ABSTRACT
Circadian rhythms regulate cell proliferation and differentiation but circadian control of tissue regeneration remains elusive at the molecular level. Here, we show that the circadian master regulators Per1 and Per2 are integral components defining the efficiency of myoblast differentiation and muscle regeneration. We found that the depletion of Per1 or Per2 suppressed myoblast differentiation in vitro and muscle regeneration in vivo, demonstrating their non-redundant functions. Both Per1 and Per2 directly activated Igf2, an autocrine promoter of myoblast differentiation, accompanied by Per-dependent recruitment of RNA polymerase II, dynamic histone modifications at the Igf2 promoter and enhancer, and the promoter-enhancer interaction. This circadian epigenetic oscillation created a preferred time window for initiating myoblast differentiation. Consistently, muscle regeneration was faster if initiated at night when Per1, Per2, and Igf2 were highly expressed compared with morning. This study reveals the circadian timing as a significant factor for effective muscle cell differentiation and regeneration.
Regulation of mammalian circadian rhythms is centered around the Clock/Bmal1 complex, a ubiquitously expressed basic helix-loop-helix Per-Arnt-Single-minded (PAS) transcription factor dimer1–4. The complex binds the E-box (5’-CANNTG-3’) in promoters and enhancers of thousands of genes to activates their transcription, including the Cry (Cry1 and Cry2) and Per (Per1-Per3) genes. Gradually accumulated Cry and Per in turn bind Clock/Bmal1 on DNA and repress its transcription activity, forming a negative feedback loop. Subsequent phosphorylation and ubiquitination of Cry and Per lead to their degradation, allowing Clock/Bmal1 to resume activation of the target genes. This oscillating activity of Clock/Bmal1 creates transcriptional circadian rhythms in more than 20% of the genes in the genome in at least one tissue in the body. In addition, Clock/Bmal1 activates retinoic acid receptor-related orphan receptor proteins (RORα-RORγ) and reverse orientation c-erb proteins (Rev-erbα and Rev-erbβ), which compete for the retinoic acid-related orphan receptor response element (RORE) in the Bmal1 promoter. Opposing activities of ROR as an activator and Rev-erb as a repressor of Clock/Bmal1 form the second circadian feedback loop. These feedback loops exist in every tissue examined (peripheral clocks), including skeletal muscle. In contrast, the central clock is located in the suprachiasmatic nucleus (SCN) in the hypothalamus, which is entrained by the light signal transmitted from the retina as the primary external cue (zeitgeber). The peripheral clock is entrained by various physiological factors such as body temperature, feeding time, and physical activity, in addition to the loose control by the central clock.
Circadian regulation is tightly integrated into the genetic program of muscle cell differentiation as demonstrated by several studies5–7. First, more than 2,000 genes, including the master myogenic regulators MyoD and myogenin, show circadian oscillation in abundance5, 8–10. Second, whereas Bmal1 promotes satellite cell proliferation and differentiation, and is required for muscle regeneration11, 12, Rev-erbα acts as an inhibitor of these processes13. Third, Clock/Bmal1 binds the E-box in the core enhancer of MyoD in a circadian manner; MyoD then binds the Bmal1 enhancer and increases the amplitude of Bmal1 expression, forming a feed-forward loop in myogenesis14, 15. Finally, we previously showed that Cry2 promotes myoblast proliferation and fusion during differentiation in a circadian manner through stabilization of mRNAs encoding cyclin D1, a G1/S phase transition driver, and Tmem176b, a transmembrane regulator for myogenic cell fusion16.
Mouse Per1 and Per2 share 73.4 % sequence similarity at the amino acid level but are not functionally redundant. Per1−/− mice and Per2−/− mice are grossly normal and fertile; however, both knockout (KO) mice exhibit circadian periods up to 2 hr shorter than wild-type (WT) mice and eventually become arrhythmic in constant darkness17–20. In contrast, Per1−/−:Per2−/− mice become arrhythmic immediately after transfer to constant darkness although they are morphologically normal and fertile. As for muscle phenotypes, Per2−/− mice show a 20% shorter running distance with a treadmill test compared with WT and Per1−/− mice although the length, weight, contractility, and abundance of several contractile proteins in the tibialis anterior (TA) muscle were similar in the three genotypes21. Unlike the extensively studied Bmal1’s involvement, however, virtually nothing is known about whether and how Per genes contribute to myogenic differentiation and muscle regeneration.
The present study uncovered insulin-like growth factor 2 (Igf2) as a critical link between Per1/Per2 and myoblast differentiation. Igf2 is a necessary and well-characterized autocrine differentiation promoter of myoblasts that increases in secretion levels during differentiation22–24. Igf2 is also upregulated upon muscle injury and enhances regeneration25–27. Additionally, several single nucleotide polymorphisms of the human IGF2 genes are associated with a loss of muscle strength following strenuous exercise28, 29. Igf2 null mice display impaired growth at birth but subsequently grow normally30. Igf2 binds the type I Igf1 receptor (Igf1r) with the highest affinity among several receptors, resulting in its auto-phosphorylation and subsequent activation of the PI3K/AKT pathway and the RAS/MAP kinase pathway31, 32. In particular, activation of p38α/β MAPK by phosphorylation is an essential downstream effector for the promotion of myoblast differentiation by Igf233, 34. p38 achieves the pro-differentiation function by triggering cell cycle exit, activating myogenic transcription factors, and opening the chromatin of muscle gene promoters (see ref35, 36 for references). Most Igf2 in blood and local tissues is bound by Igf-biding proteins (IGFBP1-IGFBP7), which up- or downregulate Igf2 functions37. For example, whereas IGFBP-3 inhibits myoblast differentiation38, IGFBP-5 is induced during early myoblast differentiation and amplifies the auto-regulatory loop of Igf2 expression, resulting in promoted differentiation39. Starting from a phenotypic analysis of Per KO mice on myoblast differentiation and muscle regeneration, this study uncovered Per1/2-regulated circadian epigenetic preconditions of the Igf2 gene in myoblast differentiation.
