that transition from active (a) to inactive (i) state occurs at a rate kin , with the reverse transition occurring at a rate kan. We further assume that an active promoter can transcribe a single mRNA at a rate v0n , which degrades at a rate δn . The model is represented schematically in Figure 1C. This model of gene expression has been shown to follow a Poisson-Beta distribution (Vu et al., 2016; Kim and Marioni, 2013), where the promoter state is determined by a Beta-distributed variable p~Betakan/δn,kin/δn , and the mRNA copy number distribution conditional on the promoter state follows a Poisson distribution m|p~Poissonpv0n/δn. Under this interpretation maximum likelihood estimates (MLEs) for the normalized underlying parameters kan/δn, kin/δn and v0n/δn can be approximated from single-cell transcriptomic data (Vu et al., 2016; Kim and Marioni, 2013). We assume that the proportion of time the promoter is in the active state, multiplied by the transcription rate of the active promoter in our promoter-switching model kanv0nkan+kin, is a reasonable proxy for the average rate of transcription of a gene in our compartmental model. We employ this assumption to estimate the average transcription rates VD and VT for the compartment model. Estimating the kinetics of DUX4 mRNA Our compartment model contains two parameters governing the kinetics of DUX4 mRNA: the degradation rate d0 and the average transcription rate VD . To estimate the degradation rate d0 we employed human immortalized LHCN-M2 myoblasts expressing DUX4 under the control of a doxycycline-inducible promoter (iDUX4 myoblasts) (Choi et al., 2016). DUX4 expression was induced to a level and duration we have found sufficient to drive robust DUX4 mRNA expression, but only weak activation of DUX4 target genes and no widespread apoptosis (Ganassi et al., 2022). After induction, myoblasts were washed and supplemented with fresh medium without doxycycline. Samples were harvested for RNA extraction in triplicate immediately after washing and then at 1, 2, 3, 4, 5, 6, 8, and 10 hr post-wash. RT-qPCR was performed employing a standard curve to quantify the DUX4 mRNA copy number over our time course, to monitor DUX4 mRNA degradation in the absence of induction (Figure 2A). Figure 2 Download asset Open asset Estimation of DUX4 mRNA degradation rate d0 and average transcription rate (VD). (A) Schematic of the experiment performed to estimate the DUX4 mRNA degradation rate d0. iDUX4 myoblasts were induced to express DUX4 with 250 ng/ml of doxycycline for 7 hr before doxycycline was washed away. DUX4 mRNA was quantified at multiple time points post-wash using RT-qPCR. (B) Bar chart displays the RT-qPCR of ln(DUX4 copy number) at post-wash times 0, 1, 2, 3, 4, 5, 6, 8, and 10 hr. Bar represents the average of triplicates and the standard error of the mean is displayed. A line of best fit of ln(DUX4 copy number) against time is displayed alongside the corresponding linear regression p-value (bold) and the slope of the line corresponds to d0 (red). (C) Schematic of the estimation of average DUX4 transcription rate VD from scRNAseq data of 5133 Facioscapulohumeral muscular dystrophy (FSHD) myocytes. The maximum likelihood estimates (MLEs) for the underlying parameters of DUX4 transcription under the promoter-switching model are estimated via gradient descent and combined to estimate the average DUX4 transcription rate (red text). As anticipated DUX4 mRNA levels decayed exponentially over time in the absence of doxycycline (linear regression of ln(DUX4 mRNA) vs time, p=4.1 × 10–4, Figure 2B), allowing us to calculate the degradation rate of DUX4 mRNA, d0=0.246/hr. This estimate suggests that the half-life of DUX4 mRNA is approximately 2.8 hr, not atypical for a transcription factor, and on the faster end of the mRNA degradation distribution (Yang et al., 2003). To estimate the mean transcription rate of DUX4, VD , we applied the promoter-switching model presented above. We considered the scRNAseq dataset of FSHD single myocytes produced by van den Heuvel et al., 2019. This dataset comprises 7047 single myocytes, which were differentiated for 3 days in the presence of EGTA to inhibit fusion. 