Developmental and Cancer Genetics

A single fertilized egg builds a body with hundreds of cell types arranged in exactly the right places, and it does this with one genome that every cell carries. Cancer is what happens when the genes that run that program, the ones that tell cells when to divide, when to stop and when to die, are broken one by one in a single lineage of cells. This chapter traces the developmental tool kit found in flies, worms and plants, and how the same genes, damaged in somatic cells, drive the stepwise evolution of a tumor.

Section 1Model organisms and developmental genetics

The central puzzle of development

A human begins as one cell and ends as roughly thirty trillion cells of more than two hundred types, yet with rare exceptions (B and T lymphocytes that rearrange their antigen receptor genes, and red cells that discard their nucleus) all of these cells carry the same DNA. Development is therefore a problem of differential gene expression: different subsets of a shared genome are switched on in different cells, at different times, in response to signals and to the cell's own history.

Genomic equivalence and cloning

Early embryologists debated whether cells lose genes as they specialize. Nuclear transfer experiments settled the question. A nucleus taken from a differentiated cell and placed into an enucleated egg can, at least sometimes, direct development of an entire organism. Frog tadpole intestinal nuclei produced swimming tadpoles and some adult frogs, and later a sheep was cloned from the nucleus of an adult mammary gland cell by somatic cell nuclear transfer. The success rate is low, because the egg must erase the donor nucleus's epigenetic marks, but the principle is clear: differentiated cells keep a complete genome. This is genomic equivalence.

A more practical demonstration came when adult skin fibroblasts were reprogrammed into induced pluripotent stem cells (iPSCs) by forcing expression of four transcription factors: Oct4, Sox2, Klf4 and c-Myc. Expression of a few master regulators was enough to reset the cell's identity, which tells us that cell type is maintained by networks of transcription factors and chromatin states, not by permanent changes to the DNA sequence.

Potency, determination and differentiation

Potency describes the range of cell types a cell can still produce. The zygote and, in mammals, the blastomeres up to about the eight cell stage are totipotent: they can form every cell of the embryo and the extraembryonic tissues such as the placenta. Cells of the inner cell mass of the blastocyst are pluripotent: they can form any cell of the body (all three germ layers and the germ line) but not the trophoblast. Embryonic stem cells are cultured pluripotent cells. Adult stem cells, such as hematopoietic stem cells, are multipotent, producing several cell types within one lineage, and some progenitors are unipotent.

Two terms are often confused. Determination is the commitment of a cell to a particular fate. A determined cell may look exactly like its neighbors, but if you transplant it to a new location it still develops according to its original fate. Differentiation is the later, visible acquisition of the specialized structure and proteins of that fate. Determination is therefore tested by transplantation, while differentiation is seen by looking at the cell. Cells become determined either by inheriting cytoplasmic determinants unevenly at division (autonomous specification) or by receiving signals from neighboring cells (induction, or conditional specification).

Why model organisms

We cannot do controlled crosses or mutant screens in humans, so developmental genetics relies on model organisms: species chosen because they are easy to raise, breed quickly, produce many offspring and can be manipulated genetically. Because the core developmental genes are deeply conserved across animals, and many across all eukaryotes, discoveries in a fly or a worm routinely explain human biology.

Forward and reverse genetics

There are two directions of attack. In forward genetics you start with a phenotype and find the gene. Animals are treated with a mutagen such as ethyl methanesulfonate (EMS), their descendants are bred to make mutations homozygous, and offspring are screened for a developmental defect. The mutant gene is then mapped and cloned. Forward screens are unbiased; the Drosophila segmentation screens behind Section 2 were forward screens.

In reverse genetics you start with a known gene sequence and ask what it does by disrupting it: a knockout by homologous recombination or CRISPR Cas9, a knockdown by RNA interference or morpholino oligonucleotides, or targeted overexpression. A conditional knockout (for example Cre lox in mice) deletes a gene only in a chosen tissue or time, so genes needed for early survival can be studied later.

