Case studies

Three case studies of the Global Cat Health Survey, published in full: the breed a registry made from DBE-CEL (Roxi), how that variant reaches breeds it was not named for, and the Igor line and the variants it shares descent with

The material on this page is drawn from the Global Cat Health Survey 2026, which is where its sources and its declarations of interest are set out in full.

Three of the survey's case studies are published here in full. Case study one of the survey is not: its material is carried across the introduction, spread, health and laboratories pages of this directory. Its case studies four, five and six are not about the dominant blue eye and stay on the survey.

Case study one · the Céleste, a breed made from DBE^CEL (Roxi)

How refusing a trait in a breed spread it across different breeds, by creating a breed that must outcross to exist

This case study follows one variant, DBE^CEL (Roxi), and the breed built on it. Two questions run through it. Where did the variant actually come from, given that three published records disagree? And what happened when a registry, having refused the trait inside an existing breed, recognised it as a breed of its own?

The answer to the second is the finding. A breed whose standard forbids breeding it to itself cannot close its stud book, so the variant leaves the breed in every generation by the terms of the standard, and two of the three permitted partners are the British breeds in which the trait was refused. Whatever anyone intended, refusal did not contain the variant.

The first of those two questions is answered in full elsewhere in this directory. Where the two variants came from, which cat each is traced to, and how the two came to be confused, are set out under where the two variants came from on the dispersal page, together with what the upward trace through recorded ancestry adds to them. This page takes that history as read and turns to the second question.

What the genotyping shows about the Céleste, against what the breed narrative says

Two sets of results answer the question of what the Céleste carries, Abitbol, Couronné et al., Animal Genetics 2024;55:670-675, Table 1 and Abitbol et al., Animals 2024;14:1845, Table 2.

• Fifty-two Céleste and Céleste-crossed cats that showed the blue eye carried the DBE^CEL (Roxi) insertion, forty-five Céleste and seven Céleste-mixed, and the outbred founder of the breed was heterozygous for it, meaning he carried one copy. That figure counts the blue-eyed cats of those two groups alone. Two further counts from the same table appear below and are not in conflict with it: sixty-seven is every cat in the study carrying the insertion whatever its eye colour, and sixty-five is every carrier that showed the blue eye.
• Twenty-two Céleste cats, blue-eyed, white and latent, were tested for the DBE^ALT insertion and none of them had it. They were genotyped wild type at that position. This is the result of looking and finding nothing, not of failing to look.

So on the cats that have been tested, the Céleste today carries DBE^CEL (Roxi) and does not carry DBE^ALT. It holds one of the two variants, not both. On Roxi, the cat the variant is traced to, the published position is stated in terms: “Roxi transmitted a peculiar DBE variant: the DBE^CEL (Roxi) allele. He was probably not an Altai cat.” In plain terms: Roxi passed on a DBE variant of his own, DBE^CEL (Roxi), and he was probably not an Altai cat. Peculiar is used there in its older sense of belonging particularly to him, and not in the sense of strange.

Why twenty-two results cannot speak for the whole breed

The Céleste is defined by descent, not by eye colour. LOOF's rule is that every Céleste must trace to Lazuli. A kitten of a Céleste mating that inherits no copy of the variant, and a kitten that inherits one but shows ordinary eyes, are both still cats of the breed. So the breed contains three kinds of cat at once: those that show the blue eye, those that carry it without showing it, and those that carry nothing.

That is what makes the twenty-two a sample rather than an audit. To say that the breed does not hold DBE^ALT, rather than that these twenty-two cats do not, every registered Céleste would have to be tested, including the ones with ordinary eyes. A second variant sitting in a cat that shows nothing is invisible to any survey that selects its animals by looking at them. The published sample did include latent cats, which is to its credit, but twenty-two animals of any kind is not a stud book.

And the breed cannot close, which is why this is not a question that will settle itself. The standard requires an outcross in every generation, so new cats enter the breed continuously from the British Shorthair, the British Longhair and the domestic population. Each one is an opportunity for a second variant to enter and sit unseen, and the register that would have to record it holds no genotype for any cat in the breed.

What the published work therefore does not establish. Nothing in it tests the founding generation. Whether DBE^ALT was ever in the breed and was bred out, or was never in it, cannot be settled from the genotyping alone.

That resolves the apparent conflict. The Céleste comes out of a population that held both variants and today holds only one, and a variant absent from twenty-two tested animals was not lost by dilution: it was bred away from. That is a fact about selection by later breeders, not about ancestry, which is why the breed's own history and its genotype can both be accurate at once.

None of this is recorded by the registry. LOOF, which recognised the Céleste, does not record the DBE genotype of any cat: PAX3 appears neither in its laboratory protocol test list nor on its health page. Every genotype quoted here comes from research papers and not from a stud book. So the question of whether the French register holds one variant or two has no answer, because the register holds neither.

Céleste and Celestial British are one breed at two stages of its development

Three names attach to this breed and none denotes a separate population. Céleste is the LOOF standard's name, Celestial the literature's, and Celestial British the messybeast.com compilation's. LOOF has never published the term Celestial British, and no registry anywhere holds a breed of that name alongside the Céleste.

The genotyping is consistent with this and not with two breeds. The twenty-two wild type results mean the LOOF-registered Céleste does not carry DBE^ALT, which is what a line looks like after the second variant has been selected out.

Two things are commonly offered as evidence of two breeds and neither is. The messybeast.com compilation files its Celestial British page under the Altai and Seymour genes, which records the composition of the early programme rather than a different breed. And the 2018 World Cat Federation Altai standard distinguishes the Altai from the Celestial; it says nothing about Céleste against Celestial British.

How the variants reached the British programme, and the third route that is not a variant at all

The programme meant here is the one that became the Céleste, run at the Aerlin cattery in France on British Shorthairs and British Longhairs. It was founded on Kaena to Lazuli, with a variant on each side of that mating, so the founding mating put one copy of each variant into one litter. That is the mating of 18 July 2016, and it is why the early generations held two variants at once. The removal of what the cattery calls the Seymour gene is the removal of Kaena's contribution, that is of DBE^ALT, leaving DBE^CEL (Roxi) alone, which is the line LOOF recognised in 2023.

That is one route into the British Shorthair and it is not the only one, and the largest of them carries no identified variant at all. The breed is reached three ways in this survey's register, and only two of the three have an allele behind them.

The third row is larger than the other two together and is the one nobody can test for. Igor's own genealogy is headed “line descended from Altai gene Igor”, which is the naming hazard set out below rather than a result: no cat of that line has ever been genotyped, and it does not descend from Roxi or, but for a single record, from Nikita'l Seymour. A DBE^ALT or DBE^CEL test on one of these cats can return wild type and settle nothing, because neither variant is what is being asked about. The line is set out in full in the variant trace, and its latent carriers on the latent lines page.

One label, two variants: why “the Altai gene” cannot be used

In the source material the phrase the Altai gene names two different variants. DBE^ALT takes its name from the Altai population and traces to Fyodor and to Nikita'l Seymour. The messybeast.com compilation also writes of “the Altai gene (Roxi Darlin)”, and Roxi is the founder to whom DBE^CEL (Roxi) traces, of whom the founding paper concludes “he was probably not an Altai cat”. A sentence containing the words “the Altai gene” cannot be resolved without knowing which founder is meant.

The annotation beside the Celestial box on the breeder-compiled genealogy is exactly this problem: “The Celestial eliminated the Seymour gene and the LOOF recognised Celestial (2023) only has the Altai gene”. Read with Altai meaning DBE^ALT it is wrong; read with Altai meaning Roxi's variant it is right. The survey records the ambiguity rather than a verdict, and applies the rule it uses throughout: resolve every gene name to a named founder cat before relying on it.

The same hazard sits in the name Celestial British, and that one is about a breed rather than a gene. The name contains a breed name, so a reader working from names rather than genotypes takes it to mean that DBE^CEL (Roxi) is established in the British Shorthair, and goes on to attribute every breed holding a blue-eyed line to DBE^CEL. The breed table records DBE^CEL (Roxi) against the British Shorthair as reported and not verified; what is genotyped in that breed is DBE^ALT. That inference is the shape of the mistaken breed lists now in circulation and the most probable origin of them. Two practices prevent it.

• Record the molecular designation, not only the nickname. PAX3 c.937C>T cannot be confused with an intron 4 insertion in the way that RE and CEL can be confused with a cattery name or a breed.
• Treat a breed name containing a variant nickname as a claim to be verified rather than as evidence of which variant the cats carry.

What the two-variant history does and does not say about the Altai breed

The two-variant history says nothing against the Altai as a breed. No health problem peculiar to the Altai has been found, and none is asserted here. The harm is attached to no single variant: it arises when a kitten inherits two DBE variants, whether two copies of one or one copy each of two. The homozygotes in the Altai programme were usually deaf; the kittens born deaf in the later literature carried DBE^CEL (Roxi) and DBE^ALT together.

