AMPK-dependent and -independent coordination of mitochondrial function and muscle fiber type by FNIP1

Mitochondria are essential for maintaining skeletal muscle metabolic homeostasis during adaptive response to a myriad of physiologic or pathophysiological stresses. The mechanisms by which mitochondrial function and contractile fiber type are concordantly regulated to ensure muscle function remain poorly understood. Evidence is emerging that the Folliculin interacting protein 1 (Fnip1) is involved in skeletal muscle fiber type specification, function, and disease. In this study, Fnip1 was specifically expressed in skeletal muscle in Fnip1-transgenic (Fnip1Tg) mice. Fnip1Tg mice were crossed with Fnip1-knockout (Fnip1KO) mice to generate Fnip1TgKO mice expressing Fnip1 only in skeletal muscle but not in other tissues. Our results indicate that, in addition to the known role in type I fiber program, FNIP1 exerts control upon muscle mitochondrial oxidative program through AMPK signaling. Indeed, basal levels of FNIP1 are sufficient to inhibit AMPK but not mTORC1 activity in skeletal muscle cells. Gain-of-function and loss-of-function strategies in mice, together with assessment of primary muscle cells, demonstrated that skeletal muscle mitochondrial program is suppressed via the inhibitory actions of FNIP1 on AMPK. Surprisingly, the FNIP1 actions on type I fiber program is independent of AMPK and its downstream PGC-1α. These studies provide a vital framework for understanding the intrinsic role of FNIP1 as a crucial factor in the concerted regulation of mitochondrial function and muscle fiber type that determine muscle fitness.

1F and 1G), which was consistent with mitochondrial DNA measurements (revised Fig 1H). Notably, the WT and Fnip1 TgKO mice have an ordered array of small intermyofibrillar mitochondria with dense mitochondrial matrix along the Z-lines of the sarcomeres, whereas the Fnip1 KO mice have a larger shape and less dense mitochondria, which displaced the myofibrils (revised Fig 1F). These results could suggest that FNIP1 not only regulates mitochondria quantity, but also regulate mitochondrial quality. Indeed, during the review of this manuscript, a study was published reporting that FNIP1 regulates mitochondrial redox homeostasis (Manford et al., Cell, 2020, 183, 46-61.e21). It is also possible that the altered mitochondrial morphology is due to the changes in mitochondrial location (e.g. the proximity to the myofibrils triads where calcium is released from SR). This information has been added as revised Fig 1F and 1G (page 9, and addressed in the revised Discussion (page 20, line 434-439). Figure 2A all the complexes are increased. Are the corresponding transcripts also increased?"

"In the expression profile 27/34 genes upregulated in the FNIP1 KO muscles are from respiratory complex I, but in the western blot in
We have found that a broad array of mitochondrial genes was up-regulated in Fnip1 KO muscle (Fig 1D and original S1 Table). The regulated genes (a total of 455 genes) were displayed in the original S1 Table  (now revised S2 Table), and the mRNA levels of Ndufb8, Uqcrc2, Cox4i1 and Atp5a1 (Western blot in Fig 2A) were all induced in Fnip1 KO muscle. We have now also added a heat map showing the up-regulated subunits from the mitochondrial respiratory complexes in Fnip1 KO muscle (revised S3C Fig). Muscle oxidative metabolism biomarkers such as myoglobin and cytochrome c are known PGC-1 targets (Lin et al., Nature, 2002, 418, 797-801;Schreiber et al., Proc Natl Acad Sci U S A., 2004, 101, 6472-6477). We found that the induced expression of myoglobin and cytochrome c protein paralleled the activation of AMPK/PGC-1 signaling in Fnip1 KO muscle, and knocking out Ampk1/2 or PGC-1 blunted the effect of Fnip1 KO on the expression of myoglobin and cytochrome c protein (original Fig 4E,5B,and 7D). We have added a sentence to the text to make this point clearer (page 10, line 204).

"in Figure 2C and 2D the authors show a switch in LDH isoforms, however it is unclear if this impacts on the blood lactate levels."
