Molecular Genetic Tools and Genomics

Fifty years ago no one could read a gene, copy it in a test tube or change a single chosen base in a living cell. Today a student can amplify DNA from a drop of saliva in an afternoon, sequence a whole human genome in a day and edit a gene with a programmable enzyme borrowed from bacteria. This chapter explains the tools that made genetics a molecular science and the genomic thinking they made possible.

Section 1Restriction enzymes, cloning and PCR

Restriction enzymes: molecular scissors

Bacteria defend themselves against viruses with restriction-modification systems. A restriction endonuclease cuts DNA at a specific recognition sequence, and a partner methyltransferase methylates the same sequence in the host's own DNA so that it is protected. Invading phage DNA, lacking the methyl marks, is cut. The type II enzymes used in the lab cut within or right next to their recognition site, which is usually 4 to 8 bp long and palindromic: it reads the same 5′ to 3′ on both strands. EcoRI recognizes 5′-GAATTC-3′, whose complement read 5′ to 3′ is also GAATTC.

Enzymes leave three kinds of ends:

Sticky ends are single stranded overhangs that base pair with any other end made by the same enzyme. When a vector and an insert are cut with the same enzyme, their ends anneal, and DNA ligase seals the sugar-phosphate backbone by forming phosphodiester bonds. Cutting with two different enzymes gives incompatible ends at either side, which forces the insert to enter in one orientation (directional cloning). Blunt ends can be joined to any other blunt end, but less efficiently.

If bases were random and equally frequent, a given 4 bp site would occur on average once every 4⁴ = 256 bp, a 6 bp site once every 4⁶ = 4,096 bp and an 8 bp site once every 4⁸ = 65,536 bp. Real genomes deviate (CpG containing sites are rare in mammals), but the estimate is useful. Fragments are separated by gel electrophoresis: DNA is negatively charged, so it migrates toward the positive electrode, and smaller fragments move faster through the agarose mesh. Distance traveled is roughly proportional to the logarithm of size, so sizes are read against a ladder of known fragments.

Vectors and cloning

A vector carries a DNA fragment into a host cell and replicates it there. A typical plasmid cloning vector has an origin of replication, a selectable marker such as an ampicillin resistance gene so that only cells that took up the plasmid survive, and a multiple cloning site (polylinker) containing many unique restriction sites. In many vectors the cloning site lies within the lacZα fragment. Cells carrying an empty vector make functional β-galactosidase and form blue colonies on X-gal; an insert disrupts lacZα, giving white colonies (blue-white screening). Plasmids are introduced into bacteria by transformation using chemically treated cells or a brief electric pulse.

Plasmids carry inserts up to about 10 to 15 kb. Larger fragments need other vectors: bacteriophage lambda (about 20 kb), cosmids (about 45 kb), bacterial artificial chromosomes (BACs, 100 to 300 kb) and yeast artificial chromosomes (up to 1 Mb or more). Expression vectors add a strong regulated promoter, such as a lac derived promoter induced by IPTG, so the cloned gene's protein can be produced.

Eukaryotic genes contain introns that bacteria cannot remove, so to express a human protein in bacteria researchers clone complementary DNA (cDNA). Retroviral reverse transcriptase copies mature mRNA, primed from its poly(A) tail with oligo(dT), into DNA. A collection of clones representing all mRNAs in a tissue is a cDNA library; one representing all of an organism's DNA is a genomic library.

The polymerase chain reaction

Invented by Kary Mullis in the 1980s, the polymerase chain reaction (PCR) copies a chosen segment of DNA billions of times in a few hours without cloning. A reaction contains template DNA, two primers (short single stranded oligonucleotides, usually 18 to 25 nt) complementary to opposite strands at the two ends of the target, the four dNTPs, Mg²⁺ and a heat stable DNA polymerase, classically Taq polymerase from the hot spring bacterium Thermus aquaticus. Each cycle has three steps:

1. Denaturation at about 94 to 98 °C separates the strands.
2. Annealing at about 50 to 65 °C, set a few degrees below the primers' melting temperature, lets primers pair with their sites.
3. Extension at 72 °C, the optimum for Taq, lets polymerase add nucleotides to each primer's 3′ end, at roughly 1 kb per minute.

