Osd1 Mutants Produce Diploid Gametes by Skipping the Second Meiotic Division
As a part of an expression-profiling screen for meiotic genes using the Expression Angler tool 
with the AtGenExpress tissue set 
was selected as a good candidate due to its co-regulation with several known meiotic genes 
corresponds to the UVI4-Like
) which was briefly described in a study of its paralogue, the UVI4
. Due to its role in meiosis (see below), we renamed the At3g57860
for omission of second division
. The OSD1 and UVI4 proteins are conserved throughout the plant kingdom () but do not contain any obvious conserved known functional domains. No homologues were identified outside of the plant kingdom. We investigated the role of the OSD1
gene by isolating and characterising two mutants. The osd1-1
and the osd1-2
Ds insertional mutants are in the Nooseen (No-0) and Landsberg (Ler) backgrounds, respectively, and in both cases the insertion is in the second exon of the OSD1
gene (). In both independent osd1
mutants, the products of male meiosis were dyads (osd1-1
: 714/714; osd1-2
: 334/334) instead of tetrads (). Complementation tests between osd1-1
confirmed that these mutations are allelic (osd1-1/osd1-
2: 369 dyads/369). osd1
mutants did not show any somatic developmental defects, male and female gametophyte lethality, or reduced fertility (wild type 38±11 seeds/fruit; osd1
35±6). Next, we measured ploidy levels among the offspring of diploid osd1
mutants. Among selfed progeny, tetraploids (4n) (84%) and triploids (3n) (16%), but no diploid (2n) plants, were found (osd1-1
24). When mutant pollen was used to fertilise a wild-type plant, all the resulting progeny were triploid (osd1-1
75). When mutant ovules were fertilised with wild-type pollen grains, we isolated 12% diploid and 88% triploid plants (n
25). This demonstrated that the osd1
mutants produce high levels of male (100%) and female (~85%) diploid spores, which result in functional gametes.
To unravel the mechanisms leading to dyad production in osd1, we investigated chromosome behaviour during meiosis. Both male and female meiosis I were indistinguishable from wild type (compare with ). Notably, chiasmata, the cytological manifestation of crossovers, and bivalents were observed. However, we were unable to find any meiosis II figures (among >500 male meiocytes from prophase to spore formation), strongly suggesting that dyad production is due to an absence of the second meiotic division. If this second division does not take place, then any heterozygosity at centromeres will be lost in the diploid gametes because of sister chromatid co-segregation and homologue separation during the first division. Because of recombination, any loci that are not linked to centromeres will segregate.
We tested our assumption by taking advantage of the two different genetic backgrounds of the osd1-1
(No-0) and osd1-2
mutants (Ler). F1 plants bearing the two mutations—mutant for osd1
and heterozygous for any No-0/Ler polymorphisms—were crossed as male or female to a third genetic background, Columbia (Col-0). Karyotyping and genotyping of the obtained plants for trimorphic molecular markers provided direct information on the genetic make-up of pollen grains and female gametophytes produced by the mutant. All the diploid gametes tested had the predicted genetic characteristics (). They were systematically homozygous at centromeres and segregating—because of recombination—at other loci (n
48 for male diploid gametes and n
41 for female diploid gametes). These results confirmed that the absence of a second meiotic division is indeed the cause of 2n gametes production in osd1.
This mechanism also implies that unbalanced chromosome segregation at meiosis I would give rise to unbalanced dyads in osd1
; this was confirmed by analysing a double Atspo11-1/osd1-1
mutant (unpublished data). Such a phenotype has been already described in the maize elongate mutant in which diploid female gametes are produced because of failure to undergo meiosis II, but the corresponding gene has not been identified 
Genetic make-up of the osd1 and MiMe diploid gametes.
Interestingly, the OSD1
is necessary for maintaining the mitotic state, and loss of UVI4 function stimulates endo-reduplication, whereas OSD1
is not required for this process 
. Meiosis is not affected when UVI4
is disrupted (unpublished data). This suggests that UVI4 and OSD1 are both involved in cell cycle regulation, with specialized functions in mitosis and meiosis, respectively. The transition from meiosis I to meiosis II requires a balance in Cyclin–Cdk activity: it must be lowered sufficiently to exit meiosis I but maintained at high enough levels to suppress DNA replication and promote entry into meiosis II 
. Two protein depletions generate a phenotype similar to that of osd1
: the fission yeast Mes1 protein partially inhibits cyclin degradation by the anaphase promoting complex (APC) and thereby allows entry into meiosis II 
. Similarly, the expression of Erp1/Emi2 at the end of meiosis I is essential for entry into meiosis II in Xenopus 
and mouse 
oocytes, most likely by inhibiting cyclin degradation by APC. Erp1/Emi2-depleted oocytes and the mes1
yeast mutant fail to enter meiosis II, which is reminiscent of the osd1
phenotype. Therefore, one possible function of OSD1 may also be to fine tune APC activity/cyclin levels at the end of meiosis I to ensure the transition to meiosis II.
