Transcriptional programming of translation by BCL6 controls skeletal muscle proteostasis

Abstract

Skeletal muscle is dynamically controlled by the balance of protein synthesis and degradation. Here we discover an unexpected function for the transcriptional repressor B cell lymphoma 6 (BCL6) in muscle proteostasis and strength in mice. Skeletal muscle-specific Bcl6 ablation in utero or in adult mice results in over 30% decreased muscle mass and force production due to reduced protein synthesis and increased autophagy, while it promotes a shift to a slower myosin heavy chain fibre profile. Ribosome profiling reveals reduced overall translation efficiency in Bcl6-ablated muscles. Mechanistically, tandem chromatin immunoprecipitation, transcriptomic and translational analyses identify direct BCL6 repression of eukaryotic translation initiation factor 4E-binding protein 1 (Eif4ebp1) and activation of insulin-like growth factor 1 (Igf1) and androgen receptor (Ar). Together, these results uncover a bifunctional role for BCL6 in the transcriptional and translational control of muscle proteostasis.

The acquisition and maintenance of skeletal muscle mass are critical for motor and metabolic functions, but the mechanisms controlling muscle proteostasis remain incompletely understood^1,2. Insulin/insulin-like growth factor 1 (IGF-1) signalling has emerged as a central regulator, leading to activation of AKT and mechanistic target of rapamycin complex 1 (mTORC1), which suppresses autophagy by phosphorylating ULK1 and enhances protein synthesis by phosphorylating S6 kinase (S6K) and 4EBP³. In addition, activated AKT phosphorylates FOXO transcription factors and localizes them to the cytoplasm, inhibiting expression of atrogenic targets including the ubiquitin ligases atrogin-1 and MuRF-1 (refs. 4,5). FOXO transcription factors act in conjunction with additional signalling cascades to mediate transcriptional control of skeletal muscle homeostasis. In particular, myostatin, a member of the transforming growth factor beta superfamily expressed and secreted predominantly by skeletal muscle, is a potent negative regulator of muscle growth that activates SMAD2/SMAD3 and suppresses AKT–mTORC1 signalling⁶. Conversely, bone morphogenetic proteins induce muscle hypertrophy by activating SMAD1/SMAD5/SMAD8 and mTOR signalling⁷. Androgens also increase muscle mass by enhancing IGF-1–AKT–mTORC1 signalling while suppressing myostatin and ubiquitin ligase expression⁸. Despite progress elucidating these and other regulators, important gaps remain in our understanding of transcriptional programmes and the precise mechanisms by which they control muscle proteostasis.

BCL6 is a transcriptional repressor recognized for its critical role in germinal centres and diffuse large B cell lymphomas but of virtually unknown function in skeletal muscle^9–11. It controls transcription through sequence-specific DNA binding by its C2H2 zinc finger domain and interactions with corepressors including SMRT, NCoR, BCoR, CtBP, MTA3/NuRD and class I and II histone deacetylases (HDACs). Although initial characterization of BCL6 nearly 30 years ago detected high levels in adult skeletal muscle, its role therein has remained elusive¹⁰. In myogenic cell lines, BCL6 was shown to be induced during differentiation¹², and expression of BCL6 in human skeletal muscle was inversely correlated to maximal oxygen uptake (VO₂ max), suggesting a role for BCL6 in metabolic programming¹³. Furthermore, BCL6 cofactors including class IIa HDACs and NCoR have been implicated in myofibre metabolism and muscle mass^14,15. Collectively, these observations pointed to the possibility that BCL6 could be an important transcriptional regulator in skeletal muscle.

Here, using tissue-specific constitutive and inducible genetic deletion, we reveal a role for Bcl6 in skeletal muscle protein homeostasis. Loss of Bcl6 in myofibres during mouse embryonic development led to reduced muscle mass, and inducible deletion in adult mice resulted in muscle atrophy. In both settings, loss of Bcl6 led to decreased muscle force production. Using stable isotopic incorporation and non-linear kinetic modelling together with integrated analysis of RNA sequencing (RNA-seq), chromatin immunoprecipitation with sequencing (ChIP–seq) and ribosome sequencing (Ribo-seq) in vivo, we found that BCL6 controls muscle mass by directly regulating key genes involved in translational control and proteostatic signalling. Accordingly, Bcl6 ablation reduces skeletal muscle protein synthesis by inhibiting early phases of mRNA translation and increasing autophagy. Together, our data establish BCL6 as a potent transcriptional determinant of translation and degradation in skeletal muscle.

Results

BCL6 is highly expressed in differentiated skeletal muscle

To begin to understand the role of BCL6 in skeletal muscle, we interrogated its expression across tissues of the body. Both in male and female mice, Bcl6 was highly expressed in skeletal muscles compared to other solid organs, particularly within mixed or predominantly glycolytic muscles such as extensor digitorum longus (EDL) and quadriceps (Fig. 1a). Similarly, in human tissues from the Genotype-Tissue Expression (GTEx) project, BCL6 is expressed highest in muscle, with levels nearly equivalent to those in whole blood (Fig. 1b). In line with these in vivo observations, Bcl6 mRNA and corresponding protein were upregulated 2.4-fold and 2.9-fold, respectively, in cultured myotubes compared to myoblasts (Extended Data Fig. 1a,b). Given its robust expression, we examined active regulatory regions along the Bcl6 locus in skeletal muscle. Using ChIP–seq datasets for the active enhancer marker acetylated histone 3 lysine 27 (H3K27ac)¹⁶, we identified a putative super-enhancer within the muscle Bcl6 locus (Fig. 1c). Together, these observations established BCL6 as a highly enriched transcription factor with potentially important roles in skeletal muscle identity and disease¹⁷.

Fig. 1 |. Bcl6 controls the development of skeletal muscle mass.

a, qPCR expression of Bcl6 in C57BL/6 male and female mouse tissues relative to the housekeeping gene 36b4 (n = 4 per group). b, RNA-seq data for BCL6 in 19 different human tissues from the GTEx database. The box plot shows the minimum to maximum tags per million values; the white bar indicates the median, and the grey rectangle represents the first to third quartile values. N is different for each tissue. c, ChIP–seq track for H3K27ac in quadriceps along the Bcl6 gene revealed a super-enhancer. d, Schematic of mouse skeletal muscle-specific in utero Bcl6 deletion model. Created with BioRender.com. e, qPCR expression of Bcl6 relative to the housekeeping gene 36b4 in quadriceps, gastrocnemius, soleus and heart from Bcl6^fl/fl and Bcl6^MKO mice (n = 5 per group). Unpaired, two-tailed t-test with Welch correction was performed. f, Western blot and densitometry of BCL6 and tubulin protein in quadriceps from n = 4 Bcl6^fl/fl and n = 5 Bcl6^MKO mice. Bcl6^fl/fl versus Bcl6^MKO P = 0.0077 by unpaired, two-tailed t-test with Welch correction. g, Total body weight, fat and lean mass in n = 10 Bcl6^fl/fl and n = 6 Bcl6^MKO male mice at 8 weeks of age. Unpaired, two-tailed t-test with Welch correction was performed. h, Representative image of lower limb muscles from Bcl6^fl/fl and Bcl6^MKO mice. i, Tissue weights of quadriceps, gastrocnemius, TA, EDL and soleus from n = 8 Bcl6^fl/fl and n = 6 Bcl6^MKO male mice at 16 weeks of age. Unpaired, two-tailed t-test with Welch correction was performed. j, H&E staining of representative TA muscle from 13-week-old male Bcl6^fl/fl and Bcl6^MKO mice. Scale bar, 300 μm. k, Myofibre CSAs determined from H&E-stained sections of TA from Bcl6^fl/fl and Bcl6^MKO mice (n = 4 per group). Bcl6^fl/fl versus Bcl6^MKO P = 0.0095 by unpaired, two-tailed t-test with Welch correction. All data are represented as the mean ± s.e.m. **P < 0.01 and ***P < 0.001. BAT, brown adipose tissue; PgAT, perigonadal adipose tissue; Quad, quadriceps; ScAT, subcutaneous adipose tissue; Skm, skeletal muscle.

Fig. 8 |. Model for BCL6-mediated control of skeletal muscle proteostasis.

BCL6-mediated transcriptional regulation controls the expression of atrogenic regulators to establish and maintain skeletal muscle mass. BCL6 directly represses Eif4ebp1 and possibly Mstn, while it directly activates Igf1 and Ar. Loss of Bcl6 in muscle acutely reduces cap-dependent translation and anabolic signalling, resulting in reduced protein synthesis, increased autophagy and skeletal muscle atrophy. Created with BioRender.com.

Developmental role of BCL6 in establishing muscle mass

To address the function of Bcl6, we crossed mice harbouring a floxed allele for Bcl6 (Bcl6^fl/fl)¹⁸ with animals expressing Cre recombinase from the human alpha-skeletal actin (HSA) promoter (Fig. 1d) to constitutively delete Bcl6 in mature striated muscle fibres (Bcl6^MKO mice) beginning 9 d post coitum¹⁹. Comparing Bcl6^MKO mice to controls at 8 weeks of age, Bcl6 mRNA was reduced by over 85% in skeletal muscles including the quadriceps, gastrocnemius and soleus but unchanged in the heart (Fig. 1e). Similarly, levels of BCL6 protein were diminished by 84% in the skeletal muscles of Bcl6^MKO animals, but not changed in heart (Fig. 1f and Extended Data Fig. 1c). Remarkably, although these mice were born at expected Mendelian ratios, males exhibited nearly 16% lower body weights and 19% reduced lean mass compared to controls (Fig. 1g). These changes were grossly apparent after necropsy of male Bcl6^MKO mice, in which we found a 36–44% weight reduction in quadriceps, gastrocnemius, tibialis anterior (TA) and EDL, but no difference in the soleus (Fig. 1h,i). Females exhibited a similar but more modest 8% reduction in body weight and 10% lower lean body mass with 28–35% reduced muscle weights (Extended Data Fig. 1d,e). Histologically, we found a reduction in myofibre cross-sectional area (CSA) with no dystrophic features in Bcl6-ablated muscle (Fig. 1j). Frequency distributions of muscle fibre CSA showed a leftward shift to smaller fibres in Bcl6^MKO muscles compared to controls with an overall 35% reduction in the CSA of TA muscle (Fig. 1k). Thus, developmental loss of skeletal muscle Bcl6 reduced skeletal muscle mass, particularly in glycolytic muscle groups.

Rapid muscle loss following muscle Bcl6 deletion in adults

To determine whether the reduced muscle mass of Bcl6^MKO mice reflected a developmental defect, we next sought to test the impact of Bcl6 ablation in adult skeletal muscle. We generated an inducible muscle Bcl6-knockout model (Bcl6^i-MKO) by crossing Bcl6 floxed mice to animals expressing a gene fusion of Cre flanked by two mutated murine oestrogen receptors (AEC mice) under the control of the HSA promoter (Fig. 2a)²⁰. This system allowed us to delete Bcl6 by treating mice with four consecutive daily doses of tamoxifen. Within 1 week of tamoxifen treatment, Bcl6 mRNA was reduced by 80–95% in various skeletal muscles of Bcl6^i-MKO animals compared to corn oil-treated Cre-expressing control mice (Cre⁺ oil) or 72–90% compared to tamoxifen-treated Cre-negative control mice (Cre⁻ tmx; Fig. 2b). Although tamoxifen itself moderately reduced muscle Bcl6 mRNA compared to corn oil treatment, BCL6 protein levels were not significantly different between the control groups, whereas BCL6 was reduced by over 90% in quadriceps from tamoxifen-treated Bcl6^i-MKO mice (Fig. 2c). Levels of Bcl6 mRNA and corresponding protein were similar in the hearts of both controls and inducible muscle knockout mice (Fig. 2b,c). We tested the impact of deleting Bcl6 from adult muscles by treating male Bcl6^i-MKO mice at 12 weeks of age with tamoxifen. Body weights and lean masses of Bcl6^i-MKO mice reduced significantly compared to both controls within 2 weeks (Fig. 2d). By 12 weeks after tamoxifen treatment, muscle Bcl6-ablated mice showed an overall 5–7% reduction in body weight, and an 11% reduction in lean body mass compared to both controls with no change in fat mass (Fig. 2d). Consistent with these changes in body composition, muscle masses for quadriceps, gastrocnemius, TA and EDL muscles were 16–29% reduced compared to muscles from either control group (Fig. 2e). We observed highly similar effects after inducible muscle Bcl6 deletion in younger cohorts treated with tamoxifen at only 8 weeks of age. In these studies, Bcl6^i-MKO male mice exhibited 4–7% reduced weight and 10% reduced lean mass compared to controls (Extended Data Fig. 2a), while females exhibited 4–8% reduced weight and 7–10% reduced lean mass within 12 weeks after Bcl6 ablation (Extended Data Fig. 2b). Moreover, necropsy of quadriceps, gastrocnemius, TA and EDL muscles revealed 27–32% reductions in mass compared to muscles from either control group in males (Extended Data Fig. 2c), and 19–23% reductions compared to either control in females (Extended Data Fig. 2d), with no loss in soleus mass. Similarly to constitutive Bcl6 muscle knockouts, myofibre mass was diminished in Bcl6^i-MKO muscles compared to Cre⁻ tmx controls, as shown by an overall 33% reduction in CSA (Fig. 2f,g). Thus, these data reveal a critical role for BCL6 in the postnatal maintenance of muscle mass.