Results
Disrupted muscle regeneration and myoblast differentiation by Per depletion
To study the roles of Per1 and Per2 in skeletal muscle regeneration, TA muscle in Per KO mice was injured by barium chloride injection and their regeneration was assessed by a series of histological analyses. The mice were entrained at 12 hr-light [Zeitgeber Time (ZT) 0 - ZT12] and 12 hr-dark cycles (ZT12 - ZT24) for two weeks before experiments. TA muscle was damaged and harvested at ZT14; ZT14 (20:00) was selected to compare with the result of the antiphase time point (ZT2, 8:00) later. Hematoxylin eosin (HE) staining on day 4.5 post-injury demonstrated smaller myofibers with centrally-located nuclei, an indication of newly formed myofibers, in Per1−/−, Per2−/−, and particularly Per1−/−:Per2−/− mice compared with WT mice (Fig. 1a-1c). This trend continued at least until day 14 and was also observed in uninjured myofibers, although the smallest size of Per1−/−:Per2−/− myofibers became less obvious (Supplementary Fig. 1a,b). In addition, myofibers expressing embryonic myosin heavy chain (eMHC), a marker for newly generated myofibers, were smaller in Per2−/− and Per1−/−:Per2−/− mice than those in WT and Per1−/− mice (Fig 1d,e). Furthermore, the kinetics of cell cycle exit of satellite cells in the KO mice were different from those in WT mice. During muscle regeneration, activated satellite cells re-enter the cell cycle [EdU (5-ethynyl-2’-deoxyuridine)(+)/MyoD(+) population], followed by exit from the cell cycle [EdU(-)/MyoD(+)] before terminal differentiation. Comparison of the frequency of each population on day 4.5 showed an increased frequency of EdU(+)/MyoD(+) cells in the single and double KO mice compared with WT mice (Fig. 1f,g). This finding could suggest delayed cell cycle exit of the satellite cells in the KO mice, which translates into delayed muscle regeneration on day 4.5. Finally, the single and double KO mice contained more scars on day 14 after injury as demonstrated by Sirius red stain, suggesting more extensive damage or delayed regeneration in the KO mice (Fig. 1h,i). Uninjured Per2−/− and Per1−/−:Per2−/− mice already contained more scars than WT mice although it was less severe than in day 14 mice, implying that natural turnover of myofibers was also disrupted in the KO mice. Together, these findings provide evidence that both Per1 and Per2 are necessary for the proper regeneration of TA muscle.
To understand cell-autonomous effects of Per KO, primary myoblasts (activated satellite cells) were purified from hind limb and induced to differentiate into myotubes with 5% horse serum in vitro. Per1−/−, Per2−/−, and Per1−/−:Per2−/− myoblasts displayed delayed activation of MHC, a marker for differentiation, and increased frequency of EdU uptake compared with WT cells (Supplementary Fig. 1c-g). This result exhibited impaired differentiation of Per KO myoblasts, consistent with the delayed TA muscle regeneration.
To obtain a large number of cells for a mechanistic study, we examined whether the mouse myoblast cell line C2C12 could recapitulate the KO phenotypes of the primary myoblasts. The Per1 or Per2 gene was depleted by shRNA-mediated knockdown (KD) and CRISPR-Cas9-mediated KO (Supplementary Fig 2a,b). These cells were used in bulk without cloning because differentiation-resistant cells would have been selected by cloning. When the cells were induced to differentiate, MHC(+) cells were shorter and more sparse in the KD and KO cells than control cells on differentiation days 3 and 5 (Fig. 2a). This finding was quantified as decreased differentiation index (frequency of nuclei in MHC(+) cells among total nuclei) and fusion index (frequency of nuclei in MHC(+) cells containing more than one nuclei among total nuclei) in the KD and KO cells (Fig. 2b,c). The KO cells also demonstrated slightly delayed cell cycle exit during differentiation (Fig. 2d). Additionally, expression of differentiation-specific genes encoding myogenin (Myog), muscle creatinine kinase (Ckm), myomaker (Mymk), and MHC (Myh3) was decreased by the KD and KO (Fig. 2e). The similarity of the inhibited differentiation between Per1 KO and Per2 KO cells was further highlighted by several transcriptome data sets. The list included a heat map, a principal component analysis, Venn diagrams of differentially expressed genes (more than 2,000 genes were commonly up- or downregulated more than 2-fold compared with control cells), scatter plots (R2>0.97), and common representations of muscle-related pathways in a gene ontology analysis of the RNA-seq data on days 0, 3, and 5 (Fig. 2f,g and Supplementary Fig. 2c-f). These results indicate that both Per1 and Per2 are necessary for effective myoblast differentiation in vitro.
Downregulation of Igf2 expression is a common consequence of Per1 KO and Per2 KO
The common phenotypes of Per1 and Per2 depletion led us to search for important myogenic genes that were commonly up- or downregulated by each KO. Analysis of the RNA-seq data revealed a substantial downregulation of Igf2 by both Per1 KO and Per2 KO throughout differentiation from day 0 (undifferentiated) to day 5 (Supplementary Fig. 3a,b). This result was verified by qPCR of Per KD and KO cells (Fig. 3a). To investigate the involvement of Igf2 in the Per depletion phenotypes, Igf2 was knocked down with two shRNAs in C2C12 cells (Supplementary Fig. 3c). MHC(+) cells in the KD cells were more sparse and shorter than control KD cells, consistent with the lower differentiation index and fusion index, as well as the decreased expression of differentiation-specific genes (Fig. 3b-e). Cell cycle exit was also delayed by Igf2 KD during differentiation (Fig. 3f). Thus, Igf2 depletion recapitulated the phenotypes of Per depletion.
Igf2 was likely to be expressed in a circadian manner as a downstream effector of Per1 and Per2. This possibility was tested by western blotting with C2C12 cells harvested every 4 hr after circadian synchronization with dexamethasone. The protein level of Bmal1 reached a peak at 44 hr after synchronization, which was anti-phasic to the expression patterns of Per1 and Per2 in control cells (Fig. 3g). Igf2 expression reached peaks at 32-36 hr and 56 hr, similar to the patterns of Per1 and Per2. Phosphorylation of p38 (p-p38) followed the expression pattern of Igf2 as its downstream effector. In contrast, Igf2 and p-p38 were severely downregulated in Per KO cells. Igf2 was also expressed in TA muscle in a circadian manner but the rhythms were largely lost in Per1−/− and Per2−/− mice (Fig. 3h). These results verified that Igf2 expression is regulated by Per1 and Per2 in a circadian manner.