What the Altai's history does show is that this was understood in the founding population before any variant had a molecular name, and that a rule was written for it. That rule, and the gap left by triggering it on eye colour, is examined alongside the other registries' rules in the survey's case study on the registry rules.

Every pair of variants that meets in this register

The same question asked of each pair: did the second variant come from above the founder, in which case one descends from the other and it is an origin question, or only from below, in which case they met in a later mating and it is a compound question.

Four pairs of variants meet in this register: DBE^ALT with DBE^CEL (Roxi), DBE^ALT with DBE^RE, DBE^ALT with the Igor line, DBE^CEL (Roxi) with DBE^RE. Three of them involve DBE^ALT and one does not. In every pair, nothing sits above the founder: not one of these is an origin question. Every collision in the register happened in a mating.

Five cats stand below three founding descents at once. No figure drawn for a pair can show that, and a genotype for one variant says nothing about the other two.

A pair that can meet and has not been recorded doing so. DBE^ALT and DBE^NAD both sit in the British Shorthair and British Longhair. No cat in this register stands below both, so the figure below is not a collision. It is the mating that is available inside one breed, and neither parent need show a blue eye for it to happen.

How the variant is named, and what the name covers

The name records a decision taken at one moment by one person, and is read by everyone afterwards as a statement about scope.

Each published variant is named for a population or a line: DBE^ALT for the Altai, DBE^RE for the Rociri Elvis line, DBE^AGO for the Agostino line, and DBE^CEL (Roxi) for the Céleste.

The name is the breed's, and the breed's name is the cattery's. The laboratory work was done on Lazuli, an existing registered descendant of Roxi, and the designation, assigned by Lazuli's owner, takes the name of the Céleste rather than of Lazuli, of Roxi, or of the population either came from. There is no rule governing what a variant is named after, and this survey proposes none: naming belongs to the authors who publish a variant and to the bodies that code it afterwards.

What the name covers is a separate question from what chose it. The Céleste is one branch of Roxi's descendants and the most recent of them. Roxi founds every line in which the variant has been genotyped, including the Topaz, which used him before the Céleste existed. The Maine Coon carriers descend from the Topaz and not from the Céleste, so a reader who takes the name as a statement of scope will place those cats in the wrong population.

The finding is not about the name. It is that a breed was defined by a variant that is lethal in the homozygote, so it cannot be bred to itself, its stud book can never be closed, and every generation is by construction half non-carrier. That holds whatever the variant is called.

The founding paper genotyped 130 cats with the trait and 101 controls. Sixty-seven carried DBE^CEL (Roxi) in all, whatever their eye colour: 45 Céleste with the phenotype, one white Céleste homozygote that died at birth, one latent Céleste, seven Céleste-mixed cats, nine Maine Coons of the Topaz line and four Siberians of unrecorded origin. The variant is confirmed by genotype in the Céleste, in Céleste crosses, in the Maine Coon through the Topaz, and in the Siberian. A further twelve rows of the breed table report it without verification.

The creation of a breed from a variant: how the Céleste was made

The Céleste is the breed whose name DBE^CEL (Roxi) carries. Its origins are settled above; what follows is how it was made into a breed, from the refusal of the trait in the British Shorthair to recognition by LOOF, and what the terms of that recognition did.

Step by step, from Roxi Darlin to a recognised breed

1. The line is Roxi, not Céleste. The variant traces to Roxi Darlin, the same cat written Roksi in Kazakh sources and Roxy in French ones, a tabby male with white hind feet and a white throat, bred at the Darlin cattery in Ust-Kamenogorsk. All cats carrying DBE^CEL (Roxi) descend from him. His own pedigree is not recorded, and his first use was not in France but in the Altai programme set out above, and then in the Topaz, which is described in the survey's case study on the registry rules.
   Roxi and Lazuli. These have been reported as competing names for the cat the variant is traced to. They are not. Roxi is the source of the variant. Lazuli is a later male cat of the same line, acquired by the founding cattery from a Ukrainian breeder and in her house from June 2015, and it is from Lazuli that LOOF's official genealogy begins.
2. Introduction into the British Shorthair. A breeder of British Shorthairs under the Aerlin prefix mated one of her British cats to a blue-eyed cat of the Roxi line. The founding cattery, Les Célestes d'Aerlin, describes itself as élevage de British à la base, a British cattery to begin with. The programme began with a cat found in Ukraine. The cattery's own account: “j'ai continué longtemps mes recherches, qui m'ont conduite jusqu'en Ukraine, où je suis tombée un peu par hasard en 2014 sur une éleveuse ayant un chat brown tabby, presque sans blanc, aux deux yeux turquoise, venant lui même du Kazakhstan : Roxy”, that is, a search that led to Ukraine, where in 2014 she came upon a breeder with a brown tabby, almost without white, with two turquoise eyes, himself from Kazakhstan: Roxy. She then acquired Lazuli, a male kitten of that line, who was in her house from June 2015.
   Where the official record begins, and why that is not where the descent begins. LOOF accepted a working dossier and granted an official genealogy for her cats under the unrecognised-breed codes XSH and XLH, in the registre de filiation, “en partant de Lazuli”, starting from Lazuli. That is a statement about documentation and not about ancestry. The registration begins at Lazuli because an official pedigree could not be obtained for the generations before him, not because those generations are unknown: the line back to Roxy is known to the breeder and is recorded in her own material. A stud book starts where the paperwork starts.
3. The trait was refused within the British Shorthair. The founding cattery states the sequence itself: “A la base le projet était de travailler les yeux bleus chez le British. Les clubs de race en France s'y sont opposés, nous avons donc créé le Céleste, avec comme contrainte du LOOF d'en faire une race qui soit différente, et pas uniquement de par la couleur de ses yeux.” In English: “The project was originally to work blue eyes in the British. The breed clubs in France opposed it, so we created the Céleste, with the constraint from LOOF that it be a breed which is different, and not only by the colour of its eyes.” The French breed history puts the refusal in four words, cette nouvelle couleur de british est refusée par les clubs. There is no accepted British Shorthair standard for the trait.
4. LOOF imposed a condition that shaped the breed. On the cattery's account LOOF required the new breed to differ in more than eye colour, which is why the standard specifies conformation as well as the trait.
5. It was then registered as a separate breed. LOOF admitted the Céleste as a preliminary breed in 2022 and as a nouvelle race in 2023, making LOOF the only registry to have admitted a dominant blue eye breed rather than restricted one. The variant was subsequently named after that breed rather than after the Roxi line.
6. Dispersal did not occur solely through the Céleste. As the founders table shows, DBE^CEL (Roxi) in the Maine Coon came from Roxi through the Topaz and into the breed by an outcross, and the Céleste was created afterwards from a separate Roxi-sired male, so the two are parallel branches from one founder rather than a trunk and a branch. The paper describing the variant records that Roxi was heterozygous for the mutation, and that the same mutation was found in nine dominant blue eye Maine Coon cats related to Roxi and in four dominant blue eye Siberian cats of uncertain origin. The nine Maine Coons are described by their relationship to Roxi and not to any Céleste. The variant is also genotyped in the Sphynx. Further breeds have been reported without verification.

What followed from the refusal

Refusal did not remove the trait from British breeding. It gave the trait a breed of its own, that breed was recognised by a national stud book, and recognition carried a rule requiring an outcross to British Shorthairs and British Longhairs in every generation. Mating back to the breeds that refused it is now a requirement of the standard rather than an option within it.

No intention is imputed here. The founding cattery's purpose was to create a breed, and it says so. Dispersal is a consequence of the mating rule that recognition came with. The finding is about what the rule does, not about what anyone wanted.

The breed did not grow and the allele did. LOOF's own breed page states the ban, “Le mariage Céleste x Céleste est interdit à ce stade”, the Céleste to Céleste mating is prohibited at this stage, and gives the permitted matings as exactly three: Céleste to British Shorthair without white, Céleste to British Longhair without white, and Céleste to domestic cat without white. Two of the three permitted partners are therefore the British breeds in which the trait was refused, and the third is not a breed at all.

The third option is different, because chat domestique means an ordinary unregistered cat. A breeder exercising it is mating a pedigree cat carrying a known dominant variant to an animal with no pedigree, no recorded ancestry, no health testing and no registry oversight of any kind. The resulting kittens are registrable.

• The variant leaves the pedigree world entirely. An outcross to a British Shorthair puts the variant into a breed where, in principle, it can be traced. An outcross to a domestic cat puts half the litter's ancestry beyond any register, and any sibling or parent of that domestic cat is untraceable.
• The unregistered partner may itself be a carrier and nobody would know. Twenty named lines in this survey's register hold a blue or odd eyed cat in which no variant has been identified, and several of them run through unregistered domestic cats. A chat domestique sans blanc that carries an unmapped variant, or is a latent carrier of a known one, is indistinguishable from one that carries nothing, and no test can settle it for the unmapped lines.
• Pedigree depth is reset at every use. A breed permitted to outcross to unregistered animals in every generation, and forbidden to breed to itself, cannot accumulate recorded ancestry. The pedigree issued at the end of it is a document about one generation.

