21 — Bacterial gene regulation
Before this: Ch 05 Transcription · Ch 07 The genetic code and translation · Ch 08 Proteins and gene function · Ch 20A Bacterial and phage genetics — §5's entire method is the F′ built there · Time: ~45 min
Regulation appears here in its clearest form. A bacterium has a small genome, no nucleus, no chromatin, a doubling time of twenty minutes, and a fluctuating food supply. Everything it does about gene expression is visible, measurable, and — unusually for biology — genuinely close to a logic circuit. Learn the logic here, where it is uncluttered, and Chapter 22 will be a matter of adding complexity to a skeleton you already have.
What you'll be able to do
- Predict lac expression for any combination of sugars and any , including partial — and explain why the system cannot bootstrap without leaky basal expression
- Write the lac truth table and show why it approximates an AND gate — and where the approximation fails
- Distinguish cis-acting from trans-acting elements — including the case where the cis element is the mRNA itself — and design the test that separates a broken from a broken operator
- Explain why -mediated is mechanically possible in bacteria and impossible in eukaryotes
- Place any regulatory mechanism on the two orthogonal axes: negative/positive control, inducible/repressible
- Trace the signal path through a σ-factor regulon, a two-component His→Asp relay and a quorum-sensing loop, and say what each system is actually measuring rather than what it appears to sense
- Classify every control point a bacterium uses, from σ-factor choice down to mRNA folding, by the timescale it acts on — and choose negative or positive control for a gene from its duty cycle, noise floor and failure mode
The core idea
A gene is not a switch. It is a rate, and regulation is a set of multiplicative factors on that rate.
The rate at which fires from a depends on what else is bound nearby. A protein sitting on the DNA where polymerase must sit lowers the rate. A protein that grabs polymerase and holds it against a weak promoter raises it. That is nearly the entire mechanism, and every regulatory story in this chapter is a variation on what makes those proteins bind or let go.
The second idea is architectural, and it outlives every increase in biological complexity:
Some regulatory elements are sequences and some are diffusible molecules, and this distinction is absolute. A sequence — a promoter, an operator — can only affect the DNA molecule it physically sits on. It is cis-acting. A gene product that leaves its template and floats around the cell affects every copy of its target. It is trans-acting. In programmer's terms: cis elements are fields on an instance; trans factors are globals in shared memory. Break a global and every instance misbehaves. Break a field and only that instance does.
The classical experiments in this chapter exist to read out that distinction — by putting two copies of the same in one cell and asking which one broke.
1. Why regulate: the arithmetic of waste
Protein synthesis dominates a bacterium's energy budget. Building one peptide bond costs roughly four high-energy phosphate bonds before you count the cost of making the amino acid itself, so a thousand-residue enzyme is a five-figure ATP investment, repeated for every copy. Expressing a full sugar-catabolism that you have no use for is not free.
Suppose the waste costs a lineage 5% of its growth rate — a plausible figure for a strongly expressed unneeded operon. That sounds negligible. It isn't, because generations are cheap:
relative fitness per generation: 0.95
after 100 generations: 0.95^100 = e^(100 × ln 0.95) = e^(−5.13) ≈ 0.006
The thrifty strain outnumbers the wasteful one ~170 : 1. At a 20-minute doubling time, 100 generations is about 33 hours. Selection on regulation is not a slow background pressure; on a bacterial clock it is brutal and fast. This is why bacterial genomes are regulatorily tight in a way vertebrate genomes are not — the population sizes and generation times make a fraction of a percent visible to selection (Ch 27).
Economy is not the only reason. Some products are toxic in the wrong context, some must be built in a strict order (you cannot make flagellin before you have a flagellum to put it in), and some are only useful if enough neighbours are doing the same thing.
2. The operon: one promoter, several genes
A bacterial often serves a run of adjacent genes, transcribed into a single polycistronic mRNA — one RNA molecule carrying several complete coding sequences. That unit is an .
This works because bacterial do not have to start at the 5′ end. Each coding sequence carries its own Shine–Dalgarno sequence a few bases upstream of its start , and a ribosome binds directly there (Ch 07). A single transcript therefore exports several independently translatable symbols. Eukaryotic ribosomes instead load at the and scan forward, so in practice only the first gets translated — which is the mechanistic reason operons are essentially a bacterial device, not a matter of bacteria being "simpler".
