Flavones provide resistance to DUX4-induced toxicity via an mTor-independent mechanism

Facioscapulohumeral muscular dystrophy (FSHD) is among the most common of the muscular dystrophies, affecting nearly 1 in 8000 individuals, and is a cause of profound disability. Genetically, FSHD is linked to the contraction and/or epigenetic de-repression of the D4Z4 repeat array on chromosome 4, thereby allowing expression of the DUX4 gene in skeletal muscle. If the DUX4 transcript incorporates a stabilizing polyadenylation site the myotoxic DUX4 protein will be synthesized, resulting in muscle wasting. The mechanism of toxicity remains unclear, as many DUX4-induced cytopathologies have been described, however cell death does primarily occur through caspase 3/7-dependent apoptosis. To date, most FSHD therapeutic development has focused on molecular methods targeting DUX4 expression or the DUX4 transcript, while therapies targeting processes downstream of DUX4 activity have received less attention. Several studies have demonstrated that inhibition of multiple signal transduction pathways can ameliorate DUX4-induced toxicity, and thus compounds targeting these pathways have the potential to be developed into FSHD therapeutics. To this end, we have screened a group of small molecules curated based on their reported activity in relevant pathways and/or structural relationships with known toxicity-modulating molecules. We have identified a panel of five compounds that function downstream of DUX4 activity to inhibit DUX4-induced toxicity. Unexpectedly, this effect was mediated through an mTor-independent mechanism that preserved expression of ULK1 and correlated with an increase in a marker of active cellular autophagy. This identifies these flavones as compounds of interest for therapeutic development, and potentially identifies the autophagy pathway as a target for therapeutics.


Introduction
With an estimated prevalence of nearly 1 in 8000 individuals (1) Facioscapulohumeral Muscular Dystrophy (FSHD) is one of the most common muscular dystrophies. Pathology typically begins with weakness in the facial, scapular, and humeral muscles, but progresses to the trunk and lower extremities, resulting in profound disability (2). FSHD is most often inherited as a dominant Mendelian trait, however the genetic etiology is complex. Disease is associated with a repeat array of 3.3 kb D4Z4 elements located near the telomere of chromosome 4q (3,4). In the unaffected population the array most often contains between approximately 10 and 100 units. The most common form of the disease, FSHD1, results from contraction of the array below ~ 9 units, which allows epigenetic de-repression of the DUX4 gene contained within each repeat (reviewed in (2)). Contraction alone does not cause disease however, as it must occur on a chromosome carrying a "permissive" 4qA haplotype, which allows the transcript expressed from the last repeat to incorporate a stabilizing polyadenylation signal that enables DUX4 protein synthesis (5)(6)(7)(8)(9)(10)(11)(12). In the second, less common, form of the disease, FSHD2, trans-acting mutations in SMCHD1, DNMT3B, or LRIF1 cause epigenetic de-repression of the array, and in the presence of a permissive 4qA allele can result in transcription of the stabilized DUX4 (5,(13)(14)(15)(16). Oddly, even when all genetic determinants of disease are present, DUX4 is not uniformly expressed, but instead activation occurs in "bursts" in a small fraction of myonuclei (17)(18)(19)(20)(21).
An alternate approach is to target DUX4-induced cytopathologies that trigger apoptosis. We have previously demonstrated that inhibiting hyaluronic acid signaling (36) or inhibition of the mTor/PI3K/AKT pathway (41) can ameliorate DUX4-induced toxicity in myoblasts. Additionally, it has recently been demonstrated that inhibition of MAP kinase pathways can have a similar effect (55). We set out to leverage these observations to identify compounds that can inhibit DUX4-induced toxicity and may be developed into FSHD therapeutics. To this end, we performed two rounds of screening and characterization on compounds that were selected either due to their structural similarity to the hyaluronic acid synthesis inhibitor 4MU or its metabolite 4MUG, or based on previous reports of activity in relevant signal transduction pathways. We identi ed a set of ve avones that can inhibit DUX4-induced toxicity at low-micromolar concentrations, thereby making them candidates for further therapeutic development. Additionally, we show that these compounds function through an mTor/AKT-independent mechanism that results in activation of cellular autophagy, thereby demonstrating that targeting both mTor-dependent and mTor-independent biochemical pathways are viable approaches for FSHD therapeutic development, and speci cally identifying autophagy as a novel target for therapies.

Results
First-generation small molecules inhibit DUX4-induced apoptosis. In our previous work, we demonstrated that the hyaluronic acid synthesis inhibitor 4MU can provide resistance to DUX4-induced toxicity (36).