LOOF's own annual statistics show the result.

103 pedigrees in the three years since recognition, against 4,151 British Shorthair and 24,197 Maine Coon in 2025 alone. Every one of the 103 came from an outcross, because the standard permits nothing else. That does not make all 103 carriers. A carrier bred to a non-carrier gives about half and half, and the whole litter is registered Céleste, so the figure is an upper bound on registered carriers and not a floor. One is named: Uskya d'Ysorian, registered RIEX, is described by her breeder as non porteuse du gène DBE-CEL.

What the Céleste does that no registry requires, and what LOOF still cannot do

Two facts about this breed belong together, because one is the best voluntary practice the survey has found anywhere and the other is the limit that practice runs into.

The breed hearing-tested its first five generations. The Céleste programme records that the first five generations underwent brainstem auditory evoked response testing, confirming normal hearing, with every breeding cat additionally screened for polycystic kidney disease and hypertrophic cardiomyopathy. No registry in the survey requires a hearing test of any cat. This is the strongest voluntary control found in any dominant blue eye population, it was done in the breed whose founding animal is the one genotyped cat in the line, and it was done without being asked.

Set against that, LOOF cannot yet act on a test result even when one exists. As of March 2026 the registry had no agreed protocol for DBE^ALT testing, and laboratory results accordingly “ne seront pas pris en compte par le LOOF en l'absence de protocole d'accord”, that is will not be taken into account by LOOF in the absence of an agreed protocol. A commercial test has existed since 2024 and the registry that created the breed will not receive its results.

The mating rule holds the line in the meantime. LOOF's standard permits Céleste to British Shorthair without white, to British Longhair without white, or to a domestic cat without white, and states that “le mariage Céleste x Céleste est interdit à ce stade”, the Céleste by Céleste mating is forbidden at this stage. That is a ban on the mating rather than on the cat, which is the control this survey recommends, and it is being carried by a breed standard because no test result can currently reach the registry that wrote it.

A breed defined by descent from one cat whose ancestry is undocumented

LOOF's 2019 ruling makes descent from Lazuli the definition of the Céleste. A proposed second line was refused because its founder was unrelated to him, so every Céleste must trace to him.

His own ancestry is not documented, and the registry says so. The registration was made “en partant de Lazuli”, starting from Lazuli, because an official pedigree could not be obtained for the generations before him. Published accounts of where he came from differ: one has him a kitten acquired from a Ukrainian breeder, one has him born in France out of Roxi, one has him imported from Kazakhstan. Roxi was a Kazakh domestic cat, so the line enters France from outside it whichever account is right, and the discrepancy does not change what follows.

Two consequences follow, and they are the ones that matter. A breed in which every animal descends from one cat cannot be assessed against an ancestor-count rule at all, and its coefficient of inbreeding can only rise as the population breeds on; the obligatory British outcross is what holds it down, which makes that outcross a genetic requirement and not a matter of type. And since April 2024 a test for DBE^CEL (Roxi) has existed, so the descent the breed is defined by could be confirmed by testing every registered Céleste. Nobody does it.

One point of record. Ojos azules appears in the third account as a registration category rather than a claim about breeding: for a period it was the only heading under which such a cat could be entered anywhere. The survey holds two further instances of the same mechanism, an Altai F2 and a British Shorthair of the Igor line, both filed under that name for want of a better one. A cat filed under an extinct breed for want of a category carries a breed attribution that means nothing about its genetics.

Three ways a cat leaves this breed, and why that makes it hard to call it one

A breed is normally a population that can reproduce itself. The Céleste cannot. Cats leave it by three routes, all of them written into the rules.

1. By being sold abroad. One registry recognises the Céleste, in one country. A cat sold outside France cannot be registered as one, because there is nowhere to register it. It enters some other programme, under some other registry, as something else.
2. By the compulsory outcross. The standard forbids Céleste to Céleste and requires a mating to a British Shorthair, a British Longhair or a domestic cat in every generation. Half of every pedigree is therefore from outside the breed, by rule.
3. By not inheriting the variant. The determinant is dominant, so on average half of each litter does not carry it. Those kittens are the offspring of a Céleste mating and are not what the breed was recognised for.

So the variant stays inside the Céleste only for as long as every cat carrying it remains in France under LOOF. Nothing else holds it there, and the standard requires an outcross that puts it into British cats in every generation.

A second account of the dispersal, mapped from breeder records

A breeder-compiled lineage genealogy on messybeast.com maps the same history from breeder records rather than from genotyping, and is the most complete account of the dispersal available. It traces almost all of it to one source in Ukraine and agrees that the Topaz was founded on both variants. It is reported to differ on one point, an annotation that the Céleste eliminated the Seymour gene and carries only the Altai gene. The elimination is consistent with the genotyping; the label attached to it is not, for the reason set out above.

The annotation is recorded here as reported and not as read, and no wording from the chart is quoted. Nothing in this section depends on it.

The chart establishes the shape of the dispersal. One source population in Ukraine reaches at least a dozen breeding programmes across France, Germany, Russia, Australia and Canada, in the British Shorthair and Longhair, Maine Coon, Persian, Ragamuffin, Munchkin, Scottish Fold, Minuet, Highlander, German Longhair and Sphynx. No registry required a test at any point on that map, and no registry holds the map.

The response, part one: two registries, one gene, and rules that point in opposite directions

DBE^CEL (Roxi) and DBE^ALT are two variants of the same gene and they behave the same way: dominant, so one copy shows; harmful when a kitten inherits two DBE variants; and invisible in a proportion of the cats that carry them. Two registries have written rules for them. The rules differ in more than detail. They are structurally opposite, and neither uses the instrument that would answer the question either is asking.

One registry has written a rule that guarantees the variant leaves the breed. The other has written a rule that guarantees it stays. Both are responses to the same problem, that two copies are harmful, and both address it by controlling which cats may be mated rather than by establishing which cats carry the variant. A commercial test exists for both variants, and neither registry requires it or records the result.

The Altai rule and its 2030 closure are examined in full below.

A third position completes the picture. The GCCF neither sends the variant out of a breed nor concentrates it in one. It refuses the trait in some breeds, permits it in others and restricts it in a third group, and it is the only registry in the world that records on a cat's registration which variant that cat carries. It has never published a figure drawn from those records. Its coding has been live since June 2025, so the number of cats it has recorded under each variant is known to the GCCF and to nobody else.

Three registries have therefore taken three different approaches to a single gene. LOOF's rule moves the variant out of the breed that carries it. The World Cat Federation's rule keeps it inside a stud book that closes in 2030. The GCCF alone records which variant each cat has, and publishes none of it.

The mating rule the Altai breed standard already has, and the gap left by triggering it on eye colour

The World Cat Federation standard for the Altai, drawn up in 2018, does what this case study proposes: it prohibits the mating and not the cat. Two blue-eyed or odd-eyed Altai may not be bred together, “because it may lead to deafness and white coats in cats homozygous for this dominant gene”, and a blue-eyed or odd-eyed Altai may be bred only to a non-blue-eyed Altai, green-eyed by preference. Outcrossing to seven named breeds is permitted until 1 May 2030, after which only Altai to Altai matings are allowed, and the ban on blue to blue continues.

The World Cat Federation's Altai standard is the closest thing in the world to the control proposed at the end of this case study, and it fails at exactly one point: it is triggered on eye colour. A latent carrier is not blue-eyed or odd-eyed, so the ban does not reach it, and the same messybeast.com compilation that records the rule also records that latents exist and produce blue-eyed offspring. The rule is right in its structure and wrong in its trigger.

Why closing the stud book in 2030 makes the trigger worse, not better

The failure is not static. It gets worse on a known date, and this survey records it as a risk the standard does not appear to have considered.

Until 1 May 2030 an Altai may be outcrossed to an American or European Shorthair, a British Shorthair, a Bengal, an Abyssinian, a Somali or a Siberian. A cat brought in from one of those breeds is almost certainly not a DBE^ALT carrier, so an outcross is a reliable way of putting a non-carrier on one side of the mating whatever its eye colour. After that date only Altai to Altai matings are permitted, and the only screen retained is eye colour.

From 2030 the permitted mating is a blue-eyed or odd-eyed Altai to a non-blue-eyed Altai. If that second cat is a latent carrier, the mating is carrier to carrier and a quarter of the litter is expected to be homozygous, which in this variant means a cat that is white or nearly white and usually deaf. That is precisely the outcome the rule exists to prevent, and nothing in an eye-colour screen can detect it.

