Champagne

Crossing two of them

What it looks like on each base

Champagne on a dominant white horse is invisible.

Its gene carrot

Gene key

horsegenetics.champagne

Alleles

Ch c

Outcomes

wild, champagne

Shows when

Ch/c, Ch/Ch

Default allele

c

Wild frequency

1 in 40 per allele

Founder draws

1 × nextFloat()

Deterministic

yes

keep red

0.69

keep black

0.20

black tint

0.30

The whole gene

private static final float KEEP_RED   = 0.69f;  // 31% of red restricted
private static final float KEEP_BLACK = 0.20f;  // 80% of black restricted - the bulk of it
private static final float BLACK_TINT = 0.30f;  // removed eumelanin, fed back as pheomelanin

@Override
public PigmentField restrict(AllelePair pair, CoatBuildContext ctx, PigmentView coat) {
    if (!isChampagne(pair)) {
        return null;
    }
    PigmentField f = coat.mutableCopy();
    CoatRegions.restrictAll(ctx.skin(), f,
            (field, px, py, p) -> field.dilute(px, py, KEEP_RED, KEEP_BLACK, BLACK_TINT));
    return f;
}

The tint is what makes amber champagne work

This gene is the clearest case for PigmentField.dilute. The previous rule set red absolutely:

field.setRed(px, py, 0.45f + 0.10f * red);   // WRONG
field.restrictBlack(px, py, 0.82f);          // black *= 0.18

• Two failures, one cause. Setting red absolutely lands within 0.10 of the same value whether the texel is a red body or a black point, so a champagne bay came out flat gold with no points at all.
• And cutting black to 0.18 was hard enough that a classic champagne on black also came out gold — indistinguishable from a gold champagne on chestnut. Two of the three champagne colours collapsed into one.

Feeding 30% of the texel’s black back in as red walks the sample sideways off the gradient’s zero-red column, so a bay’s points (which carry red = 0) land in the warm browns while the body, which already had red, goes gold. One term, three distinct colours.

The eye is the one thing about it that varies

Champagne dilutes the iris along with everything else, and the shade an adult settles on is famously not one colour: amber, golden, hazel, light brown, or — occasionally — a greenish olive. So the coat is deterministic and the iris roll is epigenetic, made once from the expressing copy's seed and inherited with it.

What the roll produces is not a colour any more. Champagne does not paint an iris: it requests an allele at the two iris loci, and the allele is written onto the horse at birth, so a champagne really carries its eye colour and really passes it on. The roll therefore picks one of three EyeHue alleles rather than one of four hex values:

This is still the mod's ordinary non-magical source of a green eye. The old rank argument survives as ordinary gene order: champagne sorts before tiger eye and before the white loci, and requests merge last-writer-wins, so tiger eye over-rides it and a splashed white champagne has blue eyes — see Eye colour.

What makes it nearly the simplest gene

• No dose — Ch/c and Ch/Ch are the same horse.
• No epigenetics in the coat — the pigment work is fully deterministic; only the iris is rolled.
• No cross-gene reading — isVisible is just pair.has(Ch).
• No geometry — one restrictAll over every texel.

The coat half is almost exactly the shape the planned tier-1 declarative builder is meant to cover in five lines.

Genetic identity

The substitution changes amino acid 63 from threonine to arginine in a predicted transmembrane part of the protein, and that threonine is highly conserved across mammals — which is the main reason to think the change is consequential. The original mapping put champagne in a roughly 6-centimorgan region of ECA14 and then sequenced the candidate genes inside it: all 85 non-cream champagne horses in the phenotype-selected sample carried the variant and none of 97 non-champagne controls did. Cook et al., PLoS Genet. 4:e1000195

It dilutes both pigments, which is the unusual part

Champagne reduces the visible intensity of both principal melanins — pheomelanin (red and yellow) and eumelanin (black and brown) — from a single copy. That distinguishes it from the neighbours: one copy of cream chiefly dilutes red, and silver primarily alters black.

Champagne segregates independently of cream, dun, silver, grey, leopard complex, tobiano and the other white-patterning loci, so combinations make visual identification unreliable and the DNA test is what a breeding decision should rest on.

Skin and eyes — often more diagnostic than the hair

• Foals commonly have blue eyes at birth.
• Those usually mature to amber, green, hazel or light brown rather than staying blue.
• Foals tend to have pink skin.
• Adult skin develops the characteristic pinkish-tan to lavender-toned, mottled “pumpkin skin”, especially around the eyes, muzzle and genital region.

