Contents
20A — Bacterial and phage genetics Part 3 — Genome instability

20A — Bacterial and phage genetics

Before this: Ch 16 · Ch 18 · Ch 20 · Time: ~50 min

Statistics needed: S1 Probability §§4, 6

Everything so far has assumed sex. Two parents, , a that can be heterozygous, offspring you count. Bacteria do none of that. And yet almost every sharp idea in classical genetics was invented or proved in bacteria and their viruses, because bacteria let you run the experiment at a scale eukaryotes never will: 10⁹ individuals on one plate, a twenty-minute generation, and conditions under which a single recombinant announces itself as a colony.

This is where the gene stops being a bead on a string. By the end of it a gene will be an interval — divisible, mappable inside itself, and defined by a test on function that is not the same test as the one on position.

What you'll be able to do

The core idea

Bacteria have no , so they cannot make a zygote. What they have is one-way transfer: a fragment of one cell's genome arrives inside another, and if it recombines into the resident it stays. The recipient is briefly a merozygote — only for whatever came across, and only until recombination resolves it. Two consequences run through everything below.

Distance is measured by co-transfer, not by . In a eukaryotic cross you ask how often a crossover falls between two markers (Ch 14). In a bacterial cross you ask how often two markers arrive in the same piece of DNA. Every bacterial map unit is a statistic about whether a fragment happened to span both loci.

Selection replaces counting. You do not score a thousand progeny. You design a medium on which only the class you want can grow, spread 10⁹ cells, and count what appears.

In a eukaryotic cross you count offspring; in a bacterial cross you select them. The sensitivity of your experiment is then set by how many cells fit on a plate. Everything here — the minute map, cotransduction, Benzer's resolution to the base pair — is that one trick applied to different transfer mechanisms.


1. Genetics without sex: the working vocabulary

Markers are metabolic or resistance . A prototroph grows on minimal medium — salts and a carbon source. An auxotroph carries a in a biosynthetic gene and needs the product supplied: his⁻ needs histidine. Gene symbols are three lowercase italic letters plus a capital for the cistron (hisG, trpE), with superscripts for state (his⁺, his⁻) and Str^R, Nal^R for resistance. The point of an auxotroph is that prototrophy is selectable with zero background: plate on minimal medium and only cells that acquired the wild-type form colonies. You 10⁹ events on one dish. (Reversion is the control everyone forgets — a his⁻ strain throws a few colonies anyway, so plate each parent alone at the same density.)

A phage is scored as a plaque. Spread ~10⁸ bacteria in soft agar to grow an opaque lawn, add a few phage, and each one that lyses its host releases progeny that infect the neighbours. The hole in the lawn — a plaque — is a clone descended from one particle, and its size, edge and turbidity are heritable phenotypes with a free readout.

Bacterial crosses are asymmetric. One cell is the donor, the other the recipient, and only a fragment moves. Three natural routes exist, told apart by three controls that each disable exactly one:

graph TD Q["Donor genotype appears<br/>in recipient cells"] Q --> C1{"Blocked by DNase<br/>in the medium?"} C1 -->|yes| TF["<b>Transformation</b><br/>naked DNA crosses the medium"] C1 -->|no| C2{"Blocked by a filter<br/>between donor and recipient?"} C2 -->|yes| CJ["<b>Conjugation</b><br/>requires cell-to-cell contact"] C2 -->|no| C3{"Cell-free filtrate<br/>still transfers?"} C3 -->|yes| TD["<b>Transduction</b><br/>a phage particle is the courier"] C3 -->|no| X["re-examine: contamination,<br/>reversion, cross-feeding"]

Read the logic, not the boxes. Naked DNA is destroyed by DNase; a capsid protects its cargo. Conjugation needs contact, so a sintered-glass filter abolishes it; a phage passes straight through. That U-tube design is what let Bernard Davis show in 1950 that Lederberg–Tatum recombination required contact, and what let Zinder and Lederberg show two years later that the Salmonella case did not.

2. Transformation: naked DNA, and how the genetic material was identified

Griffith, 1928. Streptococcus pneumoniae comes in a virulent S (smooth, capsulated) and an avirulent R form. Mice given live R survive; mice given heat-killed S survive; mice given live R plus heat-killed S die, and live S is recovered from the corpse. Something in the dead cells converted R to S heritably — the transforming principle.

