The BMP signaling axis does not efficiently activate a SMAD4 transcriptional programme in SBMA skeletal muscle. In order to gain insight into the mechanisms of polyQ AR regulation of muscle homeostasis, we employed RNA sequencing of skeletal muscle samples from male SBMA transgenic mice carrying the human AR transgene with pathological glutamine expansion (AR100Q)22 and wild type littermates. Analysis of the differentially expressed genes recently performed by our group23 revealed that the TGFβ signaling, a major player in maintenance of skeletal muscle homeostasis in adulthood24, is one of the most significantly dysregulated pathways in SBMA muscle. To confirm the disease relevance of these findings, we performed transcriptomic analysis of skeletal muscles from genetically-diagnosed SBMA individuals (n=9; average CAG repeats: 46, range: 44-50) and age- and sex-matched unaffected controls (n=4). KEGG pathway enrichment studies of this dataset also identified this pathway as dysregulated in SBMA (Fig. 1a), with 2 genes of the TGFβ cascade being downregulated (ACVR1B, SMURF1) and 12 upregulated (CDKN2B, DCN, FST, GDF6, ID2, INHBE, SMAD6, SMAD9, TGFB1, TGFB2, TGFB3) (Fig. 1a). This pathway comprises two arms: the activin/myostatin and the BMP axes, which, by phosphorylation of either the SMAD2/3 or SMAD1/5/8 complex, compete over recruitment of the transcription factor SMAD4 to drive a muscle atrophy or hypertrophy transcriptional programme, respectively24. Western blot analyses of skeletal muscle samples from SBMA individuals and age- and sex-matched controls showed dramatic increased levels of phosphorylated SMAD1/5 and reduced levels of phosphorylated SMAD2/3 (Fig. 1b,c), suggesting that in SBMA skeletal muscle the BMP axis largely prevails over the competing activin/myostatin axis. Highly concordant results were observed in skeletal muscle of 8 week old male SBMA transgenic mice, carrying a normal (24Q) or pathological (100Q) polyglutamine stretch in the human AR, which is considered early disease stage in this model25(Supplementary Fig. 1a,b). We next sought to investigate the downstream effects of BMP-SMAD4 activation in skeletal muscle of SBMA patients, and found no change in expression of BMP-negatively regulated FBXO3026, as well as other E3 ubiquitin ligases commonly upregulated in muscle atrophy (FBXO21, TRIM63)27,28, the master controller of muscle homeostasis HDAC429 or other atrophy-related genes (FBXO32, ASB2), whose induction or repression directly depends on the TGFβ pathway26,30,31, as previously observed32,33 (Supplementary Fig. 2). As mice carrying a Smad4 deletion exclusively in muscle displayed features of severe primary muscle atrophy with increased protein catabolism and peculiarly no activation of atrophy-related genes when subject to fasting or denervation26, we hypothesised that activation of the BMP signalling in SBMA fails to generate an anti-atrophic response due to a functional SMAD4 deficiency. Over 45% (510/1127) of the transcripts differentially expressed in skeletal muscle of Smad4 knock-out (KO) mice upon denervation significantly overlapped with the SBMA mice transcriptomic profile (Fig. 1d), the near totality of which (483/510, 94%) following the same pattern of dysregulation, with 186 transcripts being upregulated (Representation factor: 1.6; P < 0.0001) and 297 downregulated (Representation factor: 1.6; P < 0.0001) in both data set and to a similar degree of fold change (Fig. 1e). Transcripts known to be activated or repressed by specific binding of the SMAD1/5/8-SMAD4 transcriptional complex to their promoter regions34, were respectively downregulated (Fig. 1f) and upregulated (Fig. 1g) in skeletal muscle of SBMA mice, further supporting a SMAD4 functional deficiency in this disease. Notably, mRNA levels of BMP ligands, which are subjected to SMAD4 positive feedback regulation35, were largely upregulated upon AR silencing in muscle of SBMA transgenic mice using a previously described miRNA approach36 (Fig. 1h), overall suggesting a negative effect of polyQ AR on SMAD4 transcriptional activity.
