Molecular Lessons from the Drosophila Bithorax Complex

Abstract

The Genetics Society of America’s (GSA’s) Edward Novitski Prize recognizes a single experimental accomplishment or a body of work in which an exceptional level of creativity, and intellectual ingenuity, has been used to design and execute scientific experiments to solve a difficult problem in genetics. The 2020 recipient is Welcome W. Bender of Harvard Medical School, recognizing his creativity and ingenuity in revealing the molecular nature and regulation of the bithorax gene complex.

Of all of the so-called complex loci explored by Drosophila geneticists, the bithorax complex (BX-C) stands out as most extensive and intriguing. That is thanks to decades of study by E.B. Lewis and his colleagues, summarized in his 1995 Nobel lecture “The Bithorax Complex: the First Fifty Years.” When techniques were developed to isolate Drosophila loci based on their chromosomal position, the bithorax complex was the most appealing target. The renewed description of genetic observations in molecular terms has led to insights widely applicable in molecular biology. This is an attempt to highlight the most important of those insights.

The initial recovery of molecular clones covering the bithorax complex, and mapping of mutant lesions, confirmed Lewis’ observations, based on classical genetic recombination, that the locus was large (over 300 kb), and that the chromosomal order of mutations aligned with the order of the body segments most affected (Bender et al. 1983a,b; Karch et al. 1985) (see Figure 1). What was the nature of all of this DNA? Contemporaneous saturation mutagenesis suggested that the BX-C probably contained only three lethal complementation groups (Sánchez-Herrero et al. 1985; Tiong et al. 1985), but it was still surprising that so few protein coding transcripts were found, principally those for Ultrabithorax (O’Connor et al. 1988; Kornfeld et al. 1989), abdominal-A (Karch et al. 1990), and Abdominal-B (Kuziora and McGinnis 1988; Celniker et al. 1989; Zavortink and Sakonju 1989) (see Figure 1). The comparison of the predicted proteins for Ultrabithorax, and for another segment identity gene, Antennapedia, led to the discovery of the homeobox (McGinnis et al. 1984a; Scott and Weiner 1984)—a DNA binding motif widely used in developmental regulation. These segment identity homeobox genes were strikingly conserved; homologous “HOX” clusters were discovered in many animals, including mammals (McGinnis et al. 1984b; Akam 1989). The recognition of this functional conservation suddenly elevated the importance of Drosophila as a model for human biology.

Figure 1.

Regulatory Domains of the bithorax complex. A diagram of an adult female fly shows the body segments specified by the BX-C, including the third thoracic (T3) and 10 abdominal segments (A1–A10). The 8th, 9th, and 10th abdominal segments contribute only small rudiments. The map below shows the embryonic binding sites of the CTCF protein along the 325 kb extent of the BX-C (Bowman et al. 2014). At the bottom are shown the transcription units of the three homeobox genes, Ubx, abd-A, and Abd-B. The colored blocks of the map show 11 successive regulatory domains, marked with parasegment numbers (PS5-15). The arrows connect the body segments with their corresponding parasegmental domains. Most domain boundaries correspond to CTCF binding sites. The righthand boundaries of domains PS5-7 and PS11-14 have been confirmed by mapping the extent of histone modifications (H3K27me3) in nuclei from single parasegments (Bowman et al. 2014; S. Tabor, H. Domingues, and W. Bender, unpublished data).

The dearth of proteins encoded in the BX-C suggested that the vast majority of the sequence, most of which is evolutionarily conserved, must be regulatory. That the regulation of a transcription factor could be so intricate was a novel realization, but it made sense when the expression patterns of the Hox proteins were found to be exceptionally granular in time and space (White and Wilcox 1984; Beachy et al. 1985; Kuziora and McGinnis 1988; Celniker et al. 1989; Delorenzi and Bienz 1990; Macias et al. 1990; Karch et al. 1990; Boulet et al. 1991; Sánchez-Herrero 1991).

A pleasant surprise of the early mapping was that mutant lesions were so easy to find. Point mutations outside of the coding regions were very rare, despite multiple mutagenesis screens by Lewis using the chemical mutagen ethyl methane sulfonate (EMS). Regulatory sequences, such as enhancers, are difficult to damage with single base changes. X-rays, by contrast, generated several deletions and many rearrangement breaks; mapping these made it possible to discern the layout of the genes of the BX-C and their regulatory regions.

Most spontaneous mutations in the BX-C and elsewhere were associated with mobile element insertions. Curiously, the majority of insertions in the BX-C with penetrant phenotypes were due to one particular transposable element named “gypsy.” These and other gypsy-induced mutations could be phenotypically reverted by a second-site mutation in the gene for suppressor-of-Hairy-wing [su(Hw)] (Modolell et al. 1983). Some revertants of gypsy-induced mutations had small deletions within gypsy in a region of tandem repeats (Peifer and Bender 1988), shown to be binding sites for the SU(HW) protein, a zinc-finger DNA-binding protein (Spana et al. 1988). The cluster of SU(HW) binding sites in the gypsy element was an early example of an “insulator,” able to block enhancer/promoter interactions (Geyer and Corces 1992; Gdula et al. 1996).

