The maize (Zea mays) desynaptic (dy) mutation defines a pathway for meiotic chromosome segregation, linking nuclear morphology, telomere distribution and synapsis

Meiosis is a specialized mechanism by which sexually reproducing species reduce their genomes from diploid to haploid, and the underlying mechanisms are highly conserved among eukaryotes (John, 1990). In general, the process of meiosis is under genetic control and consists of commitment and initiation, homologous chromosome pairing and synapsis, interhomolog reciprocal recombination, disjunctive segregation, and haploid gamete or gametophyte formation. Interhomolog chromosome recombination between parental chromosomes during meiotic prophase I establishes physical connections for bipolar spindle attachment between the homologs, and it generates novel allelic combinations. The study of meiosis in plants has made major contributions to our understanding of this key step in eukaryotic life cycles (for a review, see Murphy and Bass., 2012). Telomeres, the structures at the ends of eukaryotic chromosomes, consist of conserved repetitive DNA sequences and associated proteins that regulate genome stability and facilitate meiotic chromosome segregation. During meiotic prophase I, the nucleus is reorganized dramatically – the chromosomes become compact, and their telomeres attach themselves to the surface of the inner nuclear membrane of the nuclear envelope (NE), cluster into a bouquet arrangement, and finally disperse just before NE breakdown at late prophase I but remain attached to the inner nuclear membrane. The formation and dynamics of the bouquet configuration of meiotic chromosomes is thought to contribute to proper homologous chromosome pairing, synapsis, recombination, interlock resolution, and ultimately segregation (for reviews, see Bass, 2003; Harper et al., 2004; Scherthan, 2007; Scherthan, 2009).

In maize, meiotic telomere clustering occurs de novo on the NE during meiotic prophase I, and the spatial and temporal patterns are nearly identical to those in mammals (Bass et al., 1997; Harper et al., 2004; Scherthan, 2009; Scherthan et al., 1996). Telomere-interacting proteins have been identified in fungi and animals and are required for telomere bouquet formation and meiotic chromosome segregation (Chikashige and Hiraoka, 2001; Conrad et al., 2007; Cooper et al., 1998; Ding et al., 2007; Schmitt et al., 2007; Trelles-Sticken et al., 2000). More recently, highly conserved Sad1/Unc-84 (SUN) and KASH (Klarsicht, ANC-1, Syne Homology) domain proteins have been shown to reside at the NE and form a functional bridge required for nuclear migration and anchorage during development and used for meiotic chromosome movements during prophase I. (For reviews, see Hiraoka and Dernburg, 2009; Starr, 2009; Starr and Fridolfsson, 2010). Proteins with homology to the SUN proteins have recently been described for several plant species (Graumann et al., 2010; Moriguchi et al., 2005; Murphy et al., 2010; Oda and Fukuda, 2011; Van Damme et al., 2004; Zhou et al., 2012), but their function in meiosis remains largely uncharacterized.

Among the model experimental systems for analysis of meiosis, maize is an excellent example and provides one of the best sources of synchronized meiotic cells (Chang and Neuffer, 1989). The meiotic process in higher plants is fairly long (Bennett et al., 1977) and therefore allows for the collection of multiple developmental time points during meiotic prophase. The ability to observe meiotic chromosomes and nuclear organization within intact nuclei by three-dimensional (3D) optical reconstruction fluorescence microscopy has proven extremely useful in cytological analyses of meiotic mutants. Over 50 meiotic mutants in a variety of maize lines have been identified, many of them by means of forward genetic screens (for reviews, see Cande et al., 2009; Golubovskaya et al., 2003). Maize mutants have been classified as those that affect a variety of meiotic events, including meiocyte differentiation, entry into and exit from the meiotic program, sister chromatid cohesion, chromosome condensation, telomere bouquet formation, synapsis, recombination, or those that specifically perturb the meiotic cytoskeleton.

The maize desynaptic (dy) mutant was initially identified by cytological and genetic approaches as a recessive mutation is manifested in several meiotic defects particularly with respect to chromosome behavior (Nelson and Clary, 1952). The chromosomes in dy appear normal at the pachytene stage of male meiotic prophase, but many of the bivalents separate into univalents by late prophase. In previous studies, the chromosomes at the pachytene stage of prophase I appear fully synapsed at first, but they exhibit extensive precocious desynapsis and chiasma-maintenance failure (Maguire et al., 1991; Maguire et al., 1993). Others have shown dy to affect the control of meiotic recombination (Ji et al., 1999). 3D telomere FISH analysis of dy revealed that the mutation resulted in premature detachment of the telomeres from the NE at pachytene. This defect precedes and probably contributes to the production of chromosome laggards, mis-segregation, and eventual pollen semi-sterility (Bass et al., 2003).

