Contents
13 — Sex chromosomes and sex linkage Part 2 — Transmission genetics

13 — Sex chromosomes and sex linkage

Before this: Ch 09 · Ch 10 · Ch 12 · Time: ~35 min

What you'll be able to do

The core idea

Two independent things are going on, and conflating them causes most of the confusion in this topic.

The first is sex determination: which switch decides that an embryo develops as male or female. That switch is evolutionarily unstable. It has been reinvented dozens of times, sits on different in different lineages, and in many species is not genetic at all.

The second is : the inheritance consequences of one sex carrying one copy of a chromosome that the other sex carries two of. This has nothing to do with what the chromosome determines. It follows purely from the copy number.

A locus is "sex-linked" because of where it sits, not because of what it does. The gene for red–green colour vision has no connection to sex. It is sex-linked because it happens to sit on the X chromosome, and men have one X. Sex linkage is an accident of address.

The third thing, which follows from the second, is that one sex having one copy of an entire chromosome is a stoichiometry problem — and the solutions to it are strange and instructive.


1. The switch is not conserved

SystemWho is heterogameticWhere the signal comes fromExamples
XYmale (XY)presence of a Y-borne gene, or X:autosome ratiomammals, Drosophila
ZWfemale (ZW)Z dosagebirds, snakes, most butterflies and moths
X0male (X, no partner)X:autosome ratiograsshoppers, C. elegans (XX hermaphrodite / X0 male)
Haplodiploidy— itself: unfertilised egg → haploid maleants, bees, wasps
Temperature-dependent (TSD)—incubation temperature during a critical windowmost turtles, all crocodilians, tuatara
Environmental (ESD)—position, density, social contextBonellia (larvae settling on a female become dwarf males), clownfish

Two facts do the work here.

Mammalian and Drosophila XY systems are not the same system. Both label the X and Y; the logic underneath is completely different. In mammals, the Y carries a male-determining gene, so presence of a Y makes a male. In Drosophila, sex is set by the ratio of X chromosomes to autosome sets, and the Y is irrelevant to it — it is needed for sperm production only. The two systems give opposite predictions, and reality obliges:

HumanDrosophila
XXfemalefemale
XYmalemale
X0female (Turner syndrome)male, sterile
XXYmale (Klinefelter syndrome)female, fertile

Same variable names, different interpreter. When you read "XY system", ask what the switch actually reads.

The downstream pathway is far more conserved than the switch. Across vertebrates the gonad is built by two mutually antagonistic gene networks — a testis network built on SOX9 and DMRT1, an ovary network built on WNT4/RSPO1/β-catenin and FOXL2. What varies between lineages is only which upstream input tips the balance: a Y gene in mammals, Z-linked DMRT1 dosage in birds, a temperature-sensitive signal in turtles. Evolution keeps rewiring the input to a stable circuit.

2. The mammalian switch: SRY

The mammalian testis-determining factor is — a single-exon gene on the short arm of the Y at Yp11.2, identified in 1990. On GRCh38 it starts at chrY:2,786,855, which is about 5 kb outside the PAR1 boundary at chrY:2,781,479. Remember that number; it explains an entire class of exceptions in a moment.

SRY is a , and it is expressed as a pulse — in mouse, for roughly a day in the genital ridge, and then it is gone. What it does in that window is push SOX9 above a threshold, by binding an upstream of it. Past the threshold, SOX9 sustains its own expression through with FGF9 and prostaglandin D2, and simultaneously represses the ovarian programme. Below it, WNT4/RSPO1/β-catenin signalling wins and represses SOX9.

graph TD SRY["<b>SRY</b><br/>Y-linked, transient pulse"] -->|"binds SOX9 enhancer"| SOX9 SOX9["<b>SOX9</b> ↑<br/>testis programme"] --> FGF9["FGF9 / PGD2<br/>feed-forward"] FGF9 --> SOX9 SOX9 --> SERT["Sertoli cells → AMH<br/>Leydig cells → testosterone, INSL3"] RSPO["<b>RSPO1 / WNT4</b><br/>β-catenin"] --> FOXL2["<b>FOXL2</b> ↑<br/>ovary programme"] FOXL2 --> GRAN["granulosa cells → oestrogen"] SOX9 -.->|represses| RSPO FOXL2 -.->|represses| SOX9