Avery, MacLeod and McCarty, 1944. Sixteen years of purification later, the activity survived proteases and RNase, was destroyed by DNase, and had the composition and UV spectrum of nucleic acid. Their conclusion — the transforming principle is DNA — is the first direct identification of the genetic material, and it was widely disbelieved. The reason is worth carrying: Levene's tetranucleotide hypothesis had convinced the field that DNA was a monotonous repeat incapable of carrying information, and protein had twenty letters (Ch 02). A trace-protein contaminant was the standing objection, and it was not unreasonable.

Hershey and Chase, 1952 removed it in a different system. Phage T2 grown with ³⁵S labels protein only; with ³²P, DNA only. Infect, shear the empty capsids off with a kitchen blender, spin: ³²P is in the cell pellet, ³⁵S in the supernatant. What enters and directs progeny production is DNA. The separation is incomplete and it is less rigorous chemistry than Avery's, but it landed in a field already softened up. Avery's experiment proved it; Hershey–Chase's convinced people.

Natural competence is a physiological state, not an accident. Roughly 80 bacterial species take up environmental DNA through a dedicated machine: a pseudopilus that pulls duplex DNA to the membrane, a channel through which one strand enters while the other is degraded, and RecA-dependent recombination into the (Ch 18). Two features mark it as evolved. It is regulated — S. pneumoniae induces competence via a secreted competence-stimulating peptide read by a two-component receptor (ComD/ComE), so uptake switches on at high cell density. And some species prefer their own DNA: Haemophilus influenzae Rd carries 1,471 copies of the 9-mer uptake signal sequence AAGTGCGGT, and Neisseria meningitidis 1,891 copies of its own 10-mer GCCGTCTGAA. Do the null calculation — the H. influenzae Rd genome is 1,830,138 bp, so a given 9-mer is expected 1.83 × 10⁶ ÷ 4⁹ ≈ 7 times per strand, ≈ 14 in the genome, and the published count is a whole-genome one. Against that baseline 1,471 is a hundredfold enrichment, and it is a filter: it biases uptake toward conspecific DNA, which is the DNA most likely to recombine usefully.

3. Conjugation: putting a time axis on a chromosome

Lederberg and Tatum, 1946 mixed two multiply-auxotrophic E. coli K-12 strains and recovered prototrophs at about one in 10⁷; neither parent alone did. Davis's U-tube then showed the process needed contact.

The F factor. Donor capability rides on a conjugative plasmid, the fertility or F factor — 99,159 bp, one copy per , about a third of it the tra region encoding the pilus and transfer machinery. A cell carrying it is F⁺, one lacking it F⁻. Transfer is rolling-circle: a relaxase nicks F at oriT, one strand is pumped through the mating junction and made double-stranded in the recipient. Both cells end up with a complete F, so F⁺ × F⁻ converts nearly the whole recipient population to F⁺ — and moves essentially no chromosomal genes. The prototrophs at 10⁻⁷ were not coming from that.

Hfr strains, Hayes 1953. Some donors gave recombinants a thousandfold more often. In these Hfr (high frequency of recombination) strains F has integrated into the chromosome, by between the insertion sequences F carries — two IS3, one IS2, one Tn1000 — and their chromosomal counterparts, of which a laboratory E. coli has about 6 IS2 and 5 IS3 copies scattered around the genome, among ~37 IS elements in all (Ch 19). Integration can happen at any of them, in either orientation. That fact is what makes the next experiment work.

graph LR F["F plasmid<br/>IS2, IS3 ×2, Tn1000"] -->|"RecA recombination<br/>at a shared IS"| HFR["<b>Hfr</b> — F integrated,<br/><i>oriT</i> now inside<br/>the chromosome"] HFR -->|"nick at oriT,<br/>pump one strand"| T["chromosome enters<br/>recipient at constant rate,<br/>from a fixed point"] T --> B["mating pair breaks<br/>— always, eventually"] B --> R["recipient stays F⁻:<br/>the rest of F is last<br/>in the queue"]

Two predictions fall out, and both hold. Transfer is oriented — fixed start, fixed direction, set by where and how F integrated. And the recipient stays F⁻, because integration splits F and puts its trailing half at the far end.