5133/7047 (73%) single myocytes were derived from four FSHD patients (two FSHD1 and two FSHD2). Patient demographics, genotypes, and DUX4 +/-ve cell counts are displayed in Supplementary file 1. DUX4 mRNA expression was detected in 27/5133 (0.53%) single FSHD myocytes but not in control myocytes (van den Heuvel et al., 2019; Banerji and Zammit, 2019). Considering the 5133 FSHD myocytes together we implemented a gradient descent algorithm to approximate the MLEs of the normalized parameters kaDUX4/δDUX4, kiDUX4/δDUX4 and v0DUX4/δDUX4 from the Poisson-Beta interpretation of the promoter-switching model of DUX4 expression (Figure 2C). We note that δDUX4=d0, which we have already estimated, this enabled us to renormalize these parameters and compute the average DUX4 transcription rate, VD:=kaDUX4v0DUX4kaDUX4+kiDUX4=0.00211/hr. As the total number of DUX4 positive cells is low, we pooled data from the four FSHD patients to allow robust estimation of the average DUX4 transcription rate for this patient cohort. We note, however, that distinct FSHD genotypes likely underlie different DUX4 transcription rates. The majority of DUX4 +ve cells (23/27) were found in two FSHD patients for this scRNAseq data. For patient FSHD1.1 19/2226 (0.85%) cells were DUX4 +ve and for patient FSHD2.1 5/1283 (0.39%) cells were DUX4 +ve (Supplementary file 1). To investigate variability in VD , we derived individual estimates for these two ‘higher’ DUX4 expressing patients, for FSHD1.1 VD=0.00373/hr, while for FSHD2.1 VD=0.000960/hr. The pooled estimate of VD is thus comparable in order of magnitude with that of individual patients. To fully utilize the available data and prevent limiting our model to only ‘higher’ DUX4 expressing patients, we employ the pooled estimate of VD=0.00211/hr for the remainder of our calculations. Estimating kinetics of DUX4 target activation We next estimated the average transcription rate of the DUX4 target genes in the presence of DUX4 mRNA, VT. We focused on eight DUX4 target genes: ZSCAN4, TRIM43, RFPL1, RFPL2, RFPL4B, PRAMEF1, PRAMEF2, and PRAMEF12 that have been identified as direct DUX4 targets via ChIP-seq (Young et al., 2013). We have shown that these eight genes are the only features consistently up-regulated in human myoblasts expressing DUX4, across four independent studies (Banerji and Zammit, 2021; Rickard et al., 2015; Choi et al., 2016; Young et al., 2013; Geng et al., 2012). Examining mRNA levels of these eight DUX4 target genes in scRNAseq (van den Heuvel et al., 2019) and snRNAseq (Jiang et al., 2020) studies of FSHD and control differentiated myoblasts, expression was restricted to FSHD cells/nuclei and never observed in controls. This pattern of expression mirrors that of DUX4 mRNA and suggests that these targets are highly specific and are unlikely to be activated in the absence of DUX4. To confirm our hypothesis, we applied the promoter-switching model. We returned to the scRNAseq data of 5133 FSHD myocytes generated by van den Heuvel et al., 2019 and divided the data into the 27 DUX4 +ve cells and 5106 DUX4 -ve cells. For each of our eight DUX4 target genes, we implemented our gradient descent algorithm to compute the MLEs of the normalized parameters kan/δn, kin/δn, and v0n/δn, underlying a promoter-switching model for the given gene across the 27 DUX4 +ve FSHD myocytes and the 5106 DUX4 -ve FSHD myocytes separately (Figure 3A). Figure 3 Download asset Open asset Estimation of promoter-switching model parameters for DUX4 target genes and estimation of VT. (A) Schematic of estimation of promoter-switching model parameters for eight DUX4 target genes, across 5106 DUX4 -ve Facioscapulohumeral muscular dystrophy (FSHD) single myocytes and 27 DUX4 +ve FSHD single myocytes from van den Heuvel et al., 2019. Violin plots display (B) the proportion of time in active promoter state (C) normalized active promoter transcription rate and (D) the mean mRNA copy number, for the eight DUX4 target genes in the 5106 DUX4 -ve and 27 DUX4 +ve FSHD myocytes separately, p-values correspond to two-tailed paired Wilcoxon tests. (E) Violin plot displays the variance of mRNA copy number for the eight DUX4 target genes calculated from the 27 DUX4 +ve FSHD myocytes (red) and calculated assuming DUX4 up-regulates targets only via the increase in (blue) the normalized transcription rate v0nδn or (green) the ratio of active to inactive promoter transition rates kankin. Paired two-tailed Wilcoxon p-values are displayed comparing adjacent distributions. (F) Schematic displaying how the change in parameters underlying the promoter-switching models for the eight DUX4 target genes in the presence of DUX4 leads to stable target gene activation. (G) Schematic of the estimation of the average DUX4 target gene transcription rate VT, from scRNAseq data of 27 DUX4 +ve FSHD myocytes. Figure 3—source data 1 Source data for Figure 3 provides the parameters of the promoter switching model for each of the 8 DUX4 targets, derived from 5106 DUX4 -ve FSHD single cells and 27 DUX4 +vs FSHD single cells seperately. https://cdn.elifesciences.org/articles/88345/elife-88345-fig3-data1-v2.csv Download elife-88345-fig3-data1-v2.csv In the presence of DUX4 mRNA, the proportion of time the promoters of the eight DUX4 target genes remained in the active state, kankan+kin, significantly increased (paired Wilcoxon p=7.8 × 10–3, Figure 3B), as expected. Curiously, however, the normalized rate of transcription of the DUX4 target genes from the active promoter, v0n/δn, significantly decreased in the presence of DUX4 (paired Wilcoxon p=0.039, Figure 3C). To investigate further, we considered the moments of the distribution of mRNA copy number, m, under the promoter-switching model. It can be shown that the mean mRNA copy number satisfies (Vu et al., 2016): Em=kanv0n/δnkan+kin. The changes in parameters we calculate for the DUX4 target genes confirmed that in the presence of DUX4 mRNA, the mean expression of all 8 DUX4 targets increases (paired Wilcoxon p=0.039, Figure 3D), i.e., the drop in v0n/δn is over-compensated for by the rise in kankan+kin. However, it is curious that v0n/δn should drop at all. It can be shown that the variance of mRNA copy number m, satisfies (Vu et al., 2016): Varm=Em1+kinv0n/(δn)^2(1+kan/δn+kin/δn)(kan/δn+kin/δn). This expression ensures that as the mean mRNA copy number rises, so too must the variance, however, the level to which the variance rises is controlled by a term that is monotonic increasing in v0n/δn and depends on the promoter state parameters in a more complex way. We postulated the pattern of DUX4 target gene parameter changes we observe in the presence of DUX4, have the effect of raising mean target mRNA expression, while suppressing target mRNA variance, i.e., a controlled activation of target genes. To investigate this, we considered two hypothetical scenarios in which the mean expression of target mRNAs, Em in the absence of DUX4, is increased to the same level we observe in the presence of DUX4. In the first scenario, we considered only increasing v0n / δn to achieve the rise in Em , in the second we considered only increasing the ratio of active to inactive promotor kankin. Both hypothetical scenarios resulted in a significantly higher variance of target mRNA than observed in our data, with the pure rise in v0n/δn driving the most dramatic increase in target mRNA variance (Figure 3E). Taken together these results suggest that under our promoter-switching model, increasing the expression of a gene comes at the cost of increasing the variance of its expression, and that the greatest contributor to this variance comes from the normalized active promoter transcription rate v0n / δn . The parameter changes we observe in DUX4 target genes, suggest a management of this trade-off by DUX4, which increases the mean expression of target mRNA through a large increase in the proportion of time the promoter is active, kankan+kin, while offsetting