MLB7052.LT: Modern Methods in Molecular Ecology
TalTech / UT Institute of Technology
Loeng 11 — valiku lugemine, genoomi muutmine ja see, mis pärandub.
Last week’s MK test and dN/dS went past too fast. We start again from the beginning, slowly, with no equations until the idea is already obvious. Then the same treatment for LAMP and for CRISPR-Cas9 — and then phylogeography, behaviour, and what gets inherited without being written in the DNA at all.
| 1 | The MK test and dN/dS — from scratch |
| 2 | LAMP — copying DNA without a thermal cycler |
| 3 | CRISPR-Cas9 — find, cut, replace (and a mammoth) |
| 4 | Phylogeography — and why the marker’s address matters |
| 5 | Behaviour, molecularly |
| 6 | Epigenetics — inheritance without sequence change |
Every protein-coding gene is read in codons — three letters per amino acid. The code is redundant: several codons mean the same amino acid.
| Codon change | Protein | Called | |
|---|---|---|---|
GCC → GCA |
Ala → Ala (no change) | synonymous (sünonüümne) | |
GCC → GTC |
Ala → Val (changed) | non-synonymous (mittesünonüümne) |
Tähtis
This is the whole trick. Synonymous changes are invisible to natural selection — the protein is identical, so selection has nothing to grip. Non-synonymous changes are visible — they alter the protein, so selection can keep them or throw them away.
Imagine a book copied by hand, over and over, for a thousand years. Copying errors creep in. Two kinds:
“colour” → “color”
The meaning is unchanged. No editor ever stops this. It accumulates at whatever rate the scribes make mistakes.
This is a synonymous change. It is your measuring stick — it tells you how fast mistakes happen when nobody is checking.
“the king is dead” → “the king is deaf”
The meaning changed. Now an editor has an opinion. Most such errors get corrected out. Occasionally one survives because someone liked it better.
This is a non-synonymous change. It is the one selection acts on.
Märkus
So: count both kinds. The silent ones tell you the background rate of copying errors. Any excess or shortage of meaning-changing ones, measured against that background, is the fingerprint of selection.
“Per opportunity” matters: there are far more ways to change meaning than not, so both counts are divided by the number of sites at which each kind could happen.
Most genes, most of the time, sit well below 1. Purifying selection is the normal state of the genome — the editor is almost always working.
dN/dS compares two species. The MK test adds a second axis, and that is what makes it powerful:
Within one species — polymorphism
Differences that are still segregating: some individuals have A, others have B. These are changes on probation — they arose recently and have not yet been accepted or rejected.
Between two species — divergence
Differences that are fixed: every individual of species 1 has A, every individual of species 2 has B. These changes have already been accepted — permanently.
Tähtis
The neutral theory makes one sharp prediction: if nothing but drift is happening, the mix of silent-to-meaningful changes should be the same in both columns. Probation and acceptance should not change the recipe.
Left: the same recipe in both columns — nothing to explain. Right: the real Adh data. Among changes still on probation, almost none alter the protein. Among changes that got accepted, a third do. Protein-altering changes are being accepted far more readily than they are tolerated while segregating. That is positive selection.
| Fixed | Polymorphic | |
|---|---|---|
| Non-synonymous | 7 | 2 |
| Synonymous | 17 | 42 |
Read it as two sentences in plain language:
An editor that rejects changes day-to-day but has permanently accepted a pile of them was not rejecting them at random. Something was actively favouring protein change.
Roughly 88% of the accepted protein changes at Adh were driven by positive selection, not drift.
What if is bigger than — more protein-changing variants on probation than ever get accepted?
That is the common case, and it means: slightly harmful variants are floating around in the population but never make it to fixation. They are on probation and they will fail. , negative (purifying) selection.
Tähtis
The trap: a population that recently crashed, or recently expanded, produces the same pattern without any selection at all, because drift lets mildly harmful variants rise. This is why every test of selection has to be read against the genome-wide background of the same population — never a single gene on its own.
Precisely because they do nothing. They record how fast mutations arrive and drift to fixation when selection is not involved — the background rate. Without them you have a count of protein changes with nothing to compare it against, and no way to tell a fast-mutating gene from a positively selected one.
The opposite. ω = 0.05 means 95% of protein-changing mutations are being removed by selection — the gene is under strong purifying selection, which is what you expect for something essential. Low ω means important and constrained, not inactive.