This is a bistable latch with mutual repression, and SRY is a momentary pulse on the set line. Three consequences follow, and all three are experimentally confirmed:

  1. The pulse must be large enough and early enough. A weakened or delayed Sry produces ovotestes or XY females even with the gene intact.
  2. The latch must keep being held. Deleting Foxl2 in an adult mouse ovary causes granulosa cells to transdifferentiate into Sertoli-like cells. Gonadal sex is not a decision made once in the embryo and stored; it is a state actively maintained for life.
  3. Anything that trips the latch works, whatever it is. SRY is one input, not the mechanism.

The informative exceptions

The exceptions are how we know any of this.

ObservationWhat it proves
46,XX males (~1 in 20,000 male births). Most carry SRY on the tip of an X, transferred by a crossover that ran a few kb past the PAR1 boundary in the father's SRY is sufficient. And the mechanism is exactly what its position 5 kb outside PAR1 predicts
46,XY females (complete gonadal dysgenesis; SRY loss-of-function in ~15% of cases)SRY is necessary in humans
XX males with no SRY at all, carrying a duplication of an enhancer ~600 kb upstream of SOX9SRY acts through SOX9. Raise SOX9 directly and you skip the switch
XY females with a deletion of that same enhancerSame conclusion from the other direction
The Amami spiny rat has lost the Y and SRY entirely, and determines sex by a duplication that raises Sox9The Y is not required for maleness. It is required to deliver an input

That last row is the general lesson: a is not what makes a male. It is a delivery vehicle for whatever currently tips a much older switch.

3. Why the Y is a wreck

Start from the fact that the X and Y were once an ordinary homologous pair, roughly 180 million years ago in the ancestor of therian mammals. Then derive everything else.

Step 1 — a sex-determining arises on one member of the pair. That is now present only in one sex.

Step 2 — selection favours suppressing recombination around it. Alleles that are good for males but bad for females (or vice versa) accumulate near the switch, because there they are transmitted preferentially to the sex they benefit. Any that locks such an allele to the switch is favoured. Recombination is what lets selection evaluate alleles individually; an inversion switches that off for everything it spans.

Step 3 — the non-recombining region decays, for three separate reasons, all of which are consequences of losing recombination rather than of anything specific to sex:

Step 4 — the process repeated. Successive inversions expanded the non-recombining region in discrete steps, leaving evolutionary strata: blocks of X–Y gene pairs whose divergence increases with distance from the pseudoautosomal end, because they stopped recombining at different times. Four to five strata are resolvable in humans; the oldest one contains .

The result:

                PAR1                                                    PAR2
             (2.78 Mb)                                                (330 kb)
  X  |████████|─────────────────────────────────────────────────────────|██|
     0      2.78 Mb                                              155.70   156.04 Mb

  Y  |████████|▲──────────────────|▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒|██|
     0      2.78 │              ~23 Mb MSY                          56.89  57.22 Mb
                 └─ SRY at 2.787 Mb — 5 kb past the boundary
                                                          (coordinates: GRCh38)
  ████  pseudoautosomal — recombines X↔Y every male meiosis, effectively autosomal
  ────  single-copy: X-specific, or male-specific Y (MSY) — ~23 Mb of euchromatin on the Y,
        no partner in males
  ▒▒▒▒  Yq12 heterochromatin — from ~26.6 Mb; no genes, and its length varies between men

Of the 600-plus genes the ancestral pair carried, the human Y retains 19 — about 3%. The X kept them all, and today carries roughly 800–900 protein-coding genes. The complete -to-telomere Y assembly (HG002, 2023) is 62,460,029 bp and annotates 106 protein-coding genes — but most of those are copies within three amplicon families (TSPY, DAZ, RBMY), so the number of distinct proteins is a few dozen. Over half of the Y was missing from GRCh38 entirely; anything you compute on chrY should state which assembly you used (Ch 45).