Interrupted mating, and why a minute is a map unit

Wollman and Jacob's experiment is the one everybody should be able to reconstruct. Mix Hfr with F⁻. At intervals, put a sample in a blender: the shear rips mating pairs apart wherever they had got to. Plate on medium selecting for a donor marker and against the donor itself (streptomycin, with the recipient Str^R). Plot recombinant frequency against interruption time:

recombinants
per 100 Hfr
 0.30┤                    ┌───────────────  thr⁺   plateau ~0.30
     │               ┌────┘
 0.20┤         ┌─────┘ ┌──────────────────  lac⁺   plateau ~0.20
     │     ┌───┘  ┌────┘
 0.10┤   ┌─┘  ┌───┘         ┌──────────────  gal⁺   plateau ~0.13
     │   │ ┌──┘       ┌─────┘
 0.03┤   │ │     ┌────┘  ┌─────────────────  his⁺   plateau ~0.03
     └───┴─┴─────┴───────┴──────────────────────────▶  minutes of mating
         6 14    23              51
      time of entry: extrapolate each rise back to zero

Every curve has the same shape: nothing, a linear rise, a plateau. Extrapolating the rise back to the axis gives that marker's time of entry. Order the markers by time of entry and you have their order on the chromosome — no microscopy, no chemistry, no crossover statistics.

Why time is a distance. The strand is pumped at a constant rate, so position is an affine function of entry time. That is the whole argument, and it has a checkable consequence: the intercept is not zero. Every curve is displaced by a few minutes of pair formation and initiation. Absolute times are not positions; differences in entry time are proportional to differences in position, so the map is calibrated on intervals, never on the clock reading.

The unit was then fixed by convention. The E. coli map "has used the basic units of minutes and a total length of 100 minutes" since a 1976 recalibration, and a minute is now formally 1/100 of the chromosome however it was measured. With MG1655 at 4,641,652 bp:

1 minute  =  4,641,652 / 100  ≈  46.4 kb        ⟹  transfer rate ≈ 46 kb/min

Sequence-derived coordinates agree with the old genetic map to a fraction of a minute:

GeneMG1655 start MinutesClassical map
thrA3360.00
proA261,5025.65.6
lacZ366,3057.98
galE792,05517.117
trpE1,321,38328.528
hisG2,090,19145.044–45

lacZ and galE are transcribed from the complement strand, so their start codons sit at the high end of the annotated interval. Divide the low end instead and each comes out short by exactly the length of its own gene — 3,075 bp for lacZ, 1,017 for galE — which is enough to move the printed minute, and is the kind of error that never announces itself.

A minute is not a . A centimorgan is a probability of exchange — a parameter of a random process, saturating at 50% (Ch 14 §3). A minute is a time converted to a length by a constant rate: additive by construction, never saturating. The minute map is a physical map that happened to be measured genetically.

For a programmer, interrupted mating is a time-indexed oracle. You cannot read the chromosome; you can only ask a monotone predicate — has marker X arrived by time t? — at one plate per query. Monotone in t means binary search on the interruption time locates any marker in log(range) queries, and the answer is a coordinate.

The gradient of transfer, and why the chromosome is a circle

The plateaus decline with distance from the origin — thr⁺ near 0.30, his⁺ near 0.03. Mating pairs break spontaneously, and if breakage is memoryless (a constant hazard λ per minute, which is what "spontaneous" means, S1 §4), the fraction still joined at time τ is e^(−λτ). So

plateau(τ) ∝ e^(−λτ)          ⟹   ln plateau is linear in τ

Fitting 0.30 at 6 min and 0.03 at 51 min gives λ = ln(10)/45 ≈ 0.051 per minute, so the fraction of pairs surviving 100 minutes is e^(−5.1) ≈ 0.006. Complete transfer essentially never happens — which is why an Hfr donates a gradient rather than a genome, and why the far end of F never arrives. The gradient is a second distance estimate that needs no blender.

And independently isolated Hfr strains give different orders — all rotations and reflections of one cyclic sequence. A set of linear orders mutually consistent only when wrapped into a loop is evidence that the map is a circle, the same combinatorial move as assembling a circular sequence from overlapping reads (Ch 43, forward reference). Check-yourself question 1 runs the argument.

4. F′ merodiploids: making a haploid diploid enough to test

An integrated F sometimes excises imprecisely, taking a stretch of adjacent and leaving some of itself behind. The product is an F′ — a fertility factor carrying a chromosomal segment, such as F′lac. Transfer one into a cell that has its own copy of that region and you have a : for the segment on the F′, haploid for everything else. This is the enabling trick of bacterial genetics, and the reason is precise.