The payoff is coordinate regulation for free. One control point buys you the entire pathway, expressed together, in roughly fixed stoichiometry, with genes usually arranged in pathway order. In E. coli K-12 roughly 4,300 protein-coding genes are organised into on the order of 2,500 units, and around 60% of those genes sit in multigene operons.
One caveat worth carrying: an operon is not a rigid object. Internal promoters and internal terminators mean the same run of genes can be transcribed as several overlapping units, in different combinations, under different conditions. "The operon" is better thought of as the commonest transcript from a region than as a fixed record layout.
3. The lac operon in full
Lactose is a disaccharide — glucose joined to galactose. To eat it, E. coli needs to import it and cleave it. Three genes, one :
lacI CAP P O1 O2
────[=========]────[≈≈≈]──[−35 −10]───[▓▓▓]────[== lacZ ==]────────────────[▓▓▓]──[= lacY =]──[= lacA =]───
repressor ▲ ▲ ▲ β-galactosidase ▲ permease transacetylase
(own weak │ │ │ │
promoter) │ │ centred at +11 at +412, inside lacZ
│ RNA polymerase binds here
centred at −61.5 O3 at −82, overlapping the CAP site
(CRP–cAMP)
├──────────── one polycistronic mRNA ────────────────────────────▶
| Element | Product | Job | cis or trans? |
|---|---|---|---|
| lacZ | β-galactosidase | cleaves lactose → glucose + galactose | trans (enzyme) |
| lacY | lactose permease | membrane transporter, imports lactose | trans (but acts on the cell) |
| lacA | thiogalactoside transacetylase | detoxifies non-metabolisable galactosides; dispensable | trans |
| lacI | Lac | binds the operator, blocks | trans |
| P | — | binding site | cis |
| O1/O2/O3 | — | repressor binding sites | cis |
| CAP site | — | activator binding site | cis |
The repressor. LacI is a homotetramer — a dimer of dimers — present at only about ten tetramers per cell. Each half can grip an operator, so one tetramer bridges two operators and loops the DNA between them. This matters quantitatively: O1 by itself gives only about 20-fold repression; O1 with one auxiliary operator reaches ~440-fold (O3) or ~700-fold (O2), and all three together give ~1,300-fold. Looping is not decoration; it is where most of the repression comes from, and it is the first appearance of a trick that dominates eukaryotic regulation (Ch 22, Ch 50).
The inducer is not lactose. It is , an isomer produced as a side reaction by β-galactosidase itself. Read that again: the enzyme encoded by the makes the molecule that turns the operon on. The system cannot bootstrap from a true zero, which is why repression is deliberately leaky — a few molecules of LacZ and LacY persist even when fully repressed, and they are functionally required. Leakiness is a design , not a defect.
Allolactose binds LacI and shifts it allosterically, reorienting the DNA-binding domains and dropping operator affinity by around three orders of magnitude. Nothing is destroyed; an equilibrium moves (Ch 01 §5).
In the laboratory one uses IPTG, a gratuitous inducer: it binds LacI but is not a substrate, so it induces without being consumed and its concentration stays under your control. Decoupling the input from the output is what makes the lac system the workhorse of expression vectors (Ch 36).
The logic so far: negative control, inducible. Default ON; a trans-acting protein turns it off; a small molecule removes that protein.
4. Catabolite repression, and the AND gate
Glucose is a better carbon source, so a cell offered both sugars eats glucose first, pauses, then eats lactose — diauxic growth, two exponential with a lag between. The pause is the being built.
Two mechanisms enforce the preference, and neither involves glucose binding anything in the lac system. The signal is the act of transporting glucose.
Layer 1 — positive control by CRP. E. coli imports glucose through the phosphotransferase system, which consumes phosphoryl groups from a protein called EIIAGlc. Phospho-EIIAGlc activates adenylate cyclase; so while glucose is flowing, EIIAGlc is mostly unphosphorylated and cAMP falls. cAMP is the ligand for CRP (cAMP receptor protein, also called CAP, catabolite activator protein — one protein, two names from two labs). CRP–cAMP binds a site centred at −61.5, bends the DNA sharply, and contacts , recruiting it. The lac is intrinsically weak — it has a poor −35 element — so without CRP it barely fires at all. The weakness is the point: it leaves something for an activator to do.