Fig. 2 |. Rapid muscle loss following muscle Bcl6 deletion in adult mice.

a, Schematic of tamoxifen-inducible model for Bcl6 deletion in adult mice. Created with BioRender.com. b, qPCR expression of Bcl6 relative to the housekeeping gene 36b4 in quadriceps, gastrocnemius, soleus and heart from n = 2 Cre-positive oil-treated (Cre⁺ oil), n = 3 Cre-negative tamoxifen-treated (Cre⁻ tmx) and n = 3 Cre-positive tamoxifen-treated (Bcl6^i-MKO) males 1 week after treatment. One-way analysis of variance (ANOVA) with Dunnett’s multiple-comparisons test showed that in quadriceps, P(Cre⁺ oil versus Cre⁻ tmx) = 0.0194, P(Cre⁺ oil versus Bcl6^i-MKO) = 0.0011, P(Cre⁻ tmx versus Bcl6^i-MKO) = 0.0356; in gastrocnemius, P(Cre⁺ oil versus Cre⁻ tmx) = 0.0090, P(Cre⁺ oil versus Bcl6^i-MKO) = 0.0345, P(Cre⁻ tmx versus Bcl6^i-MKO) = 0.0015; in soleus, P(Cre⁺ oil versus Bcl6^i-MKO) = 0.0009, P(Cre⁻ tmx versus Bcl6^i-MKO) = 0.0042. c, Western blot and densitometry of BCL6, actin, total (Licor) protein in quadriceps (left, short exposure; middle, long exposure) and heart (right) from n = 2 Cre⁺ oil, n = 3 Cre⁻ tmx and n = 4 Bcl6^i-MKO males 1 week after tamoxifen or oil treatment. One-way ANOVA with Dunnett’s multiple-comparisons test showed that P(Cre⁺ oil versus Bcl6^i-MKO) = 0.0498 and P(Cre⁻ tmx versus Bcl6^i-MKO) = 0.0240. d, Serial measurements of total body weight, lean and fat mass in n = 10 Cre⁺ oil, n = 10 Cre⁻ tmx and n = 13 Bcl6^i-MKO male mice starting at 12 weeks of age before (pre) and up to 12 weeks after treatment. Two-way ANOVA with Tukey’s multiple-comparisons test was performed. e, Tissue weights of quadriceps, gastrocnemius, TA, EDL and soleus from Cre⁺ oil, Cre⁻ tmx and Bcl6^i-MKO male mice 12 weeks after treatment (n same as d). One-way ANOVA with Dunnett’s multiple-comparisons test was performed. f, Laminin staining of representative TA muscle from Cre⁻ tmx and Bcl6^i-MKO male mice 6 weeks after tamoxifen treatment. Scale bar, 200 μm. g, Myofibre CSAs determined from laminin-stained sections (n = 3 per group; P = 0.0025 by unpaired, two-tailed t-test with Welch correction). All data are represented as the mean ± s.e.m. ^δP < 0.05, ^δδP < 0.01 for Cre⁺ oil versus Cre⁻ tmx; ^#P < 0.05, ^##P < 0.01, ^###P < 0.001 for Cre⁺ oil versus Bcl6^i-MKO; *P < 0.05, **P < 0.01, ***P < 0.001 for Cre⁻ tmx versus Bcl6^i-MKO.

Muscle atrophy is associated with reduced BCL6

Given that genetic loss of muscle Bcl6 in adult mice led to atrophy, we next explored whether changes in BCL6 levels could play a pathophysiological role in muscle wasting conditions. C57BL/6 mice were fasted for 48 h, leading to a 15% reduction in skeletal muscle mass (Extended Data Fig. 2e), and we detected 50–70% reduced BCL6 transcript and protein in their atrophic muscles (Extended Data Fig. 2f,g). We also mined published data from mouse models of cancer-associated cachexia and found that muscle Bcl6 was significantly diminished in metastatic colon cancer (63% reduced) and trended lower in breast cancer and non-small-cell lung adenocarcinoma (55% and 18% reduced, respectively; Extended Data Fig. 2h)^21–23. Based on the suppression of muscle Bcl6 in these cancer models, we exposed differentiated C2C12 myotubes to cell-free conditioned media from colon-26 carcinoma and likewise observed significant reductions in BCL6 transcript and protein (Extended Data Fig. 2i). To determine signals that control muscle Bcl6 expression, we treated differentiated C2C12 myotubes with cytokines associated with cancer cachexia and hormones elicited by fasting (Extended Data Fig. 2j). We found that cytokines, particularly interleukin (IL)-6, and to a lesser extent IL-1B and tumour necrosis factor, reduced Bcl6 mRNA expression. The suppressive effect of IL-6 was blocked by inhibiting STAT3, the major transcriptional mediator for IL-6 signalling (Extended Data Fig. 2j). Similarly, growth hormone, which activates signal transducer and activator of transcription (STAT) proteins, reduced Bcl6 expression, while glucocorticoids (dexamethasone) had no effect. Collectively, these findings suggested that Bcl6 is regulated by circulating cytokines and growth factors and may be an important mediator of muscle loss in undernutrition or cancer-associated wasting.

Bcl6 controls a slow to fast shift in myofibres

Muscle atrophy differentially affects slow-twitch and fast-twitch myofibres and can be associated with changes in myofibre composition^24,25. Thus, to address the impact of Bcl6 ablation on muscle myofibres, we quantified slow-twitch type 1 fibres and fast-twitch oxidative (2a), intermediate (2x) and glycolytic (2b) fibres using immunofluorescence with antibodies recognizing their characteristic myosin heavy chains. In the quadriceps of Bcl6^MKO mice, we observed a 4.4-fold increase in type 1 fibres and 1.4-fold increase in 2a fibres with commensurate decreases in type 2x and 2b fibres compared to Bcl6^fl/fl controls (Fig. 3a). This shift was also apparent in other muscle groups of Bcl6^MKO mice. In the predominantly oxidative soleus muscle from Bcl6^MKO mice, we detected a twofold increase in type 1 fibres and a complete loss of type 2b fibres (Fig. 3b). Furthermore, in the highly glycolytic EDL, nearly 11% of myofibres were type 1 in Bcl6^MKO mice, while no type 1 fibres were detected in controls (Fig. 3c). We further confirmed our fibre typing results by quantitative PCR (qPCR) analysis and detected increased expression of slow troponins and myosin heavy chains (Tnni1, Tnnt1 and Myh7) and reciprocally reduced expression of fast isoforms (Tnni2, Tnnt3 and Myh4) in Bcl6^MKO muscles (Fig. 3d). We performed similar fibre-type analysis in adult inducible Bcl6^i-MKO mice 6 weeks after tamoxifen treatment (Extended Data Fig. 3a). In contrast to deletion during muscle development, adult deletion of Bcl6 in the quadriceps led to only a modest 1.5-fold increase in type 1 fibres but a more remarkable 2.3-fold increase in type 2a fibres with corresponding reductions in 2x fibres. Next, we assessed muscle mitochondrial content by quantifying mitochondrial DNA and western blotting for the mitochondrial membrane protein VDAC but found no changes in levels in Bcl6-knockout muscles (Extended Data Fig. 3b,c). We also interrogated respiratory chain mitochondrial complex subunits I–V and found that in Bcl6^MKO mice, complex II was significantly increased by 1.5-fold, while complex III and V trended higher (Extended Data Fig. 3d). Finally, we quantified changes in myofibre CSA according to myofibre type in Bcl6^MKO and Bcl6^i-MKO mice compared to their respective controls. We found that myofibre CSAs were reduced most strikingly among type 2b fibres in both our Bcl6 muscle knockout models (Extended Data Fig. 3e,f), resembling the preferential type 2b myofibre atrophy associated with starvation or cancer cachexia. Together, these results indicated that loss of muscle Bcl6 caused a shift towards myosin heavy chains typically associated with oxidative metabolism but did not significantly increase mitochondrial content and could not explain the reduced mass of knockout tissues.

Fig. 3 |. Bcl6 controls an oxidative to glycolytic shift in myofibres.

a–c, Representative fluorescent antibody-stained images of four myosin heavy chain isoforms in quadriceps (a), soleus (b) and EDL (c) from Bcl6^fl/fl and Bcl6^MKO mice. MYH7 (type 1 fibre) is blue, MYH2 (type 2a fibre) is green, MYH1 (type 2x fibre) is black and MYH4 (type 2b fibre) is red. Quantification of fibre types is expressed as percentage of the total shown on the right (a, n = 3 per group; b and c, n = 2 Bcl6^fl/fl and n = 3 Bcl6^MKO mice). Bcl6^fl/fl versus Bcl6^MKO P value for type 1 fibre in quad = 0.0041, type 1 fibre in soleus = 0.0315, type 2a fibre in soleus = 0.0002, type 2b fibre in soleus = 0.0088 and type 1 fibre in EDL = 0.0266 by unpaired, two-tailed t-test with Welch correction. d, qPCR expression of oxidative and glycolytic genes relative to the housekeeping gene 36b4 in EDL, quadriceps and soleus from Bcl6^fl/fl and Bcl6^MKO mice (n = 4 per group). Bcl6^fl/fl versus Bcl6^MKO; EDL: P value for Tnni1 = 0.0175, Tnnt1 = 0.02, Myh7 = 0.0019, Tnnt3 = 0.0334; Bcl6^fl/fl versus Bcl6^MKO; quadriceps: P value for Tnni1 = 0.0147, Tnnt1 = 0.0238, Myh7 = 0.0174, Tnni2 = 0.0001, Tnnt3 = 0.0021, Myh4 = 0.00008; Bcl6^fl/fl versus Bcl6^MKO; soleus: P value for Tnni1 = 0.0012, Tnnt1 = 0.0125, Myh7 = 0.0267, Tnni2 = 0.0242, Tnnt3 = 0.0042 by unpaired, two-tailed t-test with Welch correction. All data are represented as the mean ± s.e.m. *P < 0.05, **P < 0.01 and ***P < 0.001.

Bcl6 regulates skeletal muscle force production

To further assess the functional implications of reduced muscle mass in skeletal muscle Bcl6-deleted mice, we tested their forelimb grip strength. Male and female Bcl6^MKO mice were 26% and 11% weaker, respectively, than their sex-specific controls (Fig. 4a and Extended Data Fig. 4a). Similarly, 12 weeks after tamoxifen treatment, Bcl6^i-MKO mice exhibited reduced forelimb grip strength by 14–17% in males and 8–11% in females compared to their controls (Fig. 4b and Extended Data Fig. 4b). Normalization to lean mass abrogated these differences in both male and female Bcl6^MKO and Bcl6^i-MKO mice (Fig. 4a,b and Extended Data Fig. 4a,b), indicating that reduced strength in these muscle Bcl6-ablated animals was a consequence of decreased muscle mass. We next evaluated isometric muscle mechanics of TA muscles in live anaesthetized mice. Bcl6^MKO muscles produced 40% or 47% diminished maximum tetanic force in males or females, respectively (Fig. 4c and Extended Data Fig. 4c). Similarly, TAs of male Bcl6^i-MKO mice exhibited nearly 40% reduced tetanic force (Fig. 4d). We also calculated the specific force normalized to muscle CSA in male mice and found no difference between Bcl6^MKO or Bcl6^i-MKO mice and their respective controls, indicating a proportional relationship between mass and strength in Bcl6-deleted muscle (Fig. 4c,d). Reductions in maximum force produced by Bcl6^MKO and Bcl6^i-MKO muscle were stable for 25 consecutive contraction bouts as seen by their fatigue curves (Fig. 4e,f), but when normalized to initial force, their relative fatigue was unaltered compared to controls (Fig. 4e,f). In sum, these data demonstrated that loss of Bcl6 in skeletal muscle leads to atrophy-associated loss of strength.

Fig. 4 |. Bcl6 enhances skeletal muscle force production.

a,b, Absolute (left) and lean body mass normalized (right) forelimb grip strength measured in n = 10 Bcl6^fl/fl and n = 6 Bcl6^MKO male mice at 12 weeks of age (P value for absolute grip strength = 0.00007 by unpaired, two-tailed t-test with Welch correction) (a) and n = 6 Cre⁺ oil, n = 9 Cre⁻ tmx and n = 9 Bcl6^i-MKO male mice 12 weeks after tamoxifen or oil treatment (b). One-way ANOVA with Dunnett’s multiple-comparisons test for absolute grip strength showed that P(Cre⁺ oil versus Bcl6^iMKO) = 0.0390 and P(Cre⁻ tmx versus Bcl6^i-MKO) = 0.0001. Whiskers in the box plot show the minimum to maximum values, the line indicates the median and the box represents the first to third quartile values. c,d, Maximum tetanic force (left) and specific force (right) for TA muscle from 13-week-old Bcl6^fl/fl and Bcl6^MKO male mice (n = 4 per group; P value for max force = 0.00004 by unpaired, two-tailed t-test with Welch correction) (c) and n = 5 Cre⁻ tmx and n = 4 Bcl6^i-MKO male mice 6 weeks after tamoxifen treatment (P value for maximum force = 0.00003 by unpaired, two-tailed t-test with Welch correction) (d). e,f, Fatigue curves showing the decline in force production by the TA muscle from 13-week-old Bcl6^fl/fl and Bcl6^MKO male mice (n = 4 per group; P value for fatigue < 1 × 10⁻¹⁵ by two-way ANOVA) (e) and n = 5 Cre⁻ tmx and n = 4 Bcl6^i-MKO male mice 6 weeks after tamoxifen treatment (P value for fatigue = 0.0088 by two-way ANOVA) (f) over 25 consecutive bouts of isometric contraction. Relative fatigue calculated by dividing each value by the initial force is shown at the bottom. All data are represented as the mean ± s.e.m. ^#P < 0.05 for Cre⁺ oil versus Bcl6^i-MKO and **P < 0.01, ***P < 0.001 for Cre⁻ tmx versus Bcl6^i-MKO.