A asks whether two damage the same function. It requires both mutations in one cell, on separate DNA molecules, each supplying a full copy of everything else (Ch 11 §7). In a diploid eukaryote that is just a cross. In a haploid it is impossible — until F′ supplies a second copy of the region you care about, and nothing else.

   chromosome    ── lacI⁻   P⁺   O⁺   lacZ⁺ ──      two independent copies of
                                                     one region in one cell;
   F′lac         ── lacI⁺   P⁺   O⁺   lacZ⁻ ──      haploid everywhere else

Two instances of one object in a single address space, independently mutable — exactly what you need to separate a broken diffusible product, which the good copy can supply to both, from a broken sequence element, which can only ever affect the molecule it sits on. Ch 21 §5 (forward reference) runs this on the lac and turns it into an algorithm; the F′ is the machine that makes it possible, and the F′ exists because excision is imperfect.

Hold on to that mechanism. Imprecise excision of an integrated element, carrying flanking host DNA with it, is the same event that produces a specialized transducing phage (§5). One mechanism, two names, depending on whether the integrated element was a plasmid or a virus.

5. Transduction: the phage as a courier

Generalized transduction: a packaging error

Phage P1 (94,800 bp) packages by the headful rule: cut at a pac site, pump DNA into an empty capsid until full, cut again, start the next head from the free end. The head holds slightly more than one genome — the terminal redundancy is 10–15 kb — so a headful is roughly 100 kb, about 2 minutes of E. coli map.

The error is simple. Occasionally packaging initiates on a host sequence resembling pac and fills a head with bacterial DNA. That particle adsorbs and injects normally but delivers 100 kb of donor instead of a phage genome. The recipient survives, and if the fragment recombines it becomes a stable transductant. Three properties follow: any marker can be transduced, since pseudo-pac sites are scattered around the chromosome (hence generalized); frequency per marker is low, of order 10⁻⁵–10⁻⁶ transductants per plaque-forming unit, because a 100 kb window is a ~2% slice and mispackaging is rare; and the particle carries no phage genes, so it can neither lyse nor lysogenize — a transductant is an ordinary cell.

Cotransduction as a ruler — derived

Two markers are co-transduced only if they fit in one headful, which immediately gives a distance measure. Let L be the fragment length in minutes and d the separation, d < L.

Model the fragment as an interval of fixed length L whose position is uniform along the chromosome, and let inheritance of a marker require one crossover on each side of it inside the fragment, with crossover probability proportional to the DNA available on that side (S1 §6). Put A at 0 and B at d; let the fragment span [s, s+L].

                    ┌──────────── L ────────────┐
   fragment:        s                         s+L
   chromosome:  ────┼───────A───────B──────────┼────
                    │←  −s  →│← d →│←  s+L−d  →│
                     left flank      right flank (for B)

Fragments that can donate A. A must lie inside: −L < s < 0, with flanks −s and s+L. Weight ∝ (−s)(s+L); substituting u = −s,

∫₀ᴸ u (L − u) du  =  L³/2 − L³/3  =  L³/6

Fragments that can donate A and B. The fragment must also reach B, so −(L−d) < s < 0 and the outer flanks are −s and s+L−d. With M = L − d the same integral gives M³/6. Dividing:

              M³/6        ⎛     d ⎞³
C(d)  =  ───────────  =  ⎜ 1 − ─ ⎟          Wu (1966)
              L³/6        ⎝     L ⎠

The cube is not decoration: one factor because the fragment must span both markers, two more because each outer flank needs room for a crossover. Inverting,

d  =  L ( 1 − C^(1/3) )

and these inverted distances are additive while the raw frequencies are not — the same relation map distance has to in Ch 14 §2, for an entirely different mechanical reason. The worked example uses this.

Two regimes where the ruler returns nothing.

Specialized transduction: an excision error

Phage λ integrates at a single site, attB, which in E. coli sits between the gal and bio at about 17.4 minutes (Ch 18 §8 covers the site-specific recombination). The integrated genome is a prophage, replicated passively with the chromosome.