Maternal effect and zygotic genes

The earliest steps of animal development are directed by mRNAs and proteins that the mother loads into the egg before fertilization. Genes whose products work this way are maternal effect genes. Their defining genetic property is that the offspring's phenotype is determined by the mother's genotype, not by its own. A bicoid mutant female (genotype bcd/bcd) produces embryos that lack head and thorax and die, even if those embryos received a normal bcd⁺ allele from their father, because the father's allele is not expressed until too late. In contrast, zygotic genes are transcribed from the embryo's own genome, and their mutant phenotypes follow the embryo's genotype in the usual Mendelian way.

Lessons from the worm: lineage and programmed cell death

C. elegans is so small and transparent that every cell division can be watched under the microscope. The adult hermaphrodite has exactly 959 somatic nuclei, and its cell lineage, the complete family tree from zygote to adult, is the same in every individual. Of the 1,090 somatic cells generated, 131 die by programmed cell death (apoptosis) at predictable times and places. Screens for worms in which those 131 cells survive identified the core apoptosis genes, and every one has a human counterpart.

The worm also revealed the first microRNAs, lin-4 and let-7, which time the transitions between larval stages by blocking translation of target mRNAs.

Section 2Pattern formation and homeotic genes

Positional information and morphogens

Pattern formation is the process that organizes differentiated cells into a body plan with axes: anterior to posterior (head to tail), dorsal to ventral (back to belly) and, in limbs, proximal to distal. A cell needs to know where it is. One elegant solution is a morphogen: a substance produced at a localized source that spreads to form a concentration gradient. Cells read the local concentration and switch on different genes above different thresholds. A single gradient can thus specify several distinct zones.

The Drosophila segmentation cascade

The fruit fly embryo is the best understood example of pattern formation. For the first thirteen nuclear divisions, the nuclei share a common cytoplasm (a syncytium), so transcription factors can diffuse between nuclei without crossing membranes. The body is laid out by a hierarchy of gene classes, each class activated or repressed by the one above it, so that the pattern becomes progressively finer.

1. Maternal effect (egg polarity) genes. bicoid mRNA is anchored at the anterior pole of the egg. After fertilization it is translated, and Bicoid protein diffuses to form an anterior to posterior gradient, high at the head end. Bicoid is a transcription factor that activates hunchback in the anterior half and represses translation of caudal mRNA. At the posterior pole, nanos mRNA is localized; Nanos protein blocks translation of maternal hunchback mRNA, so Hunchback protein ends up high in front and low behind. Without bicoid embryos lack head and thorax; without nanos, abdomen.
2. Gap genes (hunchback, Krüppel, knirps, giant) are the first zygotic genes. Each is expressed in one or two broad bands. Their name comes from the mutant phenotype: loss of a gap gene removes a block of several adjacent segments, leaving a gap.
3. Pair rule genes (even-skipped, fushi tarazu, hairy) are expressed in seven stripes, each stripe two segments wide. Each stripe has its own enhancer that reads a unique combination of gap proteins (the even-skipped stripe 2 enhancer is activated by Bicoid and Hunchback and bounded by the repressors Giant and Krüppel). Mutants lose every other segment, so the larva has half the normal number.
4. Segment polarity genes (engrailed, wingless, hedgehog) are expressed in fourteen stripes, one per segment, and define the anterior and posterior compartment of each segment. Engrailed cells secrete Hedgehog, and neighboring cells answer with Wingless, which in turn maintains engrailed: a signaling loop that stabilizes each boundary. Mutants have segments in which one half is replaced by a mirror image of the other half.
5. Homeotic (Hox) genes, discussed below, then give each segment its distinct identity: antennae on the head, legs on the thorax, wings on the second thoracic segment and so on.

Dorsal and ventral

The dorsal to ventral axis uses a different maternal system. Signaling through the Toll receptor on the ventral side of the embryo causes the transcription factor Dorsal to enter nuclei, but only ventrally, forming a nuclear gradient that is high on the belly side. High Dorsal activates twist and snail, specifying mesoderm; intermediate levels specify neurogenic ectoderm; and Dorsal represses decapentaplegic (dpp), so Dpp is made only on the back. (Loss of dorsal makes embryos that are entirely dorsal, hence the name.)