In light of the increased slow myofibre content in muscle Bcl6-deleted mice, we tested their endurance and activity. Bcl6^MKO and inducibly ablated Bcl6^i-MKO mice 12 weeks after tamoxifen treatment along with their respective controls were run to exhaustion on a treadmill, but we observed no impact for loss of Bcl6 on endurance performance (Extended Data Fig. 4d,e). Furthermore, we assessed total activity via infra-red beam breaking but detected no difference among Bcl6^MKO mice or Bcl6^i-MKO mice 10 d after tamoxifen treatment compared to controls (Extended Data Fig. 4f,g). These experiments demonstrated that the increase in slow myofibres was not reflected functionally in Bcl6 muscle knockout mice. Moreover, altered activity could not explain their shifted myofibre profiles or contribute to atrophy from skeletal muscle unloading.

BCL6 directly controls proteostatic regulators in muscle

To begin to understand the gene regulatory basis by which BCL6 influences skeletal muscle atrophy, we performed mRNA sequencing of quadriceps from Bcl6^i-MKO mice 1 week after tamoxifen-induced Bcl6 deletion. Our analysis revealed 3,595 differentially expressed (DE) genes compared to corn oil-treated Bcl6^i-MKO controls (Cre⁺ oil) or 2,592 genes compared to tamoxifen-treated Bcl6^fl/fl controls (Cre⁻ tmx; Fig. 5a). Despite some tamoxifen-driven contributions to gene expression variance (Extended Data Fig. 5a), 2,114 DE genes were identified between knockouts and both oil-treated and tamoxifen-treated controls, with 2,103 high-confidence genes exhibiting strongly correlated expression changes (r_s = 0.96; Fig. 5b). Prior studies have suggested a role for BCL6 to oppose STAT proteins, particularly STAT5 (refs. 26,27). However, we detected only 79 genes in common between our high-confidence muscle Bcl6 DE gene set and published microarray data from muscle-specific Stat5-knockout mice (Extended Data Fig. 5b)²⁸. This corresponded to only 13% of STAT5-regulated genes and 4% of BCL6-regulated genes, respectively, and we found no correlation in their regulation (Extended Data Fig. 5c). These findings suggest that BCL6 controls a distinct muscle gene network.

Fig. 5 |. BCL6 directly controls the expression of anabolic regulators.

a, Venn diagram depicting DE muscle genes between n = 5 Bcl6^i-MKO and n = 4 tamoxifen-treated controls (Cre⁻ tmx) or n = 4 corn oil-treated controls (Cre⁺ oil). DE genes identified in both are in the centre. DE genes specific to Bcl6^i-MKO and Cre⁻ tmx (right) or Bcl6^i-MKO and Cre⁺ oil (left) are also shown. b, Quadrant plot of log₂ fold change in gene expression for common DE genes in Bcl6^i-MKO/Cre⁻ tmx versus log₂ fold change in Bcl6^i-MKO/Cre⁺ oil. Spearman correlation coefficient r_s = 0.96. c, Ontology analysis of DE downregulated (left column) and upregulated (right column) genes in Bcl6^i-MKO mice compared to both controls. Cumulative hypergeometric statistical test was used. d, Heat map of FPKM values for selected genes in muscle samples from Bcl6^i-MKO and Cre⁻ tmx mice. Colours represent expression from maximum (yellow) to minimum (blue) for each gene. Log₂ fold change in expression for Bcl6^i-MKO samples versus Cre⁺ oil controls are also shown (right). e, Pie chart showing genomic distribution of BCL6 binding sites in C57BL/6 mouse quadriceps. f, Top motifs enriched near BCL6 peaks and P values calculated with cumulative hypergeometric statistical testing. g, Top biological process (top) and mouse phenotype (bottom) ontologies for genes near BCL6 peaks. Cumulative hypergeometric statistical testing was used. h, Integrated RNA-seq and ChIP–seq analysis using Cistrome-GO depicting the likelihood that a gene is regulated by BCL6 (rank product score). Genes with log₂ fold change expression > 1 in Bcl6^i-MKO/Cre⁻ tmx and annotated BCL6 binding sites are shown. i, Volcano plot showing upregulated (pink) and downregulated (light blue) BCL6 target genes from nascent RNA-seq in C57BL/6 mice exposed to the BCL6 inhibitor FX-1 (50 mg per kg body weight) compared to vehicle-treated controls for 3 h (n = 5 per group). Wald test was used to calculate the P values. j, UCSC browser tracks of RNA-seq in quadriceps from Cre⁺ oil, Cre⁻ tmx and Bcl6^i-MKO mice, H3K27ac ChIP–seq in quadriceps from Cre⁻ tmx and Bcl6^i-MKO mice 1 week after oil or tamoxifen treatment, nascent RNA-seq in quadriceps from mice treated with DMSO or FX-1 for 3 h, and BCL6 ChIP–seq track in quadriceps from C57BL/6 animals along the Eif4ebp1 (left) and Igf1 (right) genes.

Ontology analysis of our high-confidence BCL6-regulated gene set identified top scoring terms of translation, ribosome biogenesis and assembly, muscle structure development, modification by small-protein conjugation, and striated muscle contraction (Fig. 5c). Within these varied categories, we found many well-documented regulators of muscle structure and proteostasis (Fig. 5d). In line with the slow myofibre enrichment in Bcl6-knockout tissue, transcript levels for myosin light and heavy chains and troponins (Myl2, Myl6b, Myh2 and Tnni1) were increased, and contraction-related genes encoding components of the sarcoplasmic reticulum (Sln, Casq2) and Z-disc (Ky) were elevated and reduced, respectively. Expression of key signalling components controlling muscle mass including myostatin (Mstn) and growth arrest and DNA damage-inducible alpha (Gadd45a) were increased, while anabolic genes including androgen receptor (Ar) and IGF-1 (Igf1) were decreased. The genes encoding critical translation inhibitor (Eif4ebp1), ribosome components (Rpl9, Rps28) and RNA binding proteins like Zfp36l1 and Cirbp were strongly induced in Bcl6-ablated muscle, while RNA binding proteins like Cpeb1 implicated in translational initiation were reduced^29–34. In addition, genes encoding synthetic enzymes for spermidine and spermine (Amd1/Amd2, Smox), which are essential polycations required for translation, were reduced in Bcl6-deleted muscle^35,36. Finally, Bcl6 ablation also elevated the levels of many ubiquitin ligases including Cul7 and Rnf217, and decreased the levels of deubiquitinases like Usp13 and Usp44. Together, these transcriptional changes suggested the potential for Bcl6 to control muscle mass and function through multiple molecular targets influencing protein synthesis and breakdown.

BCL6 has been characterized as a transcriptional repressor in immune cells, liver and adipose tissue^18,37,38, but its direct regulatory impact in skeletal muscle has not been established. To further address its epigenomic control, we performed BCL6 ChIP–seq in C57BL/6 mouse quadriceps. Using a stringent false discovery rate (FDR) cutoff of 1 × 10⁻⁷, we identified 10,925 BCL6 binding peaks, over 88% of which were promoter-distal binding in location (Fig. 5e). Motif analysis of these sites showed a striking enrichment of consensus recognition sequences for BCL6 (Fig. 5f) and, to a lesser extent, motifs for well-known muscle transcription factors including MEF2 and SIX. Top ontologies for nearby genes included muscle development and adaptation, actomyosin structure and translation (Fig. 5g). Moreover, mouse phenotypes associated with genes vicinal to BCL6 peaks included decreased skeletal muscle fibre size, impaired contractility and muscular atrophy, recapitulating the phenotype of mice lacking skeletal muscle Bcl6 (Fig. 5g). BCL6 is known to interact with many corepressors including HDAC3 (refs. 37,39). To test whether loss of BCL6 affects acetylation in skeletal muscle, we performed ChIP–seq for H3K27ac in Cre⁻ tmx and Bcl6^i-MKO mice following tamoxifen treatment and detected significantly more signal surrounding BCL6 binding sites in knockout muscles, with an overall 15% increase in H3K27ac within 1 kb of peak centres (Extended Data Fig. 5d).

Next, we integrated our RNA-seq and BCL6 ChIP–seq datasets using the Cistrome-GO pipeline to determine putative direct targets for BCL6 and define its regulatory potential⁴⁰. Interestingly, we found that among genes that were transcriptionally altered by muscle Bcl6 ablation and located in the vicinity of BCL6 binding sites, BCL6 could act both as a transcriptional activator and repressor (Extended Data Fig. 5e). However, there was a stronger association with gene repression (that is, upregulated expression in Bcl6-ablated muscle) than activation. Both negative and positive transcriptional regulation were apparent among putative direct target genes of BCL6 (Fig. 5h). For example, BCL6 was predicted to repress Eif4ebp1 and Mstn but activate the expression of Igf1 and Ar. To further confirm direct BCL6 target genes, we performed nascent RNA-seq in vivo using 5-ethynyluridine labelling while acutely exposing mice to FX-1. a selective BCL6 inhibitor that competes off its corepressor interactions with SMRT, NCoR and BCoR⁴¹. This nascent RNA-seq data revealed that Eif4ebp1 is induced within 3 h of BCL6 inhibition, while Igf1 and Ar are simultaneously suppressed (Fig. 5i). In contrast, Mstn expression was unchanged, indicating either indirect BCL6 regulation or direct BCL6 regulation mediated via corepressors that are not targeted by FX-1. Gene-level visualization of our mRNA and nascent RNA-seq data confirmed that the expression of Eif4ebp1 and Mstn was increased and the expression of Igf1, Smox and Ar was decreased in Bcl6-deleted muscles (Fig. 5j and Extended Data Fig. 5f). Moreover, BCL6 binding sites were present in the vicinity of these genes, and there were changes in the surrounding H3K27ac signals that parallelled alterations in gene transcription. These findings suggested a target-specific role for BCL6 to directly repress or activate genes controlling muscle proteostasis.

The atrophy and loss of strength caused by Bcl6 muscle deletion appeared relatively more modest in females than in males (Extended Data Figs. 2b,d and 4a,b), raising the possibility that BCL6 gene regulation could be sexually dimorphic in muscle, as has been reported for its regulation in the liver²⁶. We found that in Bcl6^fl/fl control animals, muscle levels of Bcl6 transcript were significantly lower and BCL6 protein trended lower in females than males (Extended Data Fig. 5g,h). Accordingly, the transcriptional impact of BCL6 on both repressed targets (Akt1, Eif4ebp1) and activated targets (Igf1, Ar) was less robust in females compared to males, although most of these target genes were still significantly altered in muscle Bcl6-deleted females (Extended Data Fig. 5g). These results indicated that BCL6 has an attenuated transcriptional impact in the muscle of females.

BCL6 promotes skeletal muscle protein synthesis

Conditions that cause muscle atrophy perturb the balance between protein synthesis and degradation in varying ways. For example, starvation and cancer cachexia reduce protein synthesis and increase protein degradation by both autophagy and the proteasome, whereas nerve injuries increase protein synthesis but more profoundly increase degradation^42–44. To address how loss of Bcl6 led to reduced muscle mass, we performed deuterium labelling in Bcl6^i-MKO and control mice using a non-steady-state precursor labelling strategy that we have previously validated for determining protein turnover rates in vivo^45,46. One week after tamoxifen administration, a bolus of deuterated saline was administered and Bcl6^i-MKO and Bcl6^fl/fl control mice were serially collected over the course of a week. Collected tissues were used to determine ²H-labelling of protein-bound alanine in muscle by gas chromatography–mass spectrometry (GC–MS) as well as relative alanine concentrations. Skeletal muscle-bound alanine levels were significantly reduced in Bcl6^i-MKO mice compared to Cre⁻ tmx controls 2 d after deuterium labelling (Fig. 6a). Furthermore, isotope kinetics were analysed with a mathematical model to determine absolute protein synthesis rates, which identified a 26% reduction in protein synthesis among muscle Bcl6-deleted animals (Fig. 6a). To confirm results from our deuterium studies, we further tested protein synthesis in skeletal muscle Bcl6-ablated animals. We used surface sensing of translation (SUnSET) to determine protein synthesis based on acute puromycin incorporation into elongating peptides in skeletal muscles⁴⁷ and compared tamoxifen-treated Bcl6^i-MKO with Cre⁻ tmx mice. Western blotting with anti-puromycin revealed 54% reduced puromycin incorporation into Bcl6-ablated muscles compared to controls (Fig. 6b).