SMAD4 physically associates with AR upon receptor activation. As SMAD4 is known to engage with other transcription factors and coregulators to orchestrate specific gene expression programmes in a time- and context-dependent fashion37, we proposed that SMAD4 directly interacts with AR to modulate expression of its target genes. Immunoprecipitation experiments in HEK293T cells transfected with N-terminal FLAG-tagged SMAD4 and AR vectors showed that such association occurs upon treatment with SMAD4-transcriptional activator BMP7 and dihydrotestosterone (DHT), which induces AR nuclear translocation (Fig. 2a-c). BMP7 was chosen among other BMP ligands because of its well-established role as a positive regulator of muscle mass through the SMAD1/5/8-SMAD4 pathway38. We next generated deletion variants harbouring only the MAD homology (MH) 1, 2, or the linker region of SMAD4 (Fig. 2d), and mapped this interaction to the MH1 region (Fig. 2e). This domain, upon activation of the TGFβ receptor, becomes available as a binding platform for other transcription factors and binds to DNA, while the MH2 domain is mainly responsible for the interaction with receptors and oligomer formation with other SMADs39. To further map the physical association with AR and determine whether pre-emptive interaction with the DNA template is necessary, we created a series of glutathione S-transferase (GST)-tagged AR constructs for in vitro pull-down studies (Fig. 2f) and demonstrated that the AR NTD, essential for AR transactivation, binds directly to SMAD4 (Fig. 2g). Notably, the polyQ AR maintained the ability to interact with SMAD4, although to a lesser extent with increasing size of the polyglutamine stretch (Fig. 2c, g). By immunofluorescence experiments in human myoblasts we observed nuclear co-localization of endogenous SMAD4 and AR following DHT and BMP7 treatment (Fig. 2h), further suggesting a direct, activation-dependent, interaction between the two proteins.
AR enhances SMAD4 bound fraction and residence time. We reasoned that AR-SMAD4 transcriptional complex efficiently drives expression of SMAD4 target genes. C2C12 cultured stably expressing the human AR transgene with either 24Q or 100Q were transfected with a plasmid consisting of BMP-responsive elements from the Id1 promoter fused to a luciferase reporter gene. Significantly increased transactivation was observed upon combined treatment with BMP7 and DHT, compared to BMP7 alone or co-transfection with a plasmid expressing a constitutively active BMP receptor type 1A (BMPR1A) (Fig. 3a), suggesting a cooperative role of AR. This DHT-mediated enhancing effect was lost in the presence of polyQ AR (AR100Q) (Fig. 3a). Increasing BMP7 concentration linearly correlated with increased luminescence signal (Fig. 3b). We next investigated SMAD4 promoter occupancy on target genes by chromatin immunoprecipitation (ChIP) assay and found overall increased enrichment upon combined BMP7 and DHT treatment in the context of wild type but not mutant AR (Fig. 3c). To further understand the mobility and binding patterns to chromatin in real-time and at single-molecule resolution, we tracked individual SMAD4 in muscle cells. First, we generated stable AR24Q and AR100Q C2C12 lines expressing SMAD4-Halo, under the control of a Tet-regulated promoter. Halo labelling was chosen as it is well-suited for single-molecule imaging applications40. Time-lapse movies at a rate of 4–20 frames per second (fps) allowed the identification of bright spots of SMAD4-Halo signal in the nuclei of C2C12 cells (Fig. 3d). Induction of Id1 expression upon doxycycline and BM7 treatments was observed, confirming that SMAD4-Halo is transcriptionally active (not shown). The estimated diffusion constant of SMAD4 situated in the nuclei of these cells in a bound state is 0.0018µm/s2, which we note is similar to that reported for chromatin-bound histone H2B (0.0019 µm/s2)41 (Fig. 3e). Treatment with BMP7 led to slower bound state diffusion constant, which was further reduced by addition of DHT, as expected with the increasing crowding density of the transcriptional complex (Fig. 3e). The presence of polyQ AR returned the diffusion constant to around 0.0018 µm/s2, hinting at alterations