BIOLOGICAL OVERVIEW

The quest for the gene responsible for regulation of muscle differentiation in Drosophila is a hot topic, as is the search for its vertebrate counterpart. Nautilus isn't the answer, since it is found only in somatic cell precursors, and not in heart muscle precursors. Tinman fails the test for the opposite reasons; it is found in heart precursors but not in skeletal muscle precursors. Does Mef2 provide an answer?

Mef2 is expressed earlier than nautilus, and it is expressed in presumptive heart and skeletal muscle precursors. It is dependent on twist , the gene that determines mesodermal fate. What prevents an otherwise neat answer is that nautilus and Mef2 expression are mutually independent of one another. Additionally, Mef2 mutants express mesodermal markers like even-skipped, that appear in precursors of one dorsal somatic muscle and a number of pericardial cells. There are fewer EVE-positive cells in Mef2 mutants but they don't disappear. nautilus expressing cells are present as well, but not in multinucleate syncytia, meaning that in Mef2 mutants nautilus expressing cells cannot mature to become muscle fibers.

Therefore, Mef2 is not necessary for the initial specification of muscle cells, but for their continued differentiation (Ranganayakulu, 1995). Muscle precursors undergo programmed cell death in Mef2 mutants, a sure indicator of defective differentiation (Bour, 1995).

mef2 and myoblast city mutations have been used to study requirements for neuromuscular junction (NMJ) formation during Drosophila development. myoblast city is required for myoblast fusion to establish syncytial muscles; its loss results in the formation of mononucleate muscles within a field of unfused myoblasts (Rushton, 1995). In myoblast city mutants, mononucleate muscles make functional neuromuscular synapses with correctly localized presynaptic active zones (presynaptic densities with clear synaptic vesicles). In mbc mutant embryos, as in wild type, only a fraction of the muscle surface is devoted to the NMJ. This fraction in mbc mutants represents a far smaller surface for the localization of active zones than the surface encountered by equivalent motorneurons in wild-type muscle, suggesting that muscles are capable of supporting a limited number of boutons, relative to their size. Motorneurons innervating the smaller mononucleate muscles in mbc mutant embryos fail to establish many active zones, as half the active zones are in axonal swellings not attached to muscle. It is clear that active zones are able to form in the absense of a NMJ.

In mef2 mutants, the pattern of muscle founder cells is largely normal in the absence of mef2 function. Myoblasts also fail to fuse but still attract appropriate innervation. Motorneurons that establish contact are always attracted by the correct founder cells, as judged by the branch pattern of innervation. There is no evidence of motor axons contacting any myoblasts other than those that become founders. These myoblasts, however, fail to differentiate into muscles. Concomitantly, active zones are present at frequencies comparable with wild type, however, nearly all of them are neurohemal, neuroneural, or on glial cells. Even in the rare instances where active zones are located at a neuromuscular contact, the pre- and postsynaptic membranes fail to form the tight apposition that is typical of synapse. In most cases the presynaptic active zones fail to localize at neuromuscular contacts. It is concluded that the localization of presynaptic active zones necessary to form a NMJ requires mef2-dependent muscle differentiation (Prokop, 1996).

In addition to essential myogenic functions, mutant Mef2 adult females are weakly fertile and produce defective eggs. Mef2 is expressed in nurse and follicle cells of the wild-type egg chamber. The Mef2 oogenic phenotype has been analyzed and it has been shown that the gene is required for the normal patterning and differentiation of the centripetally migrating follicle cells (CMFCs) that are crucial for development of the anterior chorionic structures. Mef2 alleles exhibit a genetic interaction with a dominant-negative allele of thick veins (tkv), which encodes a type I receptor of the Decapentaplegic-signaling pathway. TKV mRNA is overexpressed in Mef2 mutant egg chambers, and, conversely, forced expression of Mef2 represses tkv expression. These results indicate roles for Mef2 in the regulation of tkv gene expression and Decapentaplegic signal transduction that are essential for proper determination and/or differentiation of the anterior follicle cells. Mef2 is also expressed in both nurse and follicle cells. No defects have been observed in the germ line, either the number of germ cells or the location of the oocyte within the egg chamber. Therefore, a possible requirement for Mef2 in germ-line cells remains to be elucidated (Mantrova, 1999).

Mef2 appears to function in the somatic follicle cells, particularly in subpopulations of the oocyte-associated follicle cells (O-FCs), by negatively regulating TKV mRNA levels. It is not known whether Mef2 directly represses tkv gene transcription. Curiously, the expression patterns of Mef2 and TKV RNA are not mutually exclusive. Whereas Mef2 is expressed in all follicle cells, TKV is absent only in different populations of follicle cells at different times. Perhaps subtle changes in Mef2 levels can have different effects on tkv expression. For example, at stage 10A, Mef2 is more abundant in the leading CMFC than in the other O-FCs, whereas tkv is not expressed in CMFCs and expressed at a low level in the rest of the follicle cells. Alternatively, Mef2 may be a constitutive repressor of tkv, whereas other tissue-specific factors can counteract Mef2 and induce tkv expression (Mantrova, 1999).

In wild type, tkv expression is dynamic during oogenesis and appears to highlight a specific group of follicle cells, the leading front of the CMFCs. At stage 10A just before the commencement of centripetal migration, these cells form a ring marking the boundary between the oocyte and the nurse cell complex. After stage 10B, this ring of cells migrates inward until it reaches the border cells located at the center of the oocyte anterior. At stage 10A, tkv is expressed in O-FC but not in the leading CMFCs. This pattern is opposite that of the dpp expression pattern, which is highly expressed in the leading CMFCs but not in the rest of the O-FCs. It will be of interest to examine whether or not tissue-specific expression of dpp and tkv in the egg chamber is autoregulated by DPP signaling (Mantrova, 1999).

At stage 10B, tkv is expressed in the ventral half of the CMFC in addition to two short stripes in the dorsal region of the oocyte-associated follicular epithelium. This expression pattern appears to be complementary to that of the Egfr blocker argos, which forms a T-shaped pattern along the dorsal CMFCs and dorsal midline. argos expression is induced by the highest level of Egfr signaling; Egfr in turn, reduces the signaling strength by blocking the interaction between the receptor and its ligands. Thus, the initial graded distribution of Egfr signaling, extending laterally from the anterodorsal midline of the O-FCs, is transformed into two ridges of the Egfr-signaling level just lateral to the dorsal midline. These two ridges define the two lines of O-FCs that ultimately produce the two dorsal appendages. Interestingly, argos expression is diminished in the Mef2 mutant, consistent with the observed mutant egg chambers possessing broad and fused appendages. Although the notion is favored that argos expression is modulated by Mef2 through the action of Tkv, it cannot be ruled out that Mef2 may directly control the transcription of argos (Mantrova, 1999).

In addition to regulating the expression pattern of argos, Mef2 may play a more general role in modulating the Egfr-signaling level. This is suggested by the presence of Mef2 mutant egg chambers with reduced and fused dorsal appendages, a phenotype typical of hypomorphic Egfr-signaling pathway mutants. Indeed, reduced expression of Egfr-signaling components such as rhomboid has been observed in Mef2 mutants. More detailed and expansive studies are needed to elucidate the possible interaction between the Dpp- and Egfr-signaling pathways with Mef2 as a potential mediator (Mantrova, 1999).