Fig. 6 |. Loss of Bcl6 reduces the rate of muscle protein synthesis.

a, Temporal profile for alanine labelling (MPE, moles percent excess) in quadriceps of Cre⁻ tmx and Bcl6^i-MKO mice following ²H-saline injection 1 week after tamoxifen treatment (n = 4 Cre⁻ tmx mice for days 0, 1 and 2 and n = 5 Cre⁻ tmx mice for days 4 and 7; n = 4 Bcl6^i-MKO for day 0 and n = 5 Bcl6^i-MKO for days 1, 2, 4 and 7); P value on day 2 = 0.0064 by unpaired, two-tailed t-test with Welch correction. Calculated rates of muscle protein synthesis on the right (n = 5 per group). P = 0.0319 by unpaired, two-tailed t-test with Welch correction. b, SUnSET analysis of puromycin incorporation into quadriceps in Cre⁻ tmx and Bcl6^i-MKO mice 1.5 weeks after tamoxifen treatment. Puromycin labelling relative to total protein (Memcode) by densitometry (right; n = 3 per group; P = 0.0315 by unpaired, two-tailed t-test with Welch correction). c, Schematic of Bcl6^i-MKOmRiboTag mouse model and Ribo-seq procedure. Created with BioRender.com. d, Quadrant plot showing log₂ fold change of RPFs versus log₂ fold change of mRNA for Bcl6^i-MKOmRiboTag/mRiboTag quadriceps 3 weeks after tamoxifen treatment. e, Volcano plot of TE. Significantly upregulated TE genes (yellow), downregulated TE genes (blue), known mRNAs with 5′ TOP motifs (pink) and non-significantly altered genes (grey) are plotted (n = 4 per group). Wald test was used to calculate the P values. f, Metagene distribution plot showing normalized Ribo-seq RPF reads from the TSS to transcription termination site (TTS) for genes with reduced TE. g, GO for transcripts with significantly altered TE. Cumulative hypergeometric statistical test was used. h, Browser tracks showing mRNA RPKM (top) and normalized RPF density (bottom) in quadriceps from mRiboTag and Bcl6^i-MKOmRiboTag mice along Lamc1 and Col1a1 genes. i, Western blot and densitometry of LC3 and total protein (Licor) in quadriceps from Cre⁻ tmx and Bcl6^i-MKO males 1.5 weeks after tamoxifen and following colchicine or vehicle treatment (n = 3 per group). Two-way ANOVA followed by Tukey’s multiple-comparisons test was performed. Flux is the difference in average LC3-II levels in colchicine minus vehicle (bottom right). Data in a, b and i are represented as the mean ± s.e.m. *P < 0.05 and **P < 0.01. NS, not significant.

Next, we crossed Bcl6^i-MKO mice to animals encoding HA epitope-tagged ribosomes (RiboTag mice) and generated tamoxifen-inducible muscle ribosome-tagged Bcl6 knockouts (Bcl6^i-MKO mRiboTag) and controls (mRiboTag) to test the impact of Bcl6 on translation using ribosome profiling (Fig. 6c)⁴⁸. Muscle lysates were treated with T1 RNase and ribosomes were immunoprecipitated to isolate ribosome protected fragments (RPFs). Over two-thirds of the recovered RPFs were between 26 and 30 nucleotides, and over 82% of RPFs mapped to coding DNA sequences in both the Bcl6^i-MKO mRiboTag and control samples (Extended Data Fig. 6a,b). Remarkably, Ribo-seq revealed that more than three times as many genes were significantly altered at the translational level than at the transcriptional level in Bcl6-ablated muscle (Fig. 6d). We detected a significant (P value < 2.2 × 10⁻¹⁶) overall reduction in translation efficiency (TE; Extended Data Fig. 6c) with 1,887 translating mRNAs significantly reduced and 1,203 translating mRNAs enhanced following Bcl6 deletion (Fig. 6e). Moreover, we found that the distribution of RPFs was increased in the vicinity of the translation start site (TSS) but decreased along protein coding sequences in Bcl6^i-MKO mRiboTag muscle, particularly among TE-reduced mRNAs (Fig. 6f and Extended Data Fig. 6d). TE-reduced targets were enriched for functions in extracellular matrix, vascular development, locomotion and cell adhesion (Fig. 6g), including laminin subunit gamma 1 (Lamc1) and collagen 1a1 (Col1a1; Fig. 6h). In contrast, TE-enhanced transcripts encoded proteins associated with nuclear bodies, mRNA processing, ribonucleoprotein complexes, chromatin binding and protein catabolism (Fig. 6g). Among these TE-enhanced mRNAs were transcripts with various roles in autophagy including initiation signalling (Cisd2, Rab12), ubiquitin-mediated cargo selection (Rnf166), mitophagy (Fundc1, Rnf166), ATG conjugation (Atg3), the lysosome (Ubqln1, Ubqln4, Rab1b, Rab24) and membrane delivery (Supt20h). Together, these findings reveal an extensive role for BCL6 to support translation beyond its immediate impact on transcription.

We also tested whether protein degradation was altered by loss of muscle Bcl6. We first assessed for changes in ubiquitin-mediated proteolysis by measuring ubiquitin signals in muscle lysates from Bcl6-ablated and control mice. We observed no changes in ubiquitin after Bcl6 deletion when examined 10 d (Extended Data Fig. 6e) or at later times after tamoxifen treatment. We interrogated autophagy using western blotting after treating both Cre⁻ tmx and Bcl6^i-MKO mice (10 d after tamoxifen treatment) with either vehicle or colchicine, a drug that blocks autophagosome degradation^49,50. While colchicine caused accumulation of LC3-II levels in both Cre⁻ tmx and Bcl6^i-MKO muscles, flux analysis indicated higher rates of autophagy in Bcl6^i-MKO mice (1.7 fold higher than controls; Fig. 6i). Collectively, these findings indicated that acute atrophy from skeletal muscle deletion of Bcl6 was attributable to reduced protein synthesis and increased autophagy.

BCL6 controls 5′ mRNA cap-dependent translation initiation

Accumulation of ribosomes at the start codon and corresponding depletion of ribosomes towards the stop codon indicated a defect in translation initiation in Bcl6^i-MKO mRiboTag muscles (Fig. 6f). Consistent with a role for BCL6 in cap-dependent translation initiation, we observed 5′ terminal oligopyrimidine (TOP) tracts among TE-reduced mRNAs but none in TE-upregulated transcripts (Fig. 6e). Based on these observations, we next sought to understand the molecular basis for reduced translation initiation in muscle Bcl6-ablated mice. A major regulatory step in cap-dependent translation initiation is the binding of 4EBP1 to EIF4E, which represses protein synthesis by blocking interaction between EIF4G and EIF4E (Fig. 7a). Using co-immunoprecipitation and western blotting to measure the EIF4E-associated proteins, we detected a clear increase in the inhibitory 4EBP1–EIF4E complex along with a corresponding decrease in the interaction of EIF4G with EIF4E in Bcl6^i-MKO muscle (Fig. 7a and Extended Data Fig. 6f). Given the critical role of threonine phosphorylation in modulating 4EBP1 activity, the above observation may result from enhanced 4EBP1 expression or reduced phosphorylation. Western blots to measure the levels of 4EBP1 and Thr37/Thr46 phosphorylation demonstrated a significant increase in 4EBP1 accompanied by a similar boost in p4EBP1 (Fig. 7b). Most importantly, we also tested levels of non-phosphorylated 4EBP1, the form that inhibits EIF4E. We observed significantly increased levels of non-phosphorylated 4EBP1 in Bcl6^i-MKO mice (Fig. 7b), consistent with the reduced muscle protein synthesis observed following Bcl6 ablation. Similar results were obtained in western blot experiments of 4EBP1 in muscle lysates from constitutive Bcl6^MKO mice (Extended Data Fig. 6g). These observations indicated that BCL6 promotes translation by suppressing 4EBP1.

Fig. 7 |. BCL6 controls translation in skeletal muscle.

a, Left: schematic of EIF4E’s interaction with 4EBP1 and EIF4G. Right: EIF4E co-immunoprecipitation and western blots for EIF4G (top), 4EBP1 (middle) and EIF4E (bottom) in quadriceps from Cre⁻ tmx and Bcl6^i-MKO males 1 week after tamoxifen treatment (n = 2 per group). b, Western blot and densitometry of p4EBP1, non-phospho-4EBP1, total 4EBP1 and actin protein in gastrocnemius from n = 3 Cre⁻ tmx and n = 4 Bcl6^i-MKO males 1 week after treatment. Cre⁻ tmx versus Bcl6^i-MKO P value for p4EBP1 = 0.0188, total 4EBP1 = 0.0319, ratio = 0.0258 and non-phospho-4EBP1 = 0.0426 by unpaired, two-tailed t-test with Welch correction. c,d, Western blot and densitometry of myostatin, AR (c), IGF-1–AKT signalling (d; pAKT, AKT, pS6, S6, pFOXO1, FOXO1) and actin or total (Licor, Memcode) protein in quadriceps from n = 3 Cre⁻ tmx and n = 4 Bcl6^i-MKO males 1 week after treatment. Cre⁻ tmx versus Bcl6^i-MKO P value for AR = 0.0133, pAKT = 0.0383, total AKT = 0.0016, pS6 = 0.048 and pFOXO1 = 0.0126 by unpaired, two-tailed t-test with Welch correction. e, Western blot and densitometry of pSMAD1/pSMAD5/pSMAD8, total SMAD1 and total (Licor) protein in quadriceps from Cre⁻ tmx and Bcl6^i-MKO males 1 week after treatment (n = 4 per group). f, Myofibre CSAs determined from skeletal muscle of Cre⁻ tmx and Bcl6^i-MKO mice infected with MyoAAV4As encoding scramble (Scrmb) or Eif4ebp1 shRNAs and then treated with tamoxifen. CSAs of 2,000 myofibres per mouse were determined 8 weeks later (n = 3 mice per group for scramble and Eif4ebp1 shRNA1 and n = 2 mice per group for Eif4ebp1 shRNA2). Whiskers in the box plot show 5–95th percentile values, the line indicates the median, and the box represents the first to third quartile values. Two-way ANOVA followed by Tukey’s multiple-comparisons test was performed. P value < 1 × 10⁻¹⁵ for Scrmb-Cre⁻ tmx versus shRNA2-Cre⁻ tmx, Scrmb-Cre⁻ tmx versus Scrmb-Bcl6^i-MKO, Scrmb-Cre⁻ tmx versus shRNA1-Bcl6^i-MKO, Scrmb-Cre⁻ tmx versus shRNA2-Bcl6^i-MKO, shRNA1-Cre⁻ tmx versus shRNA2-Cre⁻ tmx, shRNA1-Cre⁻ tmx versus Scrmb-Bcl6^i-MKO, shRNA1-Cre⁻ tmx versus shRNA1-Bcl6^i-MKO, shRNA1-Cre⁻ tmx versus shRNA2-Bcl6^i-MKO, shRNA2-Cre⁻ tmx versus Scrmb-Bcl6^i-MKO, shRNA2-Cre⁻ tmx versus shRNA1-Bcl6^i-MKO, shRNA2-Cre⁻ tmx versus shRNA2-Bcl6^i-MKO, Scrmb-Bcl6^i-MKO versus shRNA1-Bcl6^i-MKO, Scrmb-Bcl6^i-MKO versus shRNA2-Bcl6^i-MKO and shRNA1-Bcl6^i-MKO versus shRNA2-Bcl6^i-MKO. All data are represented as the mean ± s.e.m. *P < 0.05, **P < 0.01 and ****P < 0.0001.

We next tested how BCL6 controls the eIF4E–4EBP1 axis and convergent signalling cascades. Our ChIP–seq and RNA-seq experiments predicted a direct regulatory role for BCL6 on Eif4ebp1, Akt1, Igf1, Ar, and possibly Mstn (Fig. 5h–j and Extended Data Fig. 5f). We confirmed increased muscle expression of Eif4ebp1 and Mstn as well as suppressed Ar and Igf1 using qPCR (Extended Data Fig. 6h). Protein levels of muscle myostatin were almost 60% higher in Bcl6^i-MKO mice than in controls (Fig. 7c). Conversely, we detected a greater than twofold reduction of AR protein in Bcl6-deleted muscle (Fig. 7c). To address the signalling impact of reduced Igf1 and AR in Bcl6-ablated muscle, we interrogated total and phosphorylated levels of AKT and its downstream signalling targets S6 and FOXO1. Total AKT protein was significantly increased in Bcl6-knockout tissue (Fig. 7d), consistent with our observation that Akt is a directly repressed BCL6 skeletal muscle target gene (Fig. 5i). Active, phosphorylated AKT (Ser473) levels were also increased (Fig. 7d), but phosphorylated forms of S6 (Ser 240/Ser244; Fig. 7d) and FOXO1 (Ser256) were significantly lower (Fig. 7d). Finally, we interrogated SMAD proteins, which transduce signals from transforming growth factor beta superfamily members including myostatin and modulate AKT–mTORC1 signalling⁷. We could not detect phosphorylated SMAD2/SMAD3 in muscle lysates from tamoxifen-treated Bcl6^i-MKO or control mice, but levels of the pro-anabolic regulator phospho-SMAD1/SMAD5/SMAD8 trended lower (P = 0.07) in Bcl6-knockout tissues (Fig. 7e)^51,52. These results indicated complex changes along the IGF–AKT–mTORC1 axis with loss of Bcl6 that at one level could be pro-anabolic (for example, increased phospho-AKT), yet downstream signalling was pro-atrogenic (for example, reduced phospho-S6 and phospho-FOXO1).

