DISCUSSION

The frequency of ENs is higher on SC segments compared to axial element segments: The frequency of ENs associated with SCs ranged from 0.38 to 1.61 ENs/μm of SC for the six plant species (Table 1). These numbers are comparable to those reported by Stack and Anderson (1986b) who observed 1.41 ENs/μm of zygotene SC from L. esculentum and by Albini and Jones (1987) who observed 0.53 and 0.78 ENs/μm of zygotene SC from A. cepa and A. fistulosum, respectively. For the species investigated here, the number of ENs per micrometer of SC was 2.5–7.5 times higher than the number of ENs per micrometer of AE (Table 1). Albini and Jones (1987) also found fewer ENs associated with aligned AE than with SC for A. fistulosum (0.39 and 0.78, respectively) but approximately the same mean number of ENs associated with both aligned AE and SC for A. cepa (0.50 and 0.53, respectively). We were unable to consistently distinguish between aligned and unaligned AEs for the species we examined, and this may be one reason why our value for the number of ENs per micrometer of AE for A. cepa is so much lower than that reported by Albini and Jones (1987).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Histograms showing the distribution of distances between adjacent ENs on SC segments for six different plant species. Only internodule distances between ENs on the same SC segment are included. Although the distributions are continuous, they have been presented in 0.1-μm increments. Using parameters estimated from the raw data, a predicted normal curve (thin line) and a predicted gamma curve (thick line) have been superimposed over each distribution. In each case, the predicted gamma curve best fits the observed distributions.

What could account for the lower frequency of ENs on AEs compared to SC segments for each species? One possibility is that the frequency is artificially low due to using middle-to-late zygotene bivalents in which a significant proportion of the remaining asynapsed AEs may be located in heterochromatic regions that typically synapse late and have few ENs (e.g., Stack and Anderson 1986a; Stacket al. 1993). This explanation likely accounts for some of the reduction because when we analyzed zygotene SCs that had 50% or less of their length synapsed, and therefore a larger proportion of axial elements in euchromatin, the frequency of ENs on AEs went up for only two species, L. esculentum (0.22 to 0.27) and P. nudum (0.14 to 0.19). However, these frequencies were still at least five times lower on AEs compared to SCs. Another possibility is that ENs associated with AEs are less stable during the spreading procedure compared to ENs associated with SCs. To our knowledge, no data that we could use to address this question are available from nondisrupted, serially sectioned nuclei from plants or animals. However, we can test this possibility by comparing our EN frequencies to Rad51p foci frequencies in squashes of L. longiflorum nuclei (Terasawaet al. 1995). Terasawa et al. (1995) found maximal numbers of foci during leptotene (~2000 foci per leptotene nucleus—our calculated number derived from their observations). Assuming that these nuclei have no SC [see comments by Franklin et al. (1999) concerning the problem of accurately staging leptotene and zygotene in the absence of an antibody probe for SC components] and assuming 3000 μm of SC for a complete set at pachytene (Andersonet al. 1985), then there must be at least twice this length of axial element (2 × 3000 μm = 6000 μm) per leptotene primary microsporocyte nucleus and ~0.33 foci/μm of AE (2000 foci/6000 μm AE). In comparison, we found a frequency of 0.58 ENs/μm AE at zygotene for L. longiflorum. Considering the assumptions required to estimate the Rad51p foci frequencies, the probability that two closely adjacent ENs would appear as one fluorescent focus (Freireet al. 1998), and the possibility that as many as half of the ENs may not be labeled with antibodies to Rad51p (Andersonet al. 1997), these numbers agree reasonably well. This correspondence makes it unlikely that a large proportion of ENs on AEs are being differentially lost during the spreading procedure. Thus, we conclude that the higher frequencies of ENs on SCs compared to AEs probably reflect the situation in vivo.