Because each new strand can serve as a template in the next cycle, the number of target copies ideally doubles each cycle: after n cycles, N = N₀ × 2ⁿ. Products of exactly the length between the primers appear in cycle 3 and soon dominate. Amplification slows after about 30 to 35 cycles as primers and dNTPs run low. Taq lacks proofreading and makes roughly one error per 10⁴ to 10⁵ bases, so high fidelity proofreading polymerases are used when sequence accuracy matters. PCR is so sensitive that a single contaminating molecule can be amplified, so careful technique and negative controls are essential.

Variations: RT-PCR and quantitative PCR

RT-PCR starts with reverse transcription of RNA into cDNA, then amplifies it, to detect transcripts or RNA viruses. Quantitative (real time) PCR (qPCR) follows product accumulation each cycle with a fluorescent dye or probe. The threshold cycle (C_t) is the cycle at which fluorescence crosses a set threshold; the more starting template, the earlier this happens. Each cycle earlier means twice as much template. Relative expression is usually calculated by the ΔΔC_t method, which normalizes the target gene to a stable reference gene in each sample.

Section 2Sequencing technologies

Sanger chain termination sequencing

In 1977 Frederick Sanger introduced the method that would dominate sequencing for thirty years. It exploits dideoxynucleotides (ddNTPs), which lack the 3′ hydroxyl group that DNA polymerase needs to attach the next nucleotide. A reaction contains a single stranded template, one primer, DNA polymerase, the four normal dNTPs and a small amount of the four ddNTPs. Each time polymerase happens to add a ddNTP instead of the matching dNTP, the strand stops growing. The result is a nested set of fragments, each beginning at the primer and ending at a particular position with a ddNTP that identifies the base there.

In the original method four separate reactions, one per ddNTP, were run in adjacent lanes of a gel. In modern automated sequencing each ddNTP carries a different colored fluorescent dye, so one reaction suffices; the fragments are separated by size in a thin capillary, and a laser reads the color of each fragment as it passes the detector. Fragments emerge from shortest to longest, so the sequence is read in the 5′ to 3′ direction of the newly synthesized strand, which is complementary to the template. A single Sanger read covers about 700 to 1,000 bases with very high accuracy (about 99.99 percent), so Sanger sequencing remains the standard for confirming a single mutation or checking a cloned insert.

Short read sequencing by synthesis

The Human Genome Project, completed in 2003 at a cost of about three billion dollars, relied on Sanger sequencing. Massively parallel (next generation) methods introduced after 2005 cut the cost of a human genome to a few hundred dollars by sequencing hundreds of millions of fragments at once. The most widely used approach is sequencing by synthesis with reversible terminators:

1. Library preparation. Genomic DNA is fragmented to a few hundred base pairs and short adapters are ligated to both ends.
2. Cluster amplification. Fragments bind to a glass flow cell coated with oligonucleotides matching the adapters and are copied in place by bridge amplification, producing clusters of about a thousand identical copies.
3. Sequencing. Each cycle, polymerase adds one fluorescent nucleotide whose 3′ end is reversibly blocked. A camera records the color of every cluster, the dye and block are removed, and the cycle repeats.

Reads are short, typically 100 to 300 bases, and are often paired end (both ends of each fragment are read), which helps place them. Per base accuracy is about 99.9 percent. Short reads are ideal for resequencing, where reads are aligned to a reference genome to find variants, and for counting molecules, as in RNA-seq.

Long read sequencing

Two technologies read single molecules thousands of bases long. In single molecule real time sequencing, one polymerase sits at the bottom of a tiny well and the flash of each fluorescent nucleotide is recorded as it is incorporated. In nanopore sequencing, a single DNA strand is pulled through a protein pore in a membrane, and the change in ionic current reveals the bases in the pore; reads can exceed 100 kb, and modified bases such as 5-methylcytosine can be detected directly. Raw error rates were once high but repeated reading of the same molecule now gives accuracy above 99.9 percent.

Coverage and assembly

Because fragments are sampled randomly, some positions are read many times and others not at all. Coverage (depth) is the average number of reads covering a base: C = L × N ÷ G, where L is read length, N the number of reads and G genome size. If reads land at random, the number covering a given base follows a Poisson distribution, so the fraction of bases with zero coverage is about e^−C (the Lander-Waterman estimate). Clinical genome sequencing typically aims for 30x or more so that both alleles at a heterozygous site are sampled many times.