Finally, we tested the functional impact of 4EBP1 on muscle atrophy in Bcl6-deleted mice. We developed MyoAAV4A vectors encoding Eif4ebp1 shRNAs or a scramble negative control⁵³. We infected mice with viruses and 1 week later treated all animals with tamoxifen to delete Bcl6 in Bcl6^i-MKO animals or serve as a drug control in Bcl6^fl/fl mice. Skeletal muscles were collected 8 weeks later. We found that our two Eif4ebp1 shRNAs knocked down Eif4ebp1 transcripts by 50–73% and corresponding protein levels by 20–40% in Bcl6-deleted skeletal muscles (Extended Data Fig. 6i,j). Remarkably, Bcl6 muscle knockout myofibre CSAs increased by 35–64% in mice with reduced 4EBP1, partially rescuing the atrophy observed in Bcl6^i-MKO mice (Fig. 7f). These observations highlight a key role for BCL6-mediated regulation of Eif4ebp1 and signalling pathways that converge with the IGF–AKT–mTORC1 axis to control translation and muscle mass (Fig. 8)⁵⁴.

Chronic Bcl6 loss evokes compensation to maintain muscle

While inducible deletion of Bcl6 caused acute muscle atrophy, lean mass became relatively stable after ~6 weeks (Fig. 2d and Extended Data Fig. 2a,b). To understand this compensation, we examined protein synthesis and degradation in Bcl6^i-MKO mice after 12 weeks of Bcl6 deletion. In contrast to the early effects of Bcl6 deletion that reduced protein synthesis (Fig. 6a,b), chronic deletion led to a modest increase in overall protein synthesis based on puromycin incorporation (Extended Data Fig. 7a). In addition, we observed slightly more autophagic flux in Bcl6-ablated muscles (Extended Data Fig. 7b), but these effects were more modest than the increase observed immediately after Bcl6 deletion (Fig. 6i). Finally, we examined ubiquitin-dependent proteolysis by blotting for ubiquitinated proteins, but we detected no change between genotypes (Extended Data Fig. 7c), mirroring findings at earlier timepoints after Bcl6 deletion. Thus, chronic adaptation for loss of Bcl6 involves raising protein synthesis and attenuating autophagy to preserve muscle mass.

To understand the molecular mechanisms of compensation from Bcl6 deficiency, we analysed gene regulatory and signalling changes that distinguish chronically from acutely Bcl6-deleted muscles. First, we performed mRNA-seq and detected nearly 1,300 gene expression changes unique to chronically Bcl6-deleted tissue (Extended Data Fig. 7d). Gene Ontology (GO) analysis for these altered transcripts revealed regulation of TOR signalling and negative regulation of cellular catabolism among top scoring terms (Extended Data Fig. 7e). Next, we carried out ChIP–sequencing of acetylated H3K27, a marker for enhancer and promoter activity, to find epigenomic changes associated with chronically Bcl6-deleted muscles. We analysed our mRNA-seq and H3K27ac ChIP–seq datasets from acute and chronic Bcl6 deletion using Integrated analysis of Motif Activity and Gene Expression (IMAGE)⁵⁵. This machine learning analysis identified many putative causal transcription factor motifs that were either (1) uniquely associated with gene transcription in acute or chronic Bcl6 deletion, or (2) commonly associated with gene transcription in both acute and chronic Bcl6 deletion but with opposite correlations to expression at these two times. Among the unique associations, motifs for NR4A nuclear receptors, which have been implicated in exercise performance and muscle mass, were predicted with highest confidence to be causal (Extended Data Fig. 7f)^56,57. Motifs for KLF, NKX, PBX and ZBTB factors were also predicted to contribute to time of Bcl6 deletion-dependent transcription with a modest degree of confidence and, to a lesser extent, FOXO transcription factors (Extended Data Fig. 7f). Motifs with opposing correlations to gene expression in acutely versus chronically Bcl6-deleted muscles included members of the KLF, MEF, MYOG, ZKSCAN (TFEB) and SMAD families, among several others (Extended Data Fig. 7g). Consistent with our GO and motif prediction analyses, we detected increased pro-anabolic phospho-SMAD1/SMAD5/SMAD8 signalling in chronically Bcl6-deleted muscles (Extended Data Fig. 7h). Additionally, we found that phosphorylated FOXO1, which was acutely reduced by Bcl6 deletion (Fig. 7d) was restored in knockouts 12 weeks later (Extended Data Fig. 7i), possibly mitigating the atrophy programme directed by its non-phosphorylated form. Profiles of phosphorylated and total levels of AKT, 4EBP1 and S6 in chronically Bcl6-ablated muscles resembled those observed immediately after Bcl6 deletion (Extended Data Fig. 7j–l). Overall, these findings pointed to a network of epigenomic regulators that redirected transcription and cell signalling to attenuate muscle loss from chronic Bcl6 deficiency by restoring protein synthesis and reducing catabolism.

Discussion

After nearly 30 years since the cloning and initial description of BCL6 as a proto-oncogene highly expressed in normal adult skeletal muscle¹⁰, we have discovered a potent regulatory role for BCL6 in the acquisition and maintenance of muscle mass and strength. We identified multiple anabolic regulators as direct BCL6 targets using genome-wide mRNA-seq, nascent RNA-seq and DNA binding analyses, including both directly BCL6-repressed targets such as Eif4ebp1 and BCL6-activated targets such as Igf1 and Ar. We postulate that de-repression of 4EBP1 and attenuated anabolic signalling driven by IGF-1, AR and SMAD pathways cause translational defects in Bcl6-ablated muscle. These effects are attributable, at least in part, to impaired cap-dependent translation, as evidenced by reductions in cellular EIF4G–EIF4E complexes, reduced ribosome occupancy across the coding sequences of thousands of genes, and partial rescue of myofibre atrophy with Eif4ebp1 knockdown in Bcl6-ablated skeletal muscle.

We found that BCL6 controls both the protein synthetic and degradative arms of muscle proteostasis. Based on ribosomal sequencing, BCL6 regulates the translation of many contractile proteins, which are among the most abundant in muscle, and vascular and remodelling factors that participate in the control of tissue differentiation and mass^58,59. It also directs translation of extracellular matrix components, including collagens and laminins implicated in muscular dystrophies. Although dystrophic changes were not apparent in Bcl6-deleted muscle, these findings suggest that BCL6 could be relevant in other muscle pathologies. Additionally, we identified a role for BCL6 in the suppression of autophagy in skeletal muscle, as we observed increased autophagic flux in mice after Bcl6 deletion. However, changes in the expression of autophagy-related genes were only modest in magnitude in knockout muscles, suggesting that BCL6-mediated control of autophagy may occur through a post-transcriptional mechanism. Indeed, our ribosomal sequencing data indicated that BCL6 suppresses the translational efficiency of proteins with varied functions in autophagy and raises the intriguing possibility that translational control underlies its regulation of autophagy. While de-repression of Eif4ebp1 is a likely candidate for initiating translational changes in our Bcl6 deletion models of muscle atrophy, the importance of other BCL6 targets linked to anabolic regulation is unknown. Precisely defining the effects of the IGF-1–AKT–mTOR pathway, SMAD, and androgen-induced signalling on skeletal muscle translation in vivo will be key to further frame our findings for BCL6 and understand the extent to which these pathways converge in their control of proteostasis.

Skeletal muscle atrophy frequently occurs in a fibre-type-specific manner and is accompanied by shifts in muscle myofibre composition, although mechanisms underlying these observations are incompletely understood^24,25. Atrophy from denervation or immobilization, for example, primarily affects type 1 fibres and is associated with a switch from slow to fast myofibres. Conversely, atrophy from cancer, sepsis, diabetes or aging is more pronounced in type 2 fibres and is linked to transition from fast to slow fibres. Resembling these illness-related and nutrient-related conditions, Bcl6 ablation leads to atrophy in predominantly glycolytic myofibres and promotes a shift to a slower myosin heavy chain isoform composition. Interestingly, this shift was not associated with altered mitochondrial content or fatigability in Bcl6-deleted muscles, consistent with other reports in which fibre type and mitochondrial metabolism can become disassociated⁶⁰. Furthermore, we observed marked reductions of muscle BCL6 in starvation and cancer cachexia models, and we found that BCL6 suppresses slow fibre-associated Myh2 and Myh7b but activates fast fibre-associated Myh4. These findings suggest that, in addition to a developmental role, BCL6 directs an integrated transcriptional programme controlling translation and myofibre specification in muscle wasting. The extent to which muscle BCL6 controls atrophy in pathophysiological states, particularly in humans, warrants further investigation.

Transcriptional repressors and corepressors play key roles in the regulation of muscle mass and metabolism. Despite documented physical interactions with NCoR1 and class II HDACs, which suppress muscle mass^15,61,62, we discovered an anabolic role for BCL6. Surprisingly, our integrated ChIP–seq and RNA-seq analyses suggested potential for BCL6 in muscle to function not only as a repressor but also as a gene-specific activator of transcription. These observations indicate BCL6 may engage alternative cofactor complexes to direct transcription in muscle, as has been described in B cells³⁷. It is also notable that previous studies have suggested opposing roles for BCL6 and STAT5 in the regulation of gene expression and involvement for both pathways in liver sex-specific transcription^11,26,27. Yet muscle-specific deletion of STAT5 resulted in skeletal muscle atrophy resembling the phenotype we observed in mice lacking muscle Bcl6 (ref. 28), and we found no correlation between gene expression changes in Bcl6-ablated and Stat5-ablated skeletal muscles. Consistent with its male-biased regulation in liver, we found that the transcriptional and functional impact of Bcl6 deletion on muscle mass and strength was more profound in males than in females. The basis for these sex differences is likely multifaceted, but reduced BCL6 and constitutively elevated 4EBP1 levels in female muscle are likely contributors (Extended Data Fig. 5h,i). Collectively, our findings establish a role for BCL6 in skeletal muscle protein homeostasis and contribute to an expanding recognition for BCL6 as a sexually dimorphic regulator of metabolism outside the immune system^18,38.

Methods

Animals

All strains were in the C57BL/6 genetic background. Bcl6^fl/fl mice¹⁸ were crossed with HSA-Cre mice (JAX, 006149) to generate Bcl6^fl/fl;HSA-Cre (Bcl6^MKO) mice. Bcl6^fl/fl mice were also crossed with animals expressing Cre fused to two mutated oestrogen receptors driven by the HSA promoter (JAX, 025750), to generate animals for tamoxifen-inducible Bcl6 deletion in muscle (Bcl6^i-MKO mice). Tamoxifen-inducible muscle ribosome-tagged Bcl6 knockouts (Bcl6^i-MKO mRiboTag) and controls (mRiboTag) were generated by further crossing Bcl6^i-MKO mice and controls to RiboTag mice (JAX, 011029). To induce Bcl6 deletion or expression of HA-tag on muscle ribosomes, 8- to 12-week-old mice were treated with tamoxifen (75 mg per kg body weight; Sigma, T5648) for four consecutive days via intraperitoneal injections. C57BL/6J (000664) mice were purchased from the Jackson Laboratory. All mice were maintained on a 14-h–10-h light–dark cycle with free access to standard chow and water at temperatures of 21–23 °C with 30–70% humidity. All animal care and use procedures were conducted in accordance with regulations of the Institutional Animal Care and Use Committees at Northwestern University and the Jesse Brown VA Medical Center under protocols IS00000321, IS00005071, IS00016185 and BX004898.

In vitro models

For primary myoblast culture, limbs were dissected from newborn C57BL/6J mice (postnatal day 1–3), washed in PBS, minced and incubated in 0.2% Collagenase D (Roche, 11088882001) solution (1 ml/0.3 g of muscle) for 30–45 min at 37 °C. The slurry was triturated every 10–15 min and filtered into a 50-ml conical tube through a 100-μm filter by adding the slurry followed by 10 ml PBS and 5 ml F10-Ham’s media (Thermo Fisher, 11550–043) with 20% FBS (Thermo Fisher, 11550), 1% Anti-Anti (Thermo Fisher, 15240–062) and 2.5 ng ml⁻¹ bFGF (Promega, G5071). The filtered cells were then grown on 60-mm plates coated with ECL (Sigma, 08110) at 37 °C with 8% CO₂. Once cells reached 80–90% confluence, they were differentiated in DMEM (Life Technologies, 11965–084) medium with 2% horse serum and 1% Anti-Anti with medium changes every 12 h for 5 d.

Colon-26 cells (American Type Culture Collection, CRL-2638) were grown to confluence in 10-cm plates using DMEM media supplemented with 10% FBS. Confluent cells were washed with PBS, and medium was replaced with DMEM supplemented with 2% horse serum for 26 h. Conditioned medium was collected, sterilized with a 22-μm filter, and added to C2C12 myotubes to a concentration of 33% conditioned medium.