of recruitment of biological molecules for the transcriptional machinery in SBMA (Fig. 3e). No changes were observed in the bound fraction across the different conditions, apart from a slight reduction upon polyQ AR activation (Fig. 3f). For quantitative comparison of bound state residence times, aspects of survival analysis were employed. The complement of the cumulative distribution function (1-CDF) was plotted as a function of time, corrected for photobleaching and fit to multi-exponential models. All datasets fit a double component model with higher precision than a simple exponential model (Supplementary Fig. 3). A triple exponential model does not improve the fit, suggesting that only two populations of bound molecules account for most of the variability. Expected residence times, the average time we expect a single molecule to stay in the bound state, were calculated directly from the residence time distributions (Fig. 3g). Decomposition of the bi-exponential fit of the distribution of residence times revealed that the short-lived populations of bound molecules (tau 1: ~3 s) largely dominated the residence time distributions over the slow component (tau 2: ~8 s) (Supplementary Fig. 4a). The faster residence times upon DHT treatment suggest more rapid target recognition of the transcriptional complex upon AR activation, rather than increased binding to specific response elements (Fig. 3g; Supplementary Fig. 4a-c)42. To corroborate this observation, we therefore calculated the rate of binding events per nucleus which appeared to be higher upon DHT treatment, suggesting a role of AR in facilitating the search for the cognate sequence (Fig. 3h). Notably, cells with the polyQ AR displayed the shortest residence time and the slowest rate of binding (Fig. 3g,h). The modulation of SMAD4 chromatin-bound fraction can occur upon changes in either its binding time to cognate sites and/or free time between binding events. In order to provide an unsupervised assessment of SMAD4 residence binding times, we analysed kymographs of the single-molecule movies. In this analysis, an immobilized SMAD4 molecule appears as a straight segment parallel to the temporal axis (Fig. 3i). We computed the distribution of SMAD4 track durations and found no difference between the unstimulated and stimulated conditions and with wild type and polyQ AR (Fig. 3j). Altogether these results indicate that AR facilitates the search process for SMAD4 binding sites rather than modulating the time SMAD4 remains bound on chromatin and that this cooperative function is impaired in SBMA.
AR activation correlates with SMAD4 transcription. We next tested SMAD4 transactivation ability of known target genes upon treatment with BMP7, DHT, or DHT and BMP7 in mouse C2C12 cells stably expressing human wild type (24Q) or polyQ (100Q) AR transgenes. Increased expression of these target genes upon BMP treatment was further potentiated by DHT, an effect that was not achieved in the presence of polyQ AR (Fig. 4a). Of note, no induction was observed in the selected targets with DHT alone (Fig. 4a).
We further investigated the ability of SMAD4 to induce the expression of Id1, a canonical BMP-responsive target gene43, using single-molecule fluorescence in situ hybridization (smFISH)44. This technology allows to distinguish between mature RNA, which appears as individual foci scattered throughout the cell, and nascent RNA, which localises at brighter foci at active transcription sites in the nucleus (Fig. 4b)45–47. In agreement with the bulk RNA assessment (Fig. 4a), DHT and BMP7-treated myoblasts displayed significantly higher amounts of both nascent (Fig. 4c) and mature Id1 mRNA (Fig. 4d), compared to both unstimulated cells and BMP treated cells only, supporting a transcription-dependent mechanism. This amplified effect was significantly reduced in the presence of polyQ AR (Fig. 4b-d). Overall these results led us to propose a model in which the anabolic activity of the transcription factor SMAD4 in skeletal muscle is enhanced by AR cooperation. Loss of this cooperative effect in SBMA leads to a summed SMAD4 functional transcriptional deficiency, which accounts for the primary muscle atrophy observed in this disease.