Given the multiple meiotic defects attributed to the dy mutation, we expected that further investigation would provide insight into the question of whether a SUN-KASH-type complex, known to operate in opisthokont meiosis, is conserved in plants. To examine this question, we undertook a forward genetics approach along with 3D telomere FISH and immunocytochemistry to characterize dy. Here we present evidence that implicates dy in connecting the integrity of the NE to telomere behavior and synapsis. Genetic linkage mapping provided an opportunity to examine candidate genes, and information regarding one compelling candidate, ZmSUN3, is presented.

To begin to investigate the molecular basis of the function of the wild-type Desynaptic (Dy) gene, we carried out 3D cytological phenotypic analyses using telomere FISH, ASY1 immunolocalization, and nuclear shape analyses. This, together with a forward genetic approach, should help to identify the genetic lesion underlying the dy mutant phenotypes. First described in 1952, dy is a mutant characterized by a partial male semi-sterility phenotype in which bivalent desynapsis leads to univalents by late prophase. Subsequently, others have shown dy lines of maize have a defect in chiasma maintenance (Maguire et al., 1991), recombination (Ji et al., 1999), and are prone to telomere-NE interaction defects at mid-prophase (Bass et al., 2003). Given these prior observations, we reasoned that a detailed 3D cytological analysis of dy would provide insights into the complex mechanisms that link telomere dynamics to chromosome behavior during meiosis, especially at mid-prophase when synapsis and recombination go to completion.

The effect of the dy mutation on telomere localization at the bouquet stage (early meiotic prophase) is shown in (Fig. 1). The normal telomere clustering in a wild-type maize zygotene-stage nucleus (Fig. 1A–D) shows that more than 90% of the telomere FISH signals are at the nuclear periphery. In a comparable stage in a dy/dy nucleus, only ∼60–70% of the telomeres are at the nuclear periphery (Fig. 1E–H). These values were typical of the other dy/dy nuclei. Nonperipheral telomeres, are those that are located at least 1 µm from the nuclear periphery, are indicated here as by arrows (Fig. 1F) or yellow signal (Fig. 1H). These examples are consistent with our summary observations, which were based on visual inspection (n = at least 150 cells), 3D data collection (n = at least 30 cells) and 3D modeling (n = at least 10 cells) of dy nuclei at the bouquet stage. In the case of dy, the bouquet stage is defined by telomere clustering, thin fiber appearance, and eccentric nucleolus. Our finding of a partial bouquet represents the earliest known defect reported for maize dy, occurring just before the previously reported telomere-misplacement phenotype of precocious telomere-NE detachment at pachytene (Bass et al., 2003).

The plant protein ASY1 is localized to the nucleus and associates with chromosome axes at early meiotic prophase (Armstrong et al., 2002; Caryl et al., 2000; Mikhailova et al., 2006). Anti-ASY1 antisera raised against Arabidopsis ASY1 have been shown to cross react with those of other plant species, including Brassica (Armstrong et al., 2002), rye (Mikhailova et al., 2006) and maize (Golubovskaya et al., 2006; Golubovskaya et al., 2011; Pawlowski et al., 2009; Wang et al., 2009). In maize, ASY1 antiserum produces continuous staining along the length of the meiotic chromosome fibers during early prophase, but the signal becomes undetectable by pachytene, once synapsis is complete (Golubovskaya et al., 2006). The results of the use of this ASY1 antiserum at early and middle prophase are shown in Fig. 2. Distribution of ASY1 along the chromosome fiber axes was normal at zygotene for both wild-type and dy nuclei (compare Fig. 2, B and H), but the patterns at midprophase differed markedly. Wild-type nuclei were unstained at pachytene (Fig. 2E), but ASY1 staining persisted during pachytene in dy, indicating incomplete, partial, or disrupted progression of synapsis as shown for a representative nucleus (Fig. 2K). The summary observations presented here are based on visual inspection (n = at least 200 cells) and 3D data collection and analysis (n = at least 50 cells) of nuclei at the pachytene stage of meiosis.

