Muscle-specific RNAi for components of diverse multiprotein complexes triggers a largely overlapping transcriptional response in skeletal muscle which is centered on protease induction
To determine the local stress responses induced by perturbation of multiprotein complexes, we have used the UAS/Gal4 system and transgenic RNAi to target core components of key protein complexes in Drosophila skeletal muscle (Fig. 1a). Specifically, the following complex components were knocked down via RNAi: Prosβ5 (homologous to human PSMB5, proteasome 20S subunit beta 5), which is part of the catalytic core of the proteasome; VCP (also known as Ter94 in Drosophila), which is mutated in inclusion body myopathy and is a core constituent of the VCP complex necessary for ERAD (endoplasmic reticulum-associated degradation) and for the degradation of subunits of multiprotein complexes; ND-75 (homologous to human NDUFS1), which is a component of the mitochondrial respiratory complex 1, one of the largest multiprotein complexes; and tropomyosin 1, which is a contractile protein of the sarcomere.
For these studies, a Mhc-Gal4 line that drives transgenic UAS-RNAi expression specifically in skeletal muscles was used, and knockdown of the levels of the target mRNAs was confirmed by qRT-PCR (Fig. 1a-c). Subsequently, RNA-seq of Drosophila thoraces (which consist primarily of skeletal muscle) determined the transcriptional changes associated with RNAi-mediated perturbation of the components of multiprotein complexes listed above. We also analyzed the transcriptional changes induced by a control GFP, and the RNA-seq resulting from each RNAi intervention was normalized to a control white (Fig. 1b, c).
Proteolysis was the major GO term that was enriched among upregulated genes: this category consisted primarily of proteases and peptidases. On this basis, we next examined whether protease modulation is part of the transcriptional response to RNAi for components of multiprotein protein complexes and found it to be the case. Specifically, several proteases/peptidases were significantly upregulated by RNAi for Prosβ5, VCP, ND-75, and Tm1 (normalized by control white) but not by GFP (Fig. 1c). These results therefore indicate that protease induction is a common response induced by perturbation of different multiprotein complexes. We next examined whether specific clans of proteases (as classified by the MEROPS database) are differentially modulated by stress. Analysis of the average protease gene regulation (log fold-changes for all members of each protease clan) indicates that proteases from diverse families are transcriptionally upregulated by perturbation of multiprotein complexes, with higher average expression detected for metalloproteases and mixed-type proteases, whereas aspartic, cysteine, and serine proteases are less modulated (Fig. 1d).
Altogether, these analyses indicate a convergence in the transcriptional stress responses induced in muscle by RNAi for components of different multiprotein complexes located in distinct cell compartments. Specifically, similar gene expression changes (including protease/peptidase upregulation) were caused by targeting complexes located in the nucleus and cytoplasm (proteasome), endoplasmic reticulum and mitochondria (VCP complex), mitochondria (respiratory complex 1), and the cytoplasm (sarcomere). In summary, these analyses indicate a remarkable similarity in the gene expression changes induced by perturbation of different molecular complexes (Fig.1, Supplementary Dataset 1, and Supplementary Figs. 1, 2), including the induction of proteases that belong to several classes.
Several studies in model organisms indicate that local perturbation of homeostasis in one tissue leads to the induction of a stress response not only locally but also systemically in distant tissues. To this purpose, we have examined the transcriptional changes in the heads (which are enriched for CNS tissues, i.e. retinas and brains) of flies with muscle-specific perturbation (obtained via RNAi) of multiprotein complexes (Fig. 1e, f and Supplementary Dataset 1).
GO terms that were enriched among the upregulated genes included proteolysis and associated categories (peptidase S1 and carboxypeptidase) and also chitin metabolism and metal-binding (Fig. 1e). Collectively, this transcriptional profiling indicates that the induction of proteases and peptidases is a core response that is commonly induced by the muscle-specific perturbation of multiprotein complexes not only locally in muscle but also distantly in the CNS (Fig. 1f). Specifically, further analyses indicate that several proteases and peptidases are transcriptionally induced in the CNS by muscle-specific Prosβ5, Ter94/VCP, ND-75, and Tm1 but not by control GFP. Analysis of the protease clans that are most highly modulated in the CNS indicated a partially divergent response compared to skeletal muscle (Fig. 1g). In particular, while mixed-type proteases were the clan most highly induced on average in both skeletal muscle and the CNS, metalloproteases were highly induced in the muscle but not in the CNS and, conversely, aspartic proteases were induced in the CNS but not in muscle (Fig. 1d, g). Collectively, these findings indicate that both local and systemic transcriptional responses induced by the perturbation of multiprotein complexes in muscle are centered on the induction of proteases.