Ed Lewis highlighted the sequential arrangement of his mutations on the genetic map with the anterior/posterior order of the body segments they most affect—his “colinearity rule.” He suggested a separate functional region for each segment, from the third thoracic to the eighth abdominal (Lewis 1981). His model was repeated with more molecular wording, with a cis-regulatory “domain” for each segment (Peifer et al. 1987), or, more properly, for each parasegment (Martinez-Arias and Lawrence 1985). The central idea was that individual regulatory domains were only “open for business” in specific body segments (Maeda and Karch 2015). This idea was reinforced when P element reporter constructs were recovered that were inserted within the BX-C. Their reporter proteins (LacZ, GFP, or Gal4) were expressed only in the segments specific to the domain where the insertion landed (McCall et al. 1994; Bender and Hudson 2000).

The features of a BX-C domain have been gradually elaborated. The anterior/posterior positional cues for each domain were presumed to come from the “gap” and “pair rule” genes that lay down the segments within the first few hours of embryonic development (Nüsslein-Volhard and Wieschaus 1980). Candidate “embryonic enhancers” or “initiators” were first identified in P element transgenes carrying various restriction fragments from the BX-C (Simon et al. 1990; Mihaly et al. 2006). Their positions were refined and correlated with binding sites for gap and pair rule genes (Qian et al. 1993), and their relevance was confirmed by the discovery of rare single-base-change mutations in gap gene binding sites in these enhancers (Shimell et al. 1994; Ho et al. 2009). The most definitive demonstration involved the replacement within the BX-C of the initiator for the sixth abdominal segment with that of the fifth; the developmental program for the sixth abdominal segment was turned on ectopically in the fifth abdominal segment (Iampietro et al. 2010). There are many other enhancers within each domain, first recognized in transgenes (Pirrotta et al. 1995), which specify the cell types where the homeotic genes are expressed.

Boundaries between BX-C regulatory domains were recognized after a striking class of dominant mutations was linked to small deletions. In the prototypic example, Fab-7, a 4 kb deletion between the domains for the sixth and seventh abdominal segments, transformed the sixth segment into a copy of the seventh (Gyurkovics et al. 1990). In the fused domain, the developmental instructions for the seventh abdominal segment came under the control of the positional cues for the sixth abdominal segment. The BX-C boundaries are sites of binding by the CTCF protein, another zinc-finger DNA-binding protein (Holohan et al. 2007) (see Figure 1). Many proposed boundaries in mammalian systems are also marked by CTCF, but the function of a boundary is still best demonstrated by the examples in the BX-C.

The third critical element of each domain is a “Polycomb Response Element” (PRE, Simon et al. 1993). Lewis showed that the BX-C is negatively regulated by the Polycomb gene (Lewis 1978), and studies of BX-C protein patterns showed that the segmentally limited expression in wild type embryos spread to all segments in Polycomb mutants (Beachy et al. 1985; Celniker et al. 1989; Simon et al. 1992). The POLYCOMB protein was found bound to several discrete sites within the BX-C (Chiang et al. 1995; Schwartz et al. 2006). A few of these PREs have been dissected and shown to include binding sites for Polycomb and for other gene products of the Polycomb Group (Horard et al. 2000; Sipos et al. 2007; Orsi et al. 2014). Because repression by the Polycomb Group acted on a variety of reporter constructs introduced into the BX-C, such repression seemed likely to involve a blockage to accessibility by transcription factors or polymerase. Restricted accessibility was confirmed by in vivo assays (Fitzgerald and Bender 2001), and was recently corroborated by chromatin compaction of repressed domains, as seen in super-resolution microscopy (Boettiger et al. 2016; Mateo et al. 2019). Polycomb repression is accompanied by trimethylation of lysine 27 of histone H3 (Müller et al. 2002). Assays of nuclei from single segments show loss of H3K27me3 correlates with the sequential activation of each domain (Bowman et al. 2014).

The domain model is perhaps the most important contribution from molecular studies of the BX-C. Each domain, with an initiator, a PRE, and flanking boundaries, is a subroutine in a larger program of development, with instructions reserved to be read only at a particular time and place. Genome-wide surveys of H3K27me3 show many other loci with broad regions of methylation (Schwartz et al. 2006); the number of loci with this sort of domain architecture will likely increase as more tissues and times are assayed.

Many outstanding questions in gene regulation and nuclear architecture might be best answered in future studies of the BX-C. We have no idea how initiator elements talk to PREs to impose or block repression. We know of many “noncoding” RNAs transcribed from within the BX-C (Lipshitz et al. 1987; Sánchez-Herrero and Akam 1989; Graveley et al. 2011; Pease et al. 2013), but, with rare exceptions (Gummalla et al. 2012; Maeda et al. 2018), we know nothing of their functions. We know that enhancers must act on distant promoters, even on opposite chromosomes (“transvection”; Lewis 1954, Duncan 2002), but aside from vague notions of “looping,” we do not know how target promoters are found. Domain boundaries with CTCF attached can sometimes be ignored or “bypassed” (Kyrchanova et al. 2019), but we do not know how this is regulated. Perhaps the most glaring curtain of ignorance surrounds Lewis’ initial rule of colinearity; why must the domains be lined up on the chromosome in the order of the body segments? The well is far from dry.

A current trend in molecular biology is to do genome-wide investigations, with results often distilled into a “metagene.” This serves to establish the generality of known mechanisms. A more traditional approach is to focus on a single locus or system, preferably one with a rich legacy of genetic description. This is arguably the more promising path to discovery of novel mechanisms. To the textbook examples of the Lac operon, phage lambda, and yeast mating type, perhaps we can add the Drosophila BX-C.