W hat can go wrong after an egg is fertilized by sperm from a different species? One of the most intriguing outcomes to the chromosome biologist is uniparental genome elimination. In this phenomenon, one of the parental chromosome sets is completely lost during embryonic cell divisions, creating sterile haploid offspring (1). Uniparental genome elimination occurs in diverse taxa, including several distantly related plant species and even in fish (1, 2). The mechanism by which chromosomes from one parent are selectively discarded has been studied cytologically in crosses between different grass species. However, these experiments have not revealed what is defective about the eliminated genome in a hybrid. In PNAS, Sanei et al. (3) make a major breakthrough by showing that missegregated chromosomes in a classic barley interspecies cross (Hordeum vulgare × Hordeum bulbosum) fail to assemble kinetochores, the microtubule attachment sites that mediate chromosome inheritance. The chromosomal location of the kinetochore, termed the centromere, is marked by incorporation of a histone H3 variant named CENH3. CENH3-containing nucleosomes are found at centromeres in all eukaryotes and are essential for recruiting other kinetochore proteins. Sanei et al. (3) used an antibody raised against rice CENH3 to stain embryos created by pollinating cultivated barley (H. vulgare) with its wild relative H. bulbosum—the antibody recognizes CENH3 from both parents. A fascinating observation about this cross is that elimination of the paternal H. bulbosum chromosomes is temperature-dependent. At low temperature, CENH3 was found at the predicted foci on all chromosomes, which segregated accurately. In embryos created by fertilization at higher temperature, chromosomes that lagged behind during mitosis and were eventually discarded showed a lack of CENH3 antibody staining. Fluorescence in situ hybridization with labeled genomic DNA (combined with a wealth of previous genetic data) indicates that lagging chromosomes without functional kinetochores were derived from the H. bulbosum parent. Loss of CENH3 from H. bulbosum chromosomes neatly explains their failure to segregate faithfully during embryonic mitosis but also raises important questions about how centromere identity and consequently kinetochore location are preserved when chromosomes are copied. Elimination of the H. bulbosum genome in unstable hybrids is gradual, taking place over several days after pollination (4). When chromosomes are replicated during S phase of the cell cycle, histones including CENH3 are distributed between the two sister chromatids (5). These observations suggest that H. bulbosum chromosomes enter the zygote with a normal complement of CENH3, which is gradually depleted by several rounds of DNA replication until the kinetochore is no longer able to function. Therefore, reloading of CENH3 after DNA replication is specifically defective in H. bulbosumderived chromosomes, although it is normal in chromosomes from the H. vulgare parent. What could cause CENH3 loading to fail in one of the two parental genomes? A possible clue is that H. bulbosumderived chromosomes in hybrids appear less condensed than their H. vulgare counterparts. This might indicate that replication of H. bulbosum chromosomes is poorly synchronized with the hybrid cell cycle, a factor that could explain why CENH3 reloading does not occur. If asynchronous replication of parental chromosome sets is responsible for a CENH3 loading defect, it would indicate that the time window for replenishment is quite narrow (Fig. 1). Increasing the ploidy of the H. bulbosum parent can prevent genome elimination, further suggesting that subtle differences in cell cycle timing may affect whether CENH3 reloading is effective in hybrids (6). Cell cycle asynchrony between H. vulgare and H. bulbosum chromosome replication may be testable through kinetic analysis with cytological markers in the barley interspecies cross. A wild card in the interpretation of these results is the possibility that histones are reprogrammed on a large scale during reproductive development. A substantial fraction of Arabidopsis thaliana CENH3 is unloaded after fertilization and apparently resynthesized from the zygotic genome (7). Whether such a process takes place in the grasses is unknown, but zygote-specific CENH3 dynamics could have a major influence on the segregation of parental genomes in interspecies hybrids. Cell cycle differences are not the only reason why centromere assembly could differ between H. vulgare and H. bulbosum chromosomes. However, any explanation must account for the fact that H. bulbosum chromosomes load CENH3 normally in selfed plants yet are defective when mixed with H. vulgare chromosomes (even this depends on the environmental factor of high temperature). An alternative hypothesis is that H. vulgare CENH3 cannot be loaded into H. bulbosum centromeres in unstable hybrids and has a dominantnegative effect on H. bulbosum CENH3 loading in this situation. The lack of evidence for obligate interactions between CENH3 and centromere DNA argues against this model; like conventional nucleosomes, CENH3 nucleosomes seem to package DNA with relatively little sequence specificity. First, Sanei et al. show that barley expresses two CENH3 variants, both of which can coexist at either H. vulgare centromeres or at H. bulbosum Fig 1. Cell cycle asynchrony could explain why H. bulbosum chromosomes fail to reload the centromere-specific histone CENH3 in an interspecies hybrid. In this model, DNA replication is a chromosome-intrinsic property, so H. bulbosum chromosomes proceed through S phase more slowly than H. vulgare chromosomes. This delay means that H. bulbosum chromosomes miss a narrow time window for CENH3 replenishment and show less condensation. The cell cycle might proceed more slowly at low temperature, possibly explaining why H. vulgare × H. bulbosum hybrids are stable when cultivated below 18 °C.