Differentiated C2C12 myotubes were exposed to dexamethasone (1 μM, Sigma), growth hormone (500 ng ml⁻¹), IL-6 (100 pg ml⁻¹; PeproTech), tumour necrosis factor (10 ng ml⁻¹; PeproTech), IL-1β (10 ng ml⁻¹, PeproTech) and STAT3 inhibitor galiellalactone (10 μM; Cayman Chemical).

Body composition, muscle weight measurements and activity

Body composition was determined using magnetic resonance imaging (EchoMRI 2014 Body Composition Analyzer v.140320). For Bcl6^i-MKO mice, body composition was analysed weekly or biweekly beginning just before tamoxifen treatment to progressively monitor the effects of Bcl6 deletion. At experimental endpoints, muscles were isolated and weighed. Mouse activity was measured using the Phenomaster Caging System (TSE Systems) with infra-red sensors.

Histology

For paraffin-embedded sections, muscles were fixed in 10% formalin (Fisher, 245–684) overnight and transferred to 70% ethanol. Further processing and H&E staining was performed at Northwestern University’s Mouse Histology and Phenotyping Laboratory. For frozen sections, muscles were embedded in OCT compound (Tissue-Tek, 4583) and cut with 10-μm transverse sections taken from three regions at mid-body at least 150 μm apart. Muscle sections were fixed with ice-cold acetone and incubated in blocking solution (10% FBS, 0.1% Triton X-100 in 1× PBS) for 1 h at room temperature. Primary antibodies were added and incubated overnight at 4 °C. The following dilutions of mouse monoclonal antibodies from Developmental Studies Hybridoma Bank were used for fibre-type staining: MYH7/BA-F8—type I (1:10 dilution), MYHIIA/SC-71—type 2 A (1:30 dilution) and MYH/IIB/BF-F3—type 2B (1:10 dilution). Rabbit anti-laminin from Sigma-Aldrich (L9393; 1:1,000 dilution) was used to delineate fibre borders. The next day, slides were incubated with secondary antibodies for 1 h at room temperature, washed and mounted with Prolong Antifade Diamond (Thermo Fisher, P36970). Slides were imaged on a Zeiss Axio bright-field or Keyence BZ-X810 fluorescence microscope. Muscle myofibre CSA was calculated from H&E images using Image Pro software v9.2 (Media Cybernetics) or laminin immunofluorescence (Sigma-Aldrich, L9393) using Keyence BZ-X810 software. Areas were determined for at least 800 cells per mouse.

Muscle function analysis

Forelimb strength was measured using a grip strength meter (Columbus instruments, 1027SM). Mice were given three trials, with 5 s rest on a flat surface between pulls. The maximum of the three readings was used. In situ tetanic force from TA muscles was measured using a Whole Mouse Test System (Aurora Scientific, 1300 A) with a 1 N dual-action lever arm force transducer (Aurora Scientific, 300C-LR) in anaesthetized animals (3% isoflurane in 100% O₂). Parameters for tetanic isometric contractions were as follows: initial delay, 0.1 s; frequency, 200 Hz; pulse width, 0.5 ms; duration, 0.5 s; and stimulation, 100 mA. Length was adjusted to a fixed baseline of 30 mN resting tension for all muscles. Specific force was calculated by normalizing maximum force to the CSA. Treadmill experiments were run on an Exer 3/6 Treadmill (Columbus Instruments International) at 10-degree inclines. Animals were acclimated by running at 1 m min⁻¹ for 2 min after which speed was increased every minute by 1 m min⁻¹ until exhaustion. Time to exhaustion was determined as the time at which an animal no longer kept pace with the belt and received ten shocks in 1 min.

Protein isolation and quantification

Cells were dounce homogenized in cold RIPA buffer consisting of 1 mM EDTA (Sigma, E9884), 0.5 mM EGTA (Sigma, E4378), 1% Triton X-100 (Sigma, T8787), 0.1% sodium deoxycholate (Sigma, D6750), 0.1% SDS (Sigma, L3771), 140 mM NaCl (Sigma, S9888) and protease (Roche, 05892970001) and phosphatase (Roche, 04906837001) inhibitors. Snap-frozen tissues were homogenized in cold SDS lysis buffer consisting of 50 mM Tris pH 7.5 (Teknova, T1075), 100 mM NaCl, 1 mM EDTA, and protease (Roche, 05892970001) and phosphatase (Roche, 04906837001) inhibitors using a Qiagen Tissue-Lyser. Samples were spiked with 0.5% SDS, and they were rotated for 15 min at 4 °C. Protein was collected by centrifugation at >12,000g for 15–20 min and quantified using BCA assays (Thermo Fisher) measured at 562 nm using a SpectraMax Plus spectrophotometer (Molecular Devices).

Western blotting

Protein lysates were denatured by heat treatment and separated in 4–20% or Any kD TGX Precast Gels (Bio-Rad, 4568094 or 4569034). For OxPhos blots, non-denatured protein lysates were applied to a 16.5% Tricine gel. For LC3 blots, protein lysates were applied to 12% TGX Precast Gel (Bio-Rad, 4561043). Gels were transferred to 0.2-μm PVDF membranes (Bio-Rad, 1704157) and stained with 0.2% Ponceau S (Sigma, P3504) solution to assess transfer efficiency before blocking in 5% milk (Bio-Rad, 1706404) in 1× PBS with 0.1% Tween 20 (PBST; Teknova, P2192) or Intercept Blocking Buffer (LI-COR, 927–60001) followed by overnight primary antibody incubation at 4 °C. Primary antibodies were diluted in 5% BSA (Bio-Rad, 1706404) in 1× PBST or Intercept T20 Antibody Diluent (LI-COR, 927–65001) at 1:1,000 dilution unless otherwise indicated for western blots. The following primary antibodies were used: α-phospho-AKT Ser473 (4060), α-Pan-AKT (4691), α-LC3 A/B (12741), α-phospho-4EBP1 Thr37/Thr46 (2855), α-non-phospho4EBP1 Thr46 (4923), α−4EBP1 (9644), α-eIF4G (2469), α-phospho-S6 Ser240/Ser244 (5364), α-total S6 (2217), α-phospho-FOXO1 Ser256 (9461), α-FOXO1 (2880), α-ubiquitin (3936), α-SMAD1 (6944, Cell Signaling), α-phospho-SMAD1/SMAD5/SMAD8 (Vli31, MaineHealth Institute for Research; 1:2,000 dilution), α-Myostatin (ab98337; 1:250 dilution), AR (ab108341), VDAC (ab15895, AbCam), α-BCL6 (7388), α-eIF4E (2067, Santa Cruz Biotechnology), α-Actin (A2066), α-Tubulin (T4026; Sigma-Aldrich), α-puromycin (PMY-2A4; Developmental Studies Hybridoma Bank; 1:500 dilution) and OxPhos Rodent WB Antibody (45–8099; Thermo Fisher; 1:250 dilution). After the overnight incubation, blots were washed and incubated with horseradish peroxidase-conjugated secondary antibodies (Jackson ImmunoResearch, 115–035-174 and 211–032-171) at 1:10,000 dilution for 1 h, and visualized on a Protein Simple FluorChem E v5.0.4 or LI-COR Fc digital imager v5.2 using enhanced chemiluminescent horseradish peroxidase substrate (Thermo Fisher SuperSignal West Pico Plus, 34577). Multiple targets were probed on the same blot by stripping with OneMinute Advance Western Blot Stripping Buffer (GM Biosciences, GM6031) before adding a different primary antibody. Memcode Reversible Stain (Thermo Fisher, 24585) or Revert 700 Total Protein Stain (LI-COR, 926–11016) was used to visualize total protein. Protein densitometry was quantified using ImageJ 1.51s.

Immunoprecipitation

Muscles were collected, snap frozen and dounced ~10× in lysis buffer containing 50 mM Tris-HCl, 150 mM NaCl,1% NP40 (Igepal, CA-630) and 0.5% sodium deoxycholate. After mixing at 4 °C, lysates were spun down, and supernatants were collected. For immunoprecipitations, 3 μg of eIF4E antibody (Santa Cruz, 271480) or mouse IgG antibody (Santa Cruz, 2025) were added to 750 μg of protein lysate and rotated overnight at 4 °C. The following day, antibody complexes were precipitated with 50 μl of protein G beads and washed four times with 0.1 M sodium phosphate containing 0.01% Tween and cOmplete EDTA-free protease inhibitors (Sigma). Protein complexes were eluted for western blotting using 35 μl of 2× loading buffer and boiled for 10 min.

RNA and DNA isolation and qPCR

For RNA extraction, snap-frozen tissues were homogenized in TRIzol (Thermo Fisher, 15596018) using a MoBio Powerlyzer 24 and incubated with chloroform (Sigma, C2432). RNA was isolated using RNeasy kits (Qiagen, 74106) and used for qPCR or next-generation sequencing. For qPCR, RNA was quantified using a Nanodrop spectrophotometer (ND-1000) and cDNA was synthesized using iScript cDNA synthesis kits (Bio-Rad, 170–8891). Real-time qPCR was performed with SYBR Green Master Mix (Bio-Rad, 1725124) on a Bio-Rad CFX384 Touch Real-Time PCR System and analysed using Bio-Rad CFX manager software v3.1 (184500). RNA expression was determined using the relative standard curve method and normalized to the housekeeping gene 36b4. For nuclear and mitochondrial DNA isolation, muscles were processed using DNeasy kits (Qiagen). The mitochondrial DNA/nuclear DNA ratio was calculated as the ratio of Cytb or Cox1 to Gcg or Hbb as determined by ΔΔCt method. Primer sequences are listed in Supplementary Table 1.

mRNA sequencing and analysis

RNA quality was assessed using a Bioanalyzer RNA 6000 nano kit (Agilent, 5067–1511) to ensure RNA Integrity Number scores > 7.0. Sequencing libraries were constructed using TruSeq Stranded mRNA sample prep kits (Illumina, 20020594) and quantified using Bioanalyzer High Sensitivity DNA kit (Agilent, 5067–4627) and qPCR (Kapa Biosystems, KK4824) and sequenced on Illumina NextSeq 500 and 2000 or HiSeq 4000 instruments. RNA raw sequence reads were aligned to a reference genome (mm10) using STAR version 2.5.2 with default parameters⁶³. Gene expression at exons was quantified using HOMER (version 4.11)⁶⁴. DE RNAs were then normalized and identified using DESeq2 (version 1.30.1)⁶⁵ with an adjusted FDR < 0.05. Control-specific and consensus DE genes were plotted as Venn diagrams using DiVenn⁶⁶. GO analysis was performed using Metascape⁶⁷. Heat maps were generated with GraphPad Prism versions 9 and 10. Principal component analysis was performed using ClustVis⁶⁸.

Nascent RNA-seq

For in vivo nascent RNA-seq, animals were treated with FX-1 (50 mg per kg body weight; Cayman Chemical) or DMSO (solvent control) for 3 h and then injected with 5-ethynyl-uridine 1 h before euthanasia. One half of a harvested quadriceps muscle was processed for total RNA as described above. Then, 5-ethynyl-uridine-labelled nascent RNAs were biotinylated using the Click-iT Nascent RNA Capture Kit (Invitrogen, C10365), isolated with MyOne Streptavidin T1 magnetic Dynabeads (Invitrogen, 65601), and used to produce sequencing libraries with the Ovation SoLo RNA-seq System (Tecan, 0407–32). Libraries were sequenced with 100-bp SE reads on an Illumina NextSeq 2000. Reads were aligned to the mm10 reference genome using STAR version 2.5.2 with default parameters. PCR duplicates were removed using Tecan’s Duplicate Marking Tool, NuDup (https://tecangenomics.github.io/nudup/). Expression was quantified with HOMER using the analyzerepeats.pl command with the --count genes option. Differential expression was analysed as described for mRNA sequencing.

Chromatin immunoprecipitation followed by sequencing

For H3K27ac ChIP–seq, a single quadriceps from each animal was processed and nuclei were isolated and sonicated as described previously¹⁶. Samples were incubated with 5 μg rabbit anti-H3K27ac (Active Motif, 39133) overnight. Antibody complexes were precipitated with anti-rabbit IgG-conjugated paramagnetic beads (Thermo Fisher, 11203D). For BCL6 ChIP–seq, quadriceps from five animals were pooled. Muscles were harvested, minced and processed using a Polytron tissue homogenizer in homogenization buffer (0.15 mM NaCl, 0.05 mM EDTA pH 7.5, 0.5 mM Tris, 0.5% NP40 (Sigma, 18896), 1% Triton X-100). The homogenate was then dounced (~10 strokes) and serially filtered from 150 μm to 30 μm (Sysmex 04–0042-2319, 2317, 2315). Filtrate was spun at 600g for 10 min and the isolated nuclei were crosslinked at room temperature with 2 mM disuccinimidyl glutarate (Proteochem, C1104) for 20 min and then with 1% formaldehyde for 5 min. After quenching with 125 mM glycine, crosslinked nuclei were rinsed with PBS and sheared in buffer containing 1% SDS, 10 mM EDTA and 50 mM Tris for six cycles (30 s on, 30 s off) using a Diagenode Bioruptor into 200–1,000-bp fragments. Chromatin was incubated with 5 μg of antibody against BCL6 (ref. 18) overnight, and antibody complexes were precipitated with anti-guinea pig IgG (AbCam, ab6698) conjugated paramagnetic beads (Thermo Fisher, 14203). Immunoprecipitates were washed and decrosslinked. DNA was isolated using MinElute PCR purification kits (QIAGEN, 28006) and used to generate sequencing libraries using KAPA HyperPrep kits (Roche, KK8504). Adaptor-ligated ChIP DNA was size selected to obtain inserts of 200–500 bp using a PippenHT (Sage Science, HTC2010). Libraries were assessed by both a Bioanalyzer High Sensitivity DNA kit (Agilent, 5067–4627) and qPCR-based quantification (Kapa Biosystems, KK4824) and sequenced using Illumina NextSeq 500 and 2000 instruments.