Two loose ends worth having:

PAR1 is not optional. X and Y must pair and cross over somewhere in male to segregate correctly, and PAR1 is the only place they can. So PAR1 carries an obligate crossover: one crossover in 2.78 Mb is 50 cM in 2.78 Mb ≈ 18 cM/Mb, against a male autosomal average of roughly 0.9 cM/Mb (about 2,700 cM over ~2,875 Mb). PAR1 recombines about twenty times faster than the male genome average, and genes in it — SHOX, for instance — are inherited exactly like autosomal genes.

The Y is not still crumbling. Gene loss was front-loaded into the early strata. Comparing the human Y with the rhesus macaque Y across ~25 million years finds essentially no further loss — one gene. The survivors are dosage-sensitive broadly expressed regulators under strong purifying selection, and the Y protects them with a substitute for recombination: giant palindromes whose arms undergo intrachromosomal , letting a damaged copy be repaired from its own mirror image.

4. Hemizygosity: what one copy does to Mendel

A male is for loci — one , no partner. Everything distinctive about falls out of that.

Dominance stops applying. Dominance is a statement about the heterozygote. A hemizygous male has no heterozygote to describe, so an X-linked "" allele is fully expressed in him. The recessive/ labels for X-linked loci are really statements about females only.

Reciprocal crosses differ. This is the diagnostic. For an autosomal locus, A♀ × B♂ and B♀ × A♂ give identical offspring distributions. For an X-linked locus they do not, because sons get their X from their mother and their Y from their father. If reciprocal crosses disagree, the locus is not autosomal.

There is no male-to-male transmission. A father passes his son a Y, never his X, so an X-linked allele can never travel father → son. Be precise about what that licenses: an affected father with an affected son argues strongly against X , but it does not rule it out, because the son may have had the allele from a carrier mother — routine for a high-frequency allele such as G6PD deficiency or red–green colour blindness. The impossibility is the transmission, not the co-occurrence.

Criss-cross inheritance. An X-linked allele goes father → daughter → grandson, skipping the intervening male entirely.

graph LR GF["affected<br/>grandfather<br/>X<sup>a</sup>Y"] -->|"his only X<br/>goes to daughters"| M["carrier<br/>daughter<br/>X<sup>A</sup>X<sup>a</sup>"] M -->|"½ of her sons"| GS["affected<br/>grandson<br/>X<sup>a</sup>Y"] GF -.->|"sons get Y only"| S["unaffected<br/>son"]

Frequencies differ between the sexes, and by a lot. Let q be the frequency of an X-linked allele. Under random mating (Ch 26) a male is affected if his single X carries it — probability q. A female needs two — probability q². So:

Affected males : affected females = q : q². The ratio is 1/q. The rarer the allele, the more extreme the male excess.

Red–green colour vision deficiency has q ≈ 0.08 in populations of northern European ancestry. Predicted: 8% of males, 0.64% of females, and 2·q(1−q) ≈ 15% of females as unaffected carriers. Observed: about 8% of males and about 0.5% of females. The model needs no fitting.

Haemophilia A (F8, factor VIII) affects roughly 1 in 4,000–5,000 male births; haemophilia B (F9, factor IX) roughly 1 in 20,000–30,000. Both are severe and historically reduced male to near zero, which lets you derive a fact clinicians rely on. Two-thirds of all X are in females, one-third in males. At –selection equilibrium with affected male fitness zero, every mutant X sitting in a male is removed each generation — a loss of q/3 per X — and this must be balanced by the mutation input μ. So q = 3μ. Of those 3μ affected males, the μ whose mutation arose in the maternal gamete are new. Therefore one-third of cases of a lethal X-linked recessive are de novo. That is Haldane's result from 1935.

