Finding the crosswalks on DNA

Any busy street is organized into clearly defined places where one can cross safely from one side to the other and places where crossing is discouraged. On many a busy meiotic chromosome, there appear to be similar sorts of “crosswalks” where crossing over occurs frequently (Fig. 1). We are still learning what defines these places, and why they occur where they do. Two recent papers in PNAS, from the labs of Hugo Dooner (1) and Pat Schnable (ref. 2, in this issue) address the matter further and point out that there is still much to learn.

Figure 1.

A crosswalk. Homologous DNA is shown as red helices and crossover event as the solid red line.

Early on, recombination was thought to occur outside genes, because genes were the indivisible smallest units of heredity. However, this idea also created a puzzle. Organisms that looked similar (and probably had similar numbers of genes) could have visible differences in chromosome size and clear differences in genome size but similar genetic map lengths.

The idea of the indivisible gene proved to be incorrect, of course (3), and intragenic recombination is commonplace. The puzzle, which came to be known as the “C-value paradox,” made more sense if most recombination, in fact, were to occur within or near genes (4). But why recombination occurs at one place and not another, or more often at one place than another, has remained a puzzle (reviewed in ref. 5). Both the current papers (1, 2) examine this idea as it applies to maize, where genes are generally separated from one another by long tracts of nested retrotransposons (6).

Work from the Dooner lab (1) has shown that the maize bronze1 (bz1) locus has some of the highest recombination per kilobase known in plants (7). Unlike much of the maize genome, sequencing reveals that bz1 is in a region that does not contain retrotransposons between its genes.

Earlier this year in PNAS, Fu et al. (1) reported on still more of the DNA around bz1. They compared recombination rates across the gene-dense area to one side of bz1 with those in the region of retroelement DNA located on the other side. Recombination per kilobase between bz1 and three sites in other genes of the gene-dense region was as much as 80-fold higher than the average for the genome as a whole and was nearly as high as rates within the bz1 gene itself. The genome average is calculated by dividing published estimates of the total genetic map length by published estimates of maize genome size, genes, retroelements and all.

In contrast, a marker separated from bz1 by the adjacent gene-poor retroelement-rich DNA shows recombination with bz1 similar to the genome average. These experiments were done in a homozygous inbred background, so differences in recombination can be attributed to differences in genome organization on either side of bz1 and not to sequence differences. The results support the idea that recombination is rare in the retroelement fraction of the maize genome, implying that recombination occurs within or near genes.

Yao et al., in the Schnable lab (2), take a different approach, using the anthocyaninless1 (a1)-shrunken2 (sh2) region of maize, which has the more “typical” gene arrangement of a few genes interspersed with retroelement DNA. They examine recombination in a strain with heterozygous sequence polymorphisms across the entire test region. The strategy is to use these polymorphisms to map where recombination junctions occur, then ask what the genome looks like where the junctions map.

In addition to a1 (involved in anthocyanin biosynthesis) and sh2 (starch biosynthesis), they can align additional sequences between their polymorphic chromosomes. Two of these are genes they call x1 and yz1. They also find an “interloop region” nestled between a retroelement called Machiavelli and a small group of nested repetitive elements, among them one called Gnat and the other Ozymandias. The interloop region is on both homologues, but the retroelement and “Gnat-Ozymandias group” are on only one.

As might be predicted, recombination junctions map within a1 and yz1 more frequently than average and within the retroelement and the nested repetitive elements at or below average. The rest of the region provides some interesting surprises, however.

The x1 gene is clearly expressed (it is a single-copy gene of unknown function found among expressed sequence tags), yet crossovers resolve within x1 10 times less often than within a1 or yz1. Interestingly, the number of recombination junctions that resolve in x1 is still 30–80 times higher than in the nongenic DNA on either side of it. These nongenic sequences themselves show 10-fold fewer junctions than the genome average. The x1 gene might therefore be what passes for a local hotspot, albeit in a region where crossover junctions are impaired overall.

This coordinated decrease in recombination for both x1 and adjacent DNA suggests multiple levels of control over where recombination events take place in maize. The more common 30- to 80-fold difference between adjacent sequences is overlaid with a second, 10-fold regional change (e.g., in the distal part of the a1–sh2 interval). It is unclear what causes the decrease, although similar discrepancies are seen in other systems (8). In yeast, sequences several kilobases away from a meiotic double-strand break site can influence its behavior, possibly by competing for limiting factors (9, 10), but such an explanation is unlikely to apply in maize because such a large area is affected.

Unlike x1 and yz1, the “interloop region” shows no signs that it contains genes or parts of genes, yet recombination events are resolved there even more frequently than within the a1 gene. Certainly, it is harder to prove that a sequence is not a gene, largely because the conclusion relies on negative results. However, the recombination data are compelling and suggest the rules for crossover sites in plants will not be as straightforward as that they occur only within or near genes.

The “interloop region,” at least in these experiments, represents 2.2 kb of pairing sequence set between several kilobases on either side that have more difficulty pairing. Could the transition from poorly paired to well-paired sequence signal a recombination site? In support of this idea, introducing even a small number of sequence heterologies dramatically decreases the frequency of recombinant junctions that form in a particular sequence (7). Heterology could either inhibit branch migration or at least discourage resolution, defining where recombination points will not be rather than where they are. The logical extension of that idea would require that typical intergenic DNA in maize does not pair quite as precisely at meiosis as genic DNA. Clearly, lines homozygous for tracts of retroelement sequence do not suddenly begin to show crossover resolution and genetic map expansion within those regions. However, it is possible that there is some low level of mispairing among intergenic sequences that is not found in genic sequences. It will be interesting to see whether the “interloop region” described by Schnable and coworkers (2) continues to accumulate recombinant junctions in lines that have separated Machiavelli from the nested elements (the very recombinants described in the paper).