Many pachytene-stage nuclei derived from dy lines of maize were severely warped and misshapen in appearance. Fig. 3 shows through-focus projections of DAPI-stained pachytene-stage nuclei (n = 15) from two wild-type lines (A344+, Fig. 3A–C, and KYS, Fig. 3D–F) and dy/dy (Fig. 3G–U). This figure also shows periphery tracings made on a 3-section (0.9 µm) projection from the middle region of the data (grey-filled objects) used for shape analyses. A one-way analysis of variance in conjunction with a Tukey’s multiple comparison test revealed that calculated means ± s.d. for the average degrees of circularity for A344 and KYS wild-type nuclei (0.89±0.01 and 0.88±0.01, respectively) different significantly from that of mutant dy pachytene stage nuclei (0.74±0.06) (Fig. 3V). A second measure of nuclear shape, the average aspect ratios of best-fit ellipses, for wild-type nuclei [1.16±0.11 (A344) and 1.11±0.06 (KYS)] differed than those for dy pachytene stage nuclei (1.43±0.29 m; Fig. 3W). This information represents the first report of nuclear deformation for a maize meiotic mutant.

The dy mutation is known to produce univalents during meiotic prophase, which give rise to micronuclei resulting from chromosome segregation (Nelson and Clary, 1952). At the later stages of meiosis, our observations revealed two phenotypes (laggards and micronuclei) known to result from the univalents produced by the dy mutation (Nelson and Clary, 1952), in addition to anaphase chromosome bridges, all of which are shown in Fig. 4. The anaphase bridges (Fig. 4F) presumably reflect failure of homologous chromosomes to produce normal bivalents and result from unresolved interlocks, recombined dicentric chromosomes, or missegregation of multicentric multivalent figures. The micronuclei may represent intact univalents or chromosome fragments or a combination of the two. Although our results cannot distinguish among these possibilities, the chromosome structures that do not end up at the spindle pole remain as unitary discrete micronuclei in the cytoplasm (Fig. 4H–J, arrows). Of all the phenotypes we have now associated with dy – partial bouquet, pachytene telomere internalization, distorted midprophase nuclear shape, irregular ASY1 staining and chromosome segregation defects detected after meiotic prophase – we find that the last is the simplest and clearest phenotype to obtain, requiring only general chromatin staining techniques and single optical section imaging. For subsequent phenotyping assays associated with linkage mapping of dy, we chose the univalent/micronucleus detection assay, as shown in Fig. 4.

Several desynaptic mutants have been described in maize, but only one, dsy2, has been localized to genetic bin 5.04 by B-A translocation and SSR linkage mapping (Franklin et al., 2003). Our similar forward genetic approach, employing bulked segregant analysis (BSA) and B-A translocation mapping (Beckett, 1978; Michelmore et al., 1991) is summarized in Fig. 5. The BSA mapping scheme begins with a cross between dy as the male parent and a wild-type female (+/+, KYS). KYS was chosen because it lends itself well to cytology at midprophase. Self-fertilized F1 plants produced a segregating F2 population. F2 plants were used as leaf source for DNA preparations, and pollen viability assays were attempted but found to be unsatisfactory for unambiguous distinction of mutant (dy/dy) from nonmutant (dy/+, +/dy or +/+) plants. Instead, we selfed all F2 plants and scored F3 progeny using the univalent/micronucleus-detection assay to determine which F2 plant DNAs were to be pooled for mutants and which for nonmutants. BSA mapping revealed co-segregation of markers on the long arm of chromosome 3. Additional SSR markers (supplementary material Table S1) were obtained in the area of linkage, and co-segregation analysis was carried out on DNA from individual F2 mutants (n = 22) to narrow down the mapping region. In addition, B-A translocation mapping was carried out with stocks that uncover mutants on chromosome 3L (Tb-3La and Tb-3La-2s6270). From progeny analysis of TB-3LA × dy, we observed that 7 of 13 progeny from one cross and 3 of 6 progeny from another were cytologically like dy. From progeny analysis of TB-3La-2S × dy, we observed that 6 of 15 progeny were cytologically like dy. Collectively, these observations agree with the expected detection of 50% mutant frequency from B-A translocation test crosses for mutations that reside in the hemizygous region.

The first two markers (bnlg1160 and bnlg197) that showed linkage by BSA mapping are in the chromosomal linkage bin 3.06. Subsequent follow-up mapping with additional SSR loci in the region placed dy between umc1730/zmm16 and bnlg1047a, and a total of five markers showed complete linkage to dy (Fig. 6C,D). The location of this interval is shown for several maize linkage maps including the reference map, UMC 1998 (Davis et al., 1999). Markers are shown or estimated for the dy mapping interval on the Genetic 2008 3 and IBM2 2008 Neighbors 3 maps. The approximate locations of the B-A translocation stocks are indicated on the basis of extrapolation with the Morgan-to-McClintock translator (Lawrence et al., 2006). Using the univalent/micronuclei scoring assay, we have now mapped dy to an ∼9 cM region within genetic