Analysis of ChIP–seq data

For ChIP–seq, reads were aligned to the mm10 reference genome with Bowtie (version 1.1.2)⁶⁹ using --m 1 and --best parameters. Tag directories were generated using makeTagDirectory using the --tbp 1 option. Further analysis was performed with HOMER⁶⁴. Super-enhancers were identified from H3K27ac ChIP–seq data in mouse quadriceps with the findPeaks command, specifying --style super (FDR = 0.001) using Gene Expression Omnibus (GEO) datasets GSM3515055, GSM3515056 and an additionally deposited dataset. The makeUCSCfile command was used to generate H3K27ac ChIP–seq UCSC browser tracks. BCL6 peaks were identified using the findPeaks command, specifying --style factor and --fdr 0.0000001 for stringent peak calling and annotated to nearest genes using annotatePeaks.pl. HOMER’s findMotifsGenome. pl command was used to perform motif analysis using --size 200 and --mask with standard background. Gene ontologies near BCL6 ChIP–seq peak were generated using GREAT⁷⁰ for regions 5 kb upstream and 1 kb downstream of the TSS plus distal extensions up to 1,000 kb. Cistrome-GO⁴⁰ was used to integrate RNA-seq and BCL6 ChIP–seq data and calculate regulatory potential scores. The makeBigWig.pl command was used to generate browser tracks for BCL6, H3K27ac and RNA-seq data. H3K27ac tag density was quantified within a 1-kb vicinity of BCL6 peaks by using HOMER’s annotatePeaks.pl command, with --size 1000 option for scatterplots. GraphPad Prism was used to generate box plots of H3K27ac tags in different genotypes. Integrated analysis of Motif Activity and Gene Expression (IMAGE)⁵⁵ was used to predict transcription factor activities accounting for changes in gene expression (https://github.com/JesperGrud/IMAGE/).

In vivo protein synthesis and degradation analysis

Rates of muscle protein synthesis were determined by following incorporation of ²H from ²H₂O into newly made protein-bound alanine as described previously⁴⁶. Mice were given 0.7 ml intraperitoneal injections of ²H-labelled saline (9 g of NaCl in 1 l of 99.9% ²H₂O; deuterium oxide; Sigma, 151882). Mice were euthanized at 1 (day 0), 24 (day 1), 48 (day 2), 96 (day 4) and 168 (day 7) hours after the injection, and serum and skeletal muscles were collected. This was followed by sample processing and GC–MS analysis. The deuterium decay curve of body water was used to compute turnover rates. ²H-labeling of protein-bound alanine in muscle was quantified by GC–MS (Agilent Chemstation Mass Spectrometry Software E.02.02.1431), while alanine concentration was determined using the coupled reaction of alanine transaminase and lactate dehydrogenase. Mathematical modelling was performed based on a two-compartment model with free deuterium in body water (precursor) and deuterium incorporated into protein-bound alanine (product) in muscle tissue. Non-linear least-square fitting mathematical algorithms were used to estimate rate constants and synthesis rates were determined.

Puromycin incorporation

For all in vivo measurements of protein synthesis with SUnSET⁴⁷, puromycin solution (0.04 μmol per gram body weight in 200 μl of PBS) was injected into mice via intraperitoneal injection. Muscles were collected after 30 min and processed for western blotting with α-puromycin antibodies.

Autophagic flux

For autophagic flux experiments, mice were given intraperitoneal injections of colchicine (0.4 mg per kg body weight per day) or water for 3 d before muscle collection. Harvested quadriceps muscles were processed for protein and subjected to western blot using anti-LC3A/LC3B (Cell Signaling, 12741). Flux was determined as the difference in normalized LC3-II levels in mice treated with colchicine versus vehicle for each genotype.

Ribosomal sequencing and analysis

To generate ribosome footprints, quadriceps muscles from Bcl6^i-MKO and control RiboTag animals were harvested and snap frozen in liquid nitrogen 3 weeks after tamoxifen treatment. Frozen tissue was pulverized, and approximately 50 mg was retained for matched RNA-seq libraries. The remaining material was resuspended in 1 ml per 200 mg Polysome Lysis Buffer (20 mM Tris pH 7.5, 100 mM KCl, 5 mM MgCl₂, 1% Tween 20, 0.25% Na-deoxycholate) supplemented with 1× protease inhibitors (Roche) and 100 μg ml⁻¹ cycloheximide, rested on ice for 10 min, dounced 10× with a tight pestle, incubated for 10 min on ice, then cleared by centrifugation at 12,000g for 10 min at 4 °C. Approximately 3 μl of T1 RNase (Roche, EN0541) was added and incubated at 37 °C for 45 min. After quenching with 15 μl SUPERase−In (Invitrogen, AM2696), 50 μl of magnetic anti-HA beads (Pierce 88836) was added and incubated at room temperature for 1 h with rotation. Beads were washed 3× with high-salt Polysome Wash Buffer (20 mM Tris pH 7.5, 300 mM KCl, 10 mM MgCl₂, 1% Tween 20, 0.25% Na-deoxycholate) supplemented with 1× protease inhibitor. Bead-bound RNA was extracted with TRIzol and precipitated overnight with isopropanol and NaOAc at −80 °C. RNA was pelleted and resuspended in a poly-nucleotide kinase reaction (5 μl 2× ligation buffer, 1 μl PNK, 1 μl SUPERase−In, 3 μl H₂O), incubated at 37 °C for 30 min and run out on a 15% polyacrylamide TBE-urea gel. Fragments between 34-nucleotide and 26-nucleotide markers were extracted from the gel in RNA gel extraction buffer (300 mM NaOAc, 1.0 mM EDTA, 0.25% (wt/vol) SDS) overnight. RNA footprints were precipitated overnight with isopropanol and NaOAc at −80 °C, resuspended in 7 μl RNase-free H₂O and used to produce libraries using the NEBNext Small RNA Library Prep Kit. Libraries were sequenced with 50 bp SE reads on an Illumina HiSeq 4000 instrument.

RPFs were processed as follows. Adaptors were trimmed (NEBNext Small RNA read1 adaptor – AGATCGGAAGAGCACACGTCTGAACTCCAGTCAC) and reads filtered for minimum length of 18 nucleotides using Cutadapt⁷¹. Reads were then aligned against a list of non-coding RNAs using Bowtie. Unaligned RPF reads were next aligned to the mm10 reference genome using STAR⁶³ with the following options: --outFilterScoreMinOverLread 0.3, --outFilterMatchNminOverLread 0.3. Matched RNA samples were processed with STAR using default settings. Raw RPF and RNA read count matrices were generated with HOMER using the analyzerepeats.pl command with the --noadj option. Metagene plots were generated with HOMER using the makeMetaGeneProfile.pl command with the options --gbin 50, --gRatio 5, --size 500 and --histNorm 10. Changes in TE were detected with RiboDiff (v0.2.1)⁷². Quality-control metrics for mapping and RPF length were determined using RiboToolKit⁷³.

MyoAAV infections

MyoAAV4A viruses encoding Eif4ebp1 shRNAs driven by the U6 promoter were produced by Vector Biolabs. For in vivo knockdown, mice were infected with 1 × 10¹² viral copies per mouse via retro-orbital injection. shRNA sequences are listed below.

1. CCGGGTCACAATTTGAGATGGACATCTCGAGATGTCCATCTCAAATTGTGACTTTTTG

2. CCGGAGGCGGTGAAGAGTCACAATTCTCGAGAATTGTGACTCTTCACCGCCTTTTTTG

Data collection and statistical analysis

For animal experiments, aged-matched mice with littermate controls were utilized. For experiments requiring treatments or conditions, mice or cells of a given genotype were randomized to generate experimental groups. Data collection was performed with blinding, but analysis was not performed blind to the conditions of experiments. All animals and data points are shown. We used Microsoft Excel v16 and GraphPad Prism v9 and v10 to perform the statistical analyses. Two-sided Student’s unpaired t-test with Welch correction or one-way ANOVA with Dunnett’s multiple-comparisons tests were used to compare body and tissue weights, muscle functional parameters, qPCR gene expression and protein densitometry between two or multiple groups, respectively. Two-way repeated-measures ANOVA with Tukey’s multiple-comparisons test was used to compare the time course of body composition analysis in Bcl6^i-MKO mice. One-way ANOVA with Tukey’s correction was performed on normalized ChIP–seq tag counts to compare Bcl6^i-MKO versus Cre⁻ tmx samples. Kolmogorov–Smirnov testing was used to analyse TE. No statistical methods were used to predetermine sample sizes, but our sample sizes were similar to those reported in previous publications characterizing muscle mass and function and gene regulation^{4,6,15,28,46,47,51}. Data distribution was assumed to be normal, but this was not formally tested. Sample sizes and P values are listed in the figure legends.

Extended Data

Extended Data Fig. 1 |. Bcl6 controls the development of skeletal muscle mass.

(a) qPCR expression of Bcl6 and myogenic markers in primary myoblasts and myotubes relative to the housekeeping gene Rn45s (n = 4 per group). Myoblast vs myotube P-value for Bcl6 = 0.0124, Mb = 0.0153, Mef2c = 0.0146 by unpaired, two-tailed t-test with Welch correction. (b) Western blot and densitometry of BCL6 and total (Memcode) protein in primary myoblasts and myotubes (n = 4/group). Myoblast vs myotube P = 0.0036 by unpaired, two-tailed t-test with Welch correction. (c) Western blot and densitometry of BCL6 and Actin protein in heart from Bcl6^fl/fl and Bcl6^MKO mice (n = 3/group). (d) Total body weight, fat, and lean mass in n = 7 Bcl6^fl/fl and n = 10 Bcl6^MKO female mice at 8 weeks of age. Bcl6^fl/fl vs Bcl6^MKO lean mass P = 0.0184 by unpaired, two-tailed t-test with Welch correction. (e) Tissue weights of quadriceps (quad), gastrocnemius (gastroc), tibialis anterior (TA), extensor digitorum longus (EDL) and soleus from female Bcl6^fl/fl and Bcl6^MKO mice at 16 weeks of age (n = 7/group). Unpaired, two-tailed t-test with Welch correction was performed and all p-values can be found in the source data file. All data are represented as mean ± SEM. * p < 0.05, ** p < 0.01 and *** p < 0.001.

Extended Data Fig. 2 |. Inducible deletion of Bcl6 in adult mice results in muscle loss.

(a, b) Serial measurements of total body, lean, and fat mass in (a) n = 6 Cre⁺ oil, n = 9 Cre⁻tmx and n = 9 Bcl6^i-MKO male and (b) n = 6 Cre⁺oil, n = 8 Cretmx and n = 8 Bcl6^i-MKO female mice starting at 8 weeks of age prior to (pre) and up to 12 weeks after treatment. Two-way ANOVA with Tukey’s multiple comparisons test was performed. (c, d) Weights of quadriceps (quad), gastrocnemius (gastroc), tibialis anterior (TA), extensor digitorum longus (EDL) and soleus from Cre⁺ oil, Cre⁻tmx and Bcl6^i-MKO (c) male and (d) female mice 12 weeks after treatment (n = same as (a,b)). One-way ANOVA with Dunnett’s multiple comparisons test was performed. (e) Tissue weights in n = 4 ad lib chow fed and n = 5 48-hour fasted C57BL/6 male mice. (f) qPCR of Bcl6 normalized to 36b4 in muscles from fed and 48-hour fasted mice (n = 3/group). (g) Western blot and densitometry of BCL6 and Actin in n = 4 fed and n = 5 from 48 hour fasted mouse quadriceps. Unpaired, two-tailed t-test with Welch correction was performed in (e-g). (h) Bcl6 levels in published cancer-associated cachexia syndrome (CACS) models. GEO accession numbers are listed. Control vs C26 colon cancer P = 0.0008 by unpaired, two-tailed t-test with Welch correction (n = 3/group). Boxplot shows min to max values with line at the mean. (i) Bcl6 expression in C2C12 myotubes exposed to 33% colon-26 carcinoma (C26) conditioned versus control media for 8 hours (n = 3/group; unpaired, two-tailed t-test with Welch correction was performed) and corresponding protein blots after 8 or 24 hours. (j) Bcl6 expression in C2C12 myotubes treated with cytokines (n = 3/group), growth hormone (n = 2/group) and glucocorticoids (n = 3/group) for 8 hours. Unpaired, two-tailed t-test with Welch correction and one-way ANOVA with Dunnett’s multiple comparisons were performed. Exact p-values can be found in the source data file. All data in (a-g) and (i-j) are represented as mean ± SEM. δ p < 0.05, δδ p < 0.01 for Cre⁺ oil vs Cre⁻tmx; # p < 0.05, ## p < 0.01, ### p < 0.001 for Cre⁺ oil vs Bcl6^i-MKO; * p < 0.05, ** p < 0.01, *** p < 0.001 for Cre⁻tmx vs Bcl6^i-MKO or fed vs fasted or vehicle vs treated myotubes.