The LDHB/LDHA isoenzyme ratio is increased in Fnip1 KO muscle. These results were of interest. We have conducted additional blood lactate measurement. As predicted, we have found that blood lactate levels were reduced in Fnip1 KO mice compared to the WT controls, and blood lactate levels from Fnip1 TgKO mice were increased to a level similar to WT controls. These new results were added in the revised Fig 2E ( We appreciate this important point and have discussed the significance of our study relative to the documented studies the Reviewer mentioned in the revised Discussion (page16, bottom paragraph, page 17, top paragraph). In fact, the mechanisms involved in the coordinate control of the two key determinants of skeletal muscle fitness: mitochondrial function and contractile fiber type remain largely unclear in the whole field. Our findings that muscle FNIP1 orchestrates mitochondrial function and type I muscle fiber program through both AMPK-dependent and independent signaling is an important novel aspect in this regard. The study by Reyes et al. (Proc Natl Acad Sci U S A., 2015, 112, 424-429) reported that whole body Fnip1 knockout mice exhibit increased proportion of type I fibers and resistant to muscle fatigue. While this suggests a role for FNIP1 in type I muscle fiber specification, the intrinsic roles of muscle FNIP1 in coordinating mitochondrial program and muscle fiber type, the two interrelated but distinct processes in skeletal muscles, is not understood. Moreover, the Reyes et al study was performed by whole body deletion of Fnip1, the global Fnip1 KO have very complicated phenotype. For instance, in addition to muscle changes, whole-body FNIP1 deficiency has profound effects on immune cell development, cardiomyopathy, and kidney development and function (Park et al., Immunity, 2012, 36, 769-781;Park et al., Proc Natl Acad Sci U S A., 2014, 111, 7066-7071;Reyes et al., Proc Natl Acad Sci U S A., 2015, 112, 424-429;Siggs et al., Proc Natl Acad Sci U S A., 2016, 113, E3706-3715;Centini et al., PLoS One, 2018, 13, e0197973). It is worth noting that according to published debate (Handschin and Spiegelman, Cell Metab. 2011, 13, 351;Zechner et al., Cell Metab., 2011, 13, 352), the solely use of a global Fnip1 KO model could result in confounding secondary effects driven by FNIP1 deficiency in nonmuscle tissues. Indeed, reports on the effects of FNIP1 on both AMPK and mTOR have been inconsistent and, thus, are less conclusive (Reyes et al., Proc Natl Acad Sci U S A., 2015, 112, 424-429;Hasumi et al., Proc Natl Acad Sci U S A., 2015, 112, E1624-1631Petit et al., J Cell Biol., 2013, 202, 1107-1122Ramirez et al., J Immunol., 2019, 203, 2899-2908. Therefore, the intrinsic physiological role of FNIP1 in skeletal muscle remains unclear. In this study, we used both gain-of-function and loss-of-function strategies in mice to generate a series of genetically modified mouse lines, including Fnip1-knockout (Fnip1 KO ) mice and Fnip1 TgKO mice expressing Fnip1 only in skeletal muscle but not in other tissues. Our mouse and muscle cell genetic systems afforded us the unique opportunity to define the intrinsic role of FNIP1 in skeletal muscle. We provided clear genetic data that led to the following surprising conclusions, which are different from the previous reports: 1) Basal levels of FNIP1 are sufficient to function as a "brake" to simultaneously inhibit mitochondrial function and type I fiber program in skeletal muscle, while muscle function is normal and fiber types are unchanged in muscle-specific Fnip1 transgenic mice; 2) FNIP1 in fact cell-autonomously regulates AMPK but not mTORC1 signaling in skeletal muscle cells, thus shed light on the inconsistencies in previous data; 3) FNIP1 regulates muscle mitochondrial oxidative program through AMPK, these findings are recapitulated in skeletal muscle cell culture and, thus, are cell autonomous; 4) FNIP1 regulates type I fiber program independent of AMPK/PGC-1, mitochondrial function and muscle fiber type are indeed controlled by distinct signaling pathways. Together, we have genetically established a pivotal intrinsic role for FNIP1 in orchestrating mitochondrial function and type I muscle fiber, as well as its epistatic relationship with AMPK and PGC-1 in driving muscle fiber type determination in skeletal muscle.