Layer 2 — inducer exclusion. Unphosphorylated EIIAGlc also binds LacY directly and shuts the permease. No lactose in, no , no induction. How the two layers divide the labour is still argued over — inducer exclusion is the larger contributor during glucose–lactose diauxie, but experiments show it cannot account for the full effect alone. Both are real.
The truth table, and why "AND" is an approximation
| Lactose | Glucose | Operator | CRP–cAMP | lacZYA |
|---|---|---|---|---|
| − | + | occupied | absent | off |
| − | − | occupied | present | off |
| + | + | mostly occupied (inducer exclusion) | absent | barely on |
| + | − | free | present | ON |
expression = lactose AND NOT glucose.
Now derive it rather than asserting it. Model the firing rate as a product of two occupancies:
rate ∝ f_free × A f_free = fraction of time O1 is unoccupied
A = 0.03 + 0.97·p_C (promoter is ~3% as strong
without CRP as with it)
with illustrative occupancies: f_free = 0.001 (no lactose), 0.15 (lactose + glucose, partial induction), 0.95 (lactose alone); p_C = 0.05 (glucose present), 0.90 (absent).
| Lactose | Glucose | f_free | p_C | A | rate | normalised |
|---|---|---|---|---|---|---|
| − | + | 0.001 | 0.05 | 0.079 | 7.9 × 10⁻⁵ | 0.00009 |
| − | − | 0.001 | 0.90 | 0.903 | 9.0 × 10⁻⁴ | 0.0011 |
| + | + | 0.15 | 0.05 | 0.079 | 1.2 × 10⁻² | 0.014 |
| + | − | 0.95 | 0.90 | 0.903 | 8.6 × 10⁻¹ | 1.00 |
Dynamic range ≈ 11,000-fold — genuinely switch-like. But the three "off" states differ from each other by a factor of ~150. This is an analogue system with a large dynamic range, not a digital gate, and the difference shows up as the diauxic lag not being perfectly flat.
What makes it look digital in a real cell is feedback: more permease → more lactose → more allolactose → less → more permease. That positive loop, on top of cooperative repressor binding, produces a sigmoidal response and, over a range of inducer concentrations, bistability — individual cells are fully induced or fully off, and the smooth population average is a mixture of two discrete states. A bulk measurement of this system reports a number that no single cell exhibits (Ch 48).
5. Merodiploids: how to tell cis from trans
This is the classical exam problem, and it is the clearest demonstration of the cis/trans distinction anywhere in genetics.
The F factor is a conjugative plasmid that can pick up a chunk of the . An F′lac carries a second copy of the lac region, so a cell holding one is a partial () — diploid for lac and haploid for everything else. You now have two instances of the same object in one address space, and you can break them independently.
Where the merodiploid comes from. Conjugation, the F factor, how an Hfr forms and how aberrant excision generates an F′ carrying a defined chunk of chromosome are derived in Ch 20A §§3–4. That chapter also runs the in its general bacterial form; this section is that test applied to one , and the cis/trans algorithm below is its specialisation.
| Lesion | Haploid | Behaviour in a merodiploid | |
|---|---|---|---|
| lacI⁻ | protein non-functional | constitutive | — a wild-type copy supplies repressor in trans |
| lacIˢ | repressor cannot bind inducer | uninducible | — see below |
| lacOᶜ | operator sequence altered; repressor cannot bind | constitutive | cis-dominant — affects only its own copy |
| lacP⁻ | defective | no expression | cis-dominant — affects only its own copy |
| lacZ⁻ | enzyme dead | no β-galactosidase | recessive |
lacIˢ deserves a note, because "dominant" here has a specific mechanism. LacI is a tetramer, so in a merodiploid the wild-type and superrepressor subunits assemble into mixed tetramers. Those still bind the operator but cannot be released by inducer, so a few mutant subunits poison the whole pool. This is negative complementation — dominance arising from multimerisation, not from the mutant protein being more active.
