Axis formation in Drosophila

No animal has taught us more about how body plans are laid out than the fruit fly Drosophila melanogaster. In the late 1970s Christiane Nüsslein-Volhard and Eric Wieschaus, working at the European Molecular Biology Laboratory in Heidelberg, mutagenized flies and screened tens of thousands of lines for mutations that disrupted the pattern of the larval cuticle. They found about 600 mutations in some 120 genes and sorted them into a logical hierarchy. With Edward Lewis, who had studied homeotic genes since the 1940s, they received the 1995 Nobel Prize in Physiology or Medicine. Their screen revealed a cascade of transcription factors that subdivides the embryo step by step, from broad gradients to sharp stripes to the identity of each segment. Because many of these genes have counterparts in vertebrates, including humans, the fly cascade became a template for understanding development in general.

The fly embryo and its special features

The Drosophila egg is about 0.5 mm long and already polarized along its anterior–posterior and dorsal–ventral axes before fertilization, through interactions between the oocyte and surrounding follicle cells in the ovary. After fertilization, the zygote nucleus divides 13 times without cell membranes, forming a syncytium; after the ninth division most nuclei move to the periphery, forming the syncytial blastoderm, and at the end of division 13 membranes grow in from the surface to make about 6,000 cells, the cellular blastoderm, about 2.5 hours after fertilization. Because nuclei share cytoplasm during these early stages, transcription factors can diffuse among them and form gradients. Gastrulation follows, and the embryo forms a segmented larva that hatches after about a day at 25 °C.

The larva and adult are built from repeating units. In the embryo the basic unit is the parasegment, which is offset by half a segment from the visible segments of the larva. There are 14 parasegments: three in the head region, three thoracic and eight abdominal, plus terminal structures at each end. The pattern of denticle belts on the ventral cuticle of the larva provides a convenient readout of segmentation, and it was the cuticle that Nüsslein-Volhard and Wieschaus examined.

The hierarchy of segmentation genes

The genes controlling the anterior–posterior axis fall into five classes acting in sequence. Maternal effect genes set up gradients in the egg. Gap genes, expressed in broad domains, respond to these gradients. Pair-rule genes, expressed in seven stripes, respond to combinations of gap proteins. Segment polarity genes, expressed in 14 stripes, define the boundaries and polarity of each parasegment. Homeotic selector (Hox) genes then specify the identity of each segment. Each class was named for the phenotype of its mutants: gap mutants lack a contiguous block of segments, pair-rule mutants lack every other segment, and segment polarity mutants have part of each segment deleted and replaced by a mirror image of the remaining part.

Maternal gradients: bicoid and nanos

The anterior of the embryo is organized by bicoid. During oogenesis, bicoid mRNA is transported along microtubules to the anterior pole and anchored there through its 3' untranslated region. After fertilization it is translated, and Bicoid protein spreads posteriorly through the syncytium, forming a gradient that falls roughly exponentially over the anterior half of the embryo. Embryos from mothers lacking bicoid have no head or thorax and instead develop posterior structures at both ends. Injecting bicoid mRNA into the posterior of such embryos makes a head there. Bicoid was the first morphogen to be identified molecularly, in 1988 by Wolfgang Driever and Nüsslein-Volhard, who showed that the level of Bicoid protein determines position along the axis: mothers with extra copies of bicoid produce embryos whose pattern shifts toward the posterior.

Bicoid acts in two ways. It is a homeodomain transcription factor that activates anterior genes, including the gap gene hunchback, in a concentration-dependent manner. It also binds the mRNA of caudal, a posterior determinant, and represses its translation, producing an opposing gradient of Caudal protein. The posterior is organized by nanos, whose mRNA is localized at the posterior pole along with the germ plasm. Nanos protein represses translation of maternal hunchback mRNA, so Hunchback protein is restricted to the anterior. The termini are specified separately by the Torso receptor tyrosine kinase, which is activated only at the two poles.

import math, random
random.seed(1)
# Bicoid gradient: C(x) = C0 exp(-x / lam), with x as a fraction of egg length (EL).
lam = 0.20                     # decay length about 20% of egg length (published estimates are near this)
threshold = math.exp(-0.5 / lam)   # set so the boundary lies at mid-embryo
def boundary(C0):
    return -lam * math.log(threshold / C0)
print("bicoid copies  relative C0   hunchback boundary (% EL from anterior)")
for copies, C0 in ((1, 0.5), (2, 1.0), (4, 2.0), (6, 3.0)):
    print(f"{copies:13d}  {C0:10.1f}   {100 * boundary(C0):6.1f}")
# positional error: if C0 varies 20% between embryos, how much does the boundary vary?
xs = [boundary(1.0 * random.gauss(1.0, 0.20)) for _ in range(5000)]
m = sum(xs) / len(xs)
sd = (sum((x - m) ** 2 for x in xs) / len(xs)) ** 0.5
print(f"20% embryo-to-embryo variation in C0 gives boundary SD = {100 * sd:.1f}% EL")

bicoid copies  relative C0   hunchback boundary (% EL from anterior)
            1         0.5     36.1
            2         1.0     50.0
            4         2.0     63.9
            6         3.0     72.0
20% embryo-to-embryo variation in C0 gives boundary SD = 4.2% EL

This simple model makes two testable predictions. Doubling bicoid dosage should shift the hunchback boundary posteriorly by λ ln 2, about 14 per cent of egg length, and real embryos from mothers with extra bicoid copies do shift, although by somewhat less than this simple model predicts, suggesting buffering by other inputs. Second, 20 per cent variation in Bicoid levels between embryos should produce about 4 per cent of egg length variation in the boundary. Measurements by Thomas Gregor and colleagues in 2007 found that the hunchback boundary is positioned to within about 1 per cent of egg length, roughly the width of a single nucleus. How embryos achieve such precision, through time averaging, cross-regulation among gap genes and input from the posterior system, has become a classic problem in quantitative biology.