Additionally, we and others have implicated hypoxia signaling as central to toxicity (41,52). Interestingly, these signaling pathways converge on the mTor/AKT/PI3K signal transduction axis (71,72), and inhibition of this pathway can provide resistance to toxicity (41). We sought to leverage these observations to identify small molecules that can provide resistance to DUX4-induced apoptosis. Unfortunately, 4MU itself requires millimolar doses for maximal effectiveness, and so is not suitable for use as a therapeutic. Thus, we considered other molecules that may provide a similar result at lower concentrations. Based on previous reports of their activity in a relevant pathway, we identi ed a panel of six rst-generation compounds that had the potential to meet this criteria-honokiol (Cas # 35354-74-6), its synthetic analogue C6F, magnolol (Cas # 528-43-8), epigallocatechin gallate (EGCG, Cas # 989-51-5), silibinin (CAS # 22888-70-6), and liquiritigenin (Cas # 69097-97-8) (73)(74)(75)(76)(77). Notably, EGCG, silibinin, and liquiritigenin were of particular interest because their chemical structures are built around the same fused ring structure as 4MU, and they each maintain the chemically active hydroxyl group (78) (Fig. 1A). To evaluate these compounds, we used the MB135-DUX4i myoblast model (38). Myoblasts were seeded on 96 well plates, and the following day they were pre-treated by adding the indicated compound to the media for three hours, followed by the addition of 2 mg/mL doxycycline (DOX) to the media for 24 hours to induce DUX4 expression (for a total of 27 hours of exposure to the compounds). As a positive control, we also included the mTor inhibitor rapamycin, which can inhibit DUX4-induced toxicity (41) and its nextgeneration analogue everolimus. Cell death was then visualized using the CellEvent Caspase 3/7 Green assay (Invitrogen, Waltham, MA USA). We tested a range of concentrations and found that each was able to provide resistance to toxicity when administered at proper concentrations (Fig. 1B, Figure S1). To con rm and quantitate these results, we conducted similar experiments using the Caspase-Glo 3/7 assay system (Promega, Madison, WI USA). Each compound provided at least a twofold reduction in caspase 3/7 activity relative to vehicle controls, with liquiritigenin showing the strongest effect (Fig. 1C). To validate these results we performed limited-cycle RT-PCR as described previously (41) and con rmed that these compounds did not interfere with the induction of the codon-altered DUX4 transgene (Fig. 1D). Similarly, we performed western blotting analysis to determine the effects on the levels of DUX4 protein.
As observed previously (41), rapamycin caused a drop in the abundance of DUX4 (Fig. 1E). Surprisingly however, only honokiol showed a similar decline in DUX4 protein abundance, but the remaining compounds had no effect. This con rms that the mechanism of observed resistance to DUX4-induced toxicity occurs downstream of DUX4 expression, but also unexpectedly suggests that these compounds function via a different mechanism than rapamycin. Finally, to con rm that these compounds function downstream of DUX4 expression and that they do not have a deleterious effect on mature myotubes, immortalized patient-derived 16ABic myoblasts (79) were induced to form myotubes for 4 days using established methods (80), and were then treated with compounds for 24 additional hours. The expression of three DUX4-target genes were then analyzed using qRT-PCR, and no signi cant change was observed ( Fig. 2A). We also analyzed three myogenesis markers and found no statistically signi cant effect on their expression (Fig. 2B). Similar results were observed using a second patient-derived cell line ( Figure  S2).