Nevertheless, this report does demonstrate that the dpp-expressing CMFCs are poorly defined in D-mef2 mutant egg chambers. CMFCs are responsible for forming the operculum and, together with the border cells, specifying the construction of the micropile. Formation of these structures is also essential to closing the anterior end of the egg chamber. Because Dpp is critical for specifying anterior chorion production, the disrupted patterning of CMFCs in the Mef2 mutant may explain, at least in part, the chorion phenotypes observed (Mantrova, 1999).

Earlier studies have shown that Mef2 is a late-acting component of the genetic network controlling embryonic myogenesis. In the cardiac lineage, Mef2 is a direct target of Tinman, a homeodomain transcription factor essential for heart formation. tin function is required for the specification of cardiac precursor cells within the dorsal mesoderm, and its expression within this domain is induced by Dpp, produced by cells of the dorsal ectoderm. Additionally, Mef2 expression in a broader dorsal mesodermal domain is controlled by Dpp and the transcription factor Medea. Thus, Mef2 can be considered a downstream regulator within the Dpp-signaling pathway needed for cardiogenesis in the fly (Mantrova, 1999).

In the current study, it has been shown that Mef2 modulates the dissemination of a Dpp signal in the egg chamber through its control of tkv expression levels. Because multifunctional proteins often are members of conserved gene cassettes that function in different developmental processes, it is possible that Mef2 provides an additional critical function for Dpp signaling in the mesoderm. tkv is required for, and expressed during, inductive events occurring in the dorsal mesoderm. Likewise, Mef2 is present throughout the mesoderm at this stage and may have a comparable function in regulating tkv expression in mesodermal cells as has been elucidated in follicle cells. This could occur through a possible feedback regulatory loop from Mef2 to tkv. It will be important to investigate potential interactions of the two genes in the mesoderm, perhaps providing new insights into the specificity of Dpp signaling during Drosophila development (Mantrova, 1999).

A temporal map of transcription factor activity: Mef2 directly regulates target genes at all stages of muscle development

Dissecting components of key transcriptional networks is essential for understanding complex developmental processes and phenotypes. Genetic studies have highlighted the role of members of the Mef2 family of transcription factors as essential regulators in myogenesis from flies to man. To understand how these transcription factors control diverse processes in muscle development, chromatin immunoprecipitation analysis was combined with gene expression profiling to obtain a temporal map of Mef2 activity during Drosophila embryonic development. This global approach revealed three temporal patterns of Mef2 enhancer binding, providing a glimpse of dynamic enhancer use within the context of a developing embryo. These results provide mechanistic insight into the regulation of Mef2's activity at the level of DNA binding and suggest cooperativity with the bHLH protein Twist. The number and diversity of new direct target genes indicates a much broader role for Mef2, at all stages of myogenesis, than previously anticipated (Sandmann, 2006).

To identify enhancer regions bound by Mef2 in vivo, chromatin immunoprecipitation followed by microarray analysis (ChIP-on-chip) was performed at five consecutive time points of embryogenesis spanning key stages of muscle development. To systematically identify Mef2 bound genomic regions in an unbiased manner, a Drosophila genomic tiling array was constructed, taking advantage of genomic clones generated by the Berkeley Drosophila Genome Project (BDGP) to sequence the Drosophila genome. The array consists of overlapping 3 kb fragments tiling across ~50% of the Drosophila genome (Sandmann, 2006).

These experiments identified 1015 significantly enriched Mef2 bound fragments, with less than 1% estimated false positives. Due to the overlapping nature of the array, this represents 670 nonoverlapping genomic regions bound by Mef2 at one or more developmental time points. To assess the quality of this data set, it was determined if regions previously reported to be bound by Mef2 were recovered. Eight of the previously characterized Mef2 direct target genes are covered by the arrays. Mef2 binding was identified in the proximity of all eight genes assayed: Actin 57B, Muscle LIM protein at 84B and Muscle LIM protein at 60A, β-tubulin60D, Tropomyosin I, inflated, mir-1, and Mef2 itself. In many cases, the study not only identified the previously reported Mef2 bound enhancer, but also identified additional ones (Sandmann, 2006).

Five genes are known to be genetically downstream of Mef2, though the mechanism of regulation remains unclear: Myosin heavy chain, meso18E, muscleblind, nautilus, and Chorion factor 2. The results show Mef2 binding to genomic regions close to four of these genes (Mhc, mbl, nau, and meso18E) and to a genomic region further 5′ of CF2. This indicates that these genes are directly regulated by Mef2 and identifies the location of at least one of their enhancer regions. In summary, the successful identification of Mef2 binding in the vicinity of all known and suspected target genes suggests a high accuracy of the approach taken. In addition to the 8 known enhancers, over 650 new Mef2 bound regions were identified (Sandmann, 2006).

As a complementary approach to assess the molecular function of Mef2, which genes depend on Mef2 activity for their correct expression during embryonic development was determined. The gene expression profile of wild-type embryos was compared to that of stage-matched Mef2 mutant embryos throughout a developmental time series. Pure populations of Mef2 homozygous mutant embryos were isolated. For each developmental time point assayed, four independent replicates were analyzed on microarrays containing at least one probe for every predicted gene in the genome (Sandmann, 2006).

The expression profiles of Mef2 mutant embryos was examined at 11 consecutive 1 hr time points of embryogenesis, spanning from 5 to 16 hr of development (stages 9–16). An additional time point was added at 18–19 hr, stage 17, to identify genes that are expressed in differentiated muscle. This allowed the generation of a high-resolution map of Mef2-dependent temporal changes in gene expression, spanning the stages of mesoderm subdivision, myoblast specification, myoblast fusion, and the initiation of terminal muscle differentiation (Sandmann, 2006).

This study identified 700 genes with significant changes in gene expression in Mef2 mutant embryos at 2 or more consecutive time points. The eight known Mef2 protein-coding target genes are among them, with Act57B, Mlp84B, Mlp60A, and TmI showing a greater than 4-fold decrease in expression at multiple stages of development. The differentially expressed genes are significantly enriched in genes expressed in muscle, indicating that many of the misregulated genes are expressed in the same cells as Mef2 (Sandmann, 2006).

The function of a substantial number of the differentially expressed genes is unknown. This study indicates that they are genetically downstream of Mef2 as either direct or indirect targets, and provides a useful resource to identify genes likely to be involved in muscle development. The combination of ChIP-on-chip results with this expression profiling data allowed determination of which genes are directly regulated by Mef2 (Sandmann, 2006).

Genomic tiling arrays provide an unbiased method to identify new regulatory regions independent of their distance to the gene. While this offers a great advantage over promoter arrays, it raises a new challenge for ChIP-on-chip studies: how to accurately match transcription factor bound regions to their correct target genes. Metazoan enhancers have been identified at large distances from their target genes, including within introns of neighboring loci. Assuming that the enhancer is regulating the closest proximal gene will, especially in gene-dense regions, often cause the wrong target gene to be selected (Sandmann, 2006).