Putting reads back together into a genome is assembly. De novo assembly builds the genome from overlaps among reads alone, producing contigs (continuous sequences) that are linked into scaffolds using paired reads or long reads. Reference based assembly aligns reads to an existing genome of the same species. The great obstacle is repeats: if a repeated element is longer than the reads, the assembler cannot tell which copy a read came from, and the assembly breaks. This is why the first human reference left roughly 8 percent of the genome, mostly centromeres and other long repeats, unfinished until long reads produced the first complete, telomere to telomere human sequence in 2022.

Section 3CRISPR and genome editing

From bacterial immunity to a programmable tool

Many bacteria and most archaea carry CRISPR arrays (clustered regularly interspaced short palindromic repeats): short repeats separated by unique spacers, many of which match sequences from viruses and plasmids. Nearby lie cas (CRISPR associated) genes. Together they form an adaptive immune system. When a cell survives infection, Cas proteins snip a piece of the invader's DNA and insert it as a new spacer at the front of the array, a heritable record of the encounter. The array is transcribed and processed into short CRISPR RNAs (crRNAs), each carrying one spacer. A crRNA guides Cas nucleases to destroy any matching DNA that enters later.

In 2012 Jennifer Doudna and Emmanuelle Charpentier showed that the Cas9 nuclease of Streptococcus pyogenes could be directed to cut any chosen DNA sequence in a test tube using a guide RNA, and within months others showed it worked in human cells. Earlier engineered nucleases, zinc finger nucleases and TALENs, recognized DNA through custom built proteins, which had to be redesigned for every target. Cas9 is retargeted simply by changing 20 nucleotides of RNA.

How Cas9 finds and cuts its target

In the lab, the natural crRNA and its partner tracrRNA are fused into a single guide RNA (sgRNA) of about 100 nucleotides. Its first 20 nucleotides, the spacer, match the target. Cas9 first scans DNA for a short protospacer adjacent motif (PAM), which for S. pyogenes Cas9 is 5′-NGG-3′ (N is any base). On finding a PAM, Cas9 unwinds the adjacent DNA and tests whether the guide pairs with it. If the 20 nt protospacer immediately 5′ of the PAM matches the guide, the two nuclease domains of Cas9 (HNH and RuvC) each cut one strand, producing a blunt double strand break 3 bp upstream of the PAM. The PAM requirement is how bacteria avoid cutting their own CRISPR arrays, which contain the spacer but not the PAM. Mismatches near the PAM, in the so called seed region of the guide, block cutting more than mismatches far from it.

Repairing the break: knockouts and knock-ins

Cas9 only cuts; the cell's own repair pathways decide the outcome. Nonhomologous end joining (NHEJ) rejoins the broken ends directly and often adds or removes a few bases. Because the edited site is still a target, cutting and repair repeat until an indel destroys the site. In a coding exon, frameshifting indels create premature stop codons and usually a knockout. Homology directed repair (HDR) uses a template with sequence matching both sides of the break; if researchers supply a donor DNA carrying a desired change between homology arms, the cell copies it in, giving a precise knock-in. HDR is active mainly in dividing cells in S and G2 and is less efficient than NHEJ.

Beyond cutting

CRISPR is also used as a screening tool: libraries of tens of thousands of guides, one or more per gene, are delivered to a population of cells, which are then selected (for example, for drug resistance), and sequencing the guides in survivors reveals which gene knockouts matter. In agriculture, editing has produced disease resistant and nonbrowning crops. A gene drive places the Cas9 and guide genes inside the target site itself, so a heterozygote converts its wild type allele and passes the drive to nearly all offspring; drives are being tested to suppress malaria carrying mosquitoes.

Editing raises safety and ethical issues. Off target cutting at similar sequences, large deletions at the target, and mosaicism when editing happens after the first cell division are technical risks. Editing of somatic cells affects only the patient, whereas germline editing of embryos would be inherited. When edited babies were announced in 2018, the work was widely condemned as unsafe and unjustified, and most countries restrict heritable editing.