the resulting rise in variance through a modest decrease in normalized active promotor transcription v0n/δn (Figure 3F). We formulate the mean transcription rate of at least 1 of the 8 DUX4 targets from our compartment model, VT, as the sum of the mean transcription rates of all eight target genes in the presence of DUX4 mRNA, i.e.,: VT≔ ∑j=18kajv0jkaj+kij where j indexes the eight DUX4 target genes, and the promoter-switching model parameters are estimated from the 27 DUX4 expressing FSHD single myocytes. Our promoter-switching model scRNAseq-derived MLEs are normalized parameters kan/δn, kin/δn and v0n/δn. We must, therefore, estimate δn for each target gene to compute VT . As there is a range of target genes, we approximate δn for each by the median mRNA degradation rate observed in an analysis of 5245 genes (Yang et al., 2003), and set δn=0.0693 resulting in VT=6.41/hr (Figure 3G). As with our calculation of VD data was pooled across four FSHD patients to calculate VT . We do not anticipate patient genotype to impact the average DUX4 target transcription rate, independently of its impact on the DUX4 transcription rate. However, to confirm our findings on the impact of DUX4 on target gene promoter dynamics, in a patient-specific setting, we attempted calculation of the parameters underlying the promoter switching model for the eight DUX4 target genes in DUX4 +ve and DUX4-ve cells, for patients FSHD1.1 and FSHD2.1 separately. Due to the limited number of cells for each patient, personalised estimates for all eight target genes could not be obtained. However, where patient-specific estimates have obtained the direction of parameter differences in target genes, between DUX4 +ve and DUX4 -ve cells were in line with those of pooled estimates across four FSHD patients (Supplementary file 2). The remaining two parameters of our compartment model were calculated from published data. For the translation rate of DUX4 mRNA to active protein TD , we considered our analysis of iDUX4 myoblasts (Ganassi et al., 2022). We induced DUX4 expression with 250 ng/ml doxycycline and performed RT-qPCR to assess the expression of DUX4, and 3/8 of our DUX4 target genes ZSCAN4, TRIM43, and PRAMEF1, at 7 hr, 16 hr, and 24 hr of induction. DUX4 mRNA levels peaked at 7 hr, while the expression of DUX4 target genes peaked between 16 and 24 hr (Ganassi et al., 2022). This suggests a delay between DUX4 mRNA production and the presence of active DUX4 protein of between 9 and 17 hr, on average 13 hr. We thus estimate TD=113/hr. For the death rate of DUX4 target positive cells, Dr , we consider the data of Rickard et al., 2015, in which differentiating FSHD myoblasts containing a DUX4-activated GFP reporter were imaged every 15 min for 120 hr. Following activation of the DUX4 reporter, cells died ~20.2 hr later (Rickard et al., 2015). We thus estimate Dr=120.2/hr. Compartment model simulation Having defined experimental estimates for parameters underlying the compartment model, we simulated the model forward in time to observe how an initial distribution of cells progresses through the five compartments. To provide a ground truth we considered the scRNAseq data of van den Heuvel et al., 2019. Defining a DUX4 target gene +ve cell as expressing at least one of our eight DUX4 target genes, we assign the 5133 FSHD myocytes to 1 of the 4 live cell states of our compartment model: S(3 days)=4956 (96.6%),E(3 days)=14(0.273%), I3 days=13 (0.253%), R3 days=150 (2.92%). If we assume that at the start of differentiation, all FSHD myocytes occupy state S0 being DUX4 negative and DUX4 target gene negative, simulating our model we estimated that 7% of starting cells will have died over 3 days. To replicate the starting conditions of the scRNAseq data we thus set S0=5133(1+0.07)=5488, E0=I0=R0=D0=0. Simulating our model over 3 days from this starting condition, we obtained cell state proportions statistically indistinguishable from the experimental scRNAseq data: S(3 days)=4953 (96.5%),E(3 days)=14(0.253%), I3 days=25 (0" @default.