Different sources of metadata were used to systematically link ChIP-enriched regions to their target genes. The genes in the vicinity of each Mef2 bound region received a cumulative score based on: (1) the distance between a gene and a Mef2 bound region, (2) a change in expression in Mef2 mutant embryos, and (3) supporting information, for example about the gene's expression patterns (BDGP in situ database, Flybase, literature). Genes were not assigned based on proximity alone. Using this approach, 211 Mef2 direct target genes were identified with high confidence, including all known targets covered by the tiling array (Sandmann, 2006).

To estimate the accuracy of the automated gene assignment, a collection was used of characterized enhancers from single gene studies. A total of 33 of the Mef2 bound regions assigned to target genes with a high confidence score overlap with a previously identified gene's enhancer. In 28 cases (84.8%), the regulated locus was correctly identified, illustrating the accuracy of the gene assignment strategy. The remaining five sequences map to the Enhancer of split region and were assigned to a different member of this gene cluster. A total of 29 additional Mef2 bound regions overlapped with known enhancer regions. The 12 regulated target genes were not assigned by using the automated approach, since no additional supporting evidence was available. Combining the automatic assignments with information about known regulatory relationships yielded 234 unique genes directly regulated by Mef2 (Sandmann, 2006).

Rather than potentially making an incorrect gene assignment, the remaining 574 Mef2 bound fragments were not assigned to genes, although they are equally significantly enriched with a stringent statistical cutoff. The vast majority (87.3%) of unassigned Mef2 bound fragments were in an intron or within 5 kb of one or more genes; therefore, they likely contain regulatory modules. To enable other researchers to make their own gene assignments, a searchable web site was created, displaying all Mef2 bound regions in their genomic context together with the results from the expression profiling experiments (Sandmann, 2006).

Of the 1015 Mef2 bound regions, 62 were previously identified as active enhancers. The vast majority of these regulatory regions were not known to bind to Mef2, but they have been shown to bind to a number of other transcription factors in vitro, revealing interesting insights into combinatorial gene regulation with Mef2. For example, Mef2 binding to an Antp bound enhancer region of the apterous and teashirt genes was identified. Many of these regions were shown to function as muscle enhancers in vivo, providing additional evidence that the identified Mef2 bound regions are active enhancers. To test if additional Mef2 bound regions can function as enhancers in vivo, five regions were selected with representative levels of enrichment in the ChIP-on-chip experiments that were assigned to target genes with known expression. This allowed evaluation of their ability to drive reporter gene expression in a pattern similar to that of the target gene. Using conservation in different Drosophila species as a guide, regions between 0.4 and 2.5 kb within the bound fragments were assayed (Sandmann, 2006).

All five Mef2 bound regions tested were able to drive GFP expression in Mef2-expressing cells. The enhancer region of pnt initiates expression of GFP early in the mesoderm, at stages 9–10. This mirrors the expression of the pnt transcripts in wild-type embryos. Since the GFP protein has a long half-life in Drosophila embryos, it can subsequently be detected in differentiating myoblasts. The enhancer regions of CG14687 and CG5080 initiate GFP expression slightly later. The enriched region upstream of CG14687 is sufficient to direct expression of GFP initially in the visceral muscle (stage 11) and later in the somatic muscle (stage 12), closely resembling the spatial and temporal expression pattern of the gene itself (Sandmann, 2006).

Myosin light chain 2 (Mlc2) is a target of Mef2 proteins in vertebrates. The data show that this regulation is conserved in flies. A Mef2 bound region 5′ of the Mlc2 locus reproduces the gene's expression, driving GFP expression in differentiating somatic muscle cells from stage 13 onward. Finally, the endogenous expression of CG9416 initiates in the longitudinal visceral muscle precursors at stage 10 and in the somatic muscle at stage 13. A 372 bp enhancer region is sufficient to direct GFP expression in both tissues, reproducing the full expression pattern of CG9416 (Sandmann, 2006).

In summary, all five Mef2 bound genomic regions are sufficient to direct reporter gene expression resembling the temporal and spatial patterns of the respective predicted target gene. These results indicate that the ChIP-on-chip approach was very successful in identifying new muscle enhancers in vivo, and that the assignment of target genes to Mef2 bound regions is accurate (Sandmann, 2006).

Many of the identified direct target genes are misexpressed in Mef2 mutant embryos, showing a requirement for Mef2 for their normal expression. To determine if Mef2 is sufficient to regulate its target genes in nonmuscle cells, Mef2 was ectopically expressed in the ectoderm by using the UAS/Gal4 system. As putative cofactors are likely to be absent, this assay is a very stringent test of a regulatory relationship. Remarkably, Mef2 could induce ectopic expression of 5 of the 13 genes tested in nonmuscle cells. Overexpression of Mef2 in engrailed stripes in the ectoderm is sufficient to cause ectopic expression of Him, CG9416, and CG30080. Overexpression of Mef2 with lmd, a transcription factor known to regulate Mef2, is sufficient to drive ectopic expression of CG5080 and delilah in the ectoderm. Ectopic expression of either transcription factor alone could not induce expression of these genes, suggesting that Mef2 and Lmd cooperatively regulate their expression and are sufficient to do so in nonmuscle cells. These results, in combination with the expression profiling data, confirm that Mef2 is essential and sufficient for the expression of a large percentage of its target genes (Sandmann, 2006).

While Mef2 is found in all muscle cells from gastrulation to the end of embryogenesis, its known target genes show temporally and spatially different expression patterns. For example, Act57B and β3-tub60D RNA are not transcribed until about stage 11, while Mhc and Mlp84B RNAs are not detectable until stages 13–14. Moreover, while Mef2 is expressed in the entire myogenic lineage, some of its known targets are expressed in smaller subsets of cells (Sandmann, 2006).

Clearly, there must be additional ways to control Mef2's regulatory activity. To determine if regulation occurs at the level of DNA binding, the temporal information from the ChIP time course was used to investigate if there are distinct patterns of Mef2 enhancer occupancy. K-means clustering analysis was used to subdivide the 1015 enriched genomic regions according to their temporal profile of Mef2 binding. Three major groups of temporally bound enhancers were identified (Sandmann, 2006).

Binding to the first group of enhancers was initially detected at 4–6 hr of development, after which Mef2 remained bound through the three subsequent developmental time points assayed. This group, representing almost half of the enriched fragments, suggests that these enhancers remain occupied by Mef2 throughout development. This temporal binding pattern of Mef2 matches its broad expression during all stages of muscle development (Sandmann, 2006).

The second group, representing 21% of the enhancers, was bound by Mef2 at 4–6 hr of development, but it was not bound at later developmental time points. Since Mef2 continues to be expressed, and is capable of binding to other enhancers, the transient occupancy of the early enhancers demonstrates that Mef2's ability to bind to DNA is tightly regulated (Sandmann, 2006).

The third group, containing 32% of the enhancers, is only bound by Mef2 late in development, with maximal binding at the last time point assayed. This group contains enhancers for many genes involved in late aspects of muscle differentiation, e.g., Mhc, Mlc1, Mlc2, TmI, TmII, act57B, β3-tub60D, Mlp60A, Mlp84b, Mp20, mbl, and wupA. Although Mef2 protein is present at high levels early in development, it does not occupy these enhancers until much later, demonstrating additional specificity in the regulation of Mef2 binding (Sandmann, 2006).

