Segregation in F1, F2 and F3 – Trait Combinations in Early Breeding Generations

A cross yields a uniform first filial generation only if both parents are true-breeding for the traits in question, meaning they carry the same variant twice at the place in the genome that decides the trait. In cannabis they practically never are: the species pollinates across individuals rather than itself, so it carries two different variants at very many places, and what is sold as hybrid usually comes from crossing two such plants. Inconsistent offspring are therefore not a fault of the particular batch but the inevitable consequence of the starting material. Changing that takes roughly six generations of inbreeding, and the rest of this article is the arithmetic behind that claim.

What uniformity in the first generation depends on

The expectation comes from the Mendelian rules, the law of uniformity and the law of segregation: cross two true-breeding lines that differ in one trait, and the first filial generation comes out uniform, with segregation appearing only in the second. The statement holds, but it carries a condition, and the condition sits right at the front: both parents have to be true-breeding.

In maize that can be arranged. Male and female flowers sit on the same plant there, so the plant can pollinate itself; several such generations produce inbred lines that differ from one another yet are each true-breeding in themselves. Because they carry the same variant twice at the places concerned, called loci, the cross between two particular lines always turns out the same. That reproducibility is the real prize, not the vigour of the offspring.

Cannabis is predominantly dioecious: male and female flowers usually sit on separate plants. Usually is not always. Male flowers appear spontaneously on female plants, and where they do, the plant pollinates itself without any intervention. What cannot be done that way is planning it: which plant flips, and when, is not the breeder’s decision. Selfing a particular plant therefore takes an intervention, and what that looks like is covered further down.

How far a line can sit from that even after work has gone into it shows in the classic chemotype study’s own material. It used lines that had already been inbred twice and scored them at markers, places in the genome that can be read unambiguously in the laboratory. Even so, between 5.6 and 37.5 % of those places were still mixed, depending on the combination, and segregated in the first filial generation.

For a commercial variety no such figure exists. True breeding cannot be seen on a plant; it is measured at markers or tested on the progeny. A variety looking the same for years is neither.

The three-quarters rule and its reach

Three to one describes exactly one trait with two versions of a variant in play: one that prevails and one that stays masked by it. As soon as several such places are involved the fractions multiply: with two traits 56 % of plants still show the wanted phenotype, with three 42 %, with ten 5.6 %. Fixing a trait counts a different figure: only a quarter of plants carry the wanted variant twice at a given place, so 1.6 % across three traits.

That rate decides not the odds but the number of plants. To be 99 % sure of finding at least one suitable plant across three target traits takes around 290 seedlings. Combining two properties cleanly is therefore a job for an ordinary breeding room. Combining five at once fails on floor space, not on knowledge.

The chemotype: one trait at a single place

One trait behaves differently, and the counter-case is instructive. Whether a plant makes mostly THC or mostly CBD is decided at a single place in the genome. Every plant holds two of that place, one from the father and one from the mother, and each carries either the variant for THC or the one for CBD. That gives three kinds of plant: twice THC yields pure THC, twice CBD yields pure CBD, and one of each yields a plant that makes both. Neither variant masks the other.

That is why the outcome of a cross can be worked out in advance here. Cross two plants that make both, and each parent passes on one of its two variants at random: 25 % of the offspring make pure THC, 50 % both, 25 % pure CBD. Across ten second-generation families of 35 to 118 plants, exactly that ratio came out, written 1:2:1 in the literature.

This is the exception rather than the rule. Aroma and yield do not hang on one place but on many at once, and for those there are neither three kinds nor a share known in advance. The practical difference: for the chemotype a breeder knows before sowing how many seeds to start in order to get a particular plant. For aroma, he grows them, checks each one and keeps the ones that fit.

From a single plant to a true-breeding line

Raising a next generation from one selected plant works only through that plant’s own pollen. The controlled way there is induced sex reversion: female plants form fertile male flowers after treatment with silver thiosulfate, male plants form fertile female flowers after treatment with ethephon.

What that generation is called depends on what the self-pollinated plant was. If it was the F1 from a cross between two parents, its progeny is the F2, the second filial generation of that cross. If it was an individual plant of any other kind, the selfing rounds are counted instead: S1 after the first, S2 after the second, and so on. The classic chemotype study ran both series and needed the same sex reversion for both.

For a true-breeding line, though, selfing is not the route but the faster of two. The other is sib-mating, the pairing of two plants from the same progeny; it needs male plants and therefore regular seed, and it reaches the goal more slowly. The slower pace has a flip side the time calculation leaves out: at equal levels of inbreeding, sib-mating clears out the variants that are lethal in double copy more effectively, 7.11 remaining variants in the model against 9.3 under selfing. The model describes natural populations; what transfers is the mechanism, not the number. On top of that, every round of selfing needs the sex reversion and therefore selects for the capacity to perform it, which raises the risk of unwanted male flowers in the finished line.

A single round of selfing then delivers the opposite of what is expected. Compared directly against two crosses onto a different plant, the selfed progeny reached about half the height and half the leaf area, flowered seven days later, produced 63 % less floral dry weight and spread two to three times as far as the comparison group in total and floral mass. A round of selfing unifies nothing, it exposes what was previously masked.

Uniformity arrives only at the end of the stretch. Single-seed descent, in which one individual plant is carried forward per generation, takes about six generations to push the share of mixed places below 3 %, starting from 17 to 35 % in open-pollinated source plants. The price is in the same study: abnormal flowers rose from 0.4 % in the source generation to 13.7 % in the sixth, more than thirtyfold. Only crossing two such lines delivers the consistent offspring an F1 is supposed to provide, and even then just two of five hybrid lines stayed within 20 % variation across all measured compounds. That completes the arithmetic behind the opening claim: uniformity is reachable, but it is a six-generation project rather than a property a good cross brings with it.

Backcrossing and the neighbourhood of the target variant

Backcrossing is treated as the shortcut when a single trait needs to move into an existing variety. On paper, one round restores 75 % of the recurrent parent’s genome, a second 87.5 % and a third 93.8 %; near-complete recovery takes five rounds or more. Two conditions are almost always left unsaid: that recurrent parent should itself be a true-breeding line, and the trait being moved should rest on one or a few places. The first condition leads straight back to the opening claim.

What does not disappear along with the rest is the immediate neighbourhood of the target variant. Because every round selects the plants carrying that one variant, a piece of the donor genome stays attached to it: the linkage drag. In a simulation on oilseed rape that piece measured 15.9 million base pairs after the first backcross and 5.4 million after the second. After three rounds followed by selfing, 98 % of the remaining donor genome came from that single piece. Where a variant sits therefore matters more to the outcome than how many generations are run.

That leaves the special case nobody plans for: offspring outside the range of both parents. They do not arise preferentially from maximally different parents. Among six rice crosses, only the one between two nearly identically flowering varieties produced a wide spread, a range of 18 days from a parental difference of 0.7 days. None of this has been examined in cannabis, but it works as an expectation: the unwelcome outliers and the welcome ones come from the same mechanism, and both appear where they are least suspected.