Many studies have demonstrated that the distribution of meiotic crossover events along chromosomes is non-random in plants and other species with sexual reproduction. Large differences in recombination frequencies appear at several scales. On a large scale, regions of high and low rates of crossover have been found to alternate along the chromosomes in all plant species studied. High crossover rates have been reported to be correlated with several chromosome features (e.g. gene density and distance to the centromeres). However, most of these correlations cannot be extended to all plant species. Only a few plant species have been studied on a finer scale. Hotspots of meiotic recombination (i.e. DNA fragments of a few kilobases in length with a higher rate of recombination than the surrounding DNA) have been identified in maize and rice. Most of these hotspots are intragenic. In Arabidopsis thaliana, we have identified several DNA fragments (less than 5 kb in size) with genetic recombination rates at least 5 times higher than the whole-chromosome average [4.6 cM (centimorgan)/Mb], which are therefore probable hotspots for meiotic recombination. Most crossover breakpoints lie in intergenic or non-coding regions. Major efforts should be devoted to characterizing meiotic recombination at the molecular level, which should help to clarify the role of this process in genome evolution.
- Arabidopsis thaliana
Meiotic recombination occurs during the prophase of the first meiotic division. It produces crossovers and gene conversion (or non-crossovers), creating genetic variation. Crossovers also have a mechanical role. Together with cohesion between sister chromatids, they provide physical links between homologous chromosomes, visualized as chiasmata in cytology and ensuring correct segregation during the first meiotic division. The essential dual role of crossovers is highlighted by the observation that no sexually reproducing species studied to date displays an absence of chiasmata in both sexes . Moreover, mutations reducing crossover frequency increase the frequency of chromosome non-disjunction (reviewed in ).
Hotspots of meiotic recombination have been defined in yeasts (Saccharomyces cerevisiae and S. pombe) as small (1–2 kb) DNA fragments centred around meiotic DSBs (DNA double-strand breaks) that are repaired using the homologous chromosome to produce crossovers and non-crossovers. Evidence for meiotic recombination hotspots in higher eukaryotes has been obtained from the study of male meiosis by sperm typing in mice and humans: clusters of crossovers and non-crossovers can be detected in regions less than 10 kb in size, with a Poisson distribution consistent with the current model for meiotic DSB repair (reviewed in [3,4]).
In S. cerevisiae, meiotic recombination hotspots are not evenly distributed along the chromosomes, tending instead to be clustered in chromosomal domains, called ‘hot domains or regions’ away from the centromeres and telomeres [5–7]. In all organisms studied (including plants), crossovers are not evenly distributed along the chromosomes, suggesting that their locations are tightly controlled. The mechanisms controlling the pattern of crossovers have been of great interest for decades. Correlations have been found between recombination rates and chromosome structure (chromosome size, arm size, distance from the centromere or the telomere etc.) or sequence features (GC content, CpG, simple repeats, transposable elements etc.) in various organisms, but most of these associations are weak, with the strength of the correlation varying greatly from one organism to another [8–10].
We review here what is known about ‘hot’ and ‘cold’ recombination domains in plants, together with the identification and characterization of meiotic hotspots.
Crossover distribution on chromosomes
Several striking features of plant chromosomes have often been described: high rates of recombination in gene-dense regions and next to the telomeres and a lack of recombination in centromeric regions. This lack of recombination close to the centromeres has been observed in all species, but the extent to which this lack of recombination extends to adjacent regions is variable (a few tens of megabases in large Triticeae genomes and a few megabases in Arabidopsis and rice).
In wheat, maize and barley, recombination increases with relative distance from the centromeres [11–14]. Similar results have been obtained in rice [15,16] and tomato, but recombination tends to decrease just next to the telomeres in these species . We recently obtained a very detailed genetic map of chromosome 4 of Arabidopsis thaliana, which showed no centromere-to-telomere gradient of recombination  (Figure 1A). Thus distal recombination cannot be considered to be general among plants.
The high rates of crossovers found in gene-dense regions (reviewed in ) are not completely independent of telomeric location. In some species, such as wheat and maize, genes are not evenly scattered along the chromosomes and instead tend to cluster in small regions that are themselves clustered in the distal regions of chromosomes. The nature of the relationship between gene density and recombination remains elusive. In maize, the distribution of genes accounts for 50% of the variation in recombination rates . In wheat, it has been suggested that the correlation between distance to the centromere and recombination rate may account for gene distribution . Moreover, two studies have shown that if the distal, gene-rich part of a wheat chromosome is deleted, the rate of recombination in the new terminal segment is much higher than that in the corresponding segment of the complete chromosome arm [21,22]. This suggests that the rate of recombination is determined more by the location of the DNA fragment than by its sequence. We have also shown that, in Arabidopsis, rates of recombination on chromosome 4 are not correlated with the distribution of genes, pseudogenes, transposable elements or dispersed repeats . Thus the localization of recombination in gene-rich regions is not general to all plants.