Next, the concentration of Igf2 in the culture supernatant of C2C12 cells was measured with ELISA. The concentration was approximately 0.025 ng/ml with undifferentiated cells and was increased 6-fold during differentiation as previously reported (Fig. 3i)22. Although Igf2 in the supernatant of Per KO cells was also increased, the level remained less than 20% of the control level on day 5, consistent with PCR and western blotting results. The Igf2 concentration with control cells also displayed oscillation that was similar to the western blotting result (Fig. 3j). The concentration represented the amount of accumulated Igf2 since 0 hr, when dexamethasone was replaced with fresh culture medium. The result likely reflected the gain by secretion and the loss by degradation and attachment to the culture dish and cell surface. The oscillation became more evident when an increase or a decrease between two time points were plotted (Fig. 3k).
We also examined whether exogenous Igf2 could rescue the disrupted differentiation of Per KO cells by adding Igf2 to the culture medium from day 0 onward. Igf2 raised differentiation index and fusion index 2- to 3-fold at >1 ng/ml but the indices did not reach the levels of the control cells (Supplementary Fig. 4a,b). In addition, the required concentrations were more than 10-times higher than that in the culture medium of the control cells. The high concentrations of Igf2 could also increase the Myog and Ckm mRNA close to the levels of the control cells that were not treated with Igf2 (Supplementary Fig. 4c,d). Thus, Igf2 could partially rescue the disrupted differentiation of Per KO cells if an excessive amount was provided. Note that the concentration of effective Igf2 in the culture medium was unknown due to the presence of IGFBPs.
Epigenetic regulation of Igf2 expression by Per1 and Per2
To elucidate how Per1 and Per2 promoted Igf2 expression, epigenetic changes caused by Per KO were studied with ChIP-qPCR. An Igf2 enhancer containing two E-boxes is embedded within an intron of the Nctc1 gene located 105 kb downstream of the Igf2 promoter40, 41. Publicly available ChIP-seq data obtained with non-synchronized myoblasts demonstrated binding peaks of Bmal1 and the histone acetylase p300 as well as the marker for enhancers H3K27ac (acetylation of lysine 27 in histone H3), but not another enhancer marker H3K4me1 (R3 region in Fig. 4a). Our ChIP-qPCR with synchronized control C2C12 cells detected binding peaks of Bmal1, Clock, Per1, and Per2 at similar time points and the patterns were preserved in Per KO cells (Fig. 4b,c and Supplementary Fig. 5a). However, the levels of H3K27ac, H3K4me1, and p300 were significantly decreased in Per KO cells, indicating Per1 and Per2 dependency of the histone modifications.
Muscle cells primarily utilize Promoter 3 among the three promoters of the Igf2 gene42, 43, which was verified by qPCR (Supplementary Fig. 6a,b). Downloaded ChIP-seq data with non-synchronized myoblasts did not show a specific increase or decrease of Bmal1, RNA polymerase II (Pol II), or histone markers for active genes (H3K4me3 and H3K9ac) and repressed genes (H3K27me3) at Promoter 3 (region R13 in Fig. 5a). However, synchronized C2C12 cells again demonstrated a Per-dependent increase (Pol II, H3K4me3, and H3K9ac) or decrease (H3K27me3) of these proteins at Promoters 3 and 1 (R15) (Fig. 5b,c and Supplementary Fig. 6c). Thus, both Per1 and Per2 were necessary for the circadian dynamics of multiple epigenetic markers characteristic for gene activation at the Igf2 enhancer and promoters.
To understand the functional significance of the circadian epigenetics, the temporal profile of nascent Igf2 mRNA was quantified with synchronized cells. A nuclear run-on assay demonstrated that nascent Igf2 mRNA was most abundant at time points when the Igf2 enhancer and promoters were enriched with active gene markers in control cells (Fig. 6a, 36 hr and 60 hr). However, the Igf2 level remained low throughout the process with Per KO cells as expected. Therefore, the circadian transcriptional changes of Igf2 indeed reflected the epigenetic dynamics of the gene.
The interaction between the Igf2 enhancer and Promoter 3 has been shown in differentiating myoblasts and skeletal muscle41, 44. Because chromatin interactions are also regulated by circadian rhythms66, 67, we hypothesized that the Igf2 enhancer-promoter interaction would also demonstrate circadian oscillation. This possibility was examined with Chromosome Conformation Capture (3C) by studying the interaction between Promoter 3 (anchor point in 3C) and the enhancer at 24 and 36 hr post-synchronization, which corresponded to the nadir and the peak of the Igf2 level, respectively (Fig. 6b). The interaction (crosslinking frequency) at 36 hr was approximately twice as high as it was at 24 hr in control cells (Fig. 6c). This pattern was preserved in Per KO cells but the crosslinking efficiency became less than half compared with control cells (Fig. 6c,d). Thus, both Per1 and Per2 are required for the circadian dynamics of the promoter-enhancer interaction at the Igf2 gene, just like the epigenetic modifications.
Coupling of differentiation efficiency and the circadian timing of differentiation initiation
The above findings led us to a hypothesis that myoblasts could differentiate more efficiently if differentiation cues are provided at the circadian timing when Per1, Per2, and Igf2 are highly expressed compared with other time points (precondition). This was evaluated with C2C12 cells that were induced to differentiate at different time points after synchronization. Indeed, the cells differentiated more efficiently when induced at 36 hr and 60 hr post-synchronization compared with 24 hr and 48 hr as demonstrated by higher differentiation index, fusion index, and the expression levels of differentiation-specific genes (Myh3, Myog, and Ckm) and Igf2 after 48 hr of differentiation (Fig. 6e-i). This trend was in agreement with the epigenetic modifications of the Igf2 promoters and enhancer. Note that since the cells kept proliferation between 24 hr and 60 hr before the differentiation induction, the increased cell density per se could promote differentiation in the later phase. In this sense, the drop of the differentiation efficiency at 44 - 48 hr in comparison to 36 hr was more significant than the increased differentiation from 48 hr to 60 hr. Igf2 KD cells differentiated poorly regardless of when differentiation was initiated (Supplementary Fig. 7a-c).