SC segment length is an excellent predictor of EN number: We observed a strong linear relationship (r² = 0.81–0.95) between EN number and SC segment length for each species (Figure 2 and Table 2), although there were statistically significant differences between the slopes of the regression lines for the different species. Other investigators have also noted a positive linear relationship between EN number and SC length (human, Rasmussen and Holm 1978; mouse, Moenset al. 1997; two Allium species, Albini and Jones 1987; Sordaria, Zickleret al. 1992). However, a linear relationship between EN number and SC length does not necessarily mean that ENs are evenly spaced on SCs (see below).

The frequency of ENs per unit length of zygotene SC varies between plant species: We observed up to a fourfold difference in the frequency of ENs per micrometer of SC segment among these six plant species (Tables 1 and 2), and these differences are reflected in the average distance between adjacent ENs on SC segments (Table 4). There is no obvious pattern of EN frequency with regard to phylogenetic class, genome size, or chromosome number. For example, species with higher frequencies of ENs on SCs (>1 EN/μm SC) include the lower vascular plant P. nudum (2n ≃ 200 and a large genome; S. M. Stack, unpublished observations), the dicot L. esculentum (2n = 24; 4C = 4 pg), and two monocots, L. longiflorum (2n = 24; 4C = 141 pg) and T. edwardsiana (2n = 12; 4C = ~120 pg). The two species with lower frequencies of ENs on SCs (<1 EN/μm SC) include the monocot A. cepa (2n = 16; 4C = 67 pg) and the dicot C. betacea (2n = 24; 4C = ~30 pg; DNA contents for all species except P. nudum are from Bennettet al. 2000). Albini and Jones (1987) also found a difference in EN frequency between two closely related Allium species. We do not know the basis or the possible significance of the differences in EN frequencies on zygotene SCs among plant species. However, it is possible that factors we did not examine such as genotype, ambient temperature, and/or age of plant may have some influence on the observed number of ENs associated with SCs in plants. Almost all of these factors have been shown to affect crossing over in plants (e.g., Griffing and Langridge 1963; Jones 1967; Maguire 1968; Rees and Dale 1974; and see Sherman and Stack 1995 for other references).

Plant ENs attach to AE and forming SC, but not to formed SC: Because ENs are identifiable as discrete structures only by electron microscopy, we often refer to them as if they were preformed complexes that attach to AEs and SC. However, it is more likely that ENs are progressively assembled, probably at sites of DNA double-strand breaks (DSBs). DSBs are required to form Rad51p foci (and probably ENs; Gasioret al. 1998; Romanienko and Camerini-Otero 2000), and a number of studies using immunolocalization of two or more proteins have indicated gradual changes in the protein composition of foci during zygotene and pachytene (e.g., Ashleyet al. 1995; Ashley and Plug 1998; Gasioret al. 1998; Pluget al. 1998; Agarwal and Roeder 2000; Moenset al. 2000; Santucci-Darmaninet al. 2000). However, to simplify the discussion and to reflect what we observed microscopically, we refer to binding or attachment of ENs to AEs and SCs with the understanding that ENs are probably assembling at each site.

There are three possible times/locations at which ENs could associate with zygotene bivalents during synapsis: (1) with AEs prior to synapsis, (2) with SC after synapsis, and/or (3) at synaptic forks during synapsis. In regard to the first possibility, we and others (e.g., Albini and Jones 1987) have observed ENs associated with single AEs as well as at convergences of two axial elements (although not all convergence sites have an EN). Convergence sites do not include SC, although SC formation may follow. AE convergences in plants are typically 3–9 μm apart, which represents a frequency of ~0.11–0.33 convergences/μm of SC (see references in Anderson and Stack 1988). This frequency range is similar to the range we observed for ENs associated with AE (0.06–0.58 ENs/μm AE), but lower than the frequency range of ENs associated with SCs (0.38–1.61 ENs/μm SC) for all six plant species (with the possible exception of A. cepa; Table 1). Therefore, for most if not all of these plant species, additional ENs must bind to SCs to account for the higher frequencies of ENs associated with SCs.