Why have conflicting results been obtained concerning the location of crossovers in plants? The main reasons are that our knowledge of the structure of plant genomes is still only partial and, in most cases, the experimental assays used to generate genetic maps were not designed to appreciate variations in recombination rates. The number of chiasmata per chromosome is regulated by two main rules: the obligatory chiasma per chromosome to ensure correct segregation and, in most species, the strength of interference. In addition, recombination between highly repeated sequences may be repressed to protect the genome against gross rearrangements. Organisms may have found many different ways to follow these rules. Only two plant genomes have been entirely sequenced: Oryza sativa (rice)  and A. thaliana . These species have much more compact genome structures than most cultivated plants (120 Mb for Arabidopsis and 389 Mb for rice versus around 2365 Mb for maize and 15000 Mb for wheat). Large regions (more than 100 Mb) almost devoid of genes are common in wheat and other large genomes, whereas gene-poor regions do not exceed a few megabases in Arabidopsis or rice . Transposable elements account for 15–35% of the genome in Arabidopsis and rice, but more than 60% of the genome in maize . Polyploidy may also impose additional constraints on the distribution of crossovers. Moreover, in many plant species, crossover rates differ for male and female meiosis, even in the same plant . Rules may therefore vary within a given species, and the regulation of recombination may involve factors other than sequence motifs and chromosome features.
Genetic maps have been obtained for plant species by a combination of classical genetic linkage and cytogenetic methods. Most linkage studies were carried out for selection purpose or QTL (quantitative trait locus) mapping. The markers used were therefore not initially intended for studies of genetic recombination itself. As a result, they were often too few in number or inappropriately spaced for a reliable assessment of variation in recombination. Alternative methods have involved cytogenetic analyses based on RN (recombination nodule) maps, chiasma counting or C-band maps. However, these techniques are labour-intensive and of limited precision.
Thus the main limitations of analyses of crossover distribution are the precision of genetic maps and the availability of known genome structures.
Hotspots of recombination in plants
In plants, there is strong evidence for the existence of hotspots of meiotic recombination. In barley, wheat, rice and maize, the rate of meiotic crossovers has been reported to vary more than 10-fold across regions of a few hundreds of kilobases [27–33]. Moreover, in rice and maize, regions only a few kilobases in size have been found to contain large numbers of crossovers. The most convincing results were obtained for the a1-sh2 and bronze regions of maize [31,33–35].
Across the 140 kb of the a1-sh2 region, genetic recombination varied between 0 and 11 cM (centimorgan)/Mb, with a region average of 0.0087 cM/Mb (0.5% of the genome average) . Unfortunately, experimental studies based on the reassociation of markers flanking the 140 kb region did not detect non-crossover events that would reinforce the suggestion that these DNA fragments are meiotic hotspots.
At the 1.5 kb bronze locus, the frequency of crossovers was shown to be at least 100 times higher than the average rate for the sh-bz-wx region. An analysis of recombination between various heteroalleles of the bronze gene clearly demonstrated the presence of non-crossovers and crossovers at this locus [34,36].
One gene, waxy, has been found to contain a high concentration of crossovers in two species: maize and rice [30,33]. The structure of waxy (in terms of introns and exons) and its surrounding chromosomal region is well conserved in both species, although this segment has opposite orientations with respect to the centromere in these two species [36a]. As studies in mammals have demonstrated that meiotic recombination hotspots are not generally conserved, it would be of great interest to assess the variations found in plants.
Most of the experimental systems used, except a1-sh2 in maize, were designed to screen for intragenic (coding region) recombination. Thus a strong correlation between a high level of genetic recombination and localization in a coding region was expected. In a1-sh2, 90% of crossovers are concentrated in three non-adjacent DNA fragments 1.7, 2.2 and 3.4 kb in size. Crossover breakpoints were located in coding regions in two of these fragments, whereas crossovers occurred in non-coding DNA in the third . Thus, as in mammals, recombination is not necessarily intragenic in plants.