Second-generation small molecules are more potent inhibitors of DUX4-induced apoptosis. While our rstgeneration compounds were effective at inhibiting DUX4-induced toxicity, the most potent, liquiritigenin, required a 150 mM dose for optimal effectiveness. While this is a signi cant improvement relative to concentrations required for 4MU, it is still too high for therapeutic use. To overcome this limitation, we performed a second screening of a larger library of compounds ( Table 1). Many of these were additional avone compounds that bear structural similarity to 4MU or to its bioactive metabolite 4MUG (81) (Fig. 3A). For the initial characterization, we used the protocol described above using 50, 5, or 1 mM concentrations and the CellEvent assay ( Figure S3). The best performing compounds were then selected for a secondary screening /optimization using concentrations of 30, 20, or 10 mM ( Figure S4). This screen identi ed 5 compounds-acacetin (Cas # 480-44-4), apigenin (Cas # 520-36-5), luteolin (Cas # 491-70-3), apigenin 7-glucoside (A7G, Cas # 578-74-5), and luteolin 7-glucoside (L7G, Cas # 5373-11-5), which provided resistance to DUX4-induced toxicity at optimal concentrations of 20 mM (Fig. 3B). We also identi ed acri avine (Cas # 8048-52-0), which based on phase-contrast images provided resistance at 5 mM ( Figure S3). We again quantitated these results using the Caspase-Glo assay and found that each compound provided at least four-fold reduction in DUX4-induced toxicity (Fig. 3C). As before, we validated these results using limited-cycle RT-PCR. Acri avine showed a notable decline in DUX4 transcript levels, suggesting that its effects were an artifact of inhibited transgene activation. Other compounds did not have an effect an DUX4 expression (Fig. 3D). Similarly, western blotting showed that acri avine treatment signi cantly reduced DUX4 protein levels, while the remaining compounds did not ( Fig. 3E). This again suggests that these compounds exert their effect through a different mechanism than rapamycin. We also examined the effects of these compounds in patient-derived myotubes, and we again found that there was no change in DUX4-target gene expression levels, except for acri avine, which inhibited expression of all three DUX4 target genes signi cantly (Fig. 4A). This may have been a result of acri avine having an effect on myogenesis, as it caused signi cant overexpression of CKM, while also inhibiting MYOG expression. Similar results were observed using a second patient-derived cell line ( Figure  S5). These results are again consistent with the ve avone compounds functioning downstream of DUX4 to inhibit toxicity without having negative effects on myogenesis. TAT-cyclo-CLLFVY 1446322-66-2 Misc.
Flavones and rapamycin inhibit DUX4-induced toxicity through distinct mechanisms. We next investigated the mechanism of action of the avones. First, we used phospho-speci c antibodies to determine how activation of DUX4 affects signaling in the mTor/AKT pathway. We analyzed either uninduced MB135-DUX4i myoblasts, or myoblasts induced with 2 mg/mL doxycycline for 5 or 24 hours with antibodies speci c to AKT phosphorylated on Thr308 or Ser473 or phosphorylated ribosomal S6 protein, a marker of mTor activity. After 5 hours of induction, we observed no change in phosphorylation of S6 or AKT at Ser473, and only a small but reproducible increase in AKT phosphorylation at Thr308 (Fig. 5A). However, we observed a notable decline in phosphorylation of AKT after 24 hours of DUX4 expression. In contrast, there was no change in the levels of overall AKT or S6 protein. Thus, it may be that prolonged DUX4 expression inhibits signaling along this axis, or that signaling along the mTor/AKT axis is lost at this late time point, as these populations are actively undergoing apoptosis. To determine the effects of our compounds on AKT/mTor activation, we pre-treated myoblasts with compounds for three hours and then induced DUX4 expression for ve hours. As expected, rapamycin and everolimus ablated S6 phosphorylation and triggered hyperphosphorylation of AKT, particularly on Thr308 (Fig. 5B). Surprisingly, of the compounds under study, only honokiol showed a reproducible, but partial inhibition of S6 phosphorylation, and none showed hyperphosphorylation of AKT. Thus, with the possible exception of honokiol, it appears that the rst-generation compounds inhibit toxicity through an mTor-independent mechanism. Interestingly, the effects are not DUX4-speci c, as performing the same experiments without inducing DUX4 yielded nearly identical results (Fig. 5C), suggesting that they function not by inhibiting a DUX4-activated signal transduction pathway, but rather by triggering a DUX4-independent response that protects against toxicity. To con rm that these observations also hold true for the second-generation compounds, we also pretreated MB135-DUX4i myoblasts with luteolin or L7G for three hours and either left them uninduced for ve more hours or induced them with doxycycline for ve or 24 hours and analyzed by western blotting ( Figure S6). As before, the second-generation compounds showed no inhibition of S6 phosphorylation.