Since Per1, Per2, and Igf2 were also expressed in TA muscle in a circadian manner, the timing of injury could be an important factor for muscle regeneration. To assess this possibility, TA muscle was injured at ZT2 (low Per1, Per2, and Igf2; the early inactive phase of mice) and ZT14 (high Per1, Per2, and Igf2; the early active phase) to compare the regeneration efficiency. HE staining clearly showed larger TA muscle in the ZT14 WT mice than in the ZT2 mice on day 4.5 and this difference was lost by day 14 (Fig. 7a,b and Supplementary Fig. 7d). There was no statistically significant difference between ZT2 and ZT14 injuries in the single and double KO mice of Per1 and Per2. The average diameter of eMHC(+) myofibers was also longer in WT mice damaged at ZT14 than that in the ZT2 damage and this difference was also lost in the single and double KO mice (Fig. 7c,d). Moreover, the frequency of the EdU(+)/MyoD(+) population was diminished in the ZT14 WT mice compared with ZT2 mice on day 4.5 and 5.5, suggesting early cell cycle exit (Fig. 7e,f). Finally, scar formation was also less abundant in the ZT14 WT mice than in the ZT2 mice on day 14; this difference was again erased in the single and double KO mice (Fig. 7. g,h). These results collectively indicate that circadian timing of injury affects the efficiency of TA muscle regeneration in a Per1- and Per2-dependent manner.
DISCUSSION
The central message of the present work is that the efficiency of myoblast differentiation and muscle regeneration is dependent on the circadian timing when these events are triggered. As a mechanistic explanation obtained with the myoblast model, the Igf2 gene was preconditioned toward activation in a circadian manner while the cells were still in the proliferation medium. Despite extensive studies of circadian regulation of cell proliferation and differentiation45, 46, studies focused on the mechanistic influence of circadian timing on tissue regeneration are quite limited. One of the few studies concerns fibroblast migration during skin wound healing47. Fibroblast mobilization to a mouse skin incision site, an early and essential step in wound healing, was greater when the wound was inflicted at ZT13 than at ZT5. Additionally, when a skin explant was harvested at different time points and immediately wounded by a biopsy punch, the number and volume of fibroblasts invading the wound area were higher in the explant harvested at ZT13 than at ZT5. Circadian regulation of actin polymerization, which controls migration and adhesion, is one of the mechanisms for the time-dependent difference in the wound healing efficiency. In a related phenomenon, circadian timing of physical exercise influences muscle strength and oxidative capacity48. For example, muscle atrophy in the mouse hind limb due to reduced gravity was prevented more effectively by intermittent weigh bearing at ZT12-ZT16 than at ZT20-ZT049. Based on our results, circadian timing could affect the recovery from muscle damage caused by trauma and surgery.
Our findings on the circadian timing-dependent differentiation and regeneration should be interpreted in a broader perspective of circadian metabolic regulation that defines the availability of energy and cellular building blocks50–52. A circadian transcriptome analysis of muscle uncovered clustered expression of genes with a common metabolic function at specific circadian phases in the mouse under constant darkness with ad libitum feeding53. Specifically, the genes involved in carbohydrate catabolism (the early active/dark phase), carbohydrate storage (the mid-active/dark), lipogenesis (the end of the active/dark phase), and fatty-acid uptake and β-oxidation (the mid-inactive/light phase) reached peaks at distinct circadian phases as indicated in the parentheses. Metabolomic profiling of muscle also demonstrated neutral lipid storage and decreased lipid and protein catabolism in the late inactive phase54. Given the global circadian oscillation of the numerous metabolites essential for tissue turnover, circadian timing could create a preferred time window for an effective response to major tissue disruption and repair although experimental evidence is lacking. The interaction between the Per/Igf2 axis and the global metabolic oscillation awaits further studies.
The notion of the preferred time window also needs to be assessed in relation to other muscle regulators that exhibit circadian expression patterns. For example, glucocorticoids are major circadian modulators of multiple activities, including energy metabolism and tissue regeneration. The glucocorticoid secretion is characterized by robust circadian oscillation with a daily peak at the start of the activity (early night for mouse)55, 56. Glucocorticoids promote myofiber repair after injury via increased expression of the immunomodulators annexins A1 and A6, both of which are also involved in muscular dystrophy phenotypes57, 58. It remains to be elucidated how glucocorticoids, along with other muscle regulators Bmal1, Rev-erbα, and Cry2 mentioned earlier, interact with the Per/Igf2 axis in organizing circadian timing-dependent muscle cell differentiation and regeneration.
Various histone modifications oscillate at circadian-regulated genes through the recruitment of responsible enzymes as binding partners of the Clock/Bmal1 complex4, 59. The modifications include markers for gene activation, such as acetylation by p300 and CREB-Binding Protein (CBP) as well as H3K4me3 by mixed lineage leukemia 1 (MLL1), and those for gene repression, including deacetylation by Sirtuin-1 (SIRT1) and H3K27me3 by Enhancer of Zeste 2 (EZH2). Reflecting the roles of the Per proteins as negative regulators, their binding partners include well-known chromatin repressive complexes, such as HP1γ-Suv39h (induces H3K9me2 and H3K9me3) and deacetylase complexes Sin3 and NuRD. The vast majority of these studies used mouse liver and embryonic fibroblasts as model materials, leaving tissue-specific variability under-explored. In our study, more than 1,000 genes were commonly activated by Per 1 KO and Per2 KO cells, suggesting that Per1 and Per2 can act as gene activators in a context-dependent manner. Per-induced activation has been demonstrated with several genes involved in sodium channels in the kidney60–62. Whereas Per1 directly inhibits the expression of the WNK4 gene, it activates the NCC and WNK1 genes in a mouse renal tubular cell line60. In another example, Bmal1 and Per1 are required for the circadian activation of prolactin in a rat mammotrope cell line63. Moreover, Per2 activates Cry1 by removing the Clock/Bmal1/Cry1 repressor complex from the Cry1 promoter in an ectopic expression model64. This study also showed that genes with complex promoters can be repressed or de-repressed by Per, depending on the regulatory elements at the promoters. Identification of the binding proteins of Per1 and Per2 would be an important next step to further clarify how Per1 and Per2 activate Igf2 during myoblast differentiation.
Long-range interactions between enhancers and promoters are achieved by chromatin loop formation, which is mediated by CCCTC-binding factor (CTCF) and the cohesin complex at the anchors of the loops and YY1 at the interface between promoters and enhancers65. Chromatin loop formation is one of the emerging mechanisms underlying oscillating gene activity66, 67. For example, the Cry1 gene promoter interacts with its enhancer in the first intron in a synchronized manner with oscillating gene expression in the liver68. Little is known, however, about the direct interface between the circadian master proteins and the loop proteins. One of few examples is Rev-Erbα, which prevents the loop formation by recruiting the NCoR-HDAC3 repressive complex and removing the Mediator complex, a promoter of enhancer-promoter interactions69. Unidentified binding proteins of Per proteins are undoubtedly involved in the loop formation at the Igf2 gene.