Now notice the assumption it rests on: that μ is the same in eggs and in sperm. It is not — roughly 80% of are paternal in origin (verified facts). And a paternal mutation cannot make an affected son, because fathers give sons a Y; it makes a carrier daughter, entering the one generation upstream. Redo the same equilibrium with a male rate four times the female rate and q comes out at six times the egg rate, of which just one part is a fresh mutation in the egg: one affected male in six is sporadic, not one in three. Haemophilia A is the extreme case — the -22 behind roughly 45% of severe disease arises almost exclusively in male germ cells, and nearly every mother of an isolated inversion case proves to be a carrier. So the sporadic fraction is lower than Haldane's third and the mother's carrier risk correspondingly higher, which is why carrier testing rather than pedigree inference is the standard of care.

5. Reading the patterns

X-linked Y-linked (holandric)
Sex ratio of affectedstrong male excessfemale excess, roughly 2:1males only
Affected father's sonsunaffected (from this locus)none affectedall affected
Affected father's daughtersall carriersall affectednone affected
Affected mother's sonsall affected (she is XaXa); it is a carrier mother who has ½ affected sons½ affected (she is heterozygous)—
Male-to-male transmissionneverneveralways
Skipped generationscommon, through carrier femalesrarenone
Exampleshaemophilia A/B, red–green colour blindness, Duchenne muscular dystrophy, G6PD deficiencyX-linked hypophosphataemia (PHEX); MECP2 Rett syndrome and IKBKG incontinentia pigmenti, both usually male-lethal; AZFa/b/c deletions causing spermatogenic failure

The sharpest single observation in this table: an affected father with an X-linked dominant condition has all daughters affected and no sons affected. Nothing else produces that signature, and one informative sibship can establish it.

Y-linked inheritance is genuinely rare, and the reason is structural: the male-specific Y carries few genes, most of them concerned with making sperm, so most Y-linked are self-eliminating. AZF deletions cause azoospermia — and are therefore almost never inherited; they arise de novo, or are transmitted only via assisted reproduction, in which case every son inherits them.

6. Dosage compensation: two problems, three solutions

Y degeneration created a stoichiometry problem — two of them, in fact, and they are worth keeping apart, because only one of them has been solved.

Every lineage that evolved differentiated had to fix the first problem, and they fixed it differently:

LineageMechanismNet result
Mammals (XY)One X per cell is transcriptionally silenced in XX individuals, via the XISTone active X in both sexes
Drosophila (XY)The single male X is transcriptionally doubled by the MSL complex and the roX lncRNAs~2× one-X output in both sexes
C. elegans (XX/X0)Both X in the XX hermaphrodite are halved by a condensin-like complex~1× in both sexes
Birds, Lepidoptera (ZW)Incomplete and gene-by-gene; no chromosome-wide mechanismpartial compensation only

Three lineages, three non-overlapping mechanisms — silence one, double one, halve two — arriving at the same functional endpoint. That convergence is the point:

Dosage compensation is a problem to be solved, not a mechanism to be memorised. If it were an inherited mechanism, it would look the same everywhere. It does not, which tells you the is real and the solution is contingent. And birds, which compensate only partially, tell you the constraint is also survivable.

Now notice what the mammalian and worm solutions do not do. Silencing one X leaves males and females alike at one active X against two autosomal sets: it equalises the sexes while preserving the X:autosome imbalance exactly as it was. (Drosophila is the exception — doubling the single male X gives him two-X-equivalent output against two autosome sets, settling both at once.) The candidate answer to the second problem is Ohno's 1967 hypothesis: that the active X is transcriptionally upregulated about twofold. Decades on it remains genuinely contested — microarray studies supported it, studies challenged it, and the current picture is at best partial upregulation concentrated on dosage-sensitive genes rather than a chromosome-wide doubling. So "dosage compensation" names a problem with one half solved and one half still open.