Second-generation compounds prevent loss of ULK1 and induce a marker of autophagy. To investigate the mechanism of action of the second-generation avone compounds, we considered other pathways that may protect against DUX4-induced toxicity. Luteolin has been previously reported as both a positive and negative regulator of autophagy (82), suggesting that this pathway may be relevant. Also, rapamycin is a known autophagy activator (83), and autophagy regulators integrate multiple signaling pathways (84). Therefore, both mTor-dependent and mTor-independent mechanisms that regulate autophagy may affect DUX4-induced toxicity. We investigated this possibility by analyzing ULK1 expression, a key autophagic regulator that integrates multiple signaling pathways (84). After 5 hours of induction there was no notable change in the levels of ULK1 protein (Fig. 6A). However, after 24 hours much of ULK1 protein was lost. Surprisingly, the decline in ULK1 was accompanied by an increase in LC3-II, a marker of active autophagy. To investigate the effects of the second-generation compounds on autophagy, we pretreated myoblasts for three hours and then induced DUX4 for 24 hours. We found that the avone compounds protected ULK1 from DUX4-induced loss. This effect appears to be speci c to ULK1 and not a general property of the autophagic machinery, as DUX4 induction did not cause loss of the autophagy-associated proteins ATG3, ATG5, ATG7, or ATG16L1, and the avones had no effect on these proteins ( Figure S7). Interestingly, we observed that each of the avones increased the abundance of the LC3-II autophagy marker well above the level induced by DUX4 alone (Fig. 6B), suggesting that cellular autophagy protects against DUX4-induced toxicity and that the avone compounds function by enhancing this protective mechanism. Taken together, these observations identify these avone compounds as potential drugs for further development, and the autophagy pathway in general as a target for future FSHD therapeutics.

Discussion
While signi cant advancement in the characterization of DUX4-induced cytopathologies has been made, the precise mechanism that leads to DUX4-induced apoptosis has remained elusive. Increasing evidence indicates that DUX4 causes widespread mis-regulation of signaling pathways (36, 41, 44, 52-55), making these cascades important potential targets for FSHD therapeutics. The mTor/AKT signaling axis is of particular interest as it mediates signaling through both the hyaluronic acid and HIF1a pathways. Additionally, mTor regulates energy homeostasis, and we have seen mitochondrial mis-localization in response to DUX4 expression (36), and a recent study has demonstrated metabolic disruption in FSHD (50). Despite this, most therapeutic development has focused on targeting DUX4 directly. Thus, in this study we endeavored to identify inhibitors of DUX4-induced toxicity that function downstream of DUX4. We conducted two screens of compounds predicted to function in the mTor/AKT/HIF1a pathway, and we identi ed three rst-generation and ve second-generation avone/ avonoid compounds that inhibit DUX4-induced toxicity (Figs. 1 and 3). Importantly, the second-generation compounds function at pharmacologically relevant concentrations, and none have negative effects on the expression of myogenic markers in FSHD patient-derived myotubes (Figs. 4 and S5). These compounds therefore have the potential to be investigated further as treatments for FSHD.
Unexpectedly, we have also observed that the avones do not function through the mTor/AKT axis. While rapamycin and everolimus ablated the phosphorylation of the S6 ribosomal protein and caused hyperphosphorylation of AKT, the avones did not show either of these effects. Rather, we observed that the avones protected ULK1, a regulator of autophagy (84) from DUX4-induced loss (presumably by triggering its degradation) (Fig. 6A-B). Importantly, this correlated with an increase in the autophagy marker LC3-II (Fig. 6B), which implies that the avone compounds protect against the toxic effects of DUX4 expression by inducing autophagy. It remains unknown how these compounds activate autophagy, however a plausible explanation is that they function via activation of AMPK. Luteolin has been implicated in activating AMPK in muscle previously (85), and AMPK and mTor have opposing effects on autophagy (84). Therefore, rapamycin and avones may ultimately share a separate, but converging mechanism of action, but this hypothesis requires a dedicated study to con rm.
The protective effect observed here may have implications for FSHD research beyond their therapeutic potential. There is a signi cant body of literature indicating that FSHD genetic markers do not strictly correlate with the disease phenotype, and DUX4 expression has been detected in the muscle of unaffected individuals (reviewed in (2)). Thus, it has long been hypothesized that either additional unknown genetic factors can modify the disease phenotype, or that environmental factors in uence progression (or a combination of both). Flavones may be one such environmental factor, as they can be introduced via the diet (86). Different avones can have very different effects on DUX4-induced toxicity ( Figure S3), and so it is possible that diets rich in a particular avone may have notable effects while others are negligible, and that this may contribute to disparate disease progression. Detailed studies of diets rich in the most effective avones or avone supplementation will be necessary to determine if these are viable approaches to control disease progression.

Methods
Cell Culture. MB135-DUX4i myoblasts were grown essentially as described previously (36)  were washed with PBS, and media was replaced with 15% KOSR differentiation media, essentially as described (80) (glutamax substituted for L-glutamine), and allowed to differentiate for 4 days. Media was then replaced and supplemented with the relevant compound for 24 hours. All experiments were performed on three independently grown cell cultures unless otherwise noted.