The algorithm
Any merodiploid problem is solved the same way:
- Pool the trans products. Is there functional LacI anywhere in the cell? Is there any lacIˢ? (Any lacIˢ poisons the pool. Any lacI⁺ rescues a lacI⁻.)
- Then take each operon copy separately and check its own cis elements. Dead promoter → this copy makes nothing, full stop. Oᶜ → this copy ignores the repressor pool entirely.
- Then check the structural genes on that copy — which enzymes can this copy actually produce?
- Sum over copies. The assay reports the total.
Step 2 is where people go wrong. A cis mutation is scored on the copy it sits on and nowhere else, so its visibility depends on what genes are downstream of it.
6. The trp operon: the other allosteric sign
Five genes — trpE, trpD, trpC, trpB, trpA — encoding the enzymes that synthesise tryptophan. The logic is inverted relative to lac: this should run unless tryptophan is already available.
TrpR, the , cannot bind DNA on its own. Tryptophan binds it and enables DNA binding. Tryptophan is a corepressor.
So both LacI and TrpR are negative regulators, and both are controlled by a small molecule — but the signs are opposite. takes LacI off the DNA; tryptophan puts TrpR on it. That is why negative/positive and inducible/repressible are two independent axes, not one:
| Inducible (signal turns expression ON) | Repressible (signal turns expression OFF) | |
|---|---|---|
| Negative (regulator blocks) | lac — allolactose evicts LacI | trp — Trp recruits TrpR |
| Positive (regulator recruits) | ara — arabinose converts AraC to an activator; mal — maltotriose activates MalT | genuinely rare; no clean textbook example in E. coli |
AraC is often offered for that empty cell, and it does not belong there. Without arabinose AraC binds araO2 and araI1 and loops the ~210 bp between them, occluding the — that is negative control, and the signal (arabinose) still turns expression on, which is inducible by definition. AraC spans the two rows of this table, not the two columns: one protein that represses by looping and activates by recruiting.
TrpR gives roughly 70-fold control. A second mechanism supplies the rest.
7. Attenuation: coupling translation to transcription
Between the and trpE sits a leader of about 160 nucleotides, trpL. It contains a 14- :
Met Lys Ala Ile Phe Val Leu Lys Gly TRP TRP Arg Thr Ser stop
1 2 3 4 5 6 7 8 9 10 11 12 13 14
Tryptophan is the rarest amino acid in most proteomes — around 1% of residues. Two consecutive Trp codons in a 14-codon peptide is a screaming statistical anomaly. It is a sensor: the peptide is not the product, its rate is the measurement.
The leader RNA has four segments that can pair in two mutually exclusive ways:
5′──[ 1 ]────[ 2 ]────[ 3 ]────[ 4 ]────▶ trpE ...
▲ └──3:4──┘ intrinsic terminator (GC stem + U-tract)
│ └──2:3──┘ antiterminator — consumes segment 3
└─ the 14-codon ORF; Trp codons at 10, 11
Now the mechanism, which works only because in a bacterium a loads onto the transcript while is still making it:
Trp PLENTIFUL — charged tRNA-Trp abundant; ribosome sails through codons 10–11,
reaches the leader stop codon, and sits ON segment 2
── 1 ──[ RIBOSOME covers 2 ]── 3 ╲___╱ 4 3:4 terminator forms
RNAP releases at the U-tract
✗ trpEDCBA never transcribed
Trp SCARCE — uncharged tRNA-Trp; ribosome STALLS at codons 10–11,
sitting on segment 1; segment 2 is left free
──[ RIBOSOME on 1 ]── 2 ╲___╱ 3 ── 4 ── 2:3 antiterminator forms
3 is used up; 3:4 cannot form
✓ RNAP reads on into trpEDCBA
A third case makes the design obvious: with no ribosome at all, 1:2 and 3:4 both form and terminates. The default is off; it takes an actively stalled ribosome to switch it on.
contributes roughly a further 8–10-fold on top of TrpR's ~70-fold, for combined control of several hundred-fold. The his runs the same trick with a leader peptide containing seven consecutive histidine codons, and several other amino-acid operons follow suit.