Champagne therefore becomes more recognisable with age, as the mottling and the eye colour develop. Going the other way, cream dilution, grey, extensive white markings, sun fading and certain base-colour variants can obscure or imitate parts of it.

Inheritance and dosage

Treated as fully dominant. There is no reliable dramatic double-dilute phenotype comparable to cream’s palomino-versus-cremello dosage effect: both Ch/N and Ch/Ch are champagne. The original study suggested homozygotes may sometimes be slightly lighter or show less skin mottling, and the phenotypes overlap enough that zygosity cannot be read off appearance — test rather than infer. Ch/Ch is viable, with no evidence of embryonic lethality or any syndrome.

The mod matches exactly: expressionOf is pair.has(Ch), one copy and two produce the same outcome, and the locus contributes one entry to the catalogue rather than two.

Health, build, performance

No adverse health condition has been established for champagne. A large 2022 breed-distribution study explicitly reported no known adverse health effects, and OMIA lists it as a pigmentation trait rather than a disease variant. OMIA 001263 That means no champagne-linked congenital eye disease, no established hearing disorder, no demonstrated skin disorder, no metabolic, neurologic, reproductive or musculoskeletal syndrome, and no evidence that the homozygote is riskier than the heterozygote.

Which does not mean champagne horses cannot be exceptional athletes — it means their ability is explained by conformation, training, management, soundness and the many performance-relevant genes, not by the SLC36A1 result. The likely source of confusion is population structure: champagne turns up in several gaited and stock-type American breeds, so champagne individuals get noticed in a particular kind of horse. Breed association, not causation. The mod agrees by construction — champagne contributes to the eye and to the coat and to nothing on the trait layer at all.

What the protein does, and where the proof runs out

Established. SLC36A1 encodes a proton-coupled small-amino-acid transporter, known in other species as PAT1. The gene is expressed in equine skin, the T63R variant is tightly associated with champagne, and the phenotype reaches hair, skin and eyes — so the locus influences pigment biology in horses. In other species the protein is known to participate in amino-acid and proton transport, described in intestinal brush-border membranes and in lysosome-associated contexts.

Plausible, unproven. The leading hypothesis is that T63R changes ion or amino-acid transport, or pH handling, in pigment-cell organelles — especially early melanosomes, which are lysosome-related. Melanin production needs correctly regulated melanosome maturation, protein trafficking and pH, and altered organelle pH could affect tyrosinase activation or pigment processing, reducing pigment in coat, skin and eyes at once. That would explain the unusual both-pigments behaviour neatly, and it remains inference. The original investigators noted they could not entirely rule out another causal non-coding variant in linkage with c.188C>G, and the field still lacks a functional experiment in an equine melanocyte system showing how T63R changes transport, melanosome pH or melanin quantity.

So: well supported that c.188C>G / T63R is the test marker and very likely causal; supported that the phenotype is dilution of red and black plus the characteristic eye and skin pigmentation; not established, the cell-biological mechanism; not supported, anything about conformation, health, jumping, speed or athleticism.

Breeds and frequency

Uncommon overall, and concentrated in North American and American-derived populations. It is documented in the American Cream Draft, Miniature Horse, American Paint Horse, American Saddlebred, American Quarter Horse, Appaloosa, Missouri Fox Trotter, Tennessee Walking Horse, Kentucky Mountain Saddle Horse, Spanish Mustang, Rocky Mountain Horse, Shetland Pony, Gypsy Cob, and related crosses and American stock or gaited populations.

The American Cream Draft is a special case: gold champagne is integral to the breed’s historical identity, so it is far more conspicuous there than in horses generally — and the large frequency study below did not include the breed, so there is no modern population estimate for it.

Those come from a UC Davis study of 11,281 samples across 28 breeds, filtered for first-degree relatives — and champagne was detected in only 8 of the 28. PMC9498372 They are allele frequencies, not the share of visibly champagne horses: since champagne is dominant and rare, visible prevalence is roughly twice the allele frequency under random mating, and real studbooks, popular sires, colour-focused breeding, registration rules, relatedness and testing bias all move that. The samples were laboratory testing records, not a census. Notably, Tennessee Walking Horses — the breed the allele was first identified in — came out at a low estimated frequency.

The mod uses one champagne allele in 40, or 2.5% — which lands inside the real range and near its middle, between the Missouri Fox Trotter at 5.4% and the several breeds under 1%. It is one of the better-calibrated frequencies in the mod, and for the same reason as PPIB: the real allele is spread thinly across many breeds rather than concentrated in one, so a single world-wide number is not far wrong.