This study revealed circadian regulation of myoblast differentiation and muscle regeneration and demonstrated epigenetic regulation of the Igf2 gene by Per1 and Per2 as one of the underlying mechanisms using a myoblast differentiation model. Future genome-wide epigenetic analysis of histone modifications and chromatin interactions would further uncover other unexpected underpinnings for the time-of-the-day-dependent regeneration of muscle and other tissues.
Author contribution
A.A. and N.K. conceived experiments and analyzed data. N,K-K., E.P, D.L.H., B.L., D.M., X.W., M.H., J.B., Y.A., A.A., and N.K. performed experiments. S.L. and C.Y. analyzed bioinformatics data. A.A. and N.K. wrote the manuscript. All edited the manuscript.
Competing interests
The authors declare no competing interests.
Supplementary Figure Legends
Online Methods
Culture of C2C12 cells
Mouse myoblast C2C12 cells (American Type Culture Collection, CRL-1772) were maintained in the growth medium [10% fetal bovine serum (FBS) in Dulbecco’s Modified Eagle Medium (DMEM)] in a 37°C and 5% CO2 incubator. Differentiation was induced on day 0 when cells were at 90% confluence by rinsing with phosphate buffered saline (PBS) twice and adding the differentiation medium [5% horse serum (HS) in DMEM]. The medium was changed every two days thereafter. Concentration of endogenous Igf2 was measured with a mouse Igf2 ELISA kit (Thermo Fisher Scientific, EMIGF2). Because of the low Igf2 concentration, 300 μl culture supernatant was added seven times for 3 hr each (2.1 ml in total). The effect of exogenous Igf2 was tested with recombinant human IGF2 (Peprotech, 100-12). To synchronize circadian rhythms, C2C12 cells were seeded on day −1 and 0.2 μM dexamethasone was added at −1 hr on day 0. Cells were washed with PBS twice and fresh 10% FBS in DMEM was added at 0 hr. Cells were harvested for PCR or fixed for immunofluorescence staining every 4 hrs at indicated time points.
Gene Knockdown in C2C12 cells
On day 1, 293FT cells (Thermo Fisher Scientific, R70007) were seeded in DMEM with 10% FBS at 3×105 cells/well in a 12-well plate. On day 2, cells were transfected with 0.5 μg pLKO.1 lentivirus vector encoding an shRNA sequence shown in Supplementary Table 1, along with 0.2 μg each of pCMV-VSV-G (Addgene, 8454), pRSV-Rev (Addgene, 12253), and pMDLg/pRRE (Addgene, 12251) using 2.75 μl Lipofectamine 2000 (Thermo Fisher Scientific, 11668019). The culture medium was replaced with fresh DMEM with 10% FBS 5 hr after transfection. On day 4, C2C12 cells were seeded at 1×105 cells/well in 12-well plates. On day 5, culture supernatant of the transfected 293FT cells was applied to a 0.45 μm syringe filter and added to C2C12 cells with 0.8 μg/ml polybrene (MilliporeSigma, H9268). The culture medium was replaced with fresh DMEM with 10% FBS on day 6. Virus-integrated cells were selected with 1 μg/ml puromycin dihydrochloride (MB Bio, 100552) between days 7 and 14. Selected cells were expanded and frozen in liquid nitrogen.
Gene Knockout in C2C12 cells
On day 1, C2C12 cells were seeded in DMEM with 10% FBS at 1.8×105 cells/well in 12-well plates. Cells were transfected with 0.5 μg single guide RNA (sgRNA, Synthego) against Igf2 (Supplementary Table 1), 0.5 μg CleanCap Cas9 mRNA (Cas9 mRNA, TriLink, L-7206), 0.2 μg CleanCap mCherry mRNA (TrLink, L-7203), and 1 μl TransIT (Mirus Bio TransIT-mRNA transfection kit, MIR-2225). More than 90% of the cells were mCherry (+) with fluorescence microscope observation on day 2. Cells were subcultured on day 3 to expand and freeze. A small aliquot of cells was used for Sanger sequencing of genomic DNA to verify the knockout efficiency.
Knockout Mice
All protocols were approved by the Institutional Animal Care and Usage Committee of the University of Minnesota (1902-36737A and 1903-36906A). Per1+/− mice (B6.129-Per1tm1Drw/J, stock # 010491) and Per2+/− mice (B6.129-Per2tm1Drw/J, stock# 010492) were purchased from Jackson Laboratory. Per1−/−, Per2−/−, Per1−/−:Per2−/−, and WT mice were obtained by breeding and identified by genotyping according to Jackson Laboratory protocols. The same number of male and female mice at the age of 8-10 weeks old were mixed in each group. Mice were entrained at 12 hr-light and 12 hr-dark cycles (6:00-18:00 light and 18:00-6:00 dark) for two weeks before experiments. ZT0 corresponds to 6:00 and ZT12, 18:00. TA muscle was injected with 50 μl 1.2% BaCl2 in 0.9% NaCl \at ZT2 or ZT14 to induce muscle injury and subsequent regeneration. Mice were euthanized on day 4.5, 7, and 14 after injury and the TA muscle was extracted at ZT2 or ZT14. In addition, uninjured TA muscle was isolated every 4 hr for qPCR. Mice were monitored by the Research Animal Resources staff of the University of Minnesota in specific pathogen free housing. Mice were given standard chow and access to drinking water without restrictions. Mice were euthanized via CO2 inhalation. All methods align with the Panel of Euthanasia of the American Veterinary Medical Association recommendations.
Hematoxylin Eosin (HE) Staining
Cryosections of the TA muscle were fixed with 2% paraformaldehyde for 10 min. The sections were then treated as follows: deionized water for 1 min, Harris Modified Hematoxylin (Thermo Fisher Scientific, SH26-500D) for 2 min, tap water for 1 min, deionized water for 1 min, Eosin-Y (Thermo Fisher Scientific, 22-220-104) for 5 min, 95% ethanol for 30 sec, 100% ethanol for 2 min twice, and xylene for 10 min twice. The sections were mounted with Permount (Thermo Fisher Scientific, SP15-100). Images were captured with the cellSens Entry 1.11 software (Olympus) and a DP26 camera (Olympus) attached to an IX73 microscope (Olympus). The minimal Feret’s diameter of each myofiber was quantified with Fiji (NIH).
