Developmental Biology: How One Cell Becomes a Body

Look at your hand. Five fingers, each with the right number of bones, the thumb on the correct side, the whole thing the right size for the arm it hangs from. Nobody drew a plan and nobody supervised the building. It began as a single fertilised egg, a cell about a tenth of a millimetre across, and the egg divided, and the descendants somehow worked out what to become.

Now look at a different puzzle: a sliced-up flatworm. Cut a planarian into pieces and each piece grows back the missing end, head where a head belongs, tail where a tail belongs. Cut a salamander's leg off and it builds a new one, bones, muscle and nerves in the right order. Something in the tissue knows where it is, and what is missing.

A developmental biologist looks at both and sees one question: how does a cell find out where it is, and what does it do with the answer? The answers turned out to be few, surprisingly general, and shared between a fruit fly and you. That shared set of answers is what a graduate of this course still just sees.

One genome, many cells: the egg already knows

The first thing to get straight is what does not change. A skin cell and a neuron hold the same DNA. In 1962 John Gurdon showed this in frogs: he took the nucleus out of an intestinal cell of a Xenopus tadpole, put it into an egg whose own nucleus had been destroyed, and sometimes got a whole new tadpole. The "specialised" nucleus still held every instruction. Gurdon shared the 2012 Nobel Prize in Physiology or Medicine for this, with Shinya Yamanaka. In 1996 the same logic produced Dolly the sheep, cloned from a mammary-gland cell.

So cells differ not by what they contain but by which genes are on. Development is a long sequence of switches. A cell that finds itself in one place switches on one set of transcription factors (proteins that read DNA); those switch on others; and each choice narrows what the cell can later become. The question becomes: what tells the first switch which way to go?

Gradients: a cell reads its position from a concentration

The classic answer was set out by Lewis Wolpert in 1969, in a paper on positional information. Imagine a row of identical cells. At one end, a source releases a chemical, a morphogen, that spreads and is slowly broken down, so its concentration falls smoothly with distance. Every cell can measure the local concentration. Give the cells two threshold rules ("above level 1 become blue; between levels 1 and 2 become white; below level 2 become red") and a smooth slope turns into three sharp bands, like the French tricolour. Wolpert's "French flag model" is the shortest possible account of how a pattern comes from nothing but a slope and a few thresholds.

The real example is the fly. The mother deposits the messenger RNA for a protein called Bicoid at the front end of the egg; it is translated and the protein spreads backward, forming a gradient. Nobel-winning work by Christiane Nüsslein-Volhard and Eric Wieschaus (1995 prize, shared with Edward Lewis) showed that this gradient switches on different genes at different positions. The mathematics is simple: a steady source and steady decay give an exponential,

Every λ of distance divides the concentration by e (about 2.7). Try it. Below, the embryo is 500 µm long (about the size of a fly embryo; the decay length λ and the thresholds are illustrative numbers in the right range, not measured values for any one gene).

At the defaults the blue band ends at 69 µm, the white at 161 µm, and the rest of the 500 µm is red. Notice two things the sliders teach. Double the source and every boundary slides back by λ·ln 2 (about 69 µm): the pattern moves. Shorten the embryo and the bands keep their absolute size but the red region shrinks, so the proportions are wrong. Real embryos are often properly proportioned at different sizes, so something must scale the gradient; how remains an active research question, and not every system solves it the same way.

A smooth slope plus a few thresholds is enough to draw sharp stripes. The cell only has to measure and compare.

Gradients are not the only pattern-maker. Alan Turing proposed in 1952 that two chemicals which diffuse at different speeds and react with each other can create spots and stripes spontaneously (reaction-diffusion). Real tissues usually mix the two: gradients set coarse positions, and local interactions sharpen and repeat them.

A clock and a wavefront: building a spine segment by segment

Your backbone is a repeat: vertebra after vertebra, each from a block of tissue called a somite. Somites form one pair at a time, head to tail, like a conveyor belt that stamps out beads. The leading model, the clock and wavefront, is neat. Cells in the tail-bud oscillate, switching key genes on and off with a fixed period, like a ticking clock. As the embryo extends, a wavefront moves backwards through the tissue at some speed. Each time the clock completes a cycle a stripe of cells gets "locked in" as one somite. The segment length is therefore:

Species set the period. A somite pair forms about every 30 minutes in zebrafish, every 90 in a chick and every 100 in a snake; in the mouse it is around two hours, and in human cells grown in a dish the clock period comes out near five hours (two 2020 studies, Matsuda et al. and Diaz-Cuadros et al.). A snake makes hundreds of vertebrae because its clock runs fast compared with its body growth.

At 90 minutes and 2 µm/min, one somite is 180 µm and 16 form in a day. Slow the clock to a human's 300 minutes at the same speed and each segment is 600 µm, with only 4.8 per day. Change the period and the number of segments changes; change the speed and the size changes. Both are things evolution can tune.

Hox genes: one toolkit for every animal's body plan

A gradient says where. Something else must say what lives here. In the fruit fly, Edward Lewis found mutants in which a whole body part was built in the wrong place: a fly with a second pair of wings where its small balancing organs (haltere) should be, or legs growing where antennae belong. These were not broken parts; they were correct parts in the wrong segment. The genes responsible are the Hox genes, a family of transcription factors that tell each segment its identity.

Two facts make them remarkable. First, in a fly the eight Hox genes sit on the chromosome in the same order as the body regions they control, head to tail (a property called colinearity). Second, mice and humans have the same kind of genes: 39 Hox genes in four clusters, which also pattern the head-to-tail axis, including our spine. Mammals duplicated the clusters; the logic survived. A fly and a mouse last shared an ancestor over 500 million years ago, and both still build their long axis with a related code. Use the tool below to see how the code reads out in the fly.