In A. thaliana, we precisely mapped 140 crossovers in the first 800 kb of the short arm of chromosome 4  (Figure 1B). We observed huge variations in crossover rate, with some regions entirely devoid of crossovers (0 cM/Mb) and others containing large numbers of crossovers concentrated in a few kilobases (>100 cM/Mb). This patchy distribution of small DNA fragments with very high rates of genetic recombination is very similar to the distribution of hotspots reported for yeast, mice and humans. An example is shown in Figure 1(C). All the crossover breakpoints of five hotspots were located in non-coding regions, whereas one-third of the crossovers were in coding regions for the other two hotspots. Thus, in an experimental system not based on strong intragenic selection, meiotic recombination is not necessarily associated with coding regions.
Non-crossover events in plants
Non-crossovers have been characterized molecularly for only one gene: the bronze locus (see above). The ratio of non-crossover to crossover events varied considerably, depending on the type of heteroalleles combined, but in most of the crosses tested, an excess of crossovers over non-crossovers has been reported . The conversion tract length has been determined for only two events, and was shown to be approx. 1 kb . Additional measurements will be required for precise estimation of average tract length compared with that in humans (<300 bp) .
There is a lack of information on non-crossovers in plants due to difficulties in observing these effects in the absence of phenotypic selection. However, several lines of evidence indirectly suggest that non-crossovers should be present in large excess in plants, as previously observed in mammals .
Electron microscopy of meiotic chromosome spreads led to the observation of RNs as proteinaceous ellipsoids approx. 100 nm in diameter appearing early in prophase I of meiosis; the molecular events of recombination are thought to occur in these structures [37,38]. Two types of RN, ENs (early nodules) and LNs (late nodules), have been identified, differing in shape, size, relative numbers, protein components and order of appearance–disappearance (leptotene to pachytene stages for ENs, and pachytene to diplotene stages for LNs). In plants, as in other organisms, the number of LNs is strongly correlated with the number of chiasmata, and these two elements display interference [37–40]. ENs do not show interference. About half of all ENs contain DSB repair proteins (Rad51/Dmc1). The timing of their appearance, their distribution and their protein content suggest that ENs play a role in searching for DNA homology, synapsis, gene conversion and/or crossover. In tomato, onion and maize, ENs are present in excess over LNs by a factor of up to 20 [37–40].
In Arabidopsis and maize, the number of Rad51 foci has been determined by immunolocalization during prophase of the first meiotic division. The number of foci reaches a peak at mid-zygotene (500 in maize and 150–250 in Arabidopsis) and then decreases at mid-pachytene, when only a few foci (10–50) persist [41–43]. The observed number of Rad51 foci and ENs is much higher than that required for crossover formation: 7–10 in Arabidopsis and 17–23 in maize [44,45]. It is tempting to suggest that the excess of Rad51 foci and at least some of the ENs may mark the sites of non-crossovers, which should thus be present in a 10–20 times excess over crossovers.
Further indirect evidence for the existence of non-crossovers comes from analysis of the pattern of DNA polymorphisms in 96 accessions of A. thaliana [46,47]. Models developed to infer genetic recombination rates from population genetic data generally fit better if non-crossovers are taken into account.
The molecular characterization of recombination events in plants is still in its infancy. However, an understanding of recombination patterns may affect the choice of markers for association studies or modelling the evolution of a region at a fine scale. The release of increasing numbers of plant genome sequences and the development of effective genotyping techniques will facilitate the study of other model plants, including polyploids. These studies will generate information essential to our understanding of meiotic recombination at a mechanistic level.
This review benefited from the reflections of the ‘Méiose et Recombinaison’ group at INRA (Versailles cedex, France). I thank Sandrine Bonhomme, Jan Drouaud and Mathilde Grelon (all at Institut Jean-Pierre Bourgin) for a critical reading of this paper before submission. Special thanks also go to Eric Jenczewski (Institut Jean-Pierre Bourgin) for challenging discussions.
Meiosis and the Causes and Consequences of Recombination: Biochemical Society Focused Meeting held at University of Warwick, U.K., 29–31 March 2006. Organized by R. Borts (Leicester, U.K.), D. Charlesworth (Edinburgh, U.K.), A. Eyre-Walker (Sussex, U.K.), A. Goldman (Sheffield, U.K.), G. McVean (Oxford, U.K.), D. Monckton (Glasgow, U.K.), G. Moore (John Innes Centre, U.K.), J. Richards (Roslin Biocentre, U.K.) and M. Stark (Glasgow, U.K.). Edited by D. Monckton.
Abbreviations: DSB, DNA double-strand break; EN, early nodule; LN, late nodule; RN, recombination nodule
- © 2006 The Biochemical Society