Section 4Genome structure and annotation

Genome size and gene number

Genomes vary enormously in size, but the number of genes varies much less, and neither tracks organismal complexity well. The lack of correlation between genome size (the C-value, the DNA content of a haploid genome) and complexity is called the C-value paradox. Some amoebae, lungfish and lilies have genomes many times larger than ours. The paradox is resolved by noncoding DNA: genome size differences come mostly from repetitive sequences and transposable elements, not from extra genes.

In bacteria, genes are packed densely, with about one gene per kilobase and short intergenic spacers. In humans, the average protein coding gene spans about 27 kb, but its mature mRNA is only about 3 kb, because most of the gene is introns; the average gene has about 9 exons. Only about 1.5 percent of the human genome encodes protein. Roughly another 5 to 10 percent is under evolutionary constraint, including regulatory elements and noncoding RNA genes. Humans also have about 20,000 noncoding RNA genes, including those for rRNA, tRNA, snRNA, microRNAs and long noncoding RNAs.

Repetitive DNA and transposable elements

About half of the human genome consists of transposable elements, sequences that move or once moved within the genome (Chapter 7), almost all of them now inactive.

• LINEs (long interspersed nuclear elements) are retrotransposons of about 6 kb that encode their own reverse transcriptase. LINE-1 makes up about 17 percent of the genome, and a few dozen copies remain active.
• SINEs (short interspersed nuclear elements), about 300 bp, are copied using LINE enzymes. The Alu element alone has over a million copies, about 10 percent of the genome.
• LTR retrotransposons, including endogenous retroviruses, make up about 8 percent, and DNA transposons about 3 percent.

Other repeats include satellite DNA (tandem arrays at centromeres, such as the 171 bp alpha satellite), minisatellites and microsatellites (short tandem repeats used in forensic profiling), and segmental duplications, long blocks over 1 kb with more than 90 percent identity that make up about 5 percent of the genome and promote rearrangements by unequal crossing over. Gene families arise by duplication, and many duplicates decay into pseudogenes; processed pseudogenes, lacking introns and ending in a poly(A) stretch, are mRNA copies reinserted by LINE reverse transcriptase.

Annotation: finding the genes

A raw genome sequence is a string of letters. Annotation identifies the genes and other functional elements in it. Structural annotation marks where genes, exons, introns, UTRs and regulatory elements lie; functional annotation assigns probable roles, usually by similarity to known proteins and domains.

The simplest clue to a protein coding gene is an open reading frame (ORF): a stretch beginning with a start codon (ATG) and continuing without a stop codon (TAA, TAG or TGA). Each double stranded sequence has six possible reading frames, three on each strand. In random sequence, 3 of the 64 codons are stops, so stops occur on average every 64 ÷ 3 = 21 codons. A long ORF is therefore unlikely by chance and strongly suggests a real gene. In bacteria, finding long ORFs and a ribosome binding site upstream identifies most genes. In eukaryotes, introns break ORFs into pieces, so gene finding programs combine several kinds of evidence:

• Ab initio predictions use statistical signals: splice site consensus sequences (GU at the 5′ end of an intron and AG at the 3′ end), codon usage, start and stop codons and promoter features.
• Transcript evidence from cDNA and RNA-seq shows exactly which sequences are present in mature mRNA, revealing exon boundaries and alternative isoforms.
• Homology evidence aligns known proteins and the genomes of related species; sequences conserved across species are likely to be functional.

Annotation also uses CpG islands, chromatin marks such as H3K4me3 at promoters and H3K27ac at enhancers (Chapter 9), and DNase or ATAC accessibility. The project that mapped these features across many human cell types concluded that a large fraction of the genome shows some biochemical activity, though how much of that activity matters for fitness is debated; evolutionary constraint suggests that under 10 percent is functional in that stricter sense.

Organelle genomes

Mitochondria and chloroplasts carry their own small circular genomes, relics of their bacterial ancestors. Human mitochondrial DNA is 16,569 bp and encodes 37 genes: 13 proteins of the respiratory chain, 22 tRNAs and 2 rRNAs. It has almost no noncoding DNA apart from the control region (D-loop), is present in hundreds to thousands of copies per cell and is maternally inherited. Most mitochondrial proteins, more than a thousand, are encoded in the nucleus and imported, showing that genes have moved from the organelle to the nucleus over evolutionary time.