Both the studies reported here look only at recombinant junctions. They do not distinguish where recombination events finish from where they start. In eukaryotes, recombination events are thought to start with a double-strand break and resolve by passing through a double Holliday structure (11). One of the recombination junctions represents an estimate of the initial double-strand break, whereas the other represents the point at which branch migration stops and heteroduplex formation ends. Even when both sequence junctions can be identified, it is difficult to tell which is which because there are no gene conversion gradients as guides. A recombination hotspot in yeast is a site where initiation occurs more frequently than the average. In plants, they are defined only as places where a recombinant junction forms more frequently than the average. The criteria may not prove to be the same for both.

Factors that determine where or how often recombination occurs remain unclear in plants. The need for relatively open chromatin to allow recombination proteins access to the DNA is likely to play an extremely important role (9, 12–14). Another appealing feature of chromatin as a control mechanism is that, once local differences in recombination are established for adjacent sequences, e.g., x1 and its flanking DNA, these proportional differences can be maintained even as recombination across whole regions is increased or decreased further. The chromatin of the “interloop region” described by Yao et al. (2) and recombinant derivatives that alter it will be very interesting in this regard. Unfortunately, even though there are clear examples of position effect on gene expression in plants, there are no demonstrated examples of position effect variegation yet. It is reasonable to suspect that, as it does in Drosophila (15), heterochromatin spreading could affect recombination as well as gene expression.

With regard to specific factors, plant homologues have been identified for proteins like Spo11, the yeast endonuclease that initiates recombination by causing a double-strand break (16, 17), and those in the Rad50/Mre11 complex involved in, among other things, chromatin remodeling around sites where Spo11 will cut (18). Homologues of additional recombination proteins, such as Rad51, Dmc1, and others, are also known in plants. There are already indications that some homologues carry out the same functions they do in other species, but that others do not. Determining what all these homologues, and any new factors that are found, interact with should be especially informative in understanding how recombination sites are selected (19).

Maize also provides one of the best direct looks at recombination: microscopy of recombination nodules (RNs). RNs, first reported in fungi, are particulate (20) protein bodies ≈100 μm in size that can be visualized in the electron microscope and are now associated with synaptonemal complexes in a number of plants, animals, and fungi.

Two types of these nodules, early and late, have been associated with the synaptonemal complex (21, 22). Early nodules form at sites where recombination is about to occur, tend to be found at synaptic forks as chromosomes are pairing, and are proposed to be involved in synapsis and/or early events of recombination (23). It is quite reasonable to suppose that early nodules will also be involved in somehow sensing where recombination is or is not to take place. In contrast, late nodules are found at sites where crossing over has just taken place and seem more likely to be associated with the events that conclude recombination. Late nodules also become involved in specifying where crossovers occur through the process of crossover interference. Maize late nodules display interference at much greater distances to one another than do early nodules, consistent with the high crossover interference observed in this plant. The means by which interference is accomplished remain unknown. Although the lore of the field has leaned toward the idea that interference occurs through preventing initiation of a second crossover, it is quite possible that the mechanism allows crossovers to initiate and then prevents their resolution (24). Whether early nodules contain proteins like Spo11p, late nodules contain proteins implicated in interference like Ndj1p of yeast (24); whether early and late nodules share components or derive from one another remains unclear but should be quite testable.

Specific recombination sequences like χ-sites in bacteria or M26 in Schizosaccharomyces pombe (see refs. 25, 26) have not been identified in plants. Instead, more loosely defined sites have been described. Xu et al. report a 377-bp fragment 5′ of the maize a1 gene where 20% of the recombination events they analyzed were resolved (27). Similarly, recombinant junctions were identified more often at the 5′ end of the b1 gene (28) and at the 3′ end of the r1 gene (29). Interestingly, studies at the bz1 and wx1 loci find that recombination junctions occur more evenly throughout these genes (7, 30), so the situation is likely to vary. That some genes appear to have these junction hotspots while others do not has also been described in other organisms, including humans (31).

In maize, transposons such as Mutator and Ac/Ds show a strong tendency to insert into the same regions that tend to show higher recombination levels (32, 33). It is not clear whether there is a mechanistic connection between the two phenomena, although it would come as little surprise. Open chromatin, tendencies to form DNA breaks, and proximity of repair and replication proteins would all make recombination-prone sites a tempting target for transposition. The case is easier to make for Mutator, which tends to insert around meiosis (34), than for Ac, which is less specific about timing its reinsertion. However, features that mark recombination sites may persist even in somatic cells. For example, transposons may tend to go where transposition complex proteins have access to DNA as the element tries to acquire a target. From the standpoint of the transposon, features that mark DNA as a recombination site might be more consistent than those that mark a gene for expression.

Recent papers have suggested important new revisions to the double-strand break model for how recombination takes place in eukaryotes. In particular, an idea gaining increasing support is that crossover and noncrossover products of recombination arise from two mechanisms that start out the same but then diverge halfway through (35, 36). It will be especially interesting in the next few years to see how the mechanics of recombination relate to the positioning of recombination events along the chromosome. It may turn out that how one crosses the chromosomal street will also influence where one is best able to cross it.