Sirius Red Staining
The Sirius red solution was composed of 1% Direct Red 80 (MilliporeSigma, 365548) in 1.3% picric acid. TA sections were fixed with acetone, pre-chilled at −20°C, for 10 min. The sections were then washed in deionized water for 1 min, stained with Sirius red for 15 min, and rinsed with 0.5% acetic acid for 1 min. The sections were subsequently washed with 100% ethanol for 2 min and twice with xylene for 10 min. Finally, the sections were mounted with Permount for taking images as described for HE staining. Quantification of Sirius red (+) fibrotic was done using the entire section area with Fiji1.
Immunofluorescence Staining of TA Sections
Sections were first fixed with 2% paraformaldehyde for 10 min, followed by permeabilization with 0.2% Triton X-100 in PBS for 5 min. The sections were treated with two blocking reagents: 3% Mouse-on-Mouse Blocking Reagent (Vector MKB-2213) in PBS for 1 hr and 10% bovine serum albumin (BSA) in PBS for 30 min. Primary antibodies against eMHC and laminin (see Supplementary Table 2 for catalog numbers) diluted in 10% BSA in PBS were applied overnight, followed by washing twice with 0.01% Triton-X100 in PBS. The secondary antibodies Alexa Fluor 488 donkey anti-mouse IgG and Alexa Fluor 594 donkey anti-rat IgG diluted in 10% BSA in PBS were used for 1 hr. DNA was counterstained with 4’,6’-diamidine-2’-phenylindole dihydrochloride (DAPI, MilliporeSigma, 10236276001). Sections were mounted using Fluorescent Mounting Medium (DAKO, S302380-2). To label proliferating myoblasts, mice were intraperitoneally injected with 5-ethynyl-2’-deoxyuridine (EdU) at 50 μg/g body weight on day 4 and euthanized 12 hr later on day 4.5 post-injury. EdU was detected using Click-iT EdU Alexa Fluor 488 Imaging Kit (ThermoFisher Scientific, C10337) followed by staining with anti-MyoD antibody, Alexa Fluor 594 donkey anti-rabbit IgG antibody and DAPI2. Fluorescence images were captured using Metamorph Basic software (Molecular Devices) and an ORCA-flash4.0LT camera (Hamamatsu) attached to an IX73 microscope (Olympus) with a 20X LUCPlan FL N lens. Images were processed with Photoshop and Illustrator CS6 (Adobe). The minimal Feret’s diameter of each myofiber was quantified with Fiji (NIH).
Preparation of Primary Myoblasts
Mouse muscle mononuclear cells were prepared from hind limbs of 8-10 weeks old male mice as previously described3, 4. More specifically, muscles were minced and digested with 0.2% collagenase type 2 (Worthington, CLS-2) in DMEM to dissociate muscle cells. Satellite cells were purified with LD columns (Miltenyi Biotec, 130-042-901) by negative selection with antibodies against CD31-PE, CD45-PE, and Sca1-PE, followed by anti-PE MicroBeads (Miltenyi Biotec, 130-048-801). Positive selection was subsequently applied with an antibody against biotin-conjugated integrin α7-biotin and anti-biotin MicroBeads (Miltenyi Biotec, 130-090-485), followed by MS columns (Miltenyi Biotec, 130-042-201). Isolated satellite cells were cultured on dishes coated with 0.01% rat tail collagen (BD Biosciences, 354236) in myoblast growth medium (HAM’s F-10 medium with 20% FBS, 10 ng/ml basic fibroblast growth factor (Thermo Fisher Scientific, PHG0263), penicillin (100 U/ml), and streptomycin (100 mg/ml)) at 37°C with 5% CO2. Low-passage satellite cell-derived primary myoblasts (typically less than eight passages) were used for differentiation and immunostaining. DMEM with 5% HS, penicillin, and streptomycin was used for myogenic differentiation.
Immunofluorescence Staining of Cells
Primary myoblasts were fixed with 2% paraformaldehyde for 10 min and blocked with 1% BSA in PBS for 30 min. Permeabilized cells were stained with antibodies against MHC and MyoD and then incubated with secondary antibody Alexa Fluor 488 anti-mouse IgG and Alexa Fluor 594 anti-rabbit IgG. DNA was counterstained with DAPI. To measure proliferating primary myoblasts, cells were pulsed with 1 μg/ml of EdU 3 hr before harvest. EdU was detected using a Click-iT EdU Alexa Fluor 488 Imaging Kit (ThermoFisher Scientific, C10337) followed by anti-MyoD antibody, Alexa Fluor 594 donkey anti-rabbit IgG antibody, and DAPI2. Fluorescence images were captured using Metamorph Basic software (Molecular Devices) with LUCPlanFLN 20x or 10x objective lens (Olympus) with 0.45 Ph1 aperture and a C11440-42U digital camera (Hamamatsu) attached to an IX73P2F microscope (Olympus). The images were processed with Adobe Photoshop and Illustrator CS6.
C2C12 cells were similarly stained with primary antibody against MHC, secondary antibody Alexa Fluor 488 anti-mouse IgG, and 5 μg/ml Hoechst 33342 (MilliporeSigma, B2261). Differentiation index was defined as a percentage of nuclei (Hoechst-stained structure) existing within MHC(+) cells. Fusion index is a percentage of nuclei that were located in MHC(+) cells containing more than one nuclei. Data of differentiation index, fusion index, and EdU uptake were obtained from biological triplicates in each experiment.
Western Blotting
Whole-cell extracts obtained from 2×105 cells with an NE-PER Nuclear and Cytoplasmic Extraction reagents (Thermo Fisher Scientific, 78833) were loaded into a 12% SDS-PAGE gel. After completion of electrophoresis, the proteins were transferred to an Immobilon P membrane (EMD Millipore, IPVH00010) at 25°C overnight. The next day, the membrane was blocked with 5% non-fat dry milk (BioRad, 180171A) in PBT (0.2% Tween 20 in PBS) for 1 hr at 25°C. Proteins were then labeled with the primary antibody of interest diluted in 5% milk in PBT at 25°C for 1 hr. After washing with PBT for 5 min three times, the membranes were incubated with secondary antibodies goat anti-rabbit IgG-HRP or goat anti-mouse IgG-HRP both diluted at 1:1000 in 5% milk in PBT for 1 hr at 25°C. After washing the membrane with PBT six times, the chemiluminescence signal was detected with a SuperSignal West Dura kit (Thermo Fisher Scientific, 34075) and X-ray films.