Homeotic genes and the homeobox

Homeotic mutations transform one body part into another body part that normally develops elsewhere. In the dominant gain of function Antennapedia mutation, the gene is wrongly switched on in the head, and the fly grows legs where its antennae should be. In combinations of loss of function Ultrabithorax mutations, the third thoracic segment takes on the identity of the second, so the fly develops a second pair of wings instead of its balancers (halteres). Notice the logic: losing a Hox gene usually makes a segment look like a more anterior segment, because the remaining, more anterior Hox genes take over.

The Drosophila Hox genes lie in two clusters on chromosome 3: the Antennapedia complex (labial, proboscipedia, Deformed, Sex combs reduced, Antennapedia) and the Bithorax complex (Ultrabithorax, abdominal-A, Abdominal-B). Remarkably, the order of the genes along the chromosome matches the order of the body regions in which they are expressed, from head to tail. This is colinearity.

All Hox genes share a 180 base pair sequence, the homeobox, which encodes a 60 amino acid DNA binding homeodomain folded into three alpha helices (a helix turn helix variant). Hox proteins are therefore transcription factors; each one switches on a battery of downstream "realizator" genes that build the structures of its segment.

Hox genes in vertebrates

Vertebrates have the same genes, multiplied. Mammals carry 39 Hox genes in four clusters (HOXA, HOXB, HOXC, HOXD) on four different chromosomes, arranged in 13 paralog groups. Colinearity holds in both space and time: genes at the 3' end of a cluster are expressed earlier and more anteriorly, and genes at the 5' end later and more posteriorly. Hox codes specify vertebral identity, hindbrain regions and limb segments. Excess retinoic acid, a teratogen, shifts Hox expression and disrupts the body plan.

Pattern in plants: the ABC model

Plants evolved their own homeotic system. An Arabidopsis flower has four concentric whorls: sepals (whorl 1, outermost), petals (whorl 2), stamens (whorl 3) and carpels (whorl 4, center). The ABC model explains floral organ identity with three classes of genes, most of which encode MADS box transcription factors rather than homeodomain proteins:

• A genes (APETALA1, APETALA2) act in whorls 1 and 2.
• B genes (APETALA3, PISTILLATA) act in whorls 2 and 3.
• C gene (AGAMOUS) acts in whorls 3 and 4.
• A and C repress each other, so if one is lost, the other spreads into its territory.

Section 3Cell cycle control

The phases of the cycle

Development requires cells to divide at the right times and stop at the right times, and cancer is, at its core, a failure of that control. The eukaryotic cell cycle has four phases. In G1 (gap 1) the cell grows and decides whether to divide. In S phase (synthesis) it replicates its DNA, so a human cell goes from a 2C to a 4C DNA content (C is the amount of DNA in one haploid set). In G2 it prepares for division and checks that replication is complete. In M phase it carries out mitosis (prophase, prometaphase, metaphase, anaphase, telophase) and cytokinesis. G1, S and G2 together are called interphase. A typical cultured human cell cycles in about 24 hours: roughly 11 hours in G1, 8 hours in S, 4 hours in G2 and only about 1 hour in M. Cells that stop dividing exit into quiescent G0; neurons stay there for life, whereas liver cells can re-enter the cycle after injury.

Cyclins and cyclin dependent kinases

The engine of the cycle is a family of protein kinases called cyclin dependent kinases (CDKs). A CDK is inactive by itself. It becomes active only when bound to a regulatory partner, a cyclin, whose levels rise and fall during the cycle (hence the name). CDK levels stay fairly constant; it is the oscillating cyclins, made by transcription and destroyed by ubiquitin mediated proteolysis, that give the cycle its rhythm. Each cyclin and CDK pair phosphorylates a different set of target proteins, which pushes the cell through a particular transition.

Cyclin B with CDK1 was first found biochemically as maturation promoting factor (MPF), which drove frog oocytes into meiosis; yeast cdc (cell division cycle) mutant screens identified the same kinase, and sea urchin eggs revealed cyclins as proteins destroyed at every division.

CDK activity is regulated at several levels, which is why it can act as a switch:

• Cyclin binding plus an activating phosphorylation by CDK activating kinase (CAK) are required.
• Inhibitory phosphorylation by Wee1 holds CDK1 off during G2; the phosphatase Cdc25 removes it. Active CDK1 activates Cdc25 and inhibits Wee1, so entry into mitosis is all or none.
• CDK inhibitors (CKIs): the INK4 family (p16) blocks CDK4/6; the Cip/Kip family (p21, p27) blocks CDK2 and CDK1.
• Proteolysis: SCF destroys CKIs at G1/S, and the anaphase promoting complex/cyclosome (APC/C) destroys securin and mitotic cyclins, making transitions irreversible.