Ribosome-mediated attenuation is impossible in a eukaryote, and the reason is architectural. In a eukaryote transcription happens in the nucleus and translation in the cytoplasm; the transcript is capped, spliced and exported before a ribosome ever touches it. No ribosome can be positioned on a nascent transcript to influence the polymerase making it. The nuclear envelope does not merely add a processing step — it removes an entire class of regulatory mechanism from the space of possibilities.
Be precise about which class. Attenuation in the wider sense — a leader RNA deciding whether an already-initiated transcript continues — is alive and well in eukaryotes: promoter-proximal premature termination of Pol II is widespread, and the eukaryotic TPP of §8 are exactly that decision made by RNA folding. What the envelope forbids is letting a ribosome on the nascent transcript be the thing that decides.
8. Riboswitches: regulation with no protein at all
Some mRNA leaders skip the protein intermediary entirely. A is a leader that folds into an aptamer which binds a metabolite directly, coupled to an expression platform whose fold is changed by that binding.
The specificity is not a compromise: riboswitches discriminate guanine from adenine — molecules differing at one functional group — with affinities in the nanomolar-to-micromolar range. RNA is a perfectly competent receptor.
The modularity is the elegant part. The same aptamer can be bolted to different platforms:
| Platform | Effect of ligand binding | Level controlled |
|---|---|---|
| Intrinsic terminator | ligand stabilises the terminator hairpin | (elongation) |
| Shine–Dalgarno-sequestering hairpin | ligand occludes the binding site | initiation |
| Self-cleaving ribozyme (glmS) | glucosamine-6-phosphate acts as coenzyme; mRNA cleaves itself | mRNA stability |
Known classes sense TPP (thiamine pyrophosphate — the most widespread, and the only class confirmed in eukaryotes, where plant and fungal versions act on and 3′-end processing), FMN, SAM, lysine, glycine, cobalamin, and purines. Almost all sit upstream of the biosynthetic or transport genes for the very metabolite they sense: direct end-product feedback with no in the loop, which is part of why riboswitches are read as an RNA World remnant — and why they are being pursued as antibiotic targets.
Note that the framework survives: the aptamer is cis in the purest possible sense — it is part of the very molecule it regulates — and the metabolite is trans.
RNA thermometers are the same idea with physics as the ligand. A hairpin occludes the Shine–Dalgarno sequence and melts at host body temperature, exposing it. Listeria uses one to switch on virulence genes at 37 °C. The sensor is the thermodynamics.
9. Sigma factors: switching the whole programme
Core (α₂ββ′ω) can transcribe but cannot find a . A σ factor binds core and confers promoter specificity — different σ, different consensus sequence, different gene set. Swapping σ is a global mode switch, not a per-gene decision.
E. coli has seven:
| σ | Gene | Regulon |
|---|---|---|
| σ⁷⁰ | rpoD | housekeeping — the default |
| σ³⁸ (σˢ) | rpoS | stationary and general stress |
| σ³² | rpoH | heat shock — chaperones and proteases |
| σ⁵⁴ | rpoN | nitrogen; mechanistically distinct (see below) |
| σ²⁸ | fliA | flagellar and chemotaxis genes |
| σ²⁴ (σᴱ) | rpoE | envelope / extracytoplasmic stress |
| σ¹⁹ | fecI | ferric citrate transport |
Three consequences.
σ factors compete for a shared pool of core polymerase. These are not independent channels. Inducing a stress regulon partially represses housekeeping simply by sequestering the enzyme — contention on a shared resource, exactly as in a fixed-size thread pool. Regulation happens both by changing σ abundance and by anti-σ factors that sequester them.
σ⁵⁴ is an obligate AND. Unlike the others it forms a stable closed complex that cannot melt the DNA by itself; it requires a bacterial -binding protein bound upstream, which loops around and hydrolyses ATP to open the complex. A σ⁵⁴ promoter therefore fires only when polymerase and an activator are both present — hard-wired coincidence detection.
The heat-shock loop senses demand, not temperature. σ³² is regulated two ways: its own mRNA carries an RNA thermometer that melts on heating, and free chaperones bind σ³² and deliver it to a protease. When unfolded protein accumulates, chaperones are titrated away, σ³² is stabilised, and more chaperones are made. The controller measures unmet demand for the output, which is a better sensor than temperature and requires no thermometer at all.