An investigation examined whether the temporal binding of Mef2 to a target gene's enhancer coincides with the onset of that gene's expression. Remarkably, the first time point when Mef2 binds to an enhancer is significantly correlated with the onset of that gene's expression during embryogenesis. While this trend holds for all time points assayed, the correlation is particularly strong for late bound enhancers, mirroring the coordinated expression of late muscle differentiation genes. Although additional levels of “post binding” regulation cannot be excluded, these results demonstrate that Mef2's DNA binding is tightly regulated and is a trigger for target gene expression (Sandmann, 2006).

These results provide the first evidence that, while Mef2 is broadly expressed during muscle development, its ability to bind to DNA is temporally regulated depending on the context of the enhancer. This finding is intriguingly similar to what has been shown for MyoD in fibroblasts and Pha-4 in C. elegans. While both transcription factors have broad temporal expression, they regulate temporally restricted enhancers. This highlights a potentially general mechanism for encoding spatiotemporal specificity within the context of a regulatory region (Sandmann, 2006).

To explain the three temporal patterns of Mef2 binding, regulatory motifs were sought within each group of enhancers. It was first determined whether the number of Mef2 sites was equally distributed between the three temporal groups of Mef2 bound enhancers. Interestingly, enhancers in the continuously bound and the late bound groups were significantly enriched in single and multiple Mef2 sites per fragment. This significant enrichment of Mef2 sites is conserved in the orthologous sequences of the related species, Drosophila pseudoobscura. In contrast, the early cluster of transiently bound enhancers contains a similar number of Mef2 sites as the rest of the genome (Sandmann, 2006).

The bHLH transcription factor Twist is essential for all aspects of early mesoderm development in Drosophila. Twist has a transient expression pattern, with peak expression at stage 11 (~6–7 hr) mirroring the peak binding of Mef2 to the early enhancers. In vitro studies have shown cooperative regulation between vertebrate Mef2 proteins and bHLH transcription factors via direct protein-protein intereactions (e.g., MyoD and Hand). Given the temporal expression of Twist protein and the ability of vertebrate Mef2 proteins to bind to bHLH proteins, cooperative binding of Mef2 and Twist is an attractive model to explain the transient Mef2 binding to the early group of enhancers (Sandmann, 2006).

Two lines of evidence indicate that this hypothesis is correct. First, Twist sites are significantly enriched in the early bound enhancers, and not in the continuous and late bound enhancers. This significant enrichment of Twist sites is conserved in the ortologous sequences of D. pseudoobscura. Second, Twist and Mef2 cobind to the early enhancers at the same stage of development. ChIP-on-chip studies were performed of Twist at 4–6 hr of development. Comparing the results of both studies showed in vivo binding of Twist to a large percentage of the early enhancers, demonstrating that Mef2 and Twist can cooccupy the early enhancers. Five examples were given of Mef2 bound early enhancers cobound by Twist at 4–6 hr of development (Sandmann, 2006).

While bHLH proteins have a central role in muscle development in all species examined to date, the predominant roles of individual family members have diverged. In vertebrates, MyoD family members play a central role in activating muscle gene expression, while Twist represses myogenesis. In Drosophila, the only MyoD family member, nau, is not essential for general muscle development. It has been speculated that Twist is the central bHLH regulator of Drosophila muscle development, as it is sufficient to activate the myogenic program upon ectopic expression in the ectoderm. The results provide further evidence for the evolutionary similar roles of Twist in flies and vertebrate MRFs; Drosophila Twist and Mef2 proteins may cooperatively regulate muscle gene expression in a similar manner to MyoD and Mef2 proteins in vertebrates (Sandmann, 2006).

Embryos with loss-of-function mutations in Drosophila Mef2 show a block of myoblast fusion and lack expression of a number of contractile muscle proteins. This study identified a number of Mef2 target genes involved in both processes; e.g., blow and lmd, two genes essential for myoblast fusion, as well as numerous cytoskeletal proteins (Sandmann, 2006).

In addition to these severe phenotypes, defects in neuromuscular junction (NMJ) formation and muscle attachments have been observed in Mef2 mutant embryos, the molecular basis of which is not understood. A significant enrichment was found of Mef2 target genes involved in both processes, providing a molecular understanding of the observed phenotypes (Sandmann, 2006).

The results also indicate that Mef2 is required for muscle function in differentiated myofibers. A number of Mef2 target genes involved in muscle energy production or storage were identified: Pfrx, GluRIIA, GlyP, Gpdh, Glycogenin, Pgi, and Pgk. This role may be further strengthened by the direct regulation of Ptx1, a transcription factor thought to regulate muscle physiology (Sandmann, 2006).

Drosophila body wall muscles are formed from progenitor cells that are selected through the action of Ras signaling and Notch-Delta lateral inhibition. The data show in vivo binding of Mef2 to enhancer regions of a striking number of genes that are essential for this process, in addition to expression changes for some genes in Mef2 mutant embryos. This includes components of the Notch-Delta pathway (Delta, mam, bib, E(spl) complex, and Neur), which is essential for specification of the somatic muscle and heart. Since these experiments were analyzed with tiling arrays covering 50% of the genome, Mef2 likely regulates even more genes involved in these signaling pathways, which could not be identifyied. This is in agreement with Junion (2005), who identified sfl, spen, and argos as Mef2 targets (Sandmann, 2006).

Once specified, founder cells express a characteristic set of transcription factors called identity genes. In vivo binding of Mef2 to enhancer regions of eight of the ten known muscle identity genes was identified. Importantly, the temporal binding of Mef2 correlates with the initiation of identity gene expression in founder cells (stages 9–11; time points 4–6 and 6–8 hr). Mef2 also binds to enhancer regions of a number of genes previously reported to be enriched in founder cells, suggesting that Mef2 acts in concert with the identity genes to regulate the transcriptional program within these myoblasts (Ubx, htl, CG14207, CG9520, CG17492, CG8417, and krT95D in the VO5 muscle (Sandmann, 2006).

Genetic studies looking at the interplay between motor neuron and muscle development observed a consistent reduction in the number of somatic muscles in Mef2 mutants. It has not been clear if the loss of muscles is a secondary defect due to a general failure of the muscle to differentiate, and the molecular mechanism was not been explored. The current results indicate that Mef2 directly regulates founder cell identity gene expression. It is suggested that Mef2 provides an extra layer of regulation to buffer this key step in muscle development from stochastic fluctuations in the levels of key regulators. When this regulation is absent in Mef2 mutant embryos, founder cells are not specified or maintained in a robust manner, leading to the observed low penetrance loss-of-muscles phenotype (Sandmann, 2006).

Traditionally, Mef2 is placed toward the bottom of the myogenic transcriptional hierarchy, due to its well-characterized role as a regulator of muscle effector proteins. Because of this, it was surprising to find that Mef2 regulates a large number of transcription factors, many of which are involved in early aspects of myogenesis. For example, Mef2 regulates transcription factors essential for visceral muscle development (bap, slp1, HLH54F) and cardioblast specification (nmr2, Zfh1, ush). These results implicate a role for Mef2 in the subdivision of the dorsal mesoderm and place Mef2 at the center of the transcriptional program required for Drosophila muscle development (Sandmann, 2006).

Vertebrate Mef2 proteins regulate the expression of the transcription factors MyoD and cJun. Direct regulation was found of the orthologs of these genes (nau, jra) by Drosophila Mef2, indicating that Mef2 is part of an evolutionary conserved genetic program. This suggests that many of the additional transcriptional connections identified in Drosophila may also be conserved (Sandmann, 2006).