X inactivation in detail

The mammalian solution runs on a long non-coding RNA. XIST (~17 kb, from the X-inactivation centre at Xq13) is transcribed from the X that is to be silenced, and coats that chromosome in cis — it never leaves the chromosome that made it. Coating recruits silencing machinery (SPEN, Polycomb), which lays down H3K27me3 and macroH2A, moves the chromosome to a late-replicating compartment, and finally methylates , locking the state in (Ch 23, Ch 24). The condensed inactive X is visible down a microscope as the Barr body.

Four properties matter downstream:

It is random, and then clonal. Each cell in the early epiblast chooses independently; every descendant of that cell keeps the same choice. A female mammal is therefore a mosaic of two cell populations expressing different X . This is not a subtle molecular fact — it is visible on a cat. The orange locus in cats is X-linked (a 5 kb deletion at ARHGAP36, identified in 2025), so a heterozygous cat is orange in patches descended from cells that inactivated the non-orange X and black in the others: tortoiseshell, or calico once autosomal white spotting is added. A tortoiseshell male is nearly always XXY. And the first cloned cat, in 2001, did not look like its nuclear donor — same genome, independently drawn X-inactivation mosaic.

The ratio is a small-sample binomial. Choice is made in a modest number of progenitor cells — on the order of tens. If the count is n and the choice is fair, the proportion of cells with a given X active is roughly Binomial(n, ½)/n, whose standard deviation is 1/(2√n). With n ≈ 25 that is about 0.10, so a 70:30 split is two standard deviations out and arises in about 4% of women by sampling alone, and 75:25 or more extreme in about 1.5% — much more often if the progenitor pool is nearer n ≈ 10, where the standard deviation is 0.16. Skewed needs no mechanism. Selection acting afterwards on the two cell populations then skews it further.

Skewing produces manifesting carriers. A female heterozygous for a Duchenne muscular dystrophy or G6PD allele who happens to inactivate the healthy X in most of the relevant tissue is symptomatic. Balanced X–autosome make this systematic: cells that inactivate the translocated X also silence the attached autosomal material and die out, so the normal X is the one inactivated everywhere, and the carrier expresses only the disrupted allele. Skewing runs the other way in X-linked immunodeficiencies — cells with the healthy X active outcompete the others in blood, so extreme skewing in a female's leukocytes is itself a carrier test.

Silencing is incomplete. Roughly 15–25% of X-linked genes escape, at least partly: about 15% escape consistently, and a further ~10% escape variably between individuals and tissues. Escapees are concentrated in the PARs and on Xp. This is the reason X has a at all — if inactivation were total, 45,X and 47,XXY would be silent (Ch 20). Turner syndrome (45,X, ~1 in 2,000–2,500 live female births; ~99% of 45,X conceptions miscarry) and Klinefelter syndrome (47,XXY, ~1 in 500–1,000 male births) are both diseases of escape-gene dosage. Short stature in Turner syndrome is substantially attributable to a single copy of SHOX — which sits in PAR1, and therefore escapes, because a gene that pairs and recombines on both sex chromosomes was never a candidate for silencing in the first place.


Common misconceptions

What people thinkWhat's actually true
The determines malenessThe Y delivers an input to a switch. Presence of a Y makes a mammal male; in Drosophila the Y is irrelevant to sex and XXY flies are fertile females. Spiny rats have lost the Y and still make males
Sex-linked means "related to sex"It means "located on a ". Colour vision, clotting factor VIII and dystrophin have nothing to do with sex
conditions are in malesDominance describes a heterozygote. A male has no second , so the terms simply do not apply to him. The allele is expressed because there is nothing else to express
Women are unaffected carriers of X-linked recessive diseaseOften, but not reliably. is random and its ratio is a small-sample binomial, so skewing is common; manifesting carriers of DMD, haemophilia and G6PD deficiency are well documented
The inactive X is completely silent15–25% of its genes escape. If it were completely silent, sex- aneuploidies would have no — and they do
The Y chromosome is disappearing and men will go extinctGene loss happened in a few early bursts and then stopped. Human and macaque Y chromosomes differ by about one gene over 25 million years. The surviving genes are dosage-sensitive regulators held under strong , and palindromic repairs damage
A tortoiseshell cat could be maleOnly if he is XXY, or a mosaic. Two different alleles at an X-linked locus, expressed in patches, requires two X chromosomes
is the mechanism of X silencingIt is a problem. Mammals silence one X, flies double the male X, worms halve both. Three lineages, three unrelated mechanisms, same endpoint
Hairy ear rims are the classic Y-linked human traitTextbook folklore. The evidence never held up. Real Y-linked loci are and the AZF spermatogenesis regions, and the latter mostly abolish their own transmission