Excision normally reverses integration exactly. Roughly once in 10⁵–10⁶ events it happens between mismatched sites, and the excised circle carries a chunk of flanking chromosome while leaving an equivalent chunk of phage behind. Because a head holds a fixed amount of DNA, what comes in is balanced by what goes out, so the particle is usually defective — λdgal, λdbio. The consequences mirror the generalized case:

Generalized (P1, P22)Specialized (λ)
Mechanismmispackaging of host DNAimprecise excision of a prophage
Which markersany, anywhereonly genes flanking the att site
Cargohost DNA onlyphage + host, usually defective
Made bylytic growth on the donorinduction of a lysogen
Frequency~10⁻⁵–10⁻⁶ per marker per pfu~10⁻⁶ initially; ~10⁻¹ from a double lysogen
Productrecombinant chromosomeoften a stable partial

The last two rows are the useful asymmetry. A first-round λdgal lysate is LFT (low-frequency transducing), but a cell transduced by λdgal together with a normal helper λ becomes a double lysogen: induce it and roughly half the burst is λdgal, an HFT lysate five orders of magnitude better. And that transductant carries both the resident gal and the incoming one — a , reached by a completely different route from §4 and usable for the same .

6. Plasmids: replicons, copy number, incompatibility, resistance

A plasmid is a replicon — a DNA molecule with its own origin and its own control over how often that origin fires. Everything else is a consequence of that definition.

Copy number is a set point, not a function of size. The replication control system measures its own concentration and inhibits initiation above a threshold: ColE1-type plasmids use an antisense RNA (RNA I) that pairs with the replication primer and blocks it, so more plasmid means more RNA I means less initiation. The inhibitor is diffusible and made by every copy, making this a negative-feedback controller with a shared sensor — and F sits at ~1 copy per while ColE1 derivatives sit at tens, neither number set by plasmid length.

Incompatibility is a corollary of that controller. Two plasmids are in the same incompatibility (Inc) group if they share a replication control system — same antisense RNA, same iterons, same partition apparatus. In one cell the controller cannot tell them apart: it counts the sum and throttles both, so which molecule replicates at any moment is a coin flip and at each division drives one lineage to .

Incompatibility is not sequence similarity, and it is not exclusion at the door. Two plasmids 99% identical but with different replication controls coexist happily; two sharing nothing but a control circuit do not. The test is functional — introduce one into a strain carrying the other and see whether the resident is lost. Inc typing classified plasmids by behaviour long before anyone could sequence them, and it survives because the behaviour is what matters clinically.

R factors. In late-1950s Japan, Shigella isolates appeared resistant to sulphonamide, streptomycin, chloramphenicol and tetracycline simultaneously, and the whole set transferred to a sensitive strain by contact. Watanabe's 1963 review established the picture: a conjugative plasmid — an R factor — carrying a transfer region like F's plus a cassette of resistance determinants, often assembled from transposons (Ch 19) and from integrons, site-specific capture systems that acquire resistance genes as cassettes and express them in tandem from one — a mechanism Ch 19 does not cover, and the reason a single acquisition event can deliver several resistances at once. Three consequences, and together they are the genetics behind a public-health problem. Resistance crosses species, because a broad-host-range conjugative plasmid carries its cargo over genus boundaries — so a pathogen can acquire resistance to a drug it was never exposed to. Selecting for one drug selects for all of them, because determinants on one replicon are co-replicated rather than merely linked. And the accounting is not vertical: resistance frequency is the frequency of a replicon in a community, not of an in a lineage (Ch 35 §8, forward reference). The GRAM analysis attributes 1.14 million deaths in 2021 directly to bacterial antimicrobial resistance and forecasts over 39 million cumulative attributable deaths between 2025 and 2050.

7. Phage as a genetic system

Plaque morphology is a phenotype

AppearanceLesion
r⁺ (wild-type T4)small plaque, fuzzy halolysis inhibition intact
r (rapid lysis)large plaque, sharp edgelysis inhibition lost
turbid plaque (λ)hazy centre — lysogens growing in itlysogeny functional
clear plaque (λ)fully lysed, no survivorslysogeny broken
h (host range)plates on a resistant hostaltered tail fibre

A phage cross is a mixed infection: coinfect at high multiplicity, let the genomes recombine in one cell, score the burst. Frequencies are computed as in Ch 14, with one caveat — a phage genome pairs several times per infection, so phage map units are inflated relative to a single and are not comparable to .