One of the most surprising findings of this study was the large number of enhancer regions that are bound by Mef2. Previous studies searching for Mef2 targets focused on a limited part of the genome. The present study provides a unique opportunity to get a more accurate view of the total number of Mef2-regulated enhancers and genes. Using a tiling array covering ~50% of the genome, at least 670 unique Mef2 bound genomic regions were found. A total of 600 of these enhancer regions are within 5 kb of a gene locus and are therefore likely to represent active enhancers of a gene. Extrapolating to the whole genome, this indicates that Mef2 regulates as many as 1000 genes during the course of muscle development (Sandmann, 2006).

The current view of Mef2 in the literature is of a transcription factor required late in development for muscle differentiation. However, given the diversity of Mef2 target genes, its ability to regulate genes during all stages of muscle development, and the huge number of enhancer regions bound by Mef2, the view of the function of the transcription factor needs to be adjusted. Mef2 is likely to bind to enhancer regions of most, if not all, muscle genes, not just structural muscle proteins, and may thereby act as a “general” muscle transcription factor. The presence of feed-forward loops regulated by Mef2 as well as of low-penetrance phenotypes in Mef2 mutants suggests that one of the functions of this transcription factor is to provide robustness within the myogenic program (Sandmann, 2006).

Simultaneous cellular and molecular phenotyping of embryonic mutants using single-cell regulatory trajectories

P>Developmental progression and cellular diversity are largely driven by transcription factors (TFs); yet, characterizing their loss-of-function phenotypes remains challenging and often disconnected from their underlying molecular mechanisms. This study combined single-cell regulatory genomics with loss-of-function mutants to jointly assess both cellular and molecular phenotypes. Performing sci-ATAC-seq at eight overlapping time points during Drosophila mesoderm development could reconstruct the developmental trajectories of all major muscle types and reveal the TFs and enhancers involved. To systematically assess mutant phenotypes, a single-nucleus genotyping strategy was developed to process embryo pools of mixed genotypes. Applying this to four TF mutants could identify and quantify their characterized phenotypes de novo and discover new ones, while simultaneously revealing their regulatory input and mode of action. This approach is a general framework to dissect the functional input of TFs in a systematic, unbiased manner, identifying both cellular and molecular phenotypes at a scale and resolution that has not been feasible before (Secchia, 2022).

The continuous temporal resolution of the time course was exploited to reconstruct regulatory trajectories for three muscle lineages, starting from unspecified mesodermal cells. Ordering cells along pseudotime revealed extensive and dynamic temporal changes in accessibility for both regulatory elements and genes along each lineage's trajectory. The loci of many upstream identity genes change in accessibility (in both directions) as the development of each lineage progresses. This includes lmd, kah, NK7.1, Pdp1, tx, and nau (dMyoD) in the somatic trajectory, in addition to more downstream effector genes required for differentiated muscle function (Mhc, Mlc, and Tropomyosin). Cardiomyocytes are specified by a highly conserved set of TFs from flies to humans, including members of the NKx2.5 (tinman [tin] in Drosophila), GATA (pannier [pnr]), T-box (Dorsocross-3 [Doc3]), and islet 1 (tailup [tup]) TFs. The dynamic usage of all factors and many more are observed along the cardiac lineage. Both the somatic and visceral muscles are formed from two populations of cells-founder cells (FCs), which give the muscle its identity, and fusion-competent myoblasts (FCMs), which fuse to FCs during differentiation to form a multinucleated syncytium. Two trajectories were observed for the visceral-muscle lineage. Starting from common precursors, one lineage branches toward the somatic body wall muscle, while the other has a distinctive lineage that remains separate from the rest of the muscle populations (Secchia, 2022).

To explore the visceral-muscle lineages further, all annotated visceral-muscle cells from the Mef2+ time course were combined with the Bin+ FAC sorted cells, and these these 4,187 visceral-muscle cells were reclustered separately, revealing a more complex multi-branched structure. The visceral muscle represents a collection of muscles with different developmental origins. The FCs and FCMs of the circular trunk visceral muscle (CVM) are specified at stages 10 and 11 (6-8 h), after which they migrate laterally and undergo myoblast fusion (stage 12) to form a continuous muscle that encloses the gut. The longitudinal VM (LVM) is formed from FCs in the caudal mesoderm toward the end of the embryo, which migrate on top of the circular VM, using it as a scaffold. The single-cell trajectory captures these diverse origins and the temporal delay of the LVM development. The main branch consists of both FCs and FCMs, as seen by their enrichment in markers for visceral-muscle FC specification (including Alk, numb, bap, and bin) and FCM specific genes (lmd and sns). A subset of cells progress to a more differentiated state, expressing specific myoblast-fusion genes in the intermediate state before differentiating to contractile muscle, indicated by the expression of sarcomere protein genes such as MHC, Tmod, Tm1, and Tm2. Given the timing of these transitions and the presence of bap (a TF not expressed in LVM), this likely reflects circular VM development. The LVM and hindgut VM (HVM) branches have different developmental origins, as expected, and are enriched in longitudinal (cluster 5) and hindgut (cluster 4) visceral-muscle markers. The precursor population has a second branch, expressing a broader set of myoblast-fusion genes, perhaps representing a more mature or more diverse FCM population. Some FCMs were proposed to remain unfused and wait for the LVM founder cells to migrate over the circular VM and then fuse during late stage 12 and stage 13 (~10-12 h). The results support this and suggest that a proportion of these cells may also fuse to the overlying somatic-muscle FCs, given their expression of Vrp1/wip and wasp, two genes specific to myoblast fusion of the somatic muscle, and their progressive acquisition of a chromatin state similar to the somatic muscle (Secchia, 2022).

To determine the functional impact of TF mutants on both the developmental trajectories of cells and their regulatory programs, sci-ATAC-seq to was applied to loss-of-function mutant embryos of four mesodermal TFs and assessed the cells' behavior by integrating them with the wild-type developmental trajectory (Figure 1D). Myocyte enhancer factor 2 (Mef2) is essential for myoblast fusion and terminal differentiation of all muscle types. Tinman (Nkx2-5) is essential for the subdivision of the dorsal mesoderm into cardiac and visceral muscle, through the activation of bagpipe (NKx3-2), which initiates biniou (FoxF2) expression and the visceral-muscle lineage. Although the occupancy of each TF has been examined in bulk, their contribution to enhancer accessibility and to an individual cell's state, remains unknown (Secchia, 2022).

Recessive lethal mutants in animal models must be maintained as heterozygotes; therefore, only 25% of offspring embryos are homozygous for the mutation of interest. This study first developed an easy and generalizable approach to profile single-cell genomic measurements from embryos of mixed genotypes. Rather than genotyping and hand sorting homozygous mutant embryos, all embryos were retained representing a pool of three genotypes (homozygous loss-of-function mutant, heterozygous, and homozygous nonmutant), and sci-ATAC-seq was performed on their dissociated pooled nuclei. In Drosophila, such heterozygous mutants are maintained over balancer chromosomes. The genetic background of the loss-of-function mutants and the balancer chromosome was sequenced, allowing each nucleus to be genotyped based on informative SNPs in the sci-ATAC reads. In contrast to standard allelic imbalance studies, this requires relatively few informative reads per nucleus, as described below (Secchia, 2022).