Look at what the "Lose Ubx" button does: without Ubx, the third thoracic segment behaves like the second, and the haltere becomes a wing, the four-winged fly of Lewis's famous bithorax mutants. The body parts were all in the genome; Hox genes only choose which program runs where.

Stem cells and reprogramming: rolling downhill in a landscape

The embryologist Conrad Waddington gave the picture that still organises the subject. In 1957 he described an "epigenetic landscape": a ball (a cell) starts high on a hillside and rolls down through branching valleys. Each valley is a fate; the ridges between them keep a cell that has chosen from drifting back. Noise and signals decide which valley an undecided cell enters. Cells at the top, with many options, are stem cells. Cells at the valley floors are specialised.

For fifty years that landscape was a one-way street. Gurdon's frogs showed it could be reversed by the egg's cytoplasm. In 2006 Shinya Yamanaka and Kazutoshi Takahashi in Kyoto showed that adding just four genes (Oct4, Sox2, Klf4, c-Myc) to mouse skin cells could push them all the way back up to an embryonic-like state. These are induced pluripotent stem cells (iPSCs); human ones followed in 2007 from Yamanaka's lab and, with a slightly different set of factors, from James Thomson's. In effect the cell flattens its own ridges and climbs back out of its valley. The simulation below uses a simple mathematical landscape; real reprogramming is far messier and slow.

With a mid-height ridge and a little noise, undecided cells split roughly evenly; a small tilt biases the split. A specialised A cell stays A. Press the third button and the ridge is flattened temporarily: at the default settings the cells scatter and about half end up as B once the ridge is rebuilt. That is the picture of reprogramming: not erasing the genome, but reshaping the landscape the cell sits in. iPSC-derived cells are used to model diseases in a dish and are in early clinical studies; this describes the science and isn't medical advice.

Regeneration: development you can restart

If a cell can read its position, a damaged body can reread it. Planarians are the extreme case: Thomas Hunt Morgan studied them around 1900, and fragments estimated at as little as 1/279 of the worm can rebuild a whole one, because a reservoir of stem cells is spread through the body and gradients tell each stump which end is missing. Salamanders such as the axolotl regrow whole limbs by forming a blastema, a bud of cells that reverts to a more flexible state and repeats limb development. Zebrafish regrow fins and, in an influential 2002 study, about a fifth of the heart's main chamber. Mammals do far less. Human fingertips can regrow in young children if the cut is near the tip and clean, and the liver can rebuild much of its mass, but a lost limb does not return. Why mammals lost the ability, and whether it can be coaxed back, is open.

Evo-devo: change when and where, not what

If the toolkit is shared, then bodies must differ mostly by when and where the same genes are used, not by having different genes. Evidence has piled up. The gene Pax6 is needed for eyes in flies and mice, though the eyes are built differently; in 1995 Walter Gehring's lab showed that mouse Pax6 switched on in a fly's leg triggers a fly-type eye there. Stickleback fish in lakes repeatedly lost their pelvic spines; Chan and colleagues (2010) traced this in several populations to deletions in the control region of the gene Pitx1, which switched off the gene in the pelvis only, leaving it working elsewhere. And in Darwin's finches in the Galápagos, deeper, broader beaks go with more expression of the signalling gene Bmp4 in the beak (Abzhanov and colleagues, 2004). Evolution works by editing switches, which is why this subject is the other half of Evolution (35/47).

What a graduate just sees

• Every cell has the whole genome; cell types are the pattern of what is switched on.
• A smooth gradient plus thresholds turns into sharp bands: the cell measures and compares.
• Shift a source or a threshold and the boundary moves by λ·ln of the ratio; pattern is tunable.
• Repeats like vertebrae and fingers come from clocks, waves and spacing rules, not from a blueprint of each part.
• Hox genes say "what is here", and the same code runs in flies and in you; a wrong address makes a right part in the wrong place.
• A cell's fate is a valley in a landscape: noise and signals choose, ridges lock the choice, and reprogramming reshapes the landscape.
• Regeneration is development replayed from an injury; animals differ in how much they kept.
• Evolution of form is mostly evolution of switches, not of the genes themselves.

Where it shows up next

• Evolution (35/47): every example of evo-devo, from stickleback to finches, is the mechanism that gives selection something to choose from.
• Genetics (35/46) and Molecular Biology (35/45): transcription factors and enhancers are the machinery behind each switch.
• Physiology (35/49): organs are the product of development; many adult diseases, and cancer, involve development programs running in the wrong place or time.
• In the world: birth-defect research (many arise from errors in gradients and Hox-type genes), organoids and iPSC disease models, regenerative-medicine trials, and animal breeding.

If you want to go deeper

• 📘 Textbook. Scott Gilbert and Michael Barresi, Developmental Biology (Oxford University Press / Sinauer, 12th edition, 2019). The standard course text, broad and well illustrated.
• 📘 Textbook. Lewis Wolpert and colleagues, Principles of Development (Oxford University Press, 6th edition, 2019). Shorter and more concept-driven, by the author of the French flag model.
• 📘 Gentler. Sean B. Carroll, Endless Forms Most Beautiful (2005). The story of the toolkit and evo-devo for a general reader.
• 🌐 Free course. MIT OpenCourseWare 7.013 Introductory Biology (Spring 2018) includes development and stem cells among its topics. For short films and animations, HHMI BioInteractive is free.