Caspase Assays. 15,000 myoblasts/well were plated on 96-well plates. The next day media was replaced with media containing the indicated compounds for the stated times. 24 hours after doxycycline induction 5 mL of CellEvent Caspase-3/7 Green reagent (Invitrogen R37111) was added per well, plates were incubated for 30 min at 37 o C, and were then imaged with an Echo Revolve microscope. For quantitative assays, the Caspase-Glo® 3/7 Assay (Promega G8090 or G8091) was performed in duplicate (two wells were measured for each of 3 replicates, for a total of 6 wells measured per condition) according to the manufacturer's instructions, and luminescence was measured in a BioTek Synergy LX multi-mode plate reader.
Gene Expression Analysis. Cells were grown on 6-well plates as described above and were lysed in Buffer RLT with 2-mercaptoethanol, scraped, moved into a 1.5 mL tube, homogenized by pipetting with a P200 tip > 60 times, followed by freezing at -80 o C. Total RNA was extracted with an RNEasy Mini kit (Qiagen, Venlo, Netherlands, 74106) with on-column DNase I (Qiagen, 79254) digestion. cDNA was made from up to 1 mg of total RNA with a SuperScript III rst-strand synthesis kit (Invitrogen, 18080051) with double priming, and the RNase H step was performed. Limited-cycle PCR was performed as described (41). qPCR was performed as described (36) using published primers (23, 36, 87) (Table S1). All experiments were performed in triplicate.
Protein Expression Analysis. Myoblasts were grown on 6-well plates as described, washed in PBS, lysed, and scraped in 150 mL RIPA (Pierce/Thermo Scienti c Waltham, MA USA, 89900) supplemented with PhosSTOP (Roche Penzberg, Germany, 04 906 837 001) and 1% protease inhibitor cocktail (Sigma P8340). Cells were lysed at 4 o C for 20 minutes, centrifuged at 12,000g for 15 min, and the pellet was discarded. Protein concentrations were measured by BIO-RAD DC assay, and 5-10 mg of protein was run.  First-generation compounds have minimal impact on DUX4 activity or expression of myogenic marker genes in patient-derived myotubes. A. Immortalized 16ABic myoblasts were seeded to high density on gelatin coated 6-well plates. The following day, growth media was replaced with differentiation media, and differentiation was allowed to proceed for 4 days. Media was then replaced with fresh diff. media containing the indicated compound and myotubes were incubated for an additional 24 hours, followed by Page 22/27 quanti cation of expression of three DUX4-target genes by qRT-PCR. B. As in A., but qRT-PCRs were performed with primers speci c to markers of myogenesis. Error bars are SEM. Statistical signi cance for samples vs relevant vehicle-treated controls is shown and was calculated using 1-way ANOVA with Tukey's test. *P<0.05, **P<0.01, ***P<0.001.  Second-generation compounds have minimal impact on DUX4 activity or expression of myogenic marker genes in patient-derived myotubes. Patient-derived myoblasts were analyzed as in Figure 2, but treated with second generation compounds, followed by qRT-PCR analysis using A. DUX4-target gene or B.

Figure 5
Compounds prevent DUX4-induced toxicity through distinct mechanisms. A. MB135-DUX4i myoblasts were either left un-induced or induced with doxycycline for 5 or 24 hours. At the indicated times myoblasts were lysed and analyzed by western blotting with antibodies speci c to markers of the mTor/AKT signaling axis. B. MB135-DUX4i myoblasts were pre-treated with the indicated concentrations of compounds for 3 hours, followed by induction of DUX4 expression for 5 hours and western blotting analysis as in A. C. MB135-DUX4i myoblasts were left un-induced and exposed to the indicated concentrations of compounds for 8 hours, followed by western blotting analysis as in A. Uncropped images of all blots are presented in the supplemental information.

Figure 6
Flavones protect ULK1 and cause increased levels of the autophagy marker LC3-II. A. MB135-DUX4i myoblasts were either left un-induced or induced with doxycycline for 5 or 24 hours. At the indicated times myoblasts were lysed and analyzed by western blotting with antibodies speci c to markers of autophagy. B. MB135-DUX4i myoblasts were pre-treated with the indicated concentrations of compounds for 3 hours, followed by induction of DUX4 expression for 24 hours and western blotting analysis as in A.
Uncropped images of all blots are presented in the supplemental information.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.