• Nobody has recorded what a carrier to latent-carrier mating produces in this breed. The founding programme observed that homozygotes were viable and usually deaf, and it observed that latent cats exist, but no study has put those two observations together and followed the matings. The frequency of latency in the Altai population is not published either, so the size of the risk cannot be estimated.
• The risk rises as the stud book closes. A closed population becomes more related over time, so the proportion of cats carrying the variant without showing it can be expected to rise rather than fall. An eye-colour screen therefore becomes less effective in the same years in which it becomes the only screen.

The remedy is available and inexpensive. DBE^ALT has a commercial test, sold as DBE3 on the five-variant panel, so a breeder can establish whether a green-eyed Altai is a latent carrier for the cost of one swab. Substituting genotype for eye colour would leave the rule's structure untouched and close the gap entirely. On the evidence available this survey treats the 2030 closure as a red flag: a mating restriction that will lose its outcross safeguard on a fixed date while continuing to screen on a trait that does not identify the carriers.

The Altai standard and the founding-programme observations are taken from Altai / Altay at messybeast.com, a self-published compilation worked up from breeder records and pedigrees, and not peer reviewed. Where it is quoted above it either agrees with the published genotyping or is marked as reported. Roxi's own pedigree is described there as unknown, so his relationship to the Altai founder line cannot be stated.

The response, part two: four registries, one trait, and four different answers

Everything above describes the first of the four answers: LOOF refused the trait in the British and then recognised a breed built on it. Three other registries have since faced the same trait in the same breed and answered differently, and one of them arrived independently at almost exactly the French solution. The three are taken in turn below and all four are set side by side at the end, because they began from the same point and did not end in the same place.

The second answer: what the GCCF decided

The British Shorthair Breed Advisory Committee amended the British Shorthair Registration Policy, and the Council approved the revision on 27 June 2026. The operative words are these:

“Blue-eyed/odd-eyed cats of BSH appearance in any pattern other than epistatic dominant white, colourpointed or colourpointed & white shall be registered on the Reference Register with no progression as no recognised breed (XSH) and are not acceptable in the pedigrees of British Shorthairs.”

• The cat is registered, and registered as nothing. XSH is the code for an unrecognised shorthair. The cat exists on the Reference Register, cannot progress to the British Shorthair register, cannot be shown, and cannot be bred from, because unrecognised cats may not be used for breeding.
• The exemptions are three. Epistatic dominant white, colourpointed, and colourpointed and white. In those three patterns a blue eye has an established cause other than PAX3, so the rule is directed at blue eyes that cannot be otherwise explained.
• Descent is cut, not restricted. The cats are “not acceptable in the pedigrees of British Shorthairs”, so the exclusion reaches backwards through any pedigree containing one, and not only forwards from the individual animal.
• The trigger is the eye. The rule reads on “blue-eyed/odd-eyed cats of BSH appearance”. It is a phenotype rule. It names no variant and requires no test.

The same code, the same breed, two destinations

The comparison with France is exact enough to be worth stating in a table, because both registries reached for the same instrument and put it to opposite use.

The two registries began from the same refusal and the outcomes differ completely. Nothing about refusing a trait determines what happens next. What determined it was what each registry allowed the unrecognised code to lead to. France left a route from XSH to recognition open and the trait took it. The GCCF closed that route.

What the rule reaches, and what it does not

• A latent carrier passes through it. The rule triggers on a blue or odd eye. A cat carrying a DBE variant without blue eyes is not caught by any part of the wording, and remains registrable and breedable as a British Shorthair. The founding Kazakh programme described latent carriers before any variant had a molecular name, so this is not a hypothetical class of animal.
• The registry that could have written it on genotype did not. As set out above, the GCCF alone records a DBE variant on the cat's own registration record, through codes 162, 163 and 164. It was therefore the single organisation in a position to make this rule turn on a test result rather than on an eye colour, and the amended policy names no variant.
• The cats are removed from the register, not from breeding. The rule governs what the GCCF will register. It does not govern what exists. A cat refused British Shorthair registration by the GCCF remains eligible for registration with other organisations operating in the same country and internationally, whose British Shorthair policies differ. The effect of the rule is that the GCCF no longer records these cats, and a cat the GCCF does not record is a cat about which the GCCF holds nothing.

No outcome is asserted here. The policy took effect on 27 June 2026, it applies only to registrations submitted from that date, and what follows from it is not yet observable. The French sequence set out above is recorded as what happened in one jurisdiction and not as a forecast for another. What can be said now is what the instrument does and what it leaves untouched.

The policy has been revised three times in eight months: effective 12 November 2025, revised 29 April 2026, and revised again with Council approval on 27 June 2026. The wording quoted above is the June 2026 text.

The third answer: a new breed, with the rule this survey recommends written into its standard

A second new breed has been created for the trait, and it was created independently of the Céleste, in Germany, from British Shorthairs and Longhairs carrying the Igor line. It is called the Azul, in shorthair and longhair forms, and it is recognised as an experimental breed by the Association of Rare Breeds of Cats. The name means blue, for the eyes, and is unconnected with the American Ojos Azules.

The sequence that produced it is the Céleste's sequence, arrived at separately. The trait was refused by most clubs and British breeders did not want their breed's name attached to it. A new breed was therefore created. To register it the association required a standard differing in more than eye colour, which is the identical condition LOOF imposed on the Céleste. And the resulting standard prohibits Azul to Azul matings and permits an outcross to a British Shorthair or Longhair, which is the Céleste's mating rule in different words.

Two registries on two continents, working without reference to one another, produced the same instrument. That is worth stating plainly, because it means the structure described in this case study is not a peculiarity of French practice. It is what happens when a trait is refused inside a breed and then given a breed of its own.

The Azul standard, however, does three things the Céleste standard does not, and each appears in the prevention control set out at the end of the survey's case study on the registry rules.

• It requires one variant per breed. The standard states that British Shorthairs and Longhairs with a dominant blue eye gene may not be recorded as Azul, expressly because there are different dominant blue eye genes in different British-derived breeds and Azul cats must all have the same one. That is the prevention control set out at the end of the survey's case study on the registry rules, written into a breed standard by the people using it. No registry examined in the main survey has such a rule.
• It closes the return path in both directions. Azul cats and their offspring may not be recorded as British on a pedigree. The Céleste standard requires the British outcross and says nothing equivalent, so the Azul is the only instrument found that names the direction of travel and blocks it.
• It codes the trait itself. An ARBC pedigree certificate codes a cat ny 11 08 60 and glosses that on the document, in English, as “black gold, shaded, dominant blue eyes, little white spots”. So 60 is confirmed as a dominant blue eye and 08 as the accompanying white, from the issuing registry's own paper. Four cats on that one certificate carry 60, across four generations. A secondary description of the same list gives 69 for latency and 68 for mosaic eyes; neither appears on the certificate, and 68 denotes a plain blue eye at WFA and albino at FIFe, so those two codes are unverified and the mosaic reading of 68 conflicts with two other registries. Two other registries record something about latency and both can now be verified: WFA issues 68c to 68e for latent carriers, and this survey holds a WFA certificate on which an entire litter of three is coded 68e; and REFR marks the registration number of every kitten of blue-eyed parents, including those with no blue eye, because it may carry the variant. All of them record it by parentage or by appearance. None records it by genotype.

The breed's machinery is otherwise conventional: a standard on the British type, minimal or no white, and a breeding ban on any cat outside the permitted colours or produced by a forbidden mating. The founding female was acquired in 2022 and the first surviving litter born on 27 March 2024. That litter is one of the few anywhere in this subject for which every animal is named and coded: five kittens, two showing a blue eye, one coded c 10 63/1 and described as “DBE odd-eyed, lilac, no visible white”, another odd-eyed with minimal white, and two showing no blue eye at all. On the breed's own rule those two may carry the variant, and nothing on their pedigrees records it.

The fourth answer: an existing breed name, used for cats it was not written for

The fourth answer is not a decision by a registry at all, which is what makes it the limiting case. No breed committee met and no rule was drafted: a breeder registered blue-eyed cats under Ojos Azules, a TICA breed whose own recognition has lapsed, and the cats were admitted under that name. The line continues with no breed committee and no health rule attached to it, and nothing anywhere records which variant, if any, those cats carry. A lapsed breed name is the one route that asks nothing of anybody, and it was available the whole time the other three registries were writing rules.

The four answers side by side

The same trait, in the same breed, has met four different registry responses, and the outcome was never determined by the refusal.

Read across the table. Two of the four refusals produced a new breed, one produced a dead end, and one produced a change of name on the paperwork. Three of the four leave the line able to continue, and in two of those the continuation runs through the British breeds in which the trait was refused. Only one registry records anything about the variant, and only one records the latent carrier, and they are not the same registry.

The Ojos Azules route is set out in full, with the cats named, in the line directory in the survey's case study on the registry rules.