The characterized mutations of these TFs were generated over 20 years ago and will have accumulated additional mutations that could impact chromatin accessibility independently of the TFs' function. To circumvent this, CRISPR-Cas9 editing with single-stranded oligo donors (ssODNs) was used to recreate the characterized loss-of-function mutations for each factor in a common isogenic genetic background. These new alleles were sequence verified and recapitulate the expected characterized muscle phenotypes. As essential factors, they are homozygous lethal, and, importantly, the new alleles do not complement the characterized loss-of-function allele when placed in trans, confirming that the lethality is due to the mutation of the TF and not a CRISPR off-target effect (Secchia, 2022).

To phenotype mutants, staged embryos were collected from the heterozygous adults (mutantCRISPR/balancer chromosome), which were formaldehyde fixed and processed for sci-ATAC-seq. The dissociated nuclei thereby come from a pool of F1 embryos which represent 25% homozygous loss-of-function mutant/mutant, 50% heterozygous mutant/balancer, and 25% homozygous balancer/balancer. After sci-ATAC-seq, each nucleus was genotyped de novo based on the fraction of reads mapping to the mutant or balancer chromosomes. The genotype assignment is based on over 450,000 genetic variants between the balancer and the mutants genetic background. With a median of roughly 1,000 variants covered per cell, it was possible to genotype 99.9% of all theoretically assignable nuclei with high confidence (>0.9 posterior probability). The genotype assignments were very robust, being 98% identical when performing the assigments using two sets of variants identified with a stringent or a more lenient filtering threshold (Secchia, 2022).

De novo single-nucleus genotyping approach has a number of advantages for single-cell profiling of mutants. It eliminates the need, and associated experimental time, to hand select embryos of the correct genotype and is therefore faster and more reliable. Profiling nuclei from homozygous and heterozygous siblings in the same experiment has an additional advantage to aid in batch correction. As the heterozygous nuclei are essentially wild type, they can be used to align mutant data from the same batch to the wild-type reference trajectory, avoiding 'over fitting' by batch aligners of biologically real mutant phenotypes (Secchia, 2022).

Mef2 regulates differentiation of all major muscle types. To determine the functional impact of Mef2 on chromatin accessibility and mesodermal cell fate, sci-ATAC-seq was performed on Mef2 mutant embryos (a pool of homozygous and heterozygous) at 10-12 h (mainly stage 13) during the initiation of terminal muscle differentiation in the somatic and visceral muscle. The de novo genotyping strategy assigned the expected proportion of profiled nuclei as homozygous mutant (expected: 25%, observed: 26%) (Secchia, 2022).

As these experiments were performed on whole embryos, a first round of clustering was performed to identify muscle cells. The profiled 12,926 cells cluster by cell type, rather than by genotype, revealing 8 broad cell states, including one large muscle population. Selecting the muscle cells, genotype and cluster-specific peaks were identified and those 2,567 cells were reclustered. This resulted in five cell clusters with distinct chromatin accessibility, three of which could be identified as somatic, cardiac, and visceral muscle. Digitally genotyped Mef2-/- mutant nuclei are almost completely absent from the somatic-muscle cluster and instead are highly enriched in two additional 'muscle clusters,' which appear close to, but distinct from, somatic-muscle cells (Mutant1 and Mutant2 clusters). Mutant1 is composed of 89%, while Mutant2 56%, Mef2-/- cells. This indicates that in the absence of Mef2, mesodermal cells are unable to establish the regulatory landscape to become somatic muscle, and instead form a new altered state. Cells in the Mutant2 cluster have lower (~2.2-fold) coverage than the other clusters. This may represent cells in a more naive state or undergoing apoptosis, although the clustering of these cells might be driven primarily by their lower coverage. The proportion of homozygous mutant cells is also partially reduced in the visceral muscle, while it is largely unaffected in the heart, at these stages. These tissue-specific differences likely reflect differences in the timing of differentiation between muscle lineages. Terminal differentiation of the somatic and visceral muscle begins after myoblast fusion at ~10-12 h (stage 13), while it cannot occur in the heart until later stages, after dorsal closure is complete at ~13 h (stage 15) (Secchia, 2022).

The somatic and visceral muscle were also the two tissues that displayed the highest relationship between Mef2 binding at 10-12 h and open chromatin signatures (chromVAR scores) in wild-type embryos, suggesting that such scores are a good indicator of the tissues and time points that will most likely be affected by the removal of a TF. For example, while Mef2 is also expressed in the cardiac muscle, Mef2-bound regions are generally less accessible compared with the somatic and visceral muscle and, concordantly, the proportion of mutant cells remains close to normal in the cardiac muscle. However, such computational inference from wild-type embryos cannot predict all mutant phenotypes. While the visceral and somatic muscle display similar accessibility (ChromVar scores) at Mef2-bound regions in wild-type embryos, the mutant analyses revealed clear differences in the response of both tissues to loss of Mef2; mutant cells are almost completely absent in the somatic muscle but only partially reduced in visceral, indicating a difference in Mef2 dependency for accessibility between these two tissues, which was not evident in wild-type conditions (Secchia, 2022).

To experimentally test the accuracy of the de novo genotyping strategy, homozygous mutant embryos were hand sorted from the Mef2 mutant based on a GFP-marked balancer chromosome, and sci-ATAC-seq was performed on these 100% Mef2-/- nuclei. The hand-sorted mutant nuclei show the same properties as the genotyped mutant nuclei; they are absent from the somatic muscle and accumulate in two mutant clusters-mainly in Mutant1, where 88% of the homozygous mutant cells reside, similar to the digital genotyping above. Both the genotyped and hand-sorted homozygous mutant nuclei display the same alterations in chromatin accessibility at individual loci, as shown for two muscle contractile proteins, Mlc1 and Msp300. Both genes have multiple Mef2-bound regions overlapping open chromatin in cells from the somatic cluster, which are almost completely closed in both the digitally genotyped and hand-sorted Mutant1 cells. Concordant gains in accessibility were also observed at regulatory regions in Mef2-/- cells, including the enhancer VT30021, which is embryonically active, but normally not in muscle tissues. This proof of principle indicates that the de novo nuclear genotyping strategy correctly assigns homozygous mutant nuclei (Secchia, 2022).