λ: lysis or lysogeny, and a switch that stays switched

On infection λ must choose. Lytic: replicate, package, lyse, ~100 progeny in under an hour. Lysogenic: integrate at attB, shut down every other phage gene, and be copied passively for as long as the host lineage survives. Decision and memory are both implemented by two proteins competing for one ~80 bp stretch carrying three operators between two divergent :

      ← P_RM  (makes cI)                    P_R →  (makes Cro)
   ───────────┬────────┬────────┬───────────
              │ O_R3   │ O_R2   │ O_R1   │
              └────────┴────────┴────────┘
   cI  affinity:  weakest  middle  STRONGEST   (cI dimers bind O_R1+O_R2 cooperatively)
   Cro affinity:  STRONGEST  ...    weakest
graph TD CI["<b>cI</b> dimer"] -->|"binds O_R1 + O_R2<br/>cooperatively"| A["blocks P_R → no Cro"] CI -->|"the O_R2-bound dimer<br/>contacts RNA polymerase"| B["<b>activates P_RM</b> → more cI"] B -.->|positive feedback| CI A --> LYS["<b>LYSOGENY</b><br/>high cI, no Cro"] CRO["<b>Cro</b> dimer"] -->|"binds O_R3 first"| C["blocks P_RM → no more cI"] C --> LYT["<b>LYSIS</b><br/>low cI, high Cro"] DMG["DNA damage → RecA*<br/>(SOS, Ch 17)"] -->|"stimulates cI<br/>self-cleavage"| C

The structure is the point. Mutual repression: cI at OR1/OR2 shuts off PR and hence Cro; Cro at OR3 shuts off PRM and hence cI. Positive autoregulation: the cI dimer on OR2 does not merely block — it contacts at PRM and recruits it, so cI activates its own gene. Negative autoregulation at the top: at very high concentration cI also fills OR3, capping its own level.

Mutual repression alone gives a switch you can flip. Mutual repression plus positive feedback on one arm gives two locally stable steady states with an unstable one between them — bistability. That is why a lysogen is stable for thousands of generations with no external input: it is not "set" to lysogeny, it is sitting in an attractor, and fluctuations in cI are pulled back rather than amplified. Spontaneous induction runs near one cell in 10⁵ per generation.

Induction is a designed escape hatch. DNA damage activates RecA, and activated RecA stimulates cI to cleave itself (Ch 17). cI collapses, Cro takes OR3, the lytic programme runs, and the phage abandons a host whose genome is falling apart. The prophage subscribes to the host's own damage signal rather than building a detector.

And the circuit was dissected genetically before any molecule was known. Kaiser (1957) collected clear-plaque mutants — phage that cannot lysogenize — and sorted them by complementation into three groups. cI⁻ mutants never lysogenize, even with a helper present; cII⁻ and cIII⁻ mutants lysogenize rarely on their own. So cI maintains lysogeny and cII/cIII establish it: same plaque phenotype, three genes, two functions, separated by a and a quantitative difference in lysogenization frequency.

Lysogens are immune to superinfection, because prophage-made cI represses an incoming λ genome exactly as it represses the resident one. Immunity is not a defence system; it is a doing its ordinary job on a second copy of its target — the cleanest possible demonstration that cI is diffusible.

8. Benzer: the gene turns out to be an interval

Before 1955 the gene was operationally a point: the unit of function, of , and of recombination at once. Benzer's rII experiments separated those three.

The system. rII mutants of phage T4 grow on E. coli B, making large sharp-edged plaques, but cannot grow on E. coli K-12(λ) at all. Wild type grows on both.

          E. coli B                   E. coli K-12(λ)
 rII      grows — large r plaque      NO GROWTH
 r⁺       grows — small fuzzy plaque  grows

That asymmetry is the whole experiment. Cross two rII mutants on B (permissive), then plate the burst on K-12(λ). Only r⁺ recombinants form plaques and the background is zero, so plating 10⁹ progeny makes a single recombinant a visible plaque and pushes the detectable to ~10⁻⁸. A Drosophila geneticist scoring 10⁴ flies bottoms out near 10⁻⁴. Benzer bought four orders of magnitude by choosing an organism where selection does the counting.

Two tests, two questions. He ran both on the same pairs, and the difference is the intellectual core of this chapter.