Section 5Functional and comparative genomics

From sequence to function

Functional genomics asks what all the genes and elements of a genome do, on a genome wide scale. Two broad strategies exist. Forward genetics starts from a phenotype, such as a mutant that cannot grow on lactose, and finds the responsible gene. Reverse genetics starts from a gene sequence and disrupts it, with knockouts, RNAi or CRISPR, to see what phenotype results. Genome wide versions of reverse genetics include collections of yeast strains with every gene deleted and pooled CRISPR screens in human cells. Such screens show, for example, that only about 1,000 of yeast's 6,000 genes are essential in rich medium, and that about 2,000 human genes are essential for growth of a typical cell line.

Measuring the transcriptome and more

The transcriptome is the set of RNAs present in a sample. RNA-seq converts RNA to cDNA, sequences it, and maps reads to the genome; the number of reads per gene estimates its expression. Because samples are sequenced to different depths and genes differ in length, counts are normalized, for example as counts per million (CPM) or transcripts per million (TPM), which also corrects for length. Changes are usually reported as log₂ fold change, so that a doubling is +1, a halving is −1 and no change is 0. Statistical tests across biological replicates determine whether a change is significant, and the large number of genes tested requires correction for multiple comparisons. Single cell RNA-seq profiles thousands of individual cells and has revealed new cell types in many tissues. Older microarrays measured hybridization of labeled cDNA to probes on a chip.

Other genome wide methods map where proteins bind and which regions are open. ChIP-seq (chromatin immunoprecipitation followed by sequencing) uses an antibody to pull down a transcription factor or histone mark with its bound DNA. ATAC-seq uses a transposase that inserts adapters preferentially into open chromatin, mapping accessible regulatory elements. Proteomics identifies and quantifies proteins by mass spectrometry. Combining these layers lets researchers trace how a variant in an enhancer changes chromatin, transcription and finally protein levels.

Comparative genomics: homologs, orthologs and paralogs

Comparative genomics compares genomes across species to infer function and evolutionary history. Genes related by common descent are homologs. Homologs come in two kinds that students often confuse:

• Orthologs are genes in different species that descend from a single gene in their last common ancestor and were separated by a speciation event. Human and mouse β-globin are orthologs. Orthologs usually keep the same function, which is why studying a gene in mice or yeast tells us about the human version.
• Paralogs are genes related by a duplication event, usually within one genome. Human α-globin and β-globin are paralogs. After duplication, one copy may keep the original function while the other is lost, divides the original functions (subfunctionalization) or gains a new one (neofunctionalization).

Synteny, the conservation of gene order on chromosomes, helps identify true orthologs and reveals rearrangements: large blocks of human chromosomes correspond to blocks of mouse chromosomes with their genes in the same order. Sequence conservation outside genes points to function. Some ultraconserved elements are identical over 200 bp or more between human, mouse and rat, and many act as developmental enhancers.

Reading selection from sequence: dN/dS

Comparing coding sequences of orthologs reveals what kind of selection has acted. Substitutions are divided into synonymous (silent, not changing the amino acid) and nonsynonymous (changing it). dS is the number of synonymous substitutions per synonymous site, and dN the number of nonsynonymous substitutions per nonsynonymous site. Because synonymous changes are nearly neutral, dS approximates the neutral rate. The ratio ω = dN/dS then indicates selection on the protein: ω < 1 means purifying selection removed most amino acid changes (the usual case); ω ≈ 1 suggests neutral evolution, as in pseudogenes; ω > 1 indicates positive selection favoring amino acid change, as in some immune and reproductive genes.

Genomes of populations and communities

Sequencing many individuals of a species reveals variation. Population projects have catalogued hundreds of millions of human single nucleotide variants, most rare, which underpin association studies (Chapter 11) and clinical interpretation of variants. A species' pangenome includes genes present in some individuals but not others; in bacteria the variable portion can be larger than the shared core. Metagenomics sequences DNA directly from environmental samples, such as soil, seawater or the human gut, identifying organisms that cannot be cultured and showing that the gut microbiome carries far more genes than the human genome.

Words worth knowing