What happened in Germany before the Azul, and why it is the clearest case of latency causing harm

The Azul was created against a background the same source records, and it is the harm this survey has argued follows from an appearance-based rule, documented at the scale of a national breeding population rather than a single litter.

In Germany a large number of latent cats were bought by breeders who did not know they carried the trait. Deaf kittens followed, because those breeders mated latent to latent, or mated two blue-eyed cats without knowing what that meant. German law forbids breeding cats likely to be deaf. The law was in force throughout, and it did not prevent any of it, because nothing about a latent carrier is visible to the person buying it.

That is the whole argument of this survey in one sequence. A rule triggered on eye colour cannot reach a cat with ordinary eyes. A law prohibiting the breeding of deaf cats operates after the kitten is born deaf. The only instrument that would have worked is a test result recorded against the cat and carried with it when it was sold, and for the Igor line no such test exists to this day.

The breeders who founded the Azul responded by neutering every dominant blue eye cat they held other than the one line, in order to keep a single variant in the programme. That is the third breeding programme recorded in this survey to adopt the one-variant control voluntarily, alongside the two Maine Coon programmes named in the line directory. No registry requires it of any of them.

What no registry records: how many carriers there are, and why the number cannot be obtained

How many cats carry this variant is a simple question, and it has no answer anywhere.

The quantity the argument needs is the number of registered cats carrying DBE^CEL (Roxi) worldwide. It does not exist, and the reason it does not is itself a finding about registry governance.

No registry outside the GCCF records a genetic variant on a registration record. Every other registry examined records either a colour, a breed or nothing: FIFe requires a test for the PAX3 gene rather than for a named variant, and its rule § 3.5.1 places the duty to register the result on the member, which produces a paper record in each member country and no central total. TICA, CFA and WCF record nothing genetic beyond parentage, and LOOF, as set out above, holds no genotype for the breed it recognised.

The laboratories necessarily hold the number, since each knows how many positive results it has issued. Three offer a DBE^CEL (Roxi) test and none publishes an aggregate figure; a fourth has withdrawn its test on the ground that false positives cannot be excluded. That such counts can be published is shown elsewhere: Wisdom's MyBreedData gives per-breed carrier counts across more than 11,000 cats, and PAX3 is not among the variants it covers.

A worldwide count is therefore not obtainable at present. The published genotyping is the only verifiable floor, and 65 blue-eyed carriers in one paper, the sixty-seven less the white homozygote and the latent cat, already exceeds the 46 Céleste pedigrees LOOF issued in 2025. Four partial counts could be assembled, in ascending order of difficulty: the published floor; a cumulative LOOF count of Céleste registrations, every one of which is a carrier; a GCCF query on code 162; and aggregate positives from the laboratories. Any count built on eye colour will undercount, because latent carriers do not show the trait.

One consequence falls on the Maine Coon, and it turns on a name

The variant is named after a breed it did not come from. DBE^CEL is called the Céleste variant because the Céleste is the breed that was later built on it. It is not called that because it arose in the Céleste, and it did not.

The order of events is the whole point.

1. Roxi, an outbred domestic cat from Kazakhstan, carries the variant. He is not a British Shorthair and has no British Shorthair in his pedigree, which is recorded as unknown.
2. The Topaz programme uses Roxi alongside Nikita'l Seymour. DBE^CEL reaches the Maine Coon through the Topaz.
3. Only afterwards is the Céleste recognised as a breed, and only then does its standard require an outcross to British cats in every generation.

So the British connection belongs to step three and the Maine Coon received the variant at step two. Because everyone now meets the variant through the Céleste, and the Céleste is mated to British cats by rule, DBE^CEL found in a Maine Coon is read as proof that Maine Coons have been outcrossed to British Shorthairs. The published record does not support that reading. The variant entered the Maine Coon from a Kazakh domestic cat by way of a discontinued Russian breed, years before the Céleste existed and before any British outcross rule was written.

The descendants are spread across breeds, which is what makes the name misleading. The register holds 131 animals descending from Roxi by unbroken parent links.

More Maine Coons than Célestes, and three British Shorthairs. Neither half of the name describes the population it is attached to. A variant named after a breed is not evidence of that breed in a pedigree. The name records who registered a breed on it, not where the variant came from.

Widen it from the 131 below Roxi to the whole DBE^CEL population of 140, which adds the eight animals the register records the variant for without a line to him, and split it by whether the cat sits inside the registered Céleste line. The registered line is the smaller half.

Lazuli is the animal the LOOF requires every Céleste to descend from, so descent from him is the test of whether a cat is in the registered line. He is himself out of a Maine Coon. The variant was in the Maine Coon before it was registered to the British side, and the 55 Maine Coons here are not in the Céleste line at all.

The rules LOOF and the World Cat Federation have written for these two variants point in opposite directions, one exporting the variant and one concentrating it. That comparison belongs with the rule-design argument and is set out in the survey's case study on the registry rules.

What a breed is for, when its own standard forbids breeding it to itself

Two dominant blue eye breeds have been created. Both are instructive and neither contained anything.

• The Topaz was founded deliberately on two different variants at once, Roxi carrying DBE^CEL (Roxi) and Nikita'l Seymour carrying DBE^ALT. A population holding two variants produces compound heterozygotes in every generation, and a compound heterozygote, meaning a cat with one copy each of two different variants, is the state in which deafness and neonatal death are documented. The Topaz has been discontinued, and DBE^CEL (Roxi) reached the Maine Coon through it.
• The Céleste was founded on a single variant which is reported lethal in the homozygous state, so no Céleste can inherit two copies and live. Its standard therefore forbids breeding the breed to itself and requires an outcross in every generation, which is why its 46 pedigrees in 2025 were all the product of an outcross.

The breed has stayed small and the variant has not. DBE^CEL (Roxi) is confirmed in six populations by genotyping and reported in twelve more without verification, while the breed it is named after registers fewer than fifty cats a year. The registry that recognised the breed records no genotype for any cat in it.

The question this raises is one of rule design rather than of genetics, and it is the question the survey's case study on the registry rules answers with a protocol. A trait whose harm arises only when two copies meet cannot be governed by making it the defining feature of a breed, because a breed is a closed population and this trait forbids closure. The instrument that fits the biology is a ban on the mating, recorded against the individual animal and carried with it when it is sold. That is what the Altai standard attempts, in the one respect in which it is right, and it is what breed recognition cannot do at all.

Sources for this case study are given in each section above. The founding cattery's breeder-compiled genealogy and the independently compiled lineage genealogy are primary records held outside any registry; the genotyping is from Abitbol et al., Animal Genetics 2024 and 2025, and Rudd Garcés et al., G3 2024; the LOOF registration figures are from LOOF's own published annual statistics for 2025. Declaration of interest: the author of this survey is a co-author of the 2025 paper.

---

Case study two · DBE^CEL (Roxi) in the Siberian, the Munchkin and the Maine Coon

How a variant arrives in a breed

A breed can acquire the same trait more than once, by routes that do not meet. Where that happens, a rule written for the breed governs cats whose genetics differ. The Siberian is the clearest documented instance in this survey, because both the genotyping and the genealogy exist and they do not say the same thing.

What the research found in the breed

DBE^CEL (Roxi) is confirmed by genotype in the Siberian. Abitbol, Couronné A, Dufaure de Citres C and Gache V, Animal Genetics 2024;55:670-675, Table 1, records four Siberians heterozygous for the variant. In the same year Abitbol M, Dufaure de Citres C, Rudd Garcés G, Lühken G, Lyons LA and Gache V, Animals 2024;14:1845, Table 2, genotyped four Siberians with the trait and found them wild type for DBE^ALT. Wild type means the ordinary, unmutated version of a stretch of DNA, so a cat genotyped wild type for a variant does not carry that variant. No Siberian has been reported positive for DBE^ALT in any published series.

One mating carried two variants into the breed

Breeder-compiled genealogy C7, headed Blue-Eyed Siberians (Altai Gene), records a single point of entry. Kaena, out of March (Marta) by Seymour (Seimur) and so of the DBE^ALT lineage, was mated to Lazuli, born 2 April 2015, a son of Roxi and so of the DBE^CEL (Roxi) lineage. The mating took place on 18 July 2016 at the D'Aerlin cattery in France and produced seven kittens. Merlin comes out of that litter, and every Siberian below him on the chart descends from a cat with one carrier parent of each variant.

The link into the current population

The branch does not stay outside the breed. The litter of 14 November 2021 is by a Maine Coon sire, Kiwano's Rock-a-Hula, MCO n 09 22, born 29 September 2012. The litter of September 2023 is by a Siberian sire, Molaska Nouky, SIB ns 24, born 13 July 2017, out of Sabrina Legende des Alpes. Two of the five kittens of that litter are blue-eyed, Uill and Uindy. The descendants of the 2016 mating are therefore being bred to registered Siberians, and the trait is passing into the current population by that route.