The phospho-p38 antibody was used as follows to avoid the sequestration of the antibody by casein in the milk. The membrane was blocked with 5% non-fat dry milk in TBST (50 mM Tris-HCl pH 7.5, 0.1% Tween 20) for 1 hr at 25°C, followed by wash with TBST for 5 min three times. The p-p38 antibody was diluted in 5% BSA in TBST and three subsequent washes were done with TBST. Goat anti-rabbit IgG-HRP was diluted in 5% milk in TBST and six subsequent washes were done with TBST.
Quantitative RT-PCR (qRT-PCR)
RNA was extracted from cells using a Quick RNA Microprep (Zymo Research, R1051) or RNeasy Plus Mini (Qiagen, 74136) kit, depending on cell number. RNA quantity and purity were assessed using a NanoDrop Lite spectrophotometer (Thermo Fisher Scientific). cDNA was synthesized with ProtoScript II Reverse Transcriptase (New England Biolabs, M0368L). Quantitative PCR (qPCR) was performed with the primers listed in Supplementary Table 3 and a GoTaq qPCR Master Mix (Promega, A6002) in a Mastercycler realplex2 thermocycler (Eppendorf). PCR conditions were as follows: initial denaturation at 95°C for 10 min, 40 cycles of 95°C for 15 sec, 30 sec at the specific annealing temperature for each set of primers, and 72°C for 30 sec, and a melting curve step to check the specificity of the reaction. mRNA expression levels were analyzed by normalizing expression values to glyceraldehyde 3-phosphate dehydrogenase (Gapdh) expression. Mean ± SEM of biological triplicates with technical triplicates each were calculated.
Nascent mRNA was measured with a Click-iT Nascent RNA Capture kit (Thermo Fisher Scientific, C10365). C2C12 cells were incubated with 0.1 mM 5-ethynyl uridine (EU) for 4 hr, followed by the isolation of EU-labeled mRNAs with magnetic beads for cDNA synthesis and qPCR.
ChIP-PCR
Two million C2C12 cells were treated with 1% paraformaldehyde for 10 min and then 125 mM glycine, followed by washing with PBS twice. Chromatin was prepared by the treatment with 300 μl cell lysis buffer (50 mM HEPES pH7.8, 85 mM NaCl, 0.5% NP-40, and cOmplete Mini Protease Inhibitor Cocktail (MilliporeSigma, 11 836 153 001)) for 15 min on ice with vortexing for 15 sec every 5 min. After centrifugation at 1,500 x g for 5 min at 4°C, the pellet was resuspended in 50 μl nuclear lysis buffer (50 mM Tris-HCl pH 8.0, 10 mM EDTA, 1% SDS, and cOmplete protease inhibitor cocktail) and incubated for 5 min on ice. Chromatin was sheared by sonication with a Bioruptor 300 (Diagenode) with 30 cycles of 30 sec-on and 30 sec-off with the high power setting at 4°C. After centrifugation at 15,000 xg for 15 min at 4°C, 5 μl supernatant (2×105 cells) was saved as input and the rest was incubated with 2 μg antibody, 2 μl Dynabeads Protein G (Thermo Fisher Scientific, 10004D), and 400 μl dilution buffer (20 mM Tris-HCl pH 8.0, 150 mM NaCl, 2 mM EDTA, 1% Triton X-100, and cOmplete protease inhibitor cocktail) for 16 hr at 4°C with rotation at 20 rpm. The beads were sequentially washed with 500 μl each of dilution buffer, LiCl buffer (20 mM Tris-HCl pH 8.0, 250 mM LiCl, 2 mM EDTA, 0.1% SDS, and 1% Triton X-100) twice, and TE (10 mM Tris-HCl and 1 mM EDTA) rotating for 5 min at 4°C each. The beads were resuspended in 100 μl elution solution (0.1 M NaHCO3, 1% SDS, and 200 μg/ml proteinase K) and incubated for 2 hr at 65°C with rotation to reverse crosslink. The input was also treated in the same condition. Proteinase K was inactivated by heating for 10 min at 95°C. DNA was purified with a ChIP DNA Clean & Concentrator (Zymo Research, D5205) and applied for qPCR as described above. Results of biological triplicates with technical duplicates each were presented as ratios in comparison to input chromatin.
The ChIP-seq data GSE25308 (Pol II, H3K4me1, H3K4me3, H3K27me3, and H3K9ac), GSE37527 (H3K27ac and p300), and GSE108650 (Bmal1) were downloaded from the UCSC Mouse Genome Browser mm9. The quality of all downloaded data was evaluated using FastQC v0.11.5 (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and adapter sequence was trimmed using Trimmomatic v0.33 (http://www.usadellab.org/cms/index.php?page=trimmomatic)5. The filtered high-quality reads were then mapped to a reference genome (GRCm38/mm10) using HISAT2 v2.0.2 (https://ccb.jhu.edu/software/hisat2/index.shtml)6. Resulting BAM files with MACS version 2 (2.2.5) were used to generate peaks7.
3C analysis
3C was performed combining two published protocols8, 9. To make a BAC control library, a mouse 110 kb BAC clone encoding Nctc1 and Igf2 were purchased from Thermo Fisher Scientific (RPCI23.C) and the DNA was prepared with a ZR BAC DNA Miniprep kit (Zymo Research (D4048) as control. Ten micrograms of DNA was digested with BamHI for 16 hr, followed by phenol-chloroform extraction and ethanol precipitation twice. DNA fragments were ligated with T4 ligase for 16 hr at 16°C. DNA was purified with phenol-chloroform extraction and ethanol precipitation twice and used in qPCR.