Extended Data Fig. 3 |. Bcl6 controls an oxidative to glycolytic shift in myofibres.

(a) Representative fluorescent antibody-stained images of four myosin heavy chain isoforms in quadriceps from Cre⁻ tmx and Bcl6^i-MKO mice (left). MYH7 (type 1 fiber) is blue, MYH2 (type 2a fiber) is green, MYH1 (type 2x fiber) is black, and MYH4 (type 2b fiber) is red. Quantification of fiber types (right) expressed as percentage of total (n = 3/group). Cre⁻ tmx vs Bcl6^i-MKO type 2a fiber P = 0.0374 by unpaired, two-tailed t-test with Welch correction. (b) qPCR DNA ratios of mitochondrial genes Cox1 and Cytb over nuclear genes Gcg and Hbb in quadriceps from Bcl6^fl/fl and Bcl6^MKO males (n = 7/group). (c) Western blot and densitometry of VDAC and total (Licor) protein in quadriceps from n = 6 Bcl6^fl/fl and n = 7 Bcl6^MKO males. (d) Protein levels of mitochondrial complex subunits and total (Memcode) protein in quadriceps from Bcl6^fl/fl and Bcl6^MKO mice. Western blot (left) and protein densitometry (right) are shown. (n = 4/group) Bcl6^fl/fl vs Bcl6^MKO complex II P = 0.0139 by unpaired, two-tailed t-test with Welch correction. (e, f) Cross-sectional areas of specific fiber types determined from fluorescent antibody-stained sections in quadriceps from (e) Bcl6^fl/fl and Bcl6^MKO mice (n = 3/group) and (f) Cre⁻ tmx and Bcl6^i-MKO mice (n = 3/group). Type 2b fiber in Bcl6^fl/fl vs Bcl6^MKO P = 0.003 and in Cre⁻ tmx vs Bcl6^i-MKO P = 0.0385 by unpaired, two-tailed t-test with Welch correction. All data are represented as mean ± SEM. * p < 0.05, ** p < 0.01.

Extended Data Fig. 4 |. Muscle loss due to Bcl6 deletion is associated with reduced muscle function.

(a, b) Absolute (left) and lean body mass normalized (right) forelimb grip strength measured in (a) 13-week old n = 7 Bcl6^fl/fl and n = 10 Bcl6^MKO female mice and (b) n = 6 Cre⁺ oil, n = 8 Cre⁻ tmx and n = 8 Bcl6^i-MKO female mice twelve weeks after tamoxifen or oil treatment. Whiskers in the boxplot show min to max values with line at the median and box representing the first to third quartile values. (c) Maximum tetanic force for the tibialis anterior (TA) muscle from 13-week old n = 3 Bcl6^fl/fl and n = 2 Bcl6^MKO female mice (P-value for maximum force = 0.0311 by unpaired, two-tailed t-test with Welch correction). (d, e) Distance and time run on a treadmill to exhaustion in (d) 13-week old n = 15 Bcl6^fl/fl and n = 14 Bcl6^MKO male mice and (e) Cre⁻ tmx and Bcl6^i-MKO female mice twelve weeks after tamoxifen treatment (n = 12/group). (f, g) Light phase, dark phase and total activity in (f) 13-week old Bcl6^fl/fl and Bcl6^MKO male mice (n = 5/group) and (g) Cre⁻ tmx and Bcl6^i-MKO male mice 1.5 weeks after tamoxifen treatment (n = 5/group). All data in (c-g) are represented as mean ± SEM. * p< 0.05.

Extended Data Fig. 5 |. BCL6 directly controls the expression of anabolic regulators.

(a) Principal component analysis of RNA-seq from n = 5 Bcl6^i-MKO and n = 4 tamoxifen-treated controls (Cre⁻ tmx) or n = 4 corn oil-treated controls (Cre⁺ oil). (b) Venn diagram depicting overlap between differentially expressed (DE) genes in muscle Stat5 knockout (Stat5^MKO) mice (GEO series GSE14710) compared to muscle Bcl6^i-MKO mice. (c) Quadrant plot of log₂ fold change in gene expression in Stat5^MKO/controls versus Bcl6^i-MKO/Cre⁻ tmx controls. Spearman correlation coefficient r_s = 0.03. (d) Scatterplot (left) of normalized H3K27ac tag counts and boxplot (right) showing quantification of H3K27ac tag densities within 1kb of BCL6 peaks in Cre⁻ tmx and Bcl6^i-MKO quadriceps (n = 3/group; P = 3.5e⁻¹⁰ by unpaired, two-tailed Mann Whitney test). Whiskers in the boxplot show min to max values with line at the median and box representing the first to third quartile values. (e) Receiver operator curve (ROC) of the association between differential expression in Bcl6-ablated muscle and BCL6 binding peaks. (f) UCSC browser tracks of RNAseq in quadriceps from Cre⁺ oil, Cre⁻ tmx and Bcl6^i-MKO mice, H3K27ac ChIP-seq in quadriceps from Cre⁻ tmx and Bcl6^i-MKO mice 1 week after oil or tamoxifen treatment, nascent RNA-seq in quadriceps from mice treated with DMSO or FX1 for 3 hours, and BCL6 ChIP-seq track in quadriceps from C57BL/6 animals along the Mstn, Ar, and Smox genes. (g) qPCR expression of Bcl6, Akt1, Eif4ebp1, Igf1 and Ar in quadriceps from males (n = 5/group) and females (n = 4 Bcl6^fl/fl and n = 5 Bcl6^MKO mice). Two-way ANOVA followed by Tukey’s multiple comparisons test was performed and all individual p-values can be found in the source data file. (h, i) Western blot and densitometry of (h) BCL6 and actin (i) 4EBP1 and total protein (licor) in quadriceps from male and female n = 3 Bcl6^fl/fl and n = 4 Bcl6^MKO mice. For BCL6, P(male Bcl6^fl/fl vs Bcl6^MKO) = 0.005, P(male Bcl6^fl/fl vs female Bcl6^MKO) = 0.005, P(male Bcl6^MKO vs female Bcl6^fl/fl) = 0.0299, P(female Bcl6^fl/fl vs Bcl6^MKO) = 0.0305 and for 4EBP1, P(male Bcl6^fl/fl vs Bcl6^MKO) = 0.0034, P(male Bcl6^fl/fl vs female Bcl6^MKO) = 0.0199 by two-way ANOVA followed by Tukey’s multiple comparisons. All data in (g-i) are represented as mean ± SEM. * p < 0.05, ** p < 0.01 and *** p < 0.001.

Extended Data Fig. 6 |. Loss of Bcl6 reduces the rate of muscle protein synthesis.

(a) Heatmap of ribosome protected fragment (RPF) sizes in sequencing libraries of tamoxifen-treated Bcl6^i-MKOmRiboTag and mRiboTag mice. (b) Pie chart distribution of mapped RPF reads. (c) Empirical cumulative distribution frequency (Ecdf) plots of the translation efficiencies (natural logarithm of TE) in quadriceps for all genes. Bcl6^i-MKO mRiboTag vs mRiboTag P = 2.2e-16 by two-sided Kolmogorov-Smirnov testing (d) Metagene distribution plot showing normalized ribo-seq RPF reads from the TSS to TTS for all genes (left) and genes with increased TE (right). (e) Western blot and densitometry of ubiquitin protein and actin in quadriceps from Cre⁻ tmx and Bcl6^i-MKO males 1.5 weeks after tamoxifen treatment (n = 4/group). (f) EIF4E co-immunoprecipitation and western blots for EIF4G (top) and 4EBP1 (bottom) in gastrocnemius from Cre⁻ tmx and Bcl6^i-MKO males 1 week after tamoxifen treatment (n = 2/group). (g) Western blot and densitometry of p4EBP1, Non-phospho 4EBP1, Total 4EBP1, and total (Memcode) protein in quadriceps from 10 week old Bcl6^fl/fl and Bcl6^MKO males (n = 4/group). P(Bcl6^fl/fl vs Bcl6^MKO) for p4EBP1 = 0.0009, total 4EBP1 = 0.0004, and nonP 4EBP1 = 0.0002 by unpaired, two-tailed t-test with Welch correction. (h) qPCR of protein synthesis and degradation regulators in n = 4 Cre⁺ oil, n = 4 Cre⁻ tmx and n = 5 Bcl6^i-MKO male mice 7 days after treatment. One-way ANOVA with Dunnett’s multiple comparisons test showed that P(Cre⁺ oil vs Bcl6^i-MKO) for Eif4ebp1 = 0.0086, Mstn = 0.0018, Igf1 = 0.0332 and Ar = 0.0420; P(Cre⁻ tmx vs Bcl6^i-MKO) for Eif4ebp1 = 0.0163, Mstn = 0.0011, Igf1 = 0.0140 and Ar = 0.0330. (i, j) Analysis of Eif4ebp1 knockdown efficiency in muscles (n = 3 mice/group for scramble and Eif4ebp1 shRNA1 and n = 2 mice/group for Eif4ebp1 shRNA2). (i) qPCR of Eif4ebp1 and (j) western blot and densitometry of 4EBP1 in Cre⁻ tmx and Bcl6^i-MKO male mice transduced with scramble (black border), Eif4ebp1 shRNA1 (green border), or Eif4ebp1 shRNA2 (yellow border) and treated with tamoxifen one week after viral infection. Tissues were analyzed eight weeks later. All bar graph data in (e, g-j) are represented as mean ± SEM. # p < 0.05, ## p < 0.01 for Cre⁺ oil vs Bcl6^i-MKO and * p < 0.05, ** p < 0.01, *** p < 0.001 for Cre⁻tmx vs Bcl6^i-MKO and Bcl6^fl/fl vs Bcl6^MKO.

Extended Data Fig. 7 |. Chronic Bcl6 loss evokes compensatory signaling to maintain muscle mass.

(a) SUnSET analysis of puromycin incorporation into quadriceps (top) in Cre⁻ tmx and Bcl6^i-MKO mice 12 weeks after tamoxifen treatment. Puromycin labeling relative to total protein (Licor) by densitometry (bottom) (n = 3/group; P = 0.0017 by unpaired, two-tailed t-test with Welch correction). (b) Western blot and densitometry in quadriceps from Cre⁻ tmx and Bcl6^i-MKO males 15 weeks after tamoxifen treatment for LC3, 3 days after colchicine or vehicle treatment. (n = 3/group; Two-way ANOVA followed by Tukey’s multiple comparisons test was performed. P (Veh-Cre⁻ tmx vs Col-Cre⁻ tmx) = 0.0174; P (Veh-Cre⁻ tmx vs Col-Bcl6^i-MKO) = 0.0005; P (Veh-Bcl6^i-MKO vs Col-Bcl6^i-MKO) = 0.0083.) Flux is the difference in average LC3-II levels in colchicine minus vehicle (bottom right). (c) Western blot and densitometry of ubiquitin and total (Licor) protein in quadriceps from Cre⁻ tmx and Bcl6^i-MKO males 12 weeks after tamoxifen treatment (n = 4/group). (d) Quadrant plot of log2 fold change in gene expression in Bcl6^i-MKO/Cre⁻ tmx 12 weeks after tamoxifen treatment (chronic) versus Bcl6^iMKO/Cre⁻ tmx 1 week after tamoxifen treatment (acute). (e) Ontology analysis of differential genes unique in chronic Bcl6 deletion. Cumulative hypergeometric statistical testing was used. (f) Log p-value versus Pearson correlation plotted for causal transcription factors identified by IMAGE that were unique to either acute or chronic Bcl6 deletion. Pink = high, green = medium, and gray = low confidence factors. Factors in acute dataset are shown as circles and factors in chronic are squares. (g) Causal transcription factors with opposing Pearson correlations to gene expression in acute versus chronic Bcl6 deletion. Acute = green circle, chronic = orange square. (h-l) Western blot and densitometry of (h) pSMAD1/5/8, Total SMAD1, and total (Licor) protein, (i) pFOXO1, Total FOXO1, and Actin, (j) p4EBP1, Non-phospho 4EBP1, Total 4EBP1, and Actin, (k) pAKT, Total AKT, and Actin and (l) pS6, Total S6, and Actin in quadriceps from Cre⁻ tmx and Bcl6^i-MKO males 12 week after tamoxifen treatment (n = 4/group). Unpaired, two-tailed t-test with Welch correction and all individual p-values can be found in the source data file. All bar graph data in (a-c, h-l) are represented as mean ± SEM. * p < 0.05 and ** p < 0.01.

Data availability

All sequencing data generated in this study are publicly available in the GEO repository SuperSeries GSE190956. We also utilized publicly accessible RNA-seq and microarray data from GEO repository entries GSE65936, GSE112204, GSE107470 and GSE14710, as well as the mm10 reference assembly for the mouse genome. Source data are provided with this paper.