Every cat below Merlin is registered XLH, experimental longhair, while the sire brought in for the 2023 litter is registered SIB. The register therefore records the branch as experimental at the same time as it records a Siberian stud in it. Which of the two variants these cats carry has never been established. Merlin is out of a carrier of each and has not been genotyped, and the chart assigns DBE^ALT throughout on no test. DBE^CEL (Roxi) is recorded in some cases below him on messybeast.com and remains officially unpublished.

The other blue eyes in the breed, and why they are not this trait

Breeder-compiled genealogy C3, headed DBE Siberians, unconfirmed if Altai gene or Barnaul gene from Altai Region, records a Russian population with a pedigree running back to Figaro Onix Gloria, SIB n 22, born 1 October 2006. It has been read as a second and older source of the trait in the breed. Its two earliest cats recorded as blue-eyed are white cats, and their certified pedigrees account for the blue eye without reference to PAX3.

Its oldest generations are ordinary registered Siberians. Figaro Onix Gloria and Blue Topaz Grilyazh, the chart's earliest dated cats, both appear as fourth-generation ancestors of registered Swiss Siberians of the Légende des Alpes cattery. Nine pedigrees of that line have been read, dated 2018 to 2025, each complete to four generations, the first eight 0.0 per cent inbred with thirty unique ancestors of thirty possible, and coded SIB n 22 or SIB n 09 24. No eye code is shown on any ancestor in any of them, which settles nothing either way, because the database does not record an eye colour on every entry. What they do settle is a matter of record: neither the chart's blue-eyed cats nor any cat of the Merlin branch appears in any of them. The early generations of C3 stand in these pedigrees as ordinary breed stock, drawn into the chart as ancestry.

• V*Grand Sineglaziy Angel, SIB w 61, born 20 April 2010. Four generations, twenty-eight ancestors of thirty possible, inbreeding 3.52 per cent. His sire, V*Rafaele Sineglaziy Angel, is NEM n 21, and the sire line is Neva Masquerade throughout, so he carries colourpoint. His dam, Lu-lu Beliy Sibirjak of Sineglaziy Angel, is SIB w 64, white with green eyes, out of CH Belosnezhka Sarji SIB w 64, out of IC Jhuliana Belosneghka SIB w 61, born 5 May 1999 to two foundation cats.
• Zephir Taiga Star, SIB w 61, his son. Four complete generations, thirty ancestors of thirty possible, inbreeding 8.01 per cent. White from his sire, and a Neva Masquerade dam, NEM n 21.

Neither pedigree holds a cat of any DBE line. Both were read from PawPeds, and both are complete enough to show it: the white and the blue eye descend the dam's line from foundation stock, skipping two generations of white green-eyed cats on the way, which is what dominant white does, and the other side of each pedigree is colourpoint.

The registrations are exactly what the breed's own standard provides for. FIFe permits a blue eye in the Siberian with EMS codes w, 01 and 02, and requires one in the Neva Masquerade, so a white Siberian coded 61 is a registration the standard anticipates rather than a finding. Both cats are recorded in this survey's register as blue eyes attributed to another cause.

Every remaining dated blue or odd-eyed cat on that chart is 2018 or later, after the mating of July 2016. A further group, the Malvina line, begins with a Siberian taken off the street in Barnaul in 2015 and has no variant established for it either. The survey therefore records one documented route by which a PAX3 variant reaches the Siberian, and it is the route through Merlin.

Every route to a blue eye recorded in the Siberian

A blue eye in a cat has more than one cause, and the Siberian holds most of them at once. Three of the causes are ordinary and have nothing to do with PAX3: the colourpoint gene, which every Neva Masquerade carries; dominant white, which makes a white cat; and white spotting, which makes a cat with white patches. Three more are routes by which a PAX3 variant, or a suspected one, has entered the breed. The table separates them, because a rule written for the breed reaches all eight and a test settles only some.

The last row is why none of this can be settled by eye. Every cat on the Merlin branch and on the Barnaul route is a tabby or a tortie with white. None is pointed and none is white, so the colourpoint and dominant white rows do not explain them. What is left is white spotting or a PAX3 variant, and the Schneesternchen cats are five Siberians of that exact description that were tested and turned out to be neither DBE^CEL (Roxi), nor DBE^ALT, nor DBE^RE.

So the position for the breed can be stated exactly. DBE^CEL (Roxi) is in the Siberian and was confirmed by genotype. No Siberian has ever returned a positive test for DBE^ALT, and four were wild type for it, meaning they carried none of it. Everything else, the Merlin branch, the Barnaul route and the Great Joy litters, is a blue eye with a documented pedigree and no genotype. That is one question, and a panel test on a handful of named living cats would answer it.

The Aktau route: an unidentified blue eye out of the Barnaul line and into three registered breeds

The Barnaul row of the table above ends with a sale to Uzbekistan. It does not end there on the ground. Velvet Slavicat, SIB f 24 03 63, out of the Malvina line, is the dam of a breeding programme at the Meowrigi cattery in Aktau, Kazakhstan, and the register carries sixteen cats below her: six household cats, two Munchkins, a Scottish Shorthair, a Scottish Fold and six British Shorthairs. Three of those are recognised breeds. Not one cat on the route has been genotyped, and no variant is attributed to any of them anywhere in this survey.

The shape of it is the same as the Merlin branch and the evidence behind it is weaker. A blue-eyed daughter, Meowrigi Cincinella, OA ns 03 24 61, is mated to four different sires of four different kinds: a household male, a Munchkin, a cat registered XSH, and a Scottish Fold. Every mating produces blue-eyed or odd-eyed kittens, and the blue-eyed son of the XSH mating, Meowrigi Karapka, is put to a British Shorthair queen and sires Achernar The Brightest Star Felicious, BRI a 09 61, a registered blue-eyed British Shorthair who is himself a sire in the generation below.

Two things follow. A blue eye of no established cause has reached the British Shorthair, the Scottish and the Munchkin by a route that touches neither Roxi, nor Nikita'l Seymour, nor the Céleste. And it runs five generations without a test at any point, so a rule triggered on eye colour reaches every animal on it while a test for the named variants would settle nothing, because the named variants are not what is being asked about. Meowrigi Cincinella, Meowrigi Karapka and Achernar The Brightest Star Felicious are the three cats a panel test would have to run on, and if all three return wild type, this is a mutation nobody has described.

Source: breeder-compiled genealogy C16, the Meowrigi cattery's own chart, with the breeder's note on Meowrigi Irisa supplying the parentage that links the British Shorthair generation to it. Graded line-traced where the parent links are given on the chart and reported where a cat appears without them. No cat on this route carries a genotype.

The rule the breed is governed by

FIFe's Siberian rule, FI and SK Proposal 1, in force 1 January 2026, allows blue eyes in the Siberian only with EMS codes w, 01 and 02, and requires any other blue-eyed or odd-eyed Siberian to be registered as XLH * 61 (SIB) or XLH * 63 (SIB), with no re-registration and no breeding. Its motivation states that in Siberians blue eyes are “likely caused by a variation of the so-called Altai gene (DBE^ALT)”. That is the one variant for which no Siberian has ever returned a positive test, and the four Siberians genotyped for it were wild type, meaning they carried none of it.

The same shape in a second breed group: the Munchkin and the Minuet

The Siberian is not the only breed group this survey can show being reached more than once. The Munchkin group is reached twice, by routes that do not meet, and the two rest on different kinds of document: one on registration certificates, the other on a breeder's account of where cats were sold.

The two routes have nothing in common. One begins with a village cat in the southern Urals a decade before any variant was described; the other begins with Roxi and reaches a related breed by documented sale. Only the second carries a variant, and only because the line it came from was genotyped.

The Céleste route into the Minuet is documented twice, by two separate sales on the founding cattery's dispersal record: Muscari of Aerlin to the Leonprime cattery, and Mistral and Neytiri to Miss Kitty in Canada. The Minuet is the Munchkin crossed to the Persian.

The Iberis Munchkins are not on that route. Certificates P3 and P4 and PawPeds pedigree P20 run them back to Semyon and to Russian household-pet foundation cats, and nothing in any of those documents touches the Céleste, Roxi, Lazuli or Nikita'l Seymour. Their blue eyes have no identified cause, and no Munchkin has ever been genotyped for this trait.

So the Munchkin group holds at least two separate sources of a blue eye. One is a named PAX3 variant with a commercial test behind it; the other is unidentified, and whether it is PAX3 at all is unknown. A DBE^CEL test clears a cat of the first and says nothing whatever about the second, which is the position a breeder in that group is in and the position a rule written on eye colour cannot distinguish.

The founding cattery's own record of where the cats went

The routes above are traced from pedigrees. A second kind of evidence comes from the breeders: the founding cattery published its own account of which cats it sold, to which catteries, in which countries and into which breeds. It is the only document in this survey that shows a variant leaving one breed for six others as a deliberate act, with the receiving catteries named.