The restriction point: Rb and E2F

In mammalian cells the most important decision is made in late G1, at the restriction point. Before it, a cell needs continuous growth factor signaling to proceed; after it, the cell is committed to complete the cycle even if growth factors are removed. The molecular switch is the retinoblastoma protein (Rb). Unphosphorylated (hypophosphorylated) Rb binds the transcription factor E2F and blocks its activation domain, and it recruits histone deacetylases that keep E2F target genes silent. Growth factor signals raise cyclin D; cyclin D with CDK4/6 begins phosphorylating Rb, which releases some E2F. E2F then turns on the gene for cyclin E, and cyclin E with CDK2 phosphorylates Rb further, releasing more E2F. This positive feedback loop converts a gradual rise in growth signals into a sharp, irreversible commitment. Free E2F activates S phase genes such as cyclin A and DNA polymerase subunits.

Checkpoints

A checkpoint is a surveillance pathway that halts the cycle until a condition is met. Checkpoints were defined genetically: yeast rad9 mutants with damaged DNA fail to pause and die. Because the pause is imposed by a separate control system, it can be lost by mutation, as happens in cancer.

Sister chromatids are held by cohesin. At anaphase, APC/C with Cdc20 destroys securin, freeing separase to cut cohesin. A single unattached kinetochore can hold this off; a weakened spindle checkpoint produces aneuploid daughter cells, common in tumors.

p53, guardian of the genome

The protein p53, encoded by TP53, is a transcription factor that integrates stress signals. In unstressed cells it is kept at very low levels by MDM2, an E3 ubiquitin ligase that tags p53 for proteasomal destruction (and whose gene is itself switched on by p53, a negative feedback loop). DNA damage activates ATM and ATR, which phosphorylate p53 and MDM2 and block their interaction; p53 accumulates, forms tetramers and binds DNA. Its targets fall into three groups:

• Arrest: p21 (CDKN1A), a CDK inhibitor that halts the cycle at G1/S and G2/M.
• Repair: genes for DNA repair proteins, giving the cell time and tools to fix the damage.
• Apoptosis: if damage is too severe, pro-apoptotic genes such as BAX, PUMA and NOXA, which release cytochrome c from mitochondria and activate the caspase cascade.

Oncogene activation can also activate p53, through the protein ARF (encoded by the same CDKN2A locus as p16), which inhibits MDM2. Thus p53 guards against both damaged DNA and inappropriate proliferation signals.

Section 4Oncogenes and tumor suppressor genes

Cancer is a genetic disease of somatic cells

Cancer is a group of diseases in which cells divide without regard to the normal controls and invade surrounding tissue. It is a genetic disease in the sense that it is caused by changes in genes, but in most cases those changes are somatic mutations, acquired during life in one cell lineage and not passed to children. A minority of cancers (roughly 5 to 10 percent overall) arise in people who inherited a germline mutation that gave every cell in the body a head start. Carcinogens are usually mutagens, tumors are clonal (descended from one altered cell) and specific inherited mutations predispose families to specific cancers.

Two broad classes of genes are involved, and the clearest way to keep them apart is the car analogy. Oncogenes are like an accelerator stuck to the floor: a gain of function in a gene that normally promotes growth. Tumor suppressor genes are like brakes that fail: a loss of function in a gene that normally restrains growth, repairs DNA or triggers apoptosis.

Proto-oncogenes and their activation

The first oncogenes were found in tumor causing retroviruses. The Rous sarcoma virus of chickens carries a gene, v-src, that transforms cells. A nearly identical gene, c-src, turned up in every normal vertebrate genome: the virus had captured a cellular gene in altered form. Such normal cellular genes are called proto-oncogenes. They encode the components of the pathways that tell a cell to divide: growth factors, growth factor receptors, signal transducers such as Ras, nuclear transcription factors such as Myc, and cell cycle proteins such as cyclin D. When mutated or overexpressed, they become oncogenes.