Because the average hides the sites. Positive selection usually acts on a handful of codons while the rest of the gene stays under purifying selection, so a genuine signal is often diluted below 1 — and an average above 1 across an entire gene is more often a sign of a pseudogene, poor alignment, or paralogues mistaken for orthologues.
— protein-changing polymorphism. Drift in a small population lets slightly deleterious variants reach appreciable frequency instead of being purged, which raises and pushes above 1, mimicking purifying selection and masking any positive selection that is really there.
PCR separates the two DNA strands by heating to 95 °C, then cools to let primers bind, then warms to extend. Thirty times. That needs a thermal cycler: mains power, moving parts, and about an hour.
LAMP never heats the DNA apart. It uses Bst polymerase, which does something Taq cannot: as it copies along, it shoves the strand in front of it out of the way. Strand displacement replaces the 95 °C step.
One temperature. A water bath, a heat block, or a cup of hot water will do.
PCR must climb to 95 °C and back, over and over. LAMP sits still.
LAMP uses six primers recognising eight regions of the target. That is why it is so specific — eight independent matches is a far higher bar than PCR’s two — and also why it is hard to design.
The key event: the inner primer FIP carries a sequence (F1c) that is complementary to a part of the new strand it is about to make. So when the strand is displaced, its own end folds back and pairs with itself.
The product becomes a dumbbell — a loop at each end. And a loop is a permanently available primer site that never needs melting open.
F3/B3 outer · FIP/BIP inner · LF/LB loop primers.
Tähtis
Once the dumbbell exists, copying never stops to be re-started. The structure keeps presenting new 3′ ends to the polymerase, so amplification is continuous rather than cyclic — which is why LAMP reaches a usable amount of DNA in 15–30 minutes.
Amplifying that hard has a convenient side effect: it makes a lot of product, and a lot of by-product.
No instrument. No gel. A result a non-specialist can read, which is exactly what citizen science and field diagnostics need.
Märkus
The cost of all that sensitivity: LAMP contaminates easily. A single opened tube of finished product aerosolises billions of copies, and every subsequent reaction in that room is positive. Colour-change LAMP is read without opening the tube, deliberately.
Think of the genome as an enormous text file and CRISPR-Cas9 as find-and-replace.
| Part | What it is | In the analogy |
|---|---|---|
| Guide RNA (sgRNA) | ~20 bases that match your target | the search term you type |
| Cas9 | a protein that cuts both DNA strands | the scissors |
PAM (NGG) |
a 3-letter motif that must sit right after the target | the rule that the search only works if the match is followed by GG |
| Repair template | DNA you supply, carrying the change | the replacement text |
Tähtis
Cas9 cannot cut without a PAM. This is not a technical nuisance — it is why the bacterium does not destroy its own CRISPR array: the stored spacers have no PAM beside them, so the scissors ignore them.
The cut is the easy part. What the cell does next decides the outcome:
NHEJ — non-homologous end joining
The cell shoves the two ends back together, usually losing or adding a few bases. The reading frame breaks, and the gene is destroyed.
Use this when you want to switch a gene off. It is the default, it is efficient, and it is sloppy.
HDR — homology-directed repair
If you supply a template whose ends match the sequence either side of the cut, the cell copies your version in.
Use this when you want to write a specific change. It is precise, much less efficient, and mostly works in dividing cells.
Remember last week: mammoth haemoglobin works at low temperature, because three amino-acid substitutions make O₂ release less temperature-sensitive. One of them is at position 12 — threonine in elephant, alanine in mammoth.
So “making a piece of mammoth genome” is not science fiction. It is one letter.
Asian elephant codon 12: ACC = Thr Mammoth codon 12: GCC = Ala
Change one A to a G.
Campbell et al. 2010 — the protein was resurrected from sequence and tested in the lab before anyone tried to edit a cell.
SCHEMATIC — the codon change is real, the flanking bases are illustrative
elephant HBB/HBD, around codon 12
... C T G A C C G A G G A G A A G ...
▲ ▲
codon 12 = ACC PAM = AGG
(threonine)
guide RNA : 20 bases matching the strand just left of that AGG
Cas9 cuts : 3 bases upstream of the PAM ──✂──
template : the same ~100 bases, but with ACC rewritten as GCC
plus a silent change in the PAM so Cas9 cannot re-cut
... C T G G C C G A G G A G A A G ...