Two things follow from reading it as a whole. The variant reaches six breeds in about two years, and every step is a sale between named catteries, not a coincidence of separate populations. And the chart labels a variant on five of the six rows and a query on the sixth, which is the same pattern the genealogies show: attribution travels with the cats, faster and further than any test does.

The same shape in the Maine Coon, and the one route that has a genotype behind it

The Maine Coon is the third worked example and the strongest, because it is the only one where the genotyping settles which of two variants passed. What follows is how the variant arrived in the breed, and that the whole breed's exposure runs through a single animal.

The animals below are named, coded and dated, the LOOF dates are the registry's own, and the dates settle the direction of transmission on their own.

One statement here is the compiler's rather than the record's and is marked where it appears: the genealogy's heading, “blue-eye line created using Roxi and Seymour genes”. No step in the table rests on it.

Blue-eyed Maine Coons carrying this variant were being bred four to five years before the breed the variant is named after existed as a breed at all. A Maine Coon of this line cannot have taken DBE^CEL (Roxi) from the Céleste. Both took it from Roxi, by separate routes, and the naming records the later of the two.

The two branches leave Roxi by two different offspring. The British Céleste descends through his son Lazuli, a Topaz; the Maine Coon line through his daughter Nikita'l Pheer-Ca, dam of Nikita'l Markiss. Neither route passes through the other, and neither breed took DBE^CEL (Roxi) from the other. They are sibling branches from one founder, which is what the published account describes when it records the Céleste as “created in France using an outbred male born to the Roxi sire” and the Maine Coon lineage as “founded by a Topaz cat”.

One consequence of the table bears on every Maine Coon of this line. Nikita'l Markiss is the son of a Seymour-line sire and a Roxi-line dam, so he may carry one copy of each variant, and every Maine Coon on the genealogy descends from him through Celebrity. A cat with one copy of each is a compound heterozygote, the combination this survey records as producing deaf and dead kittens.

How much of the breed's exposure runs through that one animal can be measured, and the answer is all of it. The register holds fifty-four Maine Coons descending by documented parent links from Nikita'l Seymour, and the same fifty-four from Roxi. Block Markiss and walk both descents again, every other route left open, and both figures go to zero. That makes him the single most informative animal in the survey: one genotype on him or on any surviving offspring would settle what fifty-four recorded Maine Coons stand to have inherited, and whether any stands to have inherited both. No genotype for DBE^ALT in a Maine Coon has been published, which is why that row of the breed table is a testing priority rather than a settled attribution.

Source: Blue Eyed Maine Coons genealogy, breeder-compiled, read 13 August 2026. Graded reported: no cat on it carries a genotype, and the genealogy marks the eye phenotype only. The LOOF dates are from the registry's own record. The published route, “the Maine Coon lineage founded by a Topaz cat”, is Abitbol et al., Animals 2024, 14(13):1845, which names no individual animals.

What this establishes about how a variant arrives in a breed

• A breed can hold the trait by more than one route, and the routes need not carry the same variant. One route into the Siberian is dated and traceable to a single mating on 18 July 2016 that held both DBE^CEL (Roxi) and DBE^ALT. A second is a Russian population documented from 2006 with no variant established. A rule written for the breed reaches all of them.
• A chart's title is not evidence of a variant. C7 is headed Altai Gene and assigns DBE^ALT to nineteen cats on the strength of that heading. The chart also enters Roxi, Nikita'l Seymour, Kaena and Lazuli as Siberians, which none of them is on any other record. A genealogy compiled before a variant was named carries the naming of its own date.
• An experimental register is not a barrier. The branch is registered XLH throughout and has taken a Maine Coon sire in 2021 and a Siberian sire in 2023. The register records the cats as experimental while the matings move the variant into two recognised breeds.
• Attribution belongs to the individual animal and is established by genotype. Twenty-one cats of this branch are recorded, none of them is genotyped, and this survey attributes no variant to any cat below Merlin. What the pedigree does establish is narrower and firmer: Merlin is out of a carrier of each variant and is blue-eyed, so he carries one of them or both, and every cat below him may carry either, both or neither. A cat with one copy of each is a compound heterozygote, the state this survey records as producing deaf and dead kittens. The cats that would settle it are alive, registered and identifiable by name.
• A whole breed's exposure can run through one animal, and that can be measured. Fifty-four Maine Coons in this register descend from Nikita'l Seymour and the same fifty-four from Roxi. Block Nikita'l Markiss and both counts fall to zero, so every one of them reaches both founders through him and through nothing else. Where a walk of the recorded ancestry produces a figure like that, it names the animal to test.
• The route with no variant on it travels as readily as the route with one. The Aktau branch out of the Barnaul line reaches the British Shorthair, the Scottish and the Munchkin in five generations without a single genotype anywhere on it, and without touching any named variant. A rule that tests for the described variants does not reach it, and a rule that triggers on eye colour reaches all of it.

Sources: breeder-compiled genealogies C3 and C7, close-up tiles supplied to this survey on 13 August 2026; Abitbol, Couronné A, Dufaure de Citres C, Gache V, Animal Genetics 2024;55:670-675, Table 1; Abitbol M, Dufaure de Citres C, Rudd Garcés G, Lühken G, Lyons LA, Gache V, Animals 2024;14:1845, Table 2; FIFe General Assembly, FI and SK Proposal 1, sections 6.22 and 6.23, in force 1 January 2026.

Case study three · the Igor line, and the variants it shares descent with

Igor, a PAX3 determinant without a named allele

Eight dominant blue eye alleles have been named and the Igor line carries none of them. It is the largest population in this survey whose allele has not been named, and the gene it sits in is not in doubt: the determinant is at PAX3. This case study sets out what establishes that, where the line meets the variants that already have names, and what happened the one time it met one.

The determinant is at PAX3, and this is what establishes it

Three things, none of which depends on the allele being named.

• Four of the eight named alleles have been excluded from the line by assay, in three rounds across three papers, every cat wild type every time. The rounds are set out in the table below. An exclusion is a test that ran and returned a negative, and a line that can be ruled out of four PAX3 alleles is a line a laboratory has been able to look at.
• The one cross this line has made to a named variant produced two white, deaf daughters. That is what two variants at one locus produce. Two variants at two different loci do not produce it, because a cat inheriting one of each is heterozygous at both and shows neither effect doubled. The mating is set out under the one time this line met a variant that has a name.
• The phenotype is the PAX3 phenotype. Blue or odd eyes with minimal to no white spotting, carriers that never show it, and a dominant that segregates one in two. Every named allele of this trait so far has been at PAX3, and nothing in this line's behaviour separates it from them.

What is not established is the allele. No sequence has been published for it and no cat of the line carries a result of its own, so every record here is graded on descent and none is graded genotyped. The locus is one question and the allele is another, and this register answers only the first.

Three published papers have used this line to exclude four named PAX3 variants

The line has been genotyped against the named alleles in three separate rounds, across three papers, excluding four of them. Every cat tested in every round was wild type, which establishes what the determinant is not and, taken with the cross, where it sits.

That a laboratory could rule the line out is itself the finding. An exclusion is a test that ran and returned a negative, not a test that was never done. Four of the eight named variants have now been excluded from this line by assay, and the line still has none of its own. Of the 58 that descend from the founder by pedigree, none carries a result of its own, and 51 records under the label are coded 61 on the strength of what a registrar saw rather than what a laboratory found.

Naming the allele is a smaller job than it has been treated as. The recommendation has been to sequence the line, which means a research group and a whole genome. The determinant being at PAX3, the first step is a targeted read of one gene in one cat. A commercial variant panel will not do it, because a panel reports the alleles already named and returns clear on an unnamed one, which is exactly what it has done here four times.

The informative animals are named on the record of the 2022 cross: the sire, the dam, and either daughter. A read on the sire alone says whether the Igor eye is a ninth PAX3 variant. A read on a daughter says whether she is a compound.

That is the shape of the problem this survey keeps meeting. The largest unidentified line in the survey, the cheapest test on the list, and, as at the census date of this survey, nobody has run it.

A documented mating points at the gene, the test costs less than a show entry, and no registry, laboratory or breed club is under any obligation to run it.

The one time this line met a variant that has a name

Two records in the register descend both from the Igor founder and from Nikita'l Seymour, the genotyped DBE^ALT founder. They are litter sisters, born 1 October 2022, from a mating made deliberately by a breeder who held both lines and wanted to see what they produced.

Neither parent is white. Neither parent is deaf. Two of two daughters were affected. The mating is documented on three stamped pedigree certificates. Neither daughter has been genotyped or hearing tested.

The sire and dam share 52 ancestors, the nearest at seven steps on the dam side and eight on the sire side. This was not an outcross, which matters when weighing what else might explain the outcome.