• W4381612243 created "2023-06-23" @default.
• W4381612243 date "2023-04-17" @default.
• W4381612243 modified "2023-09-25" @default.
• W4381612243 title "Decision letter: An in silico FSHD muscle fiber for modeling DUX4 dynamics and predicting the impact of therapy" @default.
• W4381612243 doi "https://doi.org/10.7554/elife.88345.sa1" @default.
• W4381612243 hasPublicationYear "2023" @default.
• W4381612243 type Work @default.
• W4381612243 citedByCount "0" @default.
• W4381612243 crossrefType "peer-review" @default.
• W4381612243 hasBestOaLocation W43816122431 @default.
• W4381612243 hasConcept C104317684 @default.
• W4381612243 hasConcept C126322002 @default.
• W4381612243 hasConcept C145912823 @default.
• W4381612243 hasConcept C15744967 @default.
• W4381612243 hasConcept C178790620 @default.
• W4381612243 hasConcept C185592680 @default.
• W4381612243 hasConcept C19417346 @default.
• W4381612243 hasConcept C2775905019 @default.
• W4381612243 hasConcept C2779959927 @default.
• W4381612243 hasConcept C3017897630 @default.
• W4381612243 hasConcept C41008148 @default.
• W4381612243 hasConcept C519885992 @default.
• W4381612243 hasConcept C54355233 @default.
• W4381612243 hasConcept C70721500 @default.
• W4381612243 hasConcept C71924100 @default.
• W4381612243 hasConcept C86803240 @default.
• W4381612243 hasConceptScore W4381612243C104317684 @default.
• W4381612243 hasConceptScore W4381612243C126322002 @default.
• W4381612243 hasConceptScore W4381612243C145912823 @default.
• W4381612243 hasConceptScore W4381612243C15744967 @default.
• W4381612243 hasConceptScore W4381612243C178790620 @default.
• W4381612243 hasConceptScore W4381612243C185592680 @default.
• W4381612243 hasConceptScore W4381612243C19417346 @default.
• W4381612243 hasConceptScore W4381612243C2775905019 @default.
• W4381612243 hasConceptScore W4381612243C2779959927 @default.
• W4381612243 hasConceptScore W4381612243C3017897630 @default.
• W4381612243 hasConceptScore W4381612243C41008148 @default.
• W4381612243 hasConceptScore W4381612243C519885992 @default.
• W4381612243 hasConceptScore W4381612243C54355233 @default.
• W4381612243 hasConceptScore W4381612243C70721500 @default.
• W4381612243 hasConceptScore W4381612243C71924100 @default.
• W4381612243 hasConceptScore W4381612243C86803240 @default.
• W4381612243 hasLocation W43816122431 @default.
• W4381612243 hasOpenAccess W4381612243 @default.
• W4381612243 hasPrimaryLocation W43816122431 @default.
• W4381612243 hasRelatedWork W2003194288 @default.
• W4381612243 hasRelatedWork W2062254828 @default.
• W4381612243 hasRelatedWork W2074002899 @default.
• W4381612243 hasRelatedWork W2153919606 @default.
• W4381612243 hasRelatedWork W2213189664 @default.
• W4381612243 hasRelatedWork W2513589746 @default.
• W4381612243 hasRelatedWork W2936968863 @default.
• W4381612243 hasRelatedWork W4361305867 @default.
• W4381612243 hasRelatedWork W4381684688 @default.
• W4381612243 hasRelatedWork W46094488 @default.
• W4381612243 isParatext "false" @default.
• W4381612243 isRetracted "false" @default.
• W4381612243 workType "peer-review" @default.