▲
codon 12 = GCC → alanine → mammoth-type haemoglobin
Märkus
Design the guide against the real sequence, not this sketch — the flanking bases here are made up so the mechanics are visible. What is real is the substitution: Thr12 in elephant, Ala12 in mammoth, one base apart.
The silent change in the PAM is the practical trick worth remembering. Without it, Cas9 recognises the repaired sequence and cuts again — and the second repair is usually NHEJ, which destroys the gene you just fixed.
One codon gives you an elephant cell with mammoth-type haemoglobin. That is genuinely useful — and it is nothing like a mammoth.
What you can honestly claim is a cold-adapted elephant, edited at a handful of known loci.
The same protein family, pointed the other way: Cas12, Cas13 and Cas14 are the basis of CRISPR diagnostics.
Phylogeography asks where the branches of a gene tree sit on a map. Not just who is related to whom, but where their ancestors were, and when.
In the comparative version, you reconstruct the same history in several co-distributed species. Patterns they share are attributed to a shared event — an ice age, a land bridge, a drained strait. Patterns they do not share are about the species’ own biology.
The method is the same one you already know. What changes everything is which piece of DNA you read it from.
| mtDNA | Autosome | Y chromosome | |
|---|---|---|---|
| Inherited from | mother only | both parents | father only (sons) |
| Recombines? | no | yes | almost none |
| Copies per person | 1 lineage | 2 | 1 lineage (males) |
| Relative | 1 | 4 | 1 |
| Tells you about | female history | the population as a whole | male history |
Tähtis
Because there are four copies of each autosome for every one mitochondrial genome in a breeding pair, autosomal lineages take about four times longer to coalesce. mtDNA and Y reach a common ancestor fast; autosomes remember much deeper history.
Small means drift is fast. Alleles fix or disappear locally in a few thousand generations, so populations diverge quickly and at mtDNA looks large. Large means drift is slow — the same populations may look barely differentiated at autosomal loci. Same populations, same history, two different answers. Neither is wrong.
This is the payoff, and it is why anyone bothers with three marker types.
Sex-biased dispersal leaves opposite fingerprints:
In many whales, females return to their mother’s feeding ground while males roam: mtDNA says “separate populations”, nuclear markers say “one”. Both are right — the females are separate, the genes are not.
Märkus
Humans: the mtDNA most recent common ancestor (“mitochondrial Eve”) and the Y one (“Y-chromosomal Adam”) are estimated at different dates — and neither is the ancestor of the species.
They are the coalescence points of two single loci, each with four times smaller than an autosome. Every other gene you carry has its own ancestor, at its own date.
Tähtis
A single-locus tree is not a species tree. One mtDNA gene gives you one history, with no replication. The whole point of nuclear markers is that you get hundreds of independent ones.
Cockroach baits worked by hiding poison in glucose syrup. Within a few years, in many populations, they stopped working.
The reason is not poison resistance. It is taste. In these populations, glucose now activates the bitter-sensing neurons as well as the sweet ones — so the roach tastes the bait as bitter and refuses it.
A change in a sensory receptor’s wiring, driven by human pest control, strong enough to show up as a behaviour in a few years.
The figure: wild-type roaches eat glucose happily; insecticide-selected ones avoid it.
Drosophila larvae develop in rotting fruit, which contains alcohol. Parasitoid wasps lay eggs in them.
Larvae that eat food containing 6% alcohol survive wasp attack far better than those on 0% — the alcohol is more toxic to the wasp than to the fly.
And the larvae seek out alcohol when they have been attacked. A behaviour, deployed as medicine, against a specific enemy.
Märkus
This is the shape of most molecular behavioural ecology: a behaviour that looks like a quirk turns out to be an adaptation whose mechanism is a measurable physiological tolerance.
Oldfield mice dig long burrows with an escape tunnel; deer mice dig short simple ones. Hybrids dig the long version with an escape tunnel — so the complex architecture is dominant.
Cross them, map the offspring, and the behaviour splits into a handful of genetic modules: one region controls whether an escape tunnel is built at all, three more control the length of the entrance tunnel.
A mouse has never seen a burrow before it digs one. The architecture is in the genome, and it is modular — length and escape tunnel are inherited separately.
Song sparrow parents that merely hear recordings of predators — no predator ever present — feed their young less often. The nestlings grow up colder, fewer survive, and the effect persists into the next stage.