The whole-embryo single-cell data allowed exploration of whether Mef2 mutant cells adopt another cell state, either from within the mesoderm or another germ layer. To assess this, clusterwise accessibility correlations of the mutant clusters against all cell types (both mesodermal and nonmesodermal cell clusters) were computed in embryos at 10-12 h. Both Mutant1 and Mutant2 are most highly correlated to clusters within the myogenic mesoderm, in particular the somatic muscle, and are clearly separated from the nonmyogenic mesoderm, ectoderm, and endoderm lineages, indicating that these cells are specified to become muscle, but appear blocked in their development. To determine if they are stuck in an earlier myogenic state, the mutant cell data was combined (combining the digitally genotyped and hand-sorted mutant cells, given that they appear identical) and the heterozygous cells with all cells in the wild-type reference trajectory and reclustered the data. The heterozygous cells behave indistinguishably from the reference cells, falling within the expected wild-type populations on the trajectory. In contrast, Mef2-/- Mutant1 and Mutant2 cells cluster separately, off the wild-type muscle trajectory, but roughly at the appropriate 'temporal' time point. If these cells were blocked in their developmental progression, they would be expected to cluster on the trajectory at some earlier time point in muscle development, which is not what was observe. To explore this further, clusterwise accessibility correlations were computed of the mutant and muscle clusters for each time point, which confirmed that both Mutant1 and Mutant2 are progressing to the appropriate developmental stage. This indicates that Mef2 mutant cells are not simply immature muscle cells but rather have developed a new abnormal 'muscle-like' state, which is likely defined by the inactivation or decreased expression of late muscle function genes, combined with the inappropriate activation of nonmuscle enhancers and genes. Taken together, this suggests that Mef2 is not only required as a differentiation factor to regulate the expression of muscle contractile genes, but also to prevent muscle cells from undergoing other cell-state changes (Secchia, 2022).

The same approach was applied to three other loss-of-function mutants for TFs involved in the specification of the dorsal mesoderm (tinman) and its derived visceral muscle (bagpipe and biniou) that forms the gut musculature. These TFs have a hierarchical relationship between them, where Tinman regulates Bagpipe expression at stage 10 (6-8 h), which in turn regulates Biniou expression. To examine the function of these TFs, bagpipe and biniou mutants were assessed at 6-8 h of development, which coincides with the initiation of their expression and with the specification of the visceral muscle (stages 10 and 11). As Tinman acts upstream of both bap and bin, = the time window was shifted 1 h earlier (5-7 h; stage 9, 10) to capture these events. For all three mutants, sci-ATAC-seq was performed on a pool of homozygous and heterozygous embryos, as above. Staged bagpipe and biniou mutant embryos were collected at 6-8 h (late stage 10, mainly stage 11) and the mesodermal population isolated by Mef2 FAC sorting, as in the wild-type trajectory, obtaining high-quality profiles for 6,306 and 5,833 mesodermal cells, respectively. Presorting for the mesodermal population was not possible for tinman, as it regulates Mef2 expression. Therefore sci-ATAC-seq was assessed on whole embryos of tinman mutants and a first round of clustering was performed to identify 6,786 high-quality mesodermal cells (Secchia, 2022).

To assess the fate of the mutant cells, their development was directly compared with the wild-type trajectory by co-clustering the combined mutant data, representing 18,925 homozygous and heterozygous cells, together with the wild-type mesoderm time course, correcting for batch-level effects. Reclustering and reannotation of this joint dataset , representing 40,232 cells, revealed a structure that is generally consistent with the wild-type trajectory. Nuclei from heterozygous mutant cells (+/-) for tinman are present in the cardiac and early visceral-muscle clusters. Similarly, bagpipe and biniou heterozygous mutant cells are present in the visceral mesoderm, spanning both branches, and extending to later stages of embryogenesis (Secchia, 2022).

Examining the homozygous mutant nuclei (-/-) revealed that, in contrast to Mef2, the proportion was significantly lower than the expected 25%, representing 15%, 19%, and 17% for tinman, bagpipe, and biniou, respectively. This indicates that a proportion of homozygous mutant cells is not maintained and likely undergoes apoptosis as they cannot progress in their development. The trajectories of the remaining mutant cells (-/-) are very different from their heterozygous siblings; tinman -/- cells are completely absent from the cardiac lineage and late-stage visceral muscle, with few remaining cells in the early visceral-muscle clusters. Moreover, there is a significant reduction of homozygous mutant cells in late mesoderm stages, which likely represents the dorsal mesoderm. Similarly, the bagpipe and biniou homozygous mutant nuclei are absent from visceral-muscle clusters at later stages of development. The early visceral cells are more prominently affected in tinman mutants and to a lesser extent in bagpipe and biniou mutants, reflecting the hierarchical position of these TFs with Tinman acting upstream of both factors. These findings indicate that the circular VM cells are initially specified in bagpipe and biniou mutant embryos but are blocked from further expansion and differentiation, resulting in a loss of the VM at later stages. Interestingly, the hindgut and longitudinal visceral muscles appear largely unaffected in all three mutants, reflecting their different developmental origin (Secchia, 2022).

This molecular data can therefore phenotype all three mutants de novo, identifying the gross phenotypes described by immunostaining of mutant embryos. In addition, the single-cell approach provides more fine-grained quantitative information on the proportion of missing cells at a given stage, in addition to revealing more subtle phenotypes not previously observed, including a gain of mutant cells in other muscle lineages. For example, there is a significant overrepresentation of tinman mutant cells in the early mesoderm and in the somatic lineage. The removal of these TFs thereby not only results in a loss of tissue (one cell fate) but also a more subtle gain of cells dispersed in other tissues from different mesodermal trajectories, highlighting the plasticity of cell fates within the myogenic mesoderm (Secchia, 2022).

Mef2 is required for chromatin accessibility at its high-affinity sites and for gene expression Applying single-cell ATAC-seq to TF mutants in the context of a developing embryo allowed exploration of the extent to which such single-cell data can discern regulatory properties of the TF or its enhancers. As a proof of principle, focus was placed on the Mef2-/- mutant cell cluster (Mutant1), as it contains the highest number of mutant cells (943 cells). The somatic-muscle cluster is the closet cell type to Mutant 1. Of the 8,725 accessible regions in both cell clusters, 408 have significant differential accessibility (DA) in Mef2-/- mutant cell. The majority of DA sites have reduced accessibility (67% [274/408]) and reduced sites often have a larger fold-change. Mef2-responsive sites are generally more gene distal, compared with unchanged sites and are overrepresented in muscle enhancers: Mef2 -/- DA regions more frequently overlap (1) characterized muscle enhancers, and (2) two large collections of putative muscle enhancers defined by ChIP and DNase-seq of FACS-sorted muscle cells, compared with non-DA regions (Secchia, 2022).

Ninety percent of the 8,725 accessible regions in either the somatic and/or Mutant1 cluster were identified in bulk DNase-seq from FACS-sorted muscle cells purified at different stages of embryogenesis. While this highlights the quality of the sci-ATAC-seq data, having such information at a single-cell resolution goes beyond the detection of regulatory regions, as it reveals enhancer usage along developmental trajectories of specific cell types and enables a detailed analysis of the functional input of TFs to enhancers. In addition, single-cell data have enhanced the sensitivity to identify regions that change in mutant embryos and enhanced the precision to uncover the cellular context that is susceptible to that change. To demonstrate this, the analysis was repeated in by testing for differential accessibility between nonmutant and mutant cells across the whole embryo and across the whole muscle population, in order to mimic samples profiled by bulk ATAC-seq in either whole embryos or FACS-purified muscle, respectively. Of the 408 DA sites in Mef2 mutants, only 7% (28) and 43% (176) are identified as significantly changed compared with whole-embryo and muscle mimic samples, respectively. The single-cell approach therefore has enhanced sensitivity to reveal regulatory changes that are not discoverable by traditional approaches, unless complex and extensive FAC sorting is used (Secchia, 2022).