Regarding the role of AMPK/PGC-1 signaling in the FNIP1-dependent fiber type determination described in this manuscript: The Hasumi et al. paper quoted in this query reported mainly on effects of FLCN on muscle color and mitochondrial biogenesis rather than muscle fiber type. In fact, a surprising finding of our study was dissociation between mitochondrial function and type I fiber program in Fnip1 KO , Ampkɑ1/ɑ2 f/f/Myf5-Cre (TKO) and Fnip1 KO , PGC-1 f/f/MCK-Cre (DKO) mice muscles, and we concluded that FNIP1-dependent type I muscle fiber program does not require the AMPK/PGC-1 signaling based on the formal muscle fiber typing phenotype of our multiple genetic mouse lines: First, MHC fiber typing data from two independent Ampkɑ1/ɑ2 KO lines revealed that FNIP1 deficiency induces increase proportion of type I fibers in all soleus, GC and EDL muscle in the absence of Ampkɑ1/ɑ2 (original  6 D) demonstrated that the expression of the major type I myosin isoform MHC1 (Myh7 gene) and slow-twitch troponin genes was induced by FNIP1 deficiency but not affected by the complete loss of muscle Ampkɑ1/ɑ2in GC muscle. To further corroborate this point, data from additional experiments showed that the expression of slow-twitch type I muscle fiber genes was activated by FNIP1 deficiency but not affected by disruption of Ampkɑ1/ɑ2insoleus muscle (revised S8 Fig), indicating that AMPK-independent regulation of type I fiber program by FNIP1 is relevant to multiple muscle types across a range of fiber type proportions and oxidative capacity. Third, despite the complete loss of muscle PGC-1, FNIP1 ablation can achieve type I fiber-type transformation in both soleus and GC muscle of DKO mice (original Fig 7H). It is worth noting that this is quite clear in soleus, a muscle which is enriched in PGC-1 and contains a high proportion of type I fibers. Lastly, we and others have identified independent factors, such as MEF2s and ERRs/miRNA axis, that are required for fundamental muscle fiber type determination (Potthoff et al., J Clin Invest., 2007, 117, 2459-2467van Rooij et al., Dev Cell, 2009, 17, 662-673;Gan et al., J Clin Invest., 2013, 123, 2564-2575Liu et al., EMBO Mol Med., 2016, 8, 1212-1228, consistent with our conclusions. This information has been added as new S8 Fig and addressed in the revised Discussion (page 18, line 389-408).
Thus, we think that our results have uncovered a previously unrecognized intrinsic role of FNIP1 that coordinately regulates mitochondrial function and muscle fiber type that govern muscle fitness. Specifically, FNIP1 regulates muscle mitochondrial program in an AMPK-dependent fashion, while it acts independently of AMPK/PGC-1α to induce type I myofiber formation. Indeed, this coordinate response is an important aspect of the message in this work, given that it could be an appealing target that coordinates multiple pathways towards reversing muscle diseases. We have clarified and emphasized these important points in the revised Discussion (page16, page 17, top paragraph, page 20, paragraph 2).

Itemized Responses to Reviewer #3
"Reviewer #3: The authors investigated the role of folliculin interacting protein 1 (Fnip1) in skeletal muscle phenotypes using transgenic and muscle-specific knockout mice. They report that Fnip1-KO induced a glycolytic-to-oxidative shift of the muscular metabolic profile via AMPK activation. Various experimental approaches were used, and the results are clear at a glance. The design of the experiments, however, has serious concerns." We wish to thank the Reviewer for the critical and constructive review. Our itemized responses are as follows: 1. "a. The authors analyzed various indexes using several kinds of hindlimb skeletal muscles. We think the Reviewer has raised important points. We agree with the Reviewer and have addressed these points experimentally and revised the manuscript to make those points clearer: a) The Reviewer is correct. The contractile and energy metabolic properties of the muscle differ between the different muscle types. We also agree that effect of Fnip1 KO may not the same between the different muscle types. In this study, we wanted to determine whether the concerted regulation of mitochondrial and type I muscle fiber programs by FNIP1 is a broader effect and not just restricted to specific muscle type, we thus analyzed type I muscle fibers and mitochondrial oxidative metabolism readouts in multiple muscle types of Fnip1 KO mice. Interestingly, whereas we found that FNIP1 simultaneously regulates mitochondrial oxidative program and type I muscle fibers in multiple muscle types, we did observed difference in the effect of Fnip1 KO on MHC2a and MHC2x gene expression in soleus and gastrocnemius (GC) muscle (original Fig 6D and  b) The