Turning Meiosis into Mitosis
Due to an absence of the second meiotic division, osd1
mutants produce high frequencies of viable diploid male and female gametophytes, which generate, after fecundation, viable tetraploid plants. However, this phenomenon differs from apomeiosis in that the produced gametes are genetically different from the mother plant. Previously, we reported that in double Atspo11-1/Atrec8
mutants, the first meiotic division is replaced by a mitotic-like division, followed by an unbalanced second division that leads to unbalanced spores and sterility 
. Triple osd1/Atrec8/Atspo11-1
mutants were generated and expressed an apomeiosis phenotype in which meiosis was completely replaced by a mitotic-like division. This was expected, because the Atspo11-1
mutations lead to a mitotic-like first meiotic division, and the osd1
mutation prevents the second meiotic division from taking place. We called this genotype MiMe
for mitosis instead of meiosis. MiMe
plants generate dyads (408/408) and are fertile (25±6 seeds per fruit). The osd1
mutation therefore suppressed the sterility phenotype of the Atspo11-1/Atrec8
double mutant. Observation of chromosome behaviour during male and female meiosis revealed a mitotic-like division: ten univalents aligned on the metaphase plate and sister chromatids separated at anaphase (). The selfed progeny of MiMe
plants were systematically tetraploid (n
24), and backcrosses between diploid MiMe
plants and wild-type plants generated triploid plants regardless of whether male (n
24) or female (n
gametes were used, showing that this mitotic-like division gives rise to functional diploid gametes. All the gametes (male and female), tested similarly as described above systematically retained the mother plant heterozygosity for every genetic marker tested () and were thus genetically identical to the mother plant. These results confirm that MiMe
plants undergo a mitotic-like division instead of a normal meiotic division, without affecting subsequent sexual processes.
Mitosis-like divisions instead of meiosis in MiMe plants.
When meiosis is replaced by mitosis, ploidy is expected to double with each generation. Indeed, in successive generations, we obtained tetraploid (4n, 20 chromosomes, n
26) and octoploid (8n, 40 chromosomes, n
33) ( and ). Fertility dropped from 25±6 seeds/fruit in 2n plants and 19±4 in 4n plants to <0.1 in 8n plants. Further investigations will be required to understand the cause of this reduced fertility associated with high ploidy level.
Doubling of ploidy at each generation in the MiMe line.
Male meiosis in 2n, 4n, and 8n MiMe plants.
Apomixis can be separated into three developmental components: an absence or alteration of meiosis which prevents reduction (apomeiosis), the fertilization-independent development of the embryo from the egg cell (parthenogenesis), and the initiation of endosperm development with or without fertilization 
. Here we showed that fully penetrant apomeiosis can be induced in a sexual plant, when a mitotic-like division replaces meiosis in the MiMe
genotype. A previous case of apomeiosis was reported in the Arabidopsis dyad
. However, only 0.2% of dyad
ovules generate viable gametes, which makes it practically unusable in an apomixis engineering strategy, in contrast to MiMe
plants, which produce quasi–wild type levels of viable—and apomeiotic—ovules and pollen grains. The three genes conferring the MiMe
genotype are strongly conserved among plants, suggesting that apomeiosis may be engineered in any plant species, including crops. In addition, several other genes required for recombination initiation have been described in plants 
, and the corresponding mutants could be used instead of spo11-1
to construct a MiMe
genotype. However, it remains to be confirmed that the MiMe
genotype would have the same phenotype in other plant species. To obtain apomixis, in addition to apomeiosis, parthenogenesis will have to be introduced, and the problem of endosperm formation must also be overcome. However, mutations that mimic early parthenogenesis or give rise to functional autonomous endosperm have been reported in Arabidopsis 
, suggesting that it should be ultimately feasible to introduce apomixis into a sexual plant species.