Because one activated allele is enough, oncogenes act dominantly at the level of the cell. The main routes to activation are:

Ras is the most frequently mutated oncogene family in human cancer, altered in roughly one quarter to one third of tumors. Ras, a small GTP binding protein linking receptor tyrosine kinases to the MAP kinase cascade, is on when bound to GTP and turns itself off by hydrolysis, aided by GTPase activating proteins (GAPs). Codon 12, 13 or 61 mutations cripple hydrolysis and lock Ras on. The GAP neurofibromin, encoded by NF1, is a tumor suppressor: one pathway holds both accelerator and brake.

Tumor suppressor genes and the two hit hypothesis

The classic tumor suppressor is the RB1 gene, discovered through retinoblastoma, a cancer of the developing retina in young children (roughly 1 in 15,000 to 20,000 births). About 40 percent of cases are hereditary: these children usually develop tumors in both eyes (bilateral), often several tumors, at an earlier age, and they have an increased risk of other cancers, especially osteosarcoma. The remaining 60 percent are sporadic: a single tumor in one eye (unilateral), appearing later, with no family history.

By comparing the age of onset in the two groups, Alfred Knudson proposed the two hit hypothesis. Retinoblastoma requires two mutational events in the same cell. In hereditary cases the first hit is inherited, so one somatic hit suffices, and among millions of retinal cells that happens several times; in sporadic cases both hits must strike one cell, a much rarer event. Both copies of RB1 must be lost for the tumor to form, so tumor suppressor genes are recessive at the level of the cell. Yet the predisposition is inherited as an autosomal dominant trait with about 90 percent penetrance, because nearly every carrier will experience a second hit somewhere in the retina. This recessive mechanism with dominant inheritance is a heavily tested idea.

Loss of heterozygosity

The second hit is often not a new point mutation. More commonly, the cell loses the remaining normal allele by deleting part of the chromosome, losing the whole chromosome (often followed by duplication of the mutant one), or by mitotic recombination or gene conversion that copies the mutant allele onto the homolog. Each of these converts a heterozygous cell into one that carries only the mutant allele, an event called loss of heterozygosity (LOH). LOH is detectable with nearby DNA markers: a person heterozygous for markers around RB1 in blood cells shows only one marker allele in tumor cells. The second hit can also be epigenetic: hypermethylation of CpG islands in a promoter can silence the remaining allele (common for MLH1, BRCA1 and CDKN2A).

Major tumor suppressors and their pathways

Tumor suppressors are often divided into gatekeepers, which directly control cell proliferation or death (such as RB1, APC and TP53), and caretakers, which maintain genome integrity (such as the mismatch repair genes and BRCA1 and BRCA2). Caretaker loss raises the mutation rate rather than driving growth directly.

Not every tumor suppressor obeys the strict two hit rule. Some are haploinsufficient: losing one copy already reduces protein enough to give cells an advantage (PTEN is an example). And because p53 works as a tetramer, a missense mutant subunit can poison normal ones, a dominant negative effect. TP53 is mutated in about half of all human cancers, more than any other gene.

Section 5The multistep nature of cancer

One mutation is not enough

If a single mutation could cause cancer, cancers would be far more common than they are, and their incidence would not rise so steeply with age. Several independent observations show that cancer requires the accumulation of multiple mutations in one cell lineage:

• Age incidence: the risk of most carcinomas climbs steeply with age, as expected if several rare events must accumulate.
• Experimental transformation: a single activated RAS oncogene cannot transform normal rodent cells, but RAS plus MYC (or RAS plus loss of p53) can. Transgenic mice carrying one oncogene develop occasional, late tumors, while mice carrying two develop tumors faster and more often.
• Sequencing: tumor genomes carry several driver mutations that give a growth advantage, typically a handful (often 2 to 8 in common solid tumors), along with many more passenger mutations that are along for the ride.

Age incidence and the number of steps

Suppose a cancer requires k rare, independent, rate limiting mutations in one cell. The probability that all have occurred by age t rises roughly as t^k, and the incidence (new cases per year) rises roughly as t^k-1. On a plot of log incidence against log age, the data fall on a straight line with slope k - 1. For many common carcinomas the slope is about 5 to 6, suggesting about six or seven rate limiting steps. The model is an approximation. A person who inherits one step needs one fewer somatic event, so hereditary cancers show a shallower slope and arise earlier.