majority of the Fnip1 KO muscle phenotypic and molecular endpoints, including RNA-seq profiling and RNA analyses, Immunoblotting studies, MHC immunostaining, SDH staining, and primary muscle cells isolation were all conducted in GC muscle as it contains both the fast-glycolytic and slow-oxidative muscle fibers, thus the GC muscle has been a little bit limited because we used the entire GC muscle for each kind of studies to keep consistent between different mice. We have conducted more experiments using Fnip1 KO GC muscles and demonstrated that the protein levels of OPA1, MFN1, Myoglobin and Cyt C were all increased in GC muscle (similar to other muscle type) of Fnip1 KO mice (new S5A and S5C Fig). In addition, given that the whole GC muscle is big and not suitable for the O2K-based permeabilized muscle OCR measurements, we have provided further experiments demonstrating that both pyruvate-and succinate-driven mitochondrial respiration rates were also induced in the plantaris part of the gastrocnemius/plantaris complex of Fnip1 KO mice (new  569,917,942,950,957,1035,1041,1048,1054,1063,1066,1074,1082,1090,1098,1113,1121,[1133][1134][1135]. Specifically, we did not dissect the red part from the white part of the GC muscle, and we used the entire GC muscle for various analyses, including RNA-seq profiling, RT-PCR analyses and immunoblotting studies, to make sure we are consistent between different mice. For fiber type determination and quantification, we focused on the red zone area of the medial head in the widest part of the GC muscle. We have provided more details with regard to histologic analyses in the Methods (page 23 and page 24, line 520-543). Briefly, the gastrocnemius and plantaris of each mouse were dissected as a whole and frozen in isopentane that had been cooled in liquid nitrogen. We cut 10 μm-thick serial GC muscles cross-sections from the knee cut side in a Leica CM1850 cryostat at -20 °C and mounted them on positively charged glass slides. Transverse sections were collected from the widest part (mid-belly) of the GC muscle, and we examined slides under the microscope during the cutting process to make sure that we are consistent in comparing the widest part of the GC muscle from mouse to mouse. We have also added a new Figure (new S6A Fig)

"Figure 1F: The images of Fnip1-KO are strange. Very few myofibrils appear in the upper image. This image may be derived from around the subsarcolemmal region, not from the intermyofibrils. And the mitochondria in both images appear to be swelling, with wide interstitial spaces. These are abnormal ultrastructures that may be attributed to technical errors during the fixation steps. The authors should replace the images with better-quality alternatives."
We thank the reviewer. We confirmed the TEM images in original Fig 1F shows the intermyofibrillar mitochondria in Fnip1 KO muscle. We have improved our TEM studies and replaced the original Fig 1F with low-power and high-power electron micrographs to clarify this point (revised Fig 1F). Notably, the WT and Fnip1 TgKO mice have an ordered array of small intermyofibrillar mitochondria with dense mitochondrial matrix along the Z-lines of the sarcomeres, whereas the Fnip1 KO mice have a larger shape and less dense mitochondria, which displaced the myofibrils (revised Fig 1F). These results could suggest that FNIP1 not only regulates mitochondria quantity, but also regulate mitochondrial quality. Indeed, during the review of this manuscript, a study was published reporting that FNIP1 regulates mitochondrial redox homeostasis (Manford et al., Cell, 2020, 183, 46-61.e21). It is also possible that the altered mitochondrial morphology is due to the changes in mitochondrial location (e.g. the proximity to the myofibrils triads where calcium is released from SR). This information has been added as revised  We thank the Reviewer's insightful points, our responses to the each of the points are as follows: a) We agree and this was addressed in the point#1c above. In brief, we have stated the sizes and actual portions of the GC muscle samples in Figure  b) As for type 1 fiber quantitation in soleus muscle, we have replaced the original Fig 2E with new representative images and re-quantified the percent proportion of type I fibers from additional narrow age range mouse soleus samples (revised Fig 2F and 2G Figure 1 on the left). The quantification of type 1 fiber and SDH staining in GC muscle has been a challenge, in part, due to heterogeneity. We focused on the same red zone area of the medial head of GC muscle to keep consistent between different mice. We have quantified numerous sections and present this as number of type I fibers per medial head of GC (WT, 77 ± 24 per section vs. Fnip1 KO , 337 ± 73 per