Worked example: Morgan's white-eyed fly, 1910

This is the experiment that tied genes to , and it is worth doing in full because the reasoning is entirely inferential — Morgan could not see a gene.

The observation. Among red-eyed Drosophila melanogaster, a single white-eyed male appeared.

Cross 1 — white male × red female. F₁: 1,237 red-eyed flies of both sexes (plus three white males, which Morgan set aside as sporadic). White is . So far, ordinary Mendel.

Cross 2 — F₁ × F₁. F₂:

redwhite
females2,4590
males1,011782

Total red 3,470, total white 782. Overall that is 4,252 flies at 4.44 : 1 — close enough to 3:1 to be recognisable, but the striking fact is that every white-eyed fly is male. A 3:1 autosomal ratio has no mechanism to produce that.

The hypothesis. Put the locus on the X. Write the white w and the red allele w⁺.

P     X^w Y   (white male)    ×    X^w+ X^w+   (red female)

F1    X^w+ X^w   red females         X^w+ Y   red males
             ↑ carrier                  ↑ got its only X from the white father?
                                          no — from the RED mother. Fathers give sons a Y.

F1 × F1:  X^w+ X^w  ×  X^w+ Y

          eggs:   X^w+        X^w
sperm
  X^w+          X^w+ X^w+   X^w+ X^w      →  all daughters red (½ carriers)
  Y             X^w+ Y      X^w  Y        →  ½ sons red, ½ sons white

Predicted F₂: red females ½, red males ¼, white males ¼, white females 0. Predicted counts for n = 4,252: 2,126 / 1,063 / 1,063 / 0.

Testing it (Ch 12). The zero class is already decisive: 782 white males and not one white female. Under an autosomal model, white females should be about a quarter of all females; observing 0 out of 2,459 has probability of order (3/4)^2459 ≈ 10⁻³⁰⁷. The autosomal model is dead on that observation alone.

The model survives but is not perfect: males total 1,793 against females 2,459, and white males 782 against red males 1,011. χ² for the 1:1 split among males is (1011−896.5)²/896.5 + (782−896.5)²/896.5 ≈ 29.3 on 1 df, p ≈ 6 × 10⁻⁸. The deficit is real and systematic — white flies have reduced viability, an early demonstration that a 's visible is rarely its only effect. It is not evidence against X , because no alternative model explains the zero.

The decisive test — reciprocal cross. Take white females (obtainable by crossing an F₁ carrier female to a white male) and cross to red males:

P     X^w X^w  (white female)  ×  X^w+ Y  (red male)

F1    X^w+ X^w   ALL daughters red        X^w Y   ALL sons white

Every daughter takes the opposite phenotype to her mother, every son takes his mother's. Criss-cross inheritance, and reciprocal crosses that disagree. No autosomal model of any dominance structure can produce a result that depends on which parent contributed which allele.

Why this mattered beyond flies. The w locus behaves exactly as the X chromosome behaves in — same transmission, same asymmetry between the sexes, same skipping through carrier daughters. That parallel was the first physical evidence that genes are carried on chromosomes, and it turned an abstract bookkeeping device into an object with an address. Everything in Ch 14 is built on it.

Connections