No gene changed. No animal was eaten. The environment reached the offspring through the parent’s behaviour and physiology — which is exactly the doorway into the last section.
Epigenetics (epigeneetika): heritable changes in gene activity that do not change the DNA sequence.
The sequence is the text. Epigenetic marks are the annotations — highlighter, sticky notes, and pages glued together — that decide which parts get read in which cell.
| Mark | What it does |
|---|---|
| DNA methylation | A methyl group on cytosine, usually at CG. Heavily methylated promoters are generally silent |
| Histone modification | Chemical tags on the proteins DNA is wound around, loosening or tightening access |
| Small RNAs | Short RNAs that target specific transcripts for silencing, and can direct methylation |
Tähtis
Every cell in your body has the same genome and behaves completely differently. That difference is epigenetic. Epigenetics is not exotic — it is how a liver happens.
For a mark to pass to the next generation, it must survive two rounds of deliberate erasure: once in the germ cells, once just after fertilisation. The embryo wipes the annotations so it can start from a clean text.
Within a lifetime — uncontroversial. Marks are copied faithfully from a cell to its daughters every time a cell divides. This is ordinary developmental biology.
Across generations — only where a mark escapes the reprogramming. Well documented in plants and some invertebrates; genuinely contested in mammals, where the erasure is most thorough.
Märkus
Be careful with the word “inherited”. Much of what is reported as transgenerational inheritance in mammals is a maternal effect: the mother’s physiology shaping the offspring directly, which looks identical in one generation and disappears by the third. The test is whether the effect is still there in the F3 generation, which never existed as a cell inside the exposed mother.
The mouse carries a transposon inserted beside the agouti coat-colour gene. How methylated that transposon is determines how much agouti is expressed:
Genetically identical mice, in the same cage, in a full spectrum from yellow to brown.
Feed the mothers extra methyl donors — folate, B12, choline, betaine — and the litters shift towards brown. Diet during pregnancy, changing coat colour and lifelong health, without touching the sequence.
Tähtis
This is the cleanest demonstration there is that the same genome can produce different animals, and that the switch is chemical, environmental and heritable through at least one generation.
Linnaeus described a toadflax (Linaria vulgaris) with radially symmetric flowers instead of the usual bilateral ones, and called it peloria — “monstrous”. It bred true. Darwin was fascinated by it.
In 1999 the cause was finally found: the Lcyc gene, which controls flower symmetry, is heavily methylated and silenced. The DNA sequence is unchanged.
The variant has been propagating in the wild since at least the 1740s, and it occasionally reverts when the methylation is lost.
An epiallele behaving as a classical Mendelian mutant for two and a half centuries.
Märkus
Plants make this easy for two reasons: their germline is set aside late, from ordinary somatic tissue, so marks acquired during growth can reach the gametes — and plant embryos do not perform the thorough genome-wide erasure that mammals do.
If you want to study transgenerational epigenetics, do not start with a mouse.
Tähtis
Back to last week’s question — plasticity or adaptation? Epigenetics is why that question needs an experiment and not a sequencer.
The answer still comes from a common garden: raise both populations in identical conditions for a generation or more. What persists is heritable, whether the mechanism is sequence or methylation. What disappears was plasticity.
| 1 | Synonymous changes are the measuring stick; non-synonymous changes are what selection sees. dN/dS and the MK test are just that comparison, done two ways |
| 2 | MK adds a second axis — on probation (polymorphism) vs accepted (divergence). A mismatch between the two recipes is selection |
| 3 | LAMP replaces 95 °C with a strand-displacing polymerase and a self-priming loop — one temperature, a colour change, no instrument |
| 4 | CRISPR-Cas9 is find-and-replace: guide RNA finds, Cas9 cuts, PAM licenses, and the repair pathway you get decides whether you broke a gene or rewrote it |
| 5 | mtDNA, Y and autosomes have different and different parents — so they tell different, equally true, histories |
| 6 | Epigenetic marks change which genes are read, respond to the environment within one generation, and sometimes survive into the next |
Tähtis
peeter.laas@taltech.ee · peeter.laas@ut.ee
Kodune ülesanne: the MK test and dN/dS — work through the Adh table yourself, and bring one question about it to the next session.
Seminarid algavad 06.11 — seitse ettekannet korraga, 12 minutit koos küsimustega. Kava on Drive’is.
Fall 2026