Cascades build state machines. Bacillus subtilis sporulation runs a σ cascade — σᶠ then σᴳ in the forespore, σᴱ then σᴷ in the mother cell — with signals passing between the two compartments so that each transition requires the other compartment to have completed the previous step. Flagellar assembly is a shorter and even cleaner example:
The checkpoint is physical. FlgM inhibits σ²⁸ and blocks flagellin transcription; the only way to remove FlgM is to pump it out through the completed export channel. The machine reports its own completion by whether it can flush its own inhibitor. Nothing measures anything.
10. Two-component systems: the generic sensor interface
Most environmental sensing in bacteria uses one repeated two-protein module.
The conserved core is a His→Asp phosphotransfer. Everything else is swappable: the input domain is whatever detects the signal, the output domain is whatever acts on it. This is a genuine interface, and evolution has done combinatorial assembly on it — E. coli K-12 alone carries about 30 histidine kinases and 32 response regulators, and many soil bacteria carry several times that.
Crosstalk between systems is suppressed kinetically, not structurally: cognate pairs have coevolved contact residues so a kinase phosphorylates its own partner far faster than any other. Specificity is a rate ratio, which is the same answer Chapter 01 gave for protein–DNA recognition.
Four canonical examples, chosen because their outputs differ in kind:
- EnvZ / OmpR — osmolarity sets the OmpF:OmpC porin ratio. Because the two carry binding sites of different affinity, a continuous OmpR-P concentration is decoded into a ratio of two outputs. An analogue-to-two-channel decoder, not a switch.
- PhoQ / PhoP — Mg²⁺ and antimicrobial peptides; controls Salmonella virulence.
- NtrB / NtrC — nitrogen limitation; NtrC is a σ⁵⁴ -binding protein, so §9's coincidence detector is its output stage.
- CheA / CheY — chemotaxis. Output is not at all: phospho-CheY binds the flagellar motor and reverses it, on a millisecond timescale.
That last one makes the general point. The same phosphotransfer architecture spans five orders of magnitude in response time depending on what the output domain touches.
11. Quorum sensing: counting without communicating
Some behaviours are worthless alone — bioluminescence, secreted virulence factors, biofilm matrix, taking up environmental DNA. Doing them at low density wastes the product or alerts a host prematurely.
The protocol is almost embarrassingly cheap. Each cell secretes a small autoinducer at a constant rate. In a confined volume the extracellular concentration rises with cell density. Above a threshold it re-enters and activates its receptor.
In Vibrio fischeri, LuxI synthesises an acyl-homoserine lactone; LuxR binds it and activates the luxICDABEG , which makes light — and which contains luxI itself. That positive feedback converts a gentle concentration gradient into a sharp, hysteretic transition. The bacterium lives in the light organ of the squid Euprymna scolopes and glows only at the density reached inside it. Gram-positive bacteria run the same logic with secreted oligopeptides read by two-component receptors.
Two honest caveats. First, a cell cannot distinguish "many neighbours" from "small enclosed space with poor diffusion" — both raise the local concentration identically. The diffusion sensing reading of the same data is not obviously wrong, and both readings are probably useful to the cell. Second, the AI-2 molecule made by many species via LuxS is often called a universal interspecies signal, but whether it is a signal or a metabolic by-product that some species eavesdrop on remains contested.
The evolutionary failure mode is the one a distributed-systems engineer would predict immediately: the signal is public and unauthenticated, so cheaters that respond without producing can invade.
12. Phase variation: regulated randomness
Not all regulation is a response. Some is deliberate, heritable bet-hedging against an environment the cell cannot predict — usually a host immune system.
DNA . Salmonella alternates between two flagellins. A 996-bp chromosomal segment flanked by 26-bp hixL/hixR sites is inverted by the Hin recombinase. In one orientation the segment's drives fljB (-2 flagellin) together with fljA, a of fliC; flipped, neither is made and FliC (phase-1 flagellin) appears instead. Switching occurs at roughly 10⁻³–10⁻⁵ per cell per generation. By the time the host has raised antibodies against one flagellin, a minority sub-population already displays the other. E. coli runs the same trick on type-1 pili with the FimB/FimE-inverted fim switch.