Proteases and peptidases constitute a superfamily of proteins that localize to different organelles and compartments. Mitochondrial proteases have important roles in mitochondrial proteostasis whereas cytosolic proteases maintain protein quality control during aging and can partially compensate for proteasome dysfunction. Proteases can also degrade pathogenic proteins, as found for the puromycin-sensitive aminopeptidase (Psa), which degrades mutant tau and pathogenic huntingtin with polyglutamine tract expansion which causes Huntington's disease. However, other proteases can promote neurodegeneration by generating huntingtin and tau fragments with increased pathogenicity.
On this basis, we have used a Drosophila model of Huntington's disease to determine whether the proteases that are induced in the CNS upon muscle-specific stress constitute a pathogenic or, conversely, a protective response that degrades a model aggregation-prone protein, i.e. huntingtin with poly-glutamine expansion. Specifically, 235 RNAi lines targeting 77 stress-induced proteases and controls were driven with the UAS/Gal4 system and the GMR-Gal4 driver in Drosophila retinas that express Htt-Q72-GFP (Fig. 2a), a GFP-tagged exon 1 of huntingtin with poly-glutamine expansion, which is a known target of proteases. Previous studies with this model reported the progressive, age-dependent accumulation of Htt-Q72-GFP aggregates that can be optimally examined after aging for 30 days at 29 °C. On this basis, the average area of Htt-Q72-GFP aggregates was quantified for each RNAi at this timepoint by using CellProfiler, starting from fluorescence images obtained from 5 retinas of 5 distinct flies. Subsequently, we examined the z-score (which indicates the deviation of each RNAi from the cumulative mean of all values) to identify RNAi interventions that substantially differ (either by increasing or decreasing) the area of Htt-Q72-GFP aggregates compared to the other RNAi interventions. Because we utilized RNAi lines from different collections that may differ in their genetic background, we examined separately the results obtained from the screen of RNAi lines from the BDSC (Bloomington Drosophila Stock Center), NIG (Japanese National Institute of Genetics), and the GD and KK collections from the VDRC (Vienna Drosophila Resource Center). Global representation of the RNAi screen data indicated that several proteases greatly affected the Htt-Q72-GFP aggregate area, with some RNAi lines increasing and others decreasing Htt-polyQ aggregates (Fig. 2b and Supplementary Dataset 2). For example, compared to control mcherry, knockdown of εTry (homologous to the human serine proteases PRSS1/2/3) increased the area of Htt-Q72-GFP aggregates, indicating that this protease normally contributes to the degradation of pathogenic Htt-polyQ. Conversely, RNAi for yip7 (homologous to the human chymotrypsin proteases CTRB2 and CTRL) reduced the area of Htt-Q72-GFP aggregates, indicating that this protease has an opposite role, i.e. normally promotes the accumulation of huntingtin-polyQ aggregates (Fig. 2b). While this RNAi screen uncovered proteases that increase and others that decrease the area of Htt-Q72-GFP aggregates, the phenotypes observed following protease knockdown were more striking for interventions that increased huntingtin-polyQ aggregates. Specifically, by applying a z-score cut-off of -3.5 and +3.5, there was only 1 protease knockdown (yip7) that reduced the area of protein aggregates with z-score < -3.5, whereas there were 27 protease that increased the area of protein aggregates with z-score > +3.5. Comparison to negative controls for each RNAi collection (i.e., mcherry, v, and y) further confirmed that RNAi for these proteases increases the area of Htt-Q72-GFP aggregates (apart from yip7, which decreases it), as also observed with Psa (Fig. 2c, d), which is used as positive control because Psa knockdown was previously found to impede the degradation of pathogenic huntingtin in mice and Drosophila. Although it is well known that not all RNAi lines work efficiently in Drosophila, in some cases there were multiple RNAi lines from different collections targeting the same gene that acted consistently, i.e. increased the area of Htt-Q72-GFP aggregates: this was the case for the knockdown of CG8560 (3 lines), Jon25Bi (3 lines), Jon25Biii (3 lines), and CG17109 (2 lines). Importantly, these and the other proteases that strikingly affected the area of huntingtin-polyQ (Fig. 2c, d) are evolutionarily conserved (Supplementary Dataset 3), suggesting that they may also play a role in the degradation of pathogenic huntingtin in humans. In summary, RNAi screening of stress-induced proteases identifies a predominant role of these proteases in reducing the accumulation of Htt-Q72-GFP aggregates, as inferred from the observation that their knockdown increases the area of Htt-Q72-GFP aggregates.