What that mating establishes about the locus

The harm described in the literature for these variants is the compound heterozygote: one copy each of two different variants. In the published litter, the two kittens genotyped compound heterozygous for DBE^CEL and DBE^ALT were white and deaf.

A compound heterozygote is two different alleles at one locus. Where a cross to a second line reproduces the same outcome, the second determinant must sit at the same locus as the first. A determinant at an independent locus would give a double heterozygote, which is an animal carrying one copy of each and is healthy. DBE^ALT in single copy does not cause deafness: nine genotyped carriers were hearing tested at a university veterinary school and all nine heard normally in both ears. The Igor determinant in single copy does not cause deafness either, and the sire of this litter hears.

The affected state arose from the combination, and that is the signature of two variants at one locus. That is the reasoning on which this register places the Igor determinant at PAX3.

The class of evidence, named so that it is not mistaken for something else. This rests on lineage association evidence: a documented mating, the phenotype of the offspring, and the mechanism the literature attaches to that phenotype. It is not a laboratory result and this survey does not record it as one. No animal of the line has been sequenced and no record on it carries the grade genotyped, so every cat here is graded on descent. The locus is established on this evidence; the allele is not, and the two are different questions.

What would overturn it. A PAX3 read on an Igor-line animal returning no variant in that gene. A hearing test on either daughter showing normal hearing. A further cross of the same two lines producing unaffected offspring. The principal alternative, which the pedigree alone cannot exclude, is an interaction between one PAX3 variant and a white spotting allele at KIT.

Igor and Nadeya are not one variant

The Nadeya line carries DBE^NAD, and all 39 dominant blue eyed cats of that line are heterozygous for it. Three dominant blue eyed cats of the Igor line were genotyped for the same variant and all three are wild-type. The two lines do not carry one variant. Abitbol, Eiselé and Lipovenko 2026, Table 1.

The two traces do share ancestors, and shared ancestry is not shared variant. Counted over the recorded traces, 159 animals stand above both the Igor and the Nadeya descent, 470 above both Igor and the ALT descent, and 156 above both Nadeya and ALT. 144 stand above all three alike. Not one of them carries an eye code or a genotype. These are the ordinary foundation animals of the British Shorthair, and a determinant that arose in 2016 in a cat born in Russia cannot have come down from them.

The pedigree could not have shown this, and the register holds the reason. A cross between two lines tests whether they carry one variant only when both parents carry a determinant. Asked how many such crosses it holds, the register answers:

A cross is only a test if both parents carry the determinant. In a dominant trait the carrier is the animal that shows it. All four candidate matings, and all eight of their parents, are recorded with the eye unknown. None of them tests anything.

No mating in this register tests anything between these two lines, and none was needed. Four genotypes answered what four generations of pedigree could not, and that is the argument of this whole case study in one line.

Marusya carries a variant of her own

The Marusya line carries DBE^MARU, PAX3 XM_019838731.3:c.784C>T, a nonsense variant in the fifth exon that truncates the protein and removes 46 per cent of it. The founder is heterozygous, all fourteen dominant blue eyed cats of the line tested are heterozygous, and the allele is absent from the 99 Lives dataset of 348 cats. The Igor line is wild-type for it.

Forty records are held under the Marusya label here, three of which carry any parent and thirty of which carry no EMS code at all. This register holds no pedigree for the line, so within it there is nothing to cross and nothing to trace, and the line shares no recorded ancestor with Igor, with Nadeya or with the ALT descent. What settled the line was a laboratory, not a pedigree, which is the point this case study keeps arriving at.

For the lines that still have no variant, a third kind of comparison is available and the survey is building it. Phenotype variant association compares photographs of animals of lines whose determinant is unidentified, to see whether they group together on their markings and on the proportions of the face. The rationale is that PAX3 acts on the neural crest, the embryonic cell population that gives rise both to the pigment cells of the iris and to the bone and cartilage of the face, so a determinant at that locus can show itself in pigment and in facial form at once.

It is called an association because that is its limit. It can say that two populations resemble one another. It cannot say which variant a cat carries, or whether a cat is a carrier at all, and no variant has been attributed to any animal on it. The method is set out on the method page.

A label is not a pedigree

The register holds 411 animals under the Igor line label. It holds 61 that descend from the founder, Igor Azure Dream, by unbroken parent links. The rest are there because a breeder chart grouped them, and a chart is a statement of belief about relatedness rather than a record of parentage.

Chart membership is not descent. The register keeps the two in separate fields for that reason, and every figure in this case study is computed from parent links and from nothing else.

The same gap appears on the two other lines examined here.

How the three lines connect

Shared ancestry is a statement about documents. It says two populations have animals in common in their recorded pedigrees, and it says nothing about which variant either carries. Every figure here is counted over the same three groups, each taken as the cats that descend from its founder by unbroken parent links together with the cats the register records the variant for, so the rows can be read against one another.

The three are one population, not three. 133 cats stand in the recorded ancestry of all three of them. Nadeya is no more separate from DBE^ALT than Igor is: 145 against 470, on groups of very different size.

Igor and Nadeya are collateral and not lineal. No Nadeya animal descends from Igor Azure Dream and no Igor animal descends from Ermine Trace Nadeya. They share 147 ancestors, which places both inside one breeding population of British chinchillas rather than two isolated ones.

The same register holds the case in its cleanest form, one line over. Oliver, the founder of the Nanotigr line, and Ermine Trace Nadeya share 55 recorded ancestors, among them one widely-used sire who stands above both. Nanotigr carries DBE^ALT and the Ermine Trace line carries DBE^NAD, and neither allele appears in the other line on any test. Fifty-five ancestors in common, and the shared ancestry is the route for neither variant. It says only that both are bred from the same British chinchilla stock. Two genotypes prove here what this survey elsewhere can only warn about.

One variant of the pair now has a name, and that changes what the shared ancestry is worth. While both determinants were unidentified, two lines meeting in a pedigree could always have been one thing under two labels. They are not. The Nadeya animals carry DBE^NAD and the Igor animals do not carry it, so where the two populations meet, two different determinants stand in one pedigree, and that is the arrangement that produces a compound heterozygote.

They meet below her and not above her, and the two halves say different things. The founding tree of the Ermine Trace line, 179 animals counting everything above Nadeya and above Oliver of the Nanotigr line together, contains not one record carried under an Igor label. Nothing of that material is ancestral to her, which closes the only route by which DBE^NAD could have arrived from the Igor line.

Below her it is the other way round. Of the 39 cats that descend from her by unbroken parent links, 37 also have an ancestor carried here under the Igor label, and every one of those paths runs through both of two animals, Leonardo De Cassis and Jaki Ami Sissy Crazy Blue. That is a compound question, not an origin question. It is this register's reading of the pedigrees and not a finding of any paper: the published work reports genotypes, and says nothing about what stands above a DBE^NAD cat.

What those two animals are is exactly the question, and the register cannot answer it. Both are recorded from a PawPeds chart with no eye colour and no variant, and both carry the Igor label because of the company they keep on that chart rather than because anything is recorded about their eyes. A label is not a determinant. If the Igor determinant came down through them, 28 cats stand below two variants. If it did not, they are two British Shorthairs in a shared pedigree and nothing follows. One genotype on either animal, or on any of their other descendants, decides it, and none has been run.

Marusya is the exception and it is an exception about records rather than about breeding. Three of its forty entries carry a parent, so the line has four recorded ancestors in total and can share nothing with anything. Its case is made in the photographs and nowhere else.

DBE^CEL and DBE^RE will not fit in three circles, so every line is set out together below, with the ancestors each pair shares on the line between them.

Where this line could collide with another variant, and where it already has

A second variant is in this line when it traces into it. That means an animal standing below both founders, or an animal of the line for which the register records the other variant. Shared ancestry is not that: it says two lines come out of the same breeding population, which is true of most British chinchillas, and it says nothing about an allele passing down.

Read the first column and not the last. One variant has traced into this line and it is DBE^ALT. The Igor line shares 139 ancestors with DBE^CEL and 470 with DBE^ALT, and those two figures say almost nothing: only one of the two has actually reached the line.

How the lines are counted. A cat belongs to a line if it descends from that line's founder by unbroken parent links, or if the register records that variant for the cat itself. A cat mated into a line is not added to it: only the kittens are, because only they can have inherited. The middle column is a forward-looking measure and not a trace: it counts the catteries holding cats of both, which is where a mating could still be made. Shared ancestors, in the last column, is counted upward and through mates, so it stands for common breed stock. The same count applied to every pair of variants is on the variant trace page.

Every figure in this case study is computed from recorded parent links in the register at build time and verified against it before publication. The 2022 mating is documented on three stamped pedigree certificates held by this survey. No variant is adopted for any animal of the three lines named here, and where a source attributes one it is held in a field separate from an adopted variant so that no computation can treat a claim as a result.