Gap genes

The gap genes, including hunchback, Krüppel, knirps and giant, encode transcription factors expressed in broad, partly overlapping domains along the embryo. Hunchback is activated by high Bicoid in the anterior. Other gap genes are activated by Bicoid, Caudal or the terminal system and, crucially, repress one another. Krüppel, for instance, is expressed in the middle of the embryo where Hunchback is moderate; high Hunchback represses it anteriorly, and Knirps and Giant repress it posteriorly. Mutual repression between adjacent gap genes sharpens fuzzy maternal gradients into crisp domains, a general mechanism for converting graded inputs into discrete outputs.

Pair-rule genes and modular enhancers

The first periodic pattern appears with the pair-rule genes, expressed in seven transverse stripes. Primary pair-rule genes such as even-skipped (eve) and hairy are regulated directly by gap and maternal proteins. The eve gene has separate enhancers for different stripes. The stripe 2 enhancer, a stretch of about 500 base pairs, contains binding sites for activators (Bicoid and Hunchback) and repressors (Giant anteriorly and Krüppel posteriorly). Stripe 2 is expressed only where activators are present and both repressors are low. When Michael Levine and colleagues joined this enhancer to a reporter gene, the reporter was expressed in a single stripe. This modular organization, in which each enhancer reads a different combination of inputs, is a general feature of developmental genes and allows evolution to alter one expression domain without disturbing others.

Segment polarity genes and parasegment boundaries

After cellularization, cells can no longer share transcription factors, so the pattern must be maintained by cell–cell signalling. The segment polarity genes do this. Pair-rule proteins activate engrailed in a narrow stripe at the anterior of each parasegment and wingless in the adjacent row of cells just anterior to it. Engrailed-expressing cells secrete Hedgehog, which keeps neighbouring cells expressing wingless; Wingless, the fly Wnt protein, signals back to keep engrailed on. This positive feedback loop stabilizes the boundary between parasegments, and the gradients of Wingless and Hedgehog across each segment pattern the denticles and naked cuticle. The names hedgehog and wingless come from mutant phenotypes; the vertebrate versions of these pathways, Sonic hedgehog and the Wnts, appear throughout this course.

Homeotic genes and segment identity

Once segments exist, each must acquire a distinct identity: antennae on the head, legs on the thorax, wings only on the second thoracic segment, halteres (balancing organs) on the third. This is the job of the homeotic selector genes, which lie in two clusters on chromosome 3, the Antennapedia complex and the bithorax complex. Mutations in them cause homeotic transformations, in which one body part develops like another. Dominant Antennapedia mutations, which express the gene in the head, turn antennae into legs. Edward Lewis's triple mutants in the bithorax complex transformed the third thoracic segment into a second thoracic segment, producing four-winged flies. The products of these genes share a 60-amino-acid DNA-binding homeodomain, encoded by a 180-base-pair homeobox, and the discovery in 1984 that homeobox genes exist in vertebrates, the Hox genes, opened up comparative developmental biology. Lesson 7 examines Hox genes and their role in evolution.

The dorsal–ventral axis

The dorsal–ventral axis is set by a separate system. During oogenesis, the position of the oocyte nucleus determines where the follicle cells are signalled to adopt dorsal fates, and ventral follicle cells then generate a protease cascade in the perivitelline space that activates the Toll receptor on the ventral side of the embryo. Toll signalling causes the transcription factor Dorsal, initially present throughout the embryo's cytoplasm, to enter nuclei on the ventral side, forming a nuclear gradient highest ventrally. High nuclear Dorsal activates twist and snail, specifying mesoderm; intermediate levels activate genes of the neurogenic ectoderm; and where Dorsal is absent from nuclei, genes such as decapentaplegic (dpp) are expressed dorsally. Dpp, a BMP family protein, then patterns the dorsal ectoderm in a gradient shaped by its antagonist Short gastrulation. The Toll pathway, discovered through this developmental role, later turned out to be central to innate immunity in insects and mammals, earning Jules Hoffmann a share of the 2011 Nobel Prize.

Remarkably, the dorsal–ventral system of insects mirrors that of vertebrates, but inverted: in flies, the BMP Dpp is dorsal and the nervous system ventral, while in vertebrates BMP is ventral and the nervous system dorsal. The fly antagonist Short gastrulation corresponds to vertebrate Chordin. This observation revived an idea of the nineteenth-century naturalist Geoffroy Saint-Hilaire that the vertebrate body is an inverted arthropod, and it suggests that a BMP–Chordin system patterned the dorsal–ventral axis of the common ancestor of flies and vertebrates more than 550 million years ago.