section, p < 0.001; n = 7-8 mice per group) and percentage of positive staining myofibers for SDH per medial head of GC (WT, 33.5 ± 7.8 % vs. Fnip1 KO , 76.9 ± 7.8 %, p < 0.001; n = 8 mice per group). We have provided more details with regard to histologic analyses in the Methods (page 23 and page 24, line 520-543). c) We are very interested in muscle atrophy, please also see our response in bellow #7. By gross examination, hindlimb muscle of adult Fnip1 KO mice was smaller than that of WT or Fnip1 TgKO mice. We have added new supplemental table demonstrating that the body weight, weight of GC muscles and GC muscle weight normalized to body weight were decreased in Fnip1 KO mice compared to WT or Fnip1 TgKO mice at 12 weeks of age (new S1 Table). These results suggested a potential role of muscle FNIP1 in controlling body weight and muscle mass. It is possible that the difference in muscle size could reflect the fiber type switching. It is also possible that FNIP1 directly regulates the dynamic balance between protein synthesis and degradation that determine muscle mass. We wish to establish the necessary assay systems in our independent studies to address this interesting muscle atrophy mechanistic question. This information has been added as new S1 Table ( d) The unstained fibers in the soleus images are likely type 2a given that soleus is type 1 and 2a fiber rich, whereas the unstained fibers in the medial head of GC muscle are likely type 2a and 2x. Unfortunately, we have not been very successful when trying to conduct type 2a and 2x immunostaing for technical reasons. However, as shown in the quantitative RT-PCR data in original Fig 6D, MHC2a (Myh2 gene) and MHC2x (Myh1) mRNA were significantly increased in GC muscles in the Fnip1 KO mice compared to WT controls. Interestingly, we have conducted further experiments and found that the expression of MHC2a and MHC2x were actually reduced in soleus muscles of the Fnip1 KO mice (new S8 Fig). Whereas we found that the concerted regulation of mitochondrial and type I fiber programs by FNIP1 is relevant to multiple muscle types, these new data suggest that FNIP1 exerts a different control effect on MHC2a and MHC2x expression in different muscle types. It is possible this could reflect the intrinsic differences between muscles at the baseline that affect the adaptive range of MHC transformations (Schiaffino and Reggiani, Physiol Rev., 2011, 91, 1447-1531. This information has been added as new S8 Fig and addressed in the revised Discussion (page 19, line 413-419). Figures 3F and 5D: Why did the authors evaluate the mitochondrial density using myoblasts instead of myotubes? All of the other analyses were performed using myotubes. Phase-contrast images should be shown."

"
We also found increased mitotracker immunofluorescence signal in Fnip1 KO myotubes compared to WT or TKO myotubes. These results together with phase-contrast images of myotubes were added to the new S7 Fig.   6. " Figures 3G and 2H: The origin of the primary cells should be shown." We thank the reviewer. Primary myoblasts were isolated from GC muscle of 4-week old male WT, Fnip1 KO , or TKO mice. We have provided more details with regard to primary muscle cell culture in the Methods section (page 28 and 29, line 640-652). We have also added the origin information of the primary muscle cells to the revised Figure legend ( line 968, 1012-1013, 1142). Figures 6A and 7H: a. Judging from the images, Fnip1-KO induced muscle atrophy in the soleus and EDL in Figure 6A and in the GAS in Figure 7H. We thank the Reviewer's insightful points, our responses to the each of the points are as follows:

"
a) We agree that while the image data showing a clear and significant difference in type I fibers in Fnip1 KO muscle, whether FNIP1 also regulate muscle atrophy remain unclear. We have investigated further, and we believe that the potential variation between images could reflect the age differences of mice used for the original histology study. We used a wider age range of mice for the immunostaing at the beginning of the study. We have now conducted more muscle fiber typing studies with narrow age ranges mouse muscle samples. We have replaced the original images in Fig 2E, 6A and 7H with the image results from 8-9 week old male mice to clarify this point (revised Fig 2F, 6A and 7H). We have also quantified the fiber size distribution from numerous cross-sections of the soleus, EDL and the red zone area of the medial head of GC muscle as suggested. Whereas fiber-size distribution was altered in EDL muscle from Fnip1 KO mice, with a clear shift toward a reduced cross-sectional area relative to that of WT controls, no clear changes in fiber size distribution was observed in soleus and the red zone area