BMP7 delivery overcomes polyQ AR defective cooperation and rescues SBMA muscle atrophy in vivo. We hypothesized that AR cooperates with SMAD4 to drive a muscle hypertrophy programme in vivo and that a malfunction in this enhancing activity conferred by the polyQ contributes to the primary muscle atrophy in SBMA.
To test the hypothesis that AR cooperative effect on SMAD4 is additive, we generated two adeno-associated viruses (AAVs) expressing human BMP7 cDNA driven by the muscle-specific enhanced muscle creatine kinase promoter (Enh358MCK) and enhanced green fluorescent protein (eGFP) driven by the ubiquitous cytomegalovirus (CMV) promoter (BMP7) or eGFP alone (mock). Male wild type and SBMA transgenic mice carrying the human AR transgene with pathological glutamine expansion (AR100Q) were randomized to receive BMP7-expressing AAV9 or mock by single tail-vein injection at 5 weeks of age, at a dose of 2 × 1011 to 2.5 × 1011 vector genomes (vg) (Fig. 5a, Supplementary Fig. 6a). This SBMA mouse model accurately recapitulates the main features of SBMA pathology22, including the aberrant BMP activation in skeletal muscle observed in patients (Supplementary Fig. 1a,b). Efficient transduction in quadriceps muscle was validated by RT-qPCR, showing increased expression of human BMP7 (Supplementary Fig. 5), and by immunofluorescence, showing increased phosphorylated SMAD1/5 in the nuclei of myofibers (Fig. 5b), indicative of activation of the BMP pathway. AAV9-mediated delivery of BMP7 induced muscle hypertrophy in wild type mice, as previously reported26, an effect that was lost upon chemical castration with Leuprorelin, a luteinizing hormone–releasing hormone (LHRH) agonist that reduces testosterone release (Supplementary Fig. 6b). BMP7 treatment in SBMA mice promoted near-to-complete normalization of muscle size (Fig. 5c) and cross-sectional myofiber areas (Fig. 5d), with ~35% mass increase in quadriceps (QUAD), gastrocnemius (GAS) and extensor digitorum longus (EDL) muscles (Fig. 5e). We observed increased number of Pax7+ nuclei in muscle of BMP-treated mice (Supplementary Fig. 7a,b), suggesting that the anti-atrophy effect is at least partially mediated by increased proliferation of the satellite cells pool upon BMP activation, as previously suggested48,49. Notably, muscle regeneration is not impaired in SBMA muscle50. In order to verify whether the histopathological improvement also was associated with functional amelioration, we treated a cohort of SBMA mice using the same treatment paradigm (Fig. 5a). Statistical power analysis was performed on the basis of quantitative measures of body weight, to establish the minimum number of mice required (Cohen’s d effect size: 0.8; minimum sample size for each group: 10 mice). BMP7 treatment resulted in significantly improved end-point grip strength (Fig. 5f), improved rotarod performance (Supplementary Fig. 8), and increased overall survival (Fig. 5g) in SBMA mice. To further characterize the effect of BMP7 overexpression in muscle, we next performed transcriptomic analysis in quadriceps muscles from 11-week old transgenic male mice treated with AAV9-BMP7 or control and wild type littermates. We identified 1204 significantly upregulated and 1103 downregulated genes in SBMA skeletal muscle compared to wild type (Supplementary Fig. 9a). BMP7 treatment resulted in significant restoration of 22/35 SBMA hallmark molecular signature genes (Fig. 5h), with gene expression clusters showing a pattern toward normalization (Supplementary Fig. 9b). BMP induction of representative known SMAD4 targets was blunted in wild type mice treated with Leuprorelin and restored to wild type levels or above in skeletal muscle of BMP7-treated SBMA mice (Fig. 5i), further suggesting that over-activation of the BMP-SMAD4 signaling is able to overcome AR polyQ failed enhancing effect.