Slipped-strand mispairing. Neisseria opa genes carry a pentameric repeat inside the coding sequence. Replication slippage adds or removes a unit, shifting the reading frame and switching the gene on or off. These are contingency loci — the local is elevated by the sequence itself, and that elevation is a selected property.
Two conclusions that surprise people. Mutation rate is under genetic control and is not uniform across the genome — it can be locally tuned upward where unpredictability pays (Ch 16). And this is not Lamarckian: the switching is blind and constant. The host selects among variants that already exist.
13. What to carry into Chapter 22
Regulation is combinatorial. lac integrates two inputs. A eukaryotic integrates dozens. The arithmetic is the same — occupancies multiply — the fan-in just grows.
A control point is anywhere a rate can be changed. Bacteria use all of them:
| Control point | Mechanism | Response time |
|---|---|---|
| σ-factor availability | σ competition, anti-σ | minutes — whole regulons |
| initiation | , activators, DNA looping | seconds–minutes |
| Transcription elongation | , terminators | seconds, per transcript |
| initiation | SD occlusion, riboswitches, small RNAs | seconds |
| mRNA stability | ribozymes, RNase targeting | seconds–minutes |
| Protein activity | allostery, phosphorylation, proteolysis | milliseconds–seconds |
Initiation dominates textbooks because it is the cheapest place to intervene — you avoid the whole downstream cost. It is not the only place.
cis versus trans is the distinction that survives. In eukaryotes the cis elements move hundreds of kilobases away and the trans factors multiply into the thousands, but the logic is unchanged — and it reappears directly as -specific expression and cis- analysis in Ch 52.
Negative and positive control are different engineering solutions.
| Negative control | Positive control | |
|---|---|---|
| Default state | ON | OFF |
| Regulator lost | constitutive expression | no expression at all |
| Noise floor | leaky — repressor occupancy is never 1 | low — nothing recruits polymerase |
| Underlying | must be strong | must be weak |
| Failure mode | one operator → permanently ON | one activator-site mutation → permanently OFF |
| Ease of evolving | easy — any DNA-binding protein overlapping a strong promoter represses | harder — the activator must bind DNA and productively contact polymerase |
| Suits genes that are | usually needed | rarely needed |
Bacteria pick according to duty cycle, and lac uses both at once. Note the statistical character of the difference: with only ~10 repressor tetramers per cell, operator occupancy fluctuates and expression arrives in bursts, so negative control has a higher variance floor. Positive control on a weak promoter has a lower floor but a slower rise. It is a bias–variance trade-off with molecules.
Finally, the network shape. E. coli has a few hundred , most regulating a handful of each, while a small number of global regulators — CRP, FNR, IHF, Fis, ArcA, Lrp, H-NS — control hundreds. The out-degree distribution is heavy-tailed, the hierarchy is only about three layers deep, and are strongly over-represented. Hold that shape in mind: Chapter 25 asks what changes when the hierarchy gets deep.
Common misconceptions
| What people think | What's actually true |
|---|---|
| Lactose is the inducer of the lac | is, and it is made by β-galactosidase — the enzyme the operon encodes. The system needs leaky basal expression to bootstrap. IPTG is used in the lab precisely because it bypasses this |
| A repressed operon is off | Repression is ~1,000-fold, not infinite. Residual LacZ and LacY are not a leak to be engineered away; without them induction could never start |
| Glucose represses lac by binding something in it | Glucose binds nothing in the lac system. Its transport dephosphorylates EIIAGlc, which both lowers cAMP (disabling CRP) and blocks the permease. The signal is the act of transport |
| CAP and CRP are two different proteins | One protein, two names — Catabolite Activator Protein and cAMP Receptor Protein, coined by different labs |
| A constitutive mutant means the is broken | Two entirely different lesions give constitutivity: a broken trans-acting repressor (lacI⁻, ) or a broken cis-acting operator (lacOᶜ, cis-). Telling them apart is the entire purpose of the test |
| cis-dominance is a kind of Mendelian dominance | It isn't. lacOᶜ has literally zero effect on the other copy — it isn't competing with it. Whether you can even detect it depends on which structural genes sit downstream of it on that same DNA molecule |
| TrpR is LacI running backwards | Both are negative regulators, but the allosteric signs are opposite: allolactose takes LacI off DNA, tryptophan puts TrpR on it. Inducer versus corepressor |
| is a form of repression | It is a decision about whether an already-initiated transcript is finished, made by RNA folding rather than protein binding. Different control point, different mechanism |
| Bacteria are too simple for interesting regulation | E. coli runs a few hundred , seven σ factors, ~30 two-component systems, hundreds of small RNAs, and quorum sensing. What it lacks is chromatin, not sophistication |
Worked example: solving a merodiploid
. () lacI⁻ P⁺ Oᶜ lacZ⁻ lacY⁺ / (F′) lacI⁺ P⁻ O⁺ lacZ⁺ lacY⁻
Predict β-galactosidase and permease, with and without inducer. Work the algorithm.