The colorectal paradigm

Colorectal cancer is the best mapped example of multistep progression because the intermediate stages, polyps of increasing size and dysplasia, can be removed at colonoscopy and analyzed. The classic sequence, worked out by comparing mutations in polyps and cancers, is:

1. Loss of APC (both alleles) in a colonic crypt cell. APC normally helps destroy beta catenin; without it, beta catenin accumulates, enters the nucleus and, with TCF transcription factors, keeps Wnt target genes such as MYC on. The crypt cell keeps dividing instead of differentiating, forming a small adenoma. This is the gatekeeper step and it occurs in about 80 percent of sporadic colorectal tumors.
2. Activating mutation of KRAS, which accelerates growth of the adenoma.
3. Loss of chromosome 18q, including SMAD4, which removes growth inhibition by TGF beta signaling, in late adenomas.
4. Loss of TP53, typically at the transition from adenoma to invasive carcinoma, allowing cells with damaged genomes to survive.
5. Further changes that permit invasion and metastasis.

The order is a tendency; the accumulation matters more than the exact sequence. A second route, seen in Lynch syndrome and in about 15 percent of sporadic tumors, starts with loss of mismatch repair. These tumors show microsatellite instability and carry many more mutations, including frameshifts in genes with short repeats such as the TGF beta receptor type II gene.

Clonal evolution

Tumor progression is Darwinian evolution inside the body. A cell acquires a mutation that lets it divide a little more than its neighbors; its descendants form a clone; within that clone one cell acquires a second advantageous mutation and its subclone expands; and so on. Each clonal expansion enlarges the target for the next hit. Tumors therefore become heterogeneous mixtures of related subclones, and treatment can select for resistant ones, which is why a cancer that responds at first may relapse.

Genomic instability speeds the process. Loss of caretaker genes raises the point mutation rate (mismatch repair defects) or produces chromosomal instability, with frequent gains and losses of whole chromosomes and arms (spindle checkpoint defects, telomere crisis, loss of p53). Most solid tumors are aneuploid.

The hallmarks of cancer

A useful way to organize the many mutations is by the capabilities they give the cell. The widely taught hallmarks of cancer are:

Two enabling characteristics, genome instability and tumor promoting inflammation, help cells acquire the hallmarks. Normal human cells divide only about 40 to 60 times (the Hayflick limit) as telomeres shorten; about 85 to 90 percent of cancers reactivate telomerase.

Environmental causes: mutagens and viruses

Most carcinogens cause cancer by damaging DNA. Tobacco smoke contains polycyclic aromatic hydrocarbons that form bulky DNA adducts and leave a signature of G to T transversions in TP53 of lung tumors. Ultraviolet light forms pyrimidine dimers and leaves C to T changes in skin cancers. The Ames test screens chemicals for mutagenicity, using Salmonella strains with a his mutation that cannot make histidine. Bacteria are plated without histidine plus the chemical, usually with rat liver extract (S9) for metabolic activation; only revertants form colonies.

Viruses cause an estimated 10 to 15 percent of cancers worldwide. High risk human papillomavirus types (16 and 18) cause nearly all cervical cancers and many oropharyngeal and anal cancers; their E6 protein targets p53 for degradation and their E7 protein binds and inactivates Rb, knocking out both key brakes at once. Hepatitis B and C, Epstein Barr virus and HTLV-1 are other cancer viruses, and the bacterium Helicobacter pylori drives gastric cancer.

Exploiting the genetics of cancer: synthetic lethality

Restoring a lost tumor suppressor is hard, but its absence creates a weakness. Two genes are synthetically lethal when losing either alone is tolerated but losing both kills the cell. Cells lacking BRCA1 or BRCA2 cannot repair double strand breaks by homologous recombination and depend heavily on other repair pathways. Drugs that inhibit PARP, an enzyme involved in single strand break repair, cause single strand breaks to become double strand breaks during replication; normal cells repair them, but BRCA deficient tumor cells die. PARP inhibitors are now used for BRCA associated cancers.

Words worth knowing