Do ENs bind to already formed SC during zygotene? Rasmussen and Holm (1978) thought so because, in thin sections from human primary spermatocytes, they observed fewer ENs per micrometer of SC at early zygotene compared to late zygotene. However, we found no change in the number of ENs per micrometer of euchromatic SC in L. esculentum from early to late zygotene (Figure 4), and our results from the other five species were consistent with this conclusion. Similarly, Albini and Jones (1987) also examined the frequency of ENs at different zygotene substages and found either no difference (A. fistulosum) or fewer ENs per micrometer of SC at later stages of zygotene (A. cepa). These results indicate that in plants ENs do not continue to bind to formed SC segments.

An alternative explanation for the lower frequency on ENs on AEs compared to SCs is that ENs bind to SCs at synaptic forks. Here we define a synaptic fork as the interface between synapsis (tripartite SC) and asynapsis (two separated axial elements). Rasmussen and Holm (1978) reported ENs at synaptic forks in human spermatocytes, and we observed that up to 50% of synaptic forks from six plant species have an EN at the fork (Table 3). In addition, for each plant species, the frequency of ENs at forks is higher than along AEs (by a factor of ~7–77) and also higher than along SC segments (by a factor of ~2–12). These results indicate that ENs bind to SCs at synaptic forks. However, if ENs assemble at forks without otherwise interfering with the progress of synapsis, one would expect the frequency of ENs at forks to be the same as the frequency of ENs along SC in general. Since this is not the case, synapsis may be delayed temporarily at forks while ENs are assembled. Then it is likely that synapsis would resume and progress until delayed again as another EN assembles, and so on. In this scenario, ENs that bind at synaptic forks would not be necessary to nucleate SC formation. However, this interpretation does not preclude the possibility that ENs at convergence sites are important for synaptic initiation.

The distribution of distances between adjacent ENs on SC segments is random: The continuous distribution of distances between adjacent ENs for each plant species is presented in histogram form in Figure 4. Each distribution is skewed to the left, and the extent of the shift is related to the different frequencies of ENs per micrometer of SC observed for the different plant species (Tables 1 and 4). While most adjacent ENs are <1 μm apart (except for A. cepa), a few adjacent ENs are separated by 3–18 μm. The longer distances between adjacent ENs may be due to SC in heterochromatic regions that have few ENs.

Two curves have been fitted to each of the observed distributions in Figure 4, using parameters derived from the data. One curve is a normal distribution based on the mean and standard deviation, and the other curve is a gamma distribution based on α and β parameters. Both normal and gamma distributions are random distributions for continuous variables, but a gamma distribution differs from a normal distribution in that there can be only positive numbers. This means that normal and gamma distributions will look quite different when means are close to zero, but as means become larger, gamma distributions will look more and more like normal distributions. This is similar to the behavior of Poisson distributions that are used for discrete variables. The constraint of only positive values for gamma distributions is expected when measuring distances between two adjacent ENs since this measurement will always yield a positive number. In addition, there is another constraint on the measurement of internodule distances because the minimum measurable distance between two nodules is limited to the center-to-center distance between two touching ENs. This physical limitation in measuring the distance between adjacent nodules accounts for the scarcity of measured distances ≤0.1 μm since only a few nodules are smaller than 0.1 μm in size and very few of those are touching (Figure 4). For each species, the predicted gamma distribution fits the data better than the predicted normal distribution. This result is supported by Kolmogorov-Smirnov one-sample goodness-of-fit statistical tests that show that the distribution of distances between ENs in euchromatin is random and best fits a gamma distribution for each plant species. Although we occasionally observed several adjacent ENs (up to five in one case) that appeared to be evenly spaced, we were unable to detect a recurrent separation distance (or multiple of a recurrent distance) between adjacent ENs in any of the six plant species. Therefore, we conclude that in plants, ENs are randomly spaced along SC segments with respect to one another.