of the medial head of GC muscle from Fnip1 KO mice (see Figure 2 below). Given that fiber size is in general highly related to the fiber type switching, the different fiber size change results may depend on the muscle region examined. Thus, we think that the data is not conclusive, so we hesitate to include these results in the revised manuscript. Whether FNIP1 regulate muscle atrophy is surely a very important and intriguing question that warrants serious further characterizations, as we have found that the muscle weight was decreased in Fnip1 KO mice compared to WT controls (new S1 Table). Please also see our response to the point#4 above. We primarily focused on the coordination of type I muscle fiber and mitochondrial programs by FNIP1 in this study, it is possible that FNIP1 directly regulates the dynamic balance between protein synthesis and degradation and is independent of the change in fiber type. We wish to establish the necessary assay systems in our independent studies to address this interesting muscle atrophy mechanistic question. This information has been added as new S1 Table (  We appreciate the Reviewer's point. We think it is worth noting that PGC-1 protein levels are hardly detectable in WT GC muscle. n.s. in the original Fig 7B stands for "none-specific band". We have improved the PGC-1Western blotting and replaced the original Fig 7B with the new immunoblot analysis of PGC-1using whole cell extracts from entire GC muscle and heart from indicated mice (revised Fig 7B). The data now indicate a clear increase in the expression of PGC-1proteinin Fnip1 KO muscle, and the induced expression of PGC-1 protein was completely abolished in the DKO muscle (see Figure 3 below).
We have provided more details with regard to primary muscle cell culture in the Methods section (page 28 and 29, line 640-652). In briefly, primary myoblasts were isolated from GC muscle of 4-week old male mice as previously described (Rando and Blau, J Cell Biol., 1994, 125, 1275-1287Gan et al., J Clin Invest., 2013, 123, 2564-2575Liu et al., EMBO Mol Med., 2016, 8, 1212-1228Fu et al., Cell Rep., 2018, 23, 1357-1372. GC muscles from both legs were removed. Minced tissue was digested in a collagenase/dispase/CaCl 2 solution for 1.5 hours at 37°C in a shaking bath. DMEM supplemented with 10% FBS (PPM) was added and samples were triturated gently before loading onto a Netwell filter (70 μm, BD). Cell suspension was pelleted at 1000 rpm for 5 minutes. Cells were then resuspended in PPM and plated on an uncoated plate for differential plating. Cell suspension (not-adherent) was centrifuged for 5 minutes at 1000 rpm and pellet was resuspended in Growth Medium (GM) (Ham's F-10 medium supplemented with 20% FBS and 2.5 ng/ml bFGF). Cells were plated on collagen coated flasks for expansion. Cells were fed daily with GM. For differentiation, cells were washed with PBS and re-fed with 2% Horse-serum/DMEM differentiation medium and re-fed daily. Cells were induced to differentiation for 3 days prior to various experiments, and no notable difference in myotube differentiation between Fnip1 KO myocytes and WT controls were observed (new S7 Fig). We have added the differentiation days to the revised Figure legend (lines 968, 1013 and 1143) and phase-contrast images of myotubes to the new S7 Fig.  9. " Figures 3G and 5E: What kind of analysis was applied for the statistical analyses? This type of data should be analyzed using 2-way ANOVA." The original data in Fig 3G and 5E were analyzed by one-way ANOVA, and we have re-analyzed those seahorse data using two-way ANOVA and Bonferroni post-hoc test and added a sentence to the revised Figure legend (revised Fig 3G and 5E, lines 983-984, 1026-1027). Thanks.
10. " [136][137][138]: At this point, none of the experimental data support a glycolytic-to-oxidative shift of skeletal muscles in Fnip1-KO mice. The result merely shows a shift in the color of the muscle from whitish to red" Agree. We have removed those sentences, thanks.
11. "Line 133:"FLCN" should be spelled out."IF" should be spelled out." We thank the reviewer and Folliculin (FLCN) and Immunofluores (IF) have been defined in the text and Figure legend (lines 91, 956, and 1034).

"a-tubulin is unsuitable for the loading and transfer control of Western blotting."
We thank the reviewer. For the loading and transfer control of Western blotting, we have used -tubulin for quite a few years in our lab because we find it is stably expressed and least variable across many tissues and cell lines we tested. Of note, we have also conducted -actin and Ponceau red staining for the loading or transfer control in Western blotting (revised S5C Fig), and we see a similar result. Therefore, we are confident in our Western blotting conclusion.