C2C12 cells were treated with 1% paraformaldehyde and then with 125 mM glycine as described in the ChIP-PCR method. Ten million cells were treated with lysis buffer (10 mM Tris-HCl pH7.5, 10 mM NaCl, and 0.2% NP-40) for 10 min on ice. After centrifugation, the pellet was washed once with the digestion buffer for BamHI and incubated in a series of buffers for 1 hr at 37°C each: BamHI buffer, BamHI buffer with 0.3% SDS, and BamHI buffer with 0.3% SDS and 2% Triton X-100. Finally, the cells were incubated with 400 units BamHI for 16 hr at 37°C with rotation. BamHI was inactivated with 1.6 % SDS and incubation at 65°C for 30 min. The cell suspension was diluted 12-fold with lysis buffer (10 mM Tris-HCl pH7.5, 10 mM NaCl, 0.2% NP-4, and 1% Triton X-100) and incubated for 1 hr at 37°C with rotation. After addition of 100 units T4 DNA ligase, the reaction mix was incubated for 4 hr at 16°C, followed by incubation for 1 hr at 25°C. Finally, the reaction mix was incubated for 16 hr at 65°C in the presence of 300 μg proteinase K, followed by phenol-chloroform extraction and ethanol precipitation twice as a 3C library.
PCR data was analyzed as detailed in reference8. Briefly, PCR was performed with 10 ng BAC control library and 3C library as templates using the 3C-0 kb primer in combination with each of 3C-76 kb through 3C-112 kb primers. The primer pair 3C-Clock1 and 3C-Clock2 at the Clock gene, which showed a consistent amplification throughout the circadian rhythms, was used as an internal control. PCR conditions were as it follows: initial denaturation at 95°C for 5 min, 35 cycles of 95°C for 30 sec, 70°C for 30 sec, and 72°C for 20 sec, and an extension at 72°C for 5 min. PCR products were resolved with a 2% agarose gel and stained with ethidium bromide. The images were captured with a Gel Logic 212 Pro system (CareStream Molecular Imaging) and the band intensity was quantified with Fiji. Biological triplicates with technical duplicates were analyzed for each group.
RNA-seq
Total RNA was prepared from KO C2C12 cells before differentiation (day 0) and day 3 and 5 during differentiation. RNA concentration and RNA Integrity Number (RIN) were measured with an Agilent BioAnalyzer 2100. Samples with RIN over 8 were used to create sequencing libraries at the University of Minnesota Genomics Center. One microgram of total RNA was used to create each sequencing library using a Truseq RNA Sample Preparation Kit (Illumina, RS-122-2001). Briefly, poly-adenylated RNA was first purified using oligo-dT-coated magnetic beads. RNA was then fragmented and reverse-transcribed into cDNA. The cDNA was further fragmented, blunt-ended, and ligated to barcoded adaptors and amplified. Final library size distribution was validated with capillary electrophoresis and quantified with a Quant-iT PicoGreen dsDNA Assay Kit (Thermo Fisher Scientific, P11496) and qPCR. Indexed libraries were pooled and size-selected to 320 bp ± 5% with a LabChip XT (PerkinElmer). Libraries were loaded onto a single-read flow cell and amplified on a cBot (Illumina) before sequencing using a NextSeq High (Illumina).
Bioinformatics Analysis
The demultiplexed FASTQ files were analyzed using a customized pipeline (gopher-pipelines; https://bitbucket.org/jgarbe/gopher-pipelines/overview) developed and maintained by the Minnesota Supercomputing Institute. Briefly, FastQC v0.11.5 (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/) was used to check on the sequencing quality of the FASTQ files. Then adapters and low-quality reads were trimmed using Trimmomatic v0.33 (http://www.usadellab.org/cms/index.php?page=trimmomatic)5. An additional quality check with FastQC was performed on the post-trimming sequences to ensure successful adaptor and quality trimming. The remaining sequences were then aligned to the GRCm38/mm10 reference genome using HISAT2 v2.0.2 (https://ccb.jhu.edu/software/hisat2/index.shtml) and transcript abundance was counted using subread v1.4.6 (http://subread.sourceforge.net/)6, 10. Differential gene expression analysis was performed in R v3.6.2 using edgeR package (https://bioconductor.org/packages/release/bioc/html/edgeR.html)11. Gene ontology analysis of differentially expressed genes was performed by functionally annotate the genes and perform an overrepresentation enrichment test using PANTHER (http://pantherdb.org/)12. Heat maps were generated using the log-transformed counts with pheatmap v1.0.12 (https://cran.r-project.org/web/packages/pheatmap/index.html) packages. Hierarchical clustering was performed using the average linkage clustering method with the correlation coefficient as a similarity metric.
Principal component analysis
The principal component analysis was performed to investigate the clustering of the datasets. The transformed and normalized gene expression values were used and principal components were computed using the prcomp() function in R v3.6.2. The first and second largest variance components in the data (PC1 and PC2) were visualized as a scatter plot using ggplot2 v3.2.1 (https://cran.r-project.org/web/packages/ggplot2/index.html).
Scattered plot
Regression analysis was used to evaluate correlations between two groups. Normalized expression values of each group were transformed with natural log for better linear fitting and plotted against each other. R-squared value was calculated and shown on each regression line. All plots were produced in R v3.6.2 with the add-on package ggplot2 v3.2.1 (https://cran.r-project.org/web/packages/ggplot2/index.html).
Statistical Analysis
On-sided Student’s t-tests were used in the analysis of statistical significance of the difference in the differentiation index, the fusion index, EdU uptake, qRT-PCR, and ChIP-PCR data. The mean + or ± SEM obtained from biological triplicates with technical triplicates was shown in each graph unless stated otherwise.
Data availability
The RNA-seq data have been deposited to Gene Expression Omnibus (GEO) under the accession number of GEO: GSE150785. Previously published ChIP-seq data that were re-analyzed here are available under the accession numbers GSE25308 (Pol II, H3K4me1, H3K4me3, H3K27me3, and H3K9ac), GSE37527 (H3K27ac and p300), and GSE108650 (Bmal1). Other data that support the findings of this study are available upon reasonable request from the corresponding authors.
Acknowledgements
We thank James Staats, Brian Ruis, and Allison Keith for technical supports and Karyn A Esser for critical reading of the manuscript. We acknowledge Minnesota Supercomputing Institute, University of Minnesota Informatics Institute, and University of Minnesota Genomics Center for providing high-performance computing resources and the gopher-pipelines. S.L. and C.Y. were supported by the Minnesota Stem Cell Institute. A.A. was supported by the NIH (R01 AR062142 and R21 AR070319). N.K was supported by the NIH (R01 GM137603 and R21 AR076167), Regenerative Medicine Minnesota (RMM 101617 DS 004), and Grant-in-Aid of Research University of Minnesota (291987). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.
Footnotes
↵6 Lead Contact