The general appearance of the internodule distance distributions for each plant is similar to those of Coprinus, Bombyx, and humans, as shown by Rasmussen and Holm (1978), Holm and Rasmussen (1980, 1983), and Holm et al. (1981). These researchers used a computer model to generate an expected random distribution for distances between adjacent ENs. In each case, the observed and expected distributions were similar. The researchers concluded that the observed distributions were random for Bombyx spermatocytes (Holm and Rasmussen 1980), nearly random for C. cinereus (Holmet al. 1981), and were not random for human spermatocytes (Rasmussen and Holm 1978). However, no statistical tests were used to assess the fit between the observed and expected distributions. Because of the similarity between the distributions these researchers observed and those we observed for plants, we compared their data [taken from Rasmussen and Holm (1978, Figure 43), Holm and Rasmussen (1980, Figure 59), and Holm et al. (1981, Figure 20)] to predicted gamma distributions on the basis of parameters estimated from their data and found that they were not significantly different (Kolmogorov-Smirnov one-sample goodness-of-fit test, P < 0.05). Therefore, we conclude that distances between ENs in humans and Coprinus as well as Bombyx are most likely random.

Relation of ENs to LNs: Observations that ENs are randomly distributed in euchromatin do not necessarily imply that crossovers are randomly distributed. Indeed, hot spots for recombination are well documented in a number of different organisms (Lichten and Goldman 1995). However, we do not know the exact relationship of ENs to LNs that lie at crossover sites. The available evidence suggests that some ENs are converted into LNs (e.g., Pluget al. 1998; Zickler and Kleckner 1999; Agarwal and Roeder 2000; Moenset al. 2000; Santucci-Darmaninet al. 2000). However, it is still not clear whether the large numbers of ENs observed in plants indicate a comparable number of DSBs (Franklinet al. 1999), whether all ENs are initially the same and then have different fates depending on their locations and/or time of association with the SC (perhaps related to the mysterious force called interference; e.g., Rasmussen and Holm 1980; Stack and Anderson 1986b), or whether there are initially different types of ENs that look morphologically similar but have different molecular components and different activities (e.g., Dresseret al. 1997; Pâques and Haber 1999; Allers and Lichten 2001; see discussion by Zalevskyet al. 1999).

Model for the distribution of early nodules: Our results suggest the following model for the frequency and distribution of ENs in plants. Many (perhaps all) ENs associated with AEs serve to link axial elements together at convergence sites where SC formation begins. Assuming that these ENs remain after SCs form, they represent only a minority (10–40%) of the ENs that ultimately will be associated with SCs. The majority (60–90%) of ENs associated with SC are added at synaptic forks. The higher frequency of ENs at synaptic forks is most likely due to a delay in the progress of synapsis while ENs assemble. As synapsis progresses, ENs bind randomly at forks over time so that the distribution of spacing between ENs along SCs is described by a gamma distribution. Thus, beginning at synaptic initiation sites, ENs along a segment of SC represent consecutive events in time. In the context of this model, we suggest two different possibilities for EN composition and behavior: There are two types of ENs, one that initially associates with AEs and another that associates with synaptic forks of SCs. This difference in behavior may be based on a difference in composition and function; e.g., the former might function in synaptic initiation while the latter might function in recombination. Alternatively, all ENs begin with the same basic molecular composition, but ENs progressively change over time. This means that ENs along a segment of SC would represent a developmental hierarchy with the more mature nodules lying closer to synaptic initiation sites and the less mature ENs lying closer to synaptic forks. A more mature EN is more likely to achieve a crossover and become a late nodule than a less mature EN. Once a crossover has been achieved, other nearby ENs would not be able to achieve a crossover due to interference, and these ENs would be lost from the SC en mass (Stack and Anderson 1986b). Using antibodies to different molecular components of nodules, it should be possible to test which, if either, of these two alternatives describes the situation in plants.