To explore the 408 DA sites further, they were first catagorized into Mef2-bound and -unbound sites, using bulk Mef2 ChIP data at multiple time points of embryonic development. Almost half of the DA regions (48%, 197/408) are bound by Mef2 at this stage or earlier in embryogenesis. Mef2-bound DA sites almost exclusively lose accessibility, consisting of 66% of all DA sites with reduced accessibility. In contrast, regions that gain accessibility are generally not bound by Mef2 and involve regions less related to muscle function (Secchia, 2022).

Removal of Mef2 affects the accessibility of only a specific subset (15%) of all Mef2-bound regions at 10-12 h. Mef2-bound sites that are sensitive to, or resistant to, Mef2 removal could depend on either their extent of co-occupancy by other factors or on the affinity of Mef2 binding. To distinguish between these two possibilities, the co-occupancy of ten TFs active in mesoderm was examined. Susceptible sites are generally less frequently occupied by these TFs compared to nonsusceptible sites. The fraction of DA sites tends to decrease as sites are bound by an increasing number of mesodermal TFs. Examining the occupancy of a much larger set of TFs, 280 factors from modERN also shows an inverse relationship between differential chromatin accessibility in Mef2-/- cells and the number of bound TFs: the median number of bound TFs is 3 for the DA class and 40 for non-DA sites. Therefore, sites that require Mef2 for their accessibility tend to be bound by Mef2 alone or with a small number of other factors, perhaps cooperatively, suggesting that these regions are very Mef2 dependent. To investigate this further, the quantitative Mef2 ChIP signal was used as a proxy for Mef2 affinity. Susceptible (DA) sites have significantly higher Mef2 ChIP signal compared with nonsusceptible (non-DA) sites. Moreover, the proportion of DA sites steadily increases as the Mef2 ChIP signal increases, going from 5% of DA sites in the lowest to 35% in the highest ChIP quantile. This indicates that sites bound more strongly by Mef2 are more likely to have reduced chromatin accessibility upon Mef2 removal. Although both classes are occupied by Mef2, susceptible sites have a 2.5-fold enrichment (61% in the DA group versus 17% in the non-DA group) in the presence of a Mef2 motif. These results indicate that Mef2 is required to establish and/or maintain chromatin accessibility at a large fraction of its high-affinity sites (Secchia, 2022).

Many of these regions overlap characterized or putative muscle enhancers. The loss of accessibility at these sites may therefore lead to changes in the expression of mesoderm/muscle genes, which most likely contributes to the mutant phenotype. To examine this, bulk RNA-expression data was integrated from Mef2 mutant embryos, and genes were sought with a Mef2-bound site in their vicinity (defined as 5 kb upstream and intronic regions). Using this metric, 1,705 differentially expressed genes are associated with at least one Mef2-bound open chromatin region. Of these, those with significantly downregulated, but not upregulated, expression in Mef2 -/- mutants are highly overrepresented for a loss or reduction in chromatin accessibility in at least one of their Mef2-bound associated peaks. Moreover, genes with reduced Mef2-bound sites have significantly stronger changes in both their chromatin accessibility and gene expression, compared with genes with unchanged Mef2-bound sites. Many known Mef2 target genes are among this set, including Mhc, Mlc1/2, Tm1, Mp20, Mlp60A, and Msp300. In addition, their expression changes become more severe with increasing numbers of associated regulatory regions with reduced accessibility. These findings indicate that Mef2 functions primarily as an activator and as the predominant regulator for the expression of these genes, which in turn likely leads to the muscle defects in Mef2 mutant embryos. It is also a rare example demonstrating that a single TF can affect the regulation of many genes by having a cumulative effect on their expression through the action of multiple dependent enhancers (Secchia, 2022).

This paper presents a general framework to obtain a fine-grained view of TF function at both a cellular and molecular level using a systematic, unbiased approach. Phenotypes of developmental mutants are typically assessed by immunostaining with tissue markers and often described in qualitative and somewhat arbitrary terms. There are many examples where other phenotypes were missed as the tissue was outside the interests or scope of the study, and in some cases, suitable tissue markers were not available because they are downstream of the mutated TF. When they are, translating such coarse-grained tissue defects to the underlying molecular function of the TF remains a challenge, and typically the regulatory input is only assessed by occupancy in wild-type embryos compared with gene-expression changes in the mutant. This study shows how single-cell regulatory trajectories, obtained by a dense time course of developmental stages, provide a new opportunity to map developmental mutants to much more precise cell states, thereby providing more fine-grained insights into mutant phenotypes. In the four mutants studied, this approach not only revealed the loss of the expected cell types but could also quantify the proportion of cells lost, pinpoint the development stages, and reveal more subtle phenotypes, such as a gain of some mutant cells in seemingly normal trajectories of other tissues. This highlights the plasticity of mesodermal cell states and also a high degree of canalization to developmental programming, even upon mutation of these essential TFs. This could increase the overall robustness of embryogenesis, for example, by providing an excess of cells that can partly compensate for the loss of others when defects occur. The data provide a rich resource of regulatory changes associated with each step of mesoderm specification and differentiation into different muscle types, which is provide das easy to search, interactive UMAPs for further exploration. Going forward, this approach could be applied to reassess phenotypes and regulatory programs of 'classic' developmental mutants and also to uncover phenotypes of completely uncharacterized mutants de novo, beckoning a new era for joint cellular and molecular phenotyping (Secchia, 2022).

Although the approach presented in this study can readily identify cellular phenotypes from mutant embryos, it might not be possible to study molecular phenotypes in mutants in cases where specific cell types are not specified or maintained. This limitation could be overcome by using conditional depletion/knockout strategies. Dissection of molecular phenotypes using scATAC-seq could also be masked by the binding of other TFs to the same enhancers. Using single-cell ChIP against H3K27ac and/or nascent RNA-seq in this study to measure eRNA at single-cell resolution would help, although both approaches are still very challenging to apply to embryos (Secchia, 2022).

PROTEIN STRUCTURE

Amino Acids - 516

Structural Domains

The N-terminal MADS box, a mediator for DNA binding, is adjacent to a Mef2 domain present in mouse and human homologs (Lilly, 1994).

The serum response factor (see Drosophila SRF or Blistered) and myocyte enhancer factor 2A (MEF2A) represent two human members of the MADS-box transcription factor family. Each protein has a distinct biological function that is reflected by the distinct specificities of the proteins for their coregulatory protein partners and DNA-binding sites. The mechanism of DNA binding utilized by these two related transcription factors was examined. Although SRF and MEF2A belong to the same family and contain related DNA-binding domains, their DNA-binding mechanisms differ in several key aspects. In contrast to the dramatic DNA bending induced by SRF, MEF2A induces minimal DNA distortion. A combination of loss- and gain-of-function mutagenesis identifies a single amino acid residue located at the N terminus of the recognition helices as the critical mediator of this differential DNA bending. This residue is also involved in determining DNA-binding specificity, thus indicating a link between DNA bending and DNA-binding specificity determination. Different basic residues within the putative recognition alpha-helices are critical for DNA binding, and the role of the C-terminal extensions to the MADS box in dimerization between SRF and MEF2A also differs. These important differences in the molecular interactions of SRF and MEF2A are likely to contribute to their differing roles in the regulation of specific gene transcription (West, 1997).

date revised: 10 July 2021 

Home page: The Interactive Fly © 1995, 1996 Thomas B. Brody, Ph.D.

The Interactive Fly resides on the
Society for Developmental Biology's Web server.