Complementation (cis–trans test)Recombination
Questiondo the lesions damage the same function?are the lesions at the same position?
Procedurecoinfect K-12(λ) with both; does the cell lyse?coinfect B; plate the burst on K-12(λ); count r⁺
Unit it definesthe cistronthe recon — down to a base pair
Two mutations in one genefail to complementstill recombine

That last row is the result. Two mutations that fail to complement — same function, same gene — nevertheless give r⁺ recombinants. A gene is therefore not the unit of recombination. It is an interval containing many mutable sites, and happens between them. Complementation partitioned all rII mutants into exactly two groups, rIIA and rIIB, and Benzer named that complementation unit the cistron, after the cis–trans test that defines it — the term Ch 11 §7 uses for eukaryotic , and this is where it comes from.

Deletion mapping, or how to avoid quadratic work. Mapping n point mutants pairwise costs n(n−1)/2 crosses; at ~2,000 mutants that is two million crosses. Benzer instead crossed each point mutant against a small set of deletions covering nested, overlapping intervals:

   rIIA ───────────────────────────┤├───────────── rIIB ─────────
   deletion 1  ▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓
   deletion 2       ▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓
   deletion 3                ▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓

   point mutant × deletion → any r⁺ recombinants?
        NO   ⟹ the point mutation lies INSIDE that deletion
        YES  ⟹ it lies outside

A point mutation inside a deletion can never yield a wild-type recombinant — no wild-type sequence exists at that position on either parent — so the answer is a binary predicate, not a frequency, and needs no counting. Each deletion halves the candidate interval: interval containment queried by binary search, O(n log n) crosses instead of O(n²), which is why the map got finished.

The resolution, computed. The smallest frequencies Benzer could measure reliably were ~0.02%. The T4 genome is 168,903 bp and its genetic map is of order 1,500–1,600 map units, so

168,903 bp / 1,600 units ≈ 106 bp per map unit
0.02 map units × 106 bp   ≈ 2 bp

Benzer resolved genetic distance to about the base pair, a decade before anyone sequenced a gene. Treat the arithmetic as approximate — phage genomes pair several times per infection, so a phage map unit is not a meiotic and the conversion factor is soft. The order of magnitude is not. And the scale is worth pinning: rIIA spans 2,178 bp and rIIB 939 bp on the modern T4 assembly — together ~3.1 kb, 1.8% of the genome — and Benzer resolved more than 300 distinct mutable sites inside that, an average spacing near ten base pairs, sharply non-random along it. A few hotspots accumulated hundreds of independent occurrences while most sites were hit once (Ch 16 §§7–8 — mutation is undirected but very far from uniform).


Common misconceptions

What people thinkWhat's actually true
An Hfr × F⁻ cross makes the recipient F⁺It almost never does. Integration splits F and puts most of it at the far end of the transferred , and mating pairs break long before it arrives — which is exactly why transfer is a gradient
A minute of the E. coli map is a A centimorgan is a probability of exchange and saturates at 50%. A minute is a time converted to a length by a constant rate: additive by construction, never saturating, and now defined as 1/100 of the chromosome from sequence
Cotransduction frequency is a distanceIt is a distance passed through (1 − d/L)³. Raw frequencies are not additive; L(1 − C^(1/3)) is. Beyond one headful the measure returns zero for every distance alike
Generalized and specialized transduction are two versions of one thingDifferent errors at different steps. Generalized is a packaging error and moves any marker; specialized is an excision error and moves only genes flanking the att site. One delivers host DNA, the other a phage–host hybrid
Plasmid incompatibility means the plasmids are similarIt means they share a replication controller, which counts them jointly and lets eliminate one. Near-identical plasmids with different controls coexist; unrelated plasmids with one control do not
A lysogen is a phage lying dormant until conditions improveIt is an actively maintained state held by a bistable circuit, with a specific trigger: DNA damage activates RecA, which makes cI cleave itself. An attractor with an escape hatch wired to the host's damage response
Failure to complement means the are at the same siteIt means they damage the same function. Benzer's result is that mutations failing to complement still recombine. function, recombination tests position, and a gene is an interval
Avery was doubted because scientists are conservativeThe objection was specific: DNA was believed monotonous (the tetranucleotide hypothesis), and trace contaminating protein could not be formally excluded. Hershey–Chase added a second system, not better chemistry

Worked example: mapping an unknown mutation, coarse then fine

The goal. You have a tryptophan auxotroph of E. coli — call the lesion trp⁻ — and you want its map position. You have Hfr strains, P1 phage, and plates.