Step 1 — pool the trans products. The chromosome's lacI⁻ contributes nothing. The F′ carries lacI⁺, and protein is diffusible, so functional LacI is present in the cell. No lacIˢ, so the pool responds normally to inducer.
Step 2 — chromosomal copy, cis elements. P⁺, so polymerase can initiate. Oᶜ, so the repressor pool — however healthy — cannot bind this operator. This copy is transcribed constitutively, inducer or not.
Step 3 — chromosomal copy, structural genes. lacZ⁻: no β-galactosidase, ever. lacY⁺: permease is produced constitutively.
Step 4 — F′ copy, cis elements. P⁻. No initiation. Nothing downstream is transcribed under any condition, and the state of O⁺ is irrelevant — you cannot regulate a transcript that is never made.
Step 5 — F′ copy, structural genes. lacZ⁺ is intact and never expressed. lacY⁻ is broken anyway.
Step 6 — sum.
| − inducer | + inducer | |
|---|---|---|
| β-galactosidase | none | none |
| Permease | present | present |
Read the trap. A β-galactosidase assay alone returns zero in both conditions, which is exactly the readout of a lacIˢ superrepressor — a completely different genotype. The two are separated only by assaying the second gene: this strain makes permease constitutively, a superrepressor strain makes none. The assay is not the genotype, and one reporter is not enough to specify a regulatory state.
Now check yourself against the general table, built the same way:
| Genotype | β-gal, − inducer | β-gal, + inducer | Reason |
|---|---|---|---|
I⁺ O⁺ Z⁺ | − | + | wild type: inducible |
I⁻ O⁺ Z⁺ | + | + | no repressor anywhere → constitutive |
I⁻ O⁺ Z⁺ / F′ I⁺ O⁺ Z⁻ | − | + | I⁺ supplies repressor in trans; lacI⁻ is |
I⁺ Oᶜ Z⁺ / F′ I⁺ O⁺ Z⁺ | + | + | the Oᶜ copy escapes repression → constitutive β-gal |
I⁺ Oᶜ Z⁻ / F′ I⁺ O⁺ Z⁺ | − | + | Oᶜ sits on the dead-Z copy, so it is invisible to the assay |
I ˢ O⁺ Z⁺ / F′ I⁺ O⁺ Z⁺ | − | − | mixed tetramers lock both operators: |
Iˢ O⁺ Z⁺ / F′ I⁺ Oᶜ Z⁺ | + | + | superrepressor locks the chromosomal copy, but Oᶜ makes the F′ copy immune |
Rows four and five are the whole lesson: the same , detectable or invisible depending only on which DNA molecule it shares with the reporter. That is what cis-acting means, operationally.
Connections
- Back to: Ch 05 Transcription — , σ factors and intrinsic termination · Ch 07 Translation — Shine–Dalgarno sequences and loading, without which §2 and §7 make no sense · Ch 20A — conjugation, the F factor and the F′ that §5 runs on; also the λ cI/Cro switch, the one major bacterial regulatory circuit this chapter does not build, and the bistable toggle Ch 25 analyses · Ch 01 §5 — binding is an occupancy, which is why every "switch" here is really a rate · Ch 08 — allostery
- Forward to: Ch 22 takes this skeleton and adds chromatin, distance, and combinatorial fan-in · Ch 24 develops and small RNAs · Ch 25 asks what happens when regulatory hierarchies get deep · Ch 36 uses the lac system as laboratory hardware · Ch 52 is the merodiploid experiment repeated in humans, using -specific expression instead of an F′