Western blot is a common method to examine huntingtin-polyQ aggregation. In particular, previous studies reported that Htt-Q72-GFP is detected as a monomer in both detergent-soluble and insoluble fractions but also as insoluble oligomers/aggregates that accumulate in the stacking gel. On this basis, we next utilized western blots to analyze the detergent-soluble and detergent-insoluble fractions obtained from 30 heads/sample of flies that express Htt-Q72-GFP in the retina concomitantly with RNAi for a protective stress-induced protease or control. Probing with anti-GFP antibodies identified Htt-Q72-GFP monomers (~55 kDa, that were detected in both the soluble and insoluble fractions) and high-molecular-weight oligomers/aggregates (>250 kDa), which were detected in the stacking part of the gel utilized for the analysis of insoluble fractions. In addition, myosin was utilized as a loading control.
Consistent with findings from previous studies, Psa driven concomitantly with Htt-Q72-GFP in retinas increased the amount of Htt-Q72-GFP aggregates detected in the stacking gel relative to the control mcherry (Fig. 3a, b). Comparable outcomes were observed via the western blot analysis of detergent-insoluble fractions derived from fly heads with RNAi targeting stress-induced proteases (Fig. 3a, b), which also displayed an increase in huntingtin-polyQ aggregates as evidenced by microscopy (Fig. 2). Specifically, most RNAi lines targeting stress-induced proteases increased the insoluble levels of Htt-Q72-GFP protein oligomers/aggregates that are not resolved by SDS-PAGE (Fig. 3a, b), whereas the soluble and insoluble levels of monomeric Htt-Q72-GFP were typically not affected. However, the knockdown of yip7/CG6457 reduced the levels of insoluble Htt-Q72-GFP monomers and aggregates (Fig. 3a, b), consistent with the immunofluorescence analysis that indicates that Htt-Q72-GFP aggregates decrease in response to yip7 (Fig. 2c, d). Together, these results indicate that proteases induced in the CNS in response to muscle-specific stress largely consist of protective proteases that can degrade a model aggregation-prone protein, i.e. Htt-Q72-GFP.
We previously found that induction of proteasome stress in muscle increases the levels of several secreted factors (i.e., myokines), including the amylase Amyrel, which in turn promotes protein quality control in the brain and retina during aging via maltose/SLC45 signaling and the transcriptional upregulation of chaperones and proteases.
Considering the important role of this stress-induced muscle-secreted factor in systemic signaling and its capacity to increase protease expression in the CNS, we next examined whether Amyrel is generally induced by perturbation of additional multiprotein complexes apart from the proteasome. To this purpose, we examined the RNA-seq data (Fig. 1) and determined that Amyrel expression increases in skeletal muscle in response to knockdown of Prosβ5, VCP, ND-75, and Tm1 compared to control white and GFP and to no transgene expression (Fig. 4a). These findings indicate that Amyrel is transcriptionally induced by the RNAi-mediated perturbation of multiple multiprotein complexes.
On this basis, and because of the known role of Amyrel in upregulating the expression of protective proteases in the CNS, we next examined whether Amyrel induction in skeletal muscle can influence Htt-Q72 aggregation in the retina. Because the Htt-Q72-GFP transgene is driven in the retina via the UAS/Gal4 system, we generated tools for transgenic expression in skeletal muscle with another binary expression system, i.e. the LexA/LexAop. Specifically, we utilized a ~4.4 kb Mhc promoter sequence to generate a Mhc-LexA-GAD fly line, which expresses the transcriptional activator LexA-GAD specifically in the thoracic flight muscle but not in the brain or retina, as evidenced by monitoring the fluorescence of a LexA-responsive transgene that expresses CD2-GFP (Fig. 4b). On this basis, we next utilized Mhc-LexA-GAD to drive Amyrel expression in skeletal muscle (with the LexA-responsive LOT-Amyrel) while concomitantly driving Htt-Q72-GFP expression in the retina with the UAS/Gal4 system. In response to Amyrel upregulation in skeletal muscle (as assessed by qRT-PCR, Fig. 4c), there was a significant decline in the area of Htt-Q72-GFP aggregates in the retina (Fig. 4d). Taken together, these results indicate that Amyrel is a muscle-derived signaling factor that contributes to the reduction of huntingtin-polyQ aggregates in the retina in response to the perturbation of multiple multiprotein complexes in skeletal muscle (Fig. 5).