Step 1 — coarse map by conjugation. Mate an Hfr (trp⁺, Str^S) with your mutant (trp⁻, Str^R), blend at intervals, plate on minimal medium + streptomycin. Trp⁺ recombinants first appear at 34 min. Two known markers bracket it: lac⁺ (7.9 min) enters at 14 min, his⁺ (45.0 min) at 51 min. The ~6-minute offset is the same in both, so it is the initiation lag and only differences count: 34 − 14 = 20 min after lac, i.e. 7.9 + 20 = 27.9 minutes. A window of about a minute — 46 kb, on the order of forty genes. Not enough.

Step 2 — fine map by cotransduction. Grow P1 on a wild-type donor, transduce the mutant, select Trp⁺, and score each transductant for tonB (28.2 min; tonB⁻ cells resist phage T1, so it is scorable by replica plating — press a sterile velvet pad onto the master plate and stamp the whole colony pattern onto test plates in register, so every colony is tested on several media at once without being picked) and cysB (28.7 min):

   tonB – trp    0.69
   trp  – cysB   0.66
   tonB – cysB   0.43

Step 3 — the wrong turn. The natural move is to read cotransduction as a distance: closer markers cotransduce more, so treat 1 − C as proportional to separation. Then

   (1 − 0.69) + (1 − 0.66) = 0.31 + 0.34 = 0.65
   observed for the outer pair:  1 − 0.43 = 0.57

The outer distance comes out 12% short of the sum of its parts — and there is no double crossover to blame, so this is not the Ch 14 shortfall. The measure is simply not linear in distance. Applying §5's with L = 2 minutes:

   d(tonB, trp)  = 2(1 − 0.69^⅓) = 2(1 − 0.8837) = 0.2327 min = 10.8 kb
   d(trp, cysB)  = 2(1 − 0.66^⅓) = 2(1 − 0.8707) = 0.2587 min = 12.0 kb
   d(tonB, cysB) = 2(1 − 0.43^⅓) = 2(1 − 0.7548) = 0.4904 min = 22.8 kb

   0.2327 + 0.2587 = 0.4914   against 0.4904 measured directly
                              ✓ additive to ~0.001 min — about 44 bp

The cube root does for cotransduction what Kosambi does for . Distances add; frequencies do not.

Step 4 — order the markers, and do not let the pairwise numbers do it. Additivity is consistent with tonB–trp–cysB, but so is a lucky set of estimates. Get the order from a three-factor transduction, where one class needs four crossovers instead of two. Transduce a tonB⁻ trp⁻ cysB⁻ recipient with P1 grown on wild type, select an outside marker (Cys⁺), and score the other two. Among 500 Cys⁺ transductants:

ClasstonBtrpCountCrossovers if order is tonB–trp–cysB
1⁻⁻1762
2⁻⁺2052
3⁺⁺1122
4⁺⁻74

Class 4 is diagnostic: it took the donor at the far marker while keeping the recipient allele in between, which needs two extra exchanges. Same logic as the double-crossover class in a three-point testcross (Ch 14 §5), and it identifies the middle locus the same way — in the rarest class, the marker that disagrees with the other two is the middle one. trp disagrees, so trp is in the middle.

Test the alternative rather than assuming it away. If the order were trp–tonB–cysB, the four-crossover class would be the one carrying donor trp⁺ with recipient tonB⁻ — class 2, observed 205 times. A class needing two extra crossovers cannot be the second most common, so that order is excluded by the data.

Step 5 — read the answer. cysB is at 28.7 min, trp is 12.0 kb (0.26 min) to its left at 28.4 minutes, tonB a further 10.8 kb beyond. The real trpE start in MG1655 is 1,321,383 bp = 28.5 minutes. Two experiments, no sequencing, agreement to about 5 kb.

Step 6 — generalise. The pattern is a resolution ladder, the same one Ch 20 §11 sets out for cytogenetics: choose the assay whose dynamic range brackets your uncertainty. Conjugation resolves ~1 minute across the whole 100-minute ; P1 cotransduction resolves ~1 kb but is blind past 100 kb. Neither alone can find a gene. In series, coarse then fine, they converge in two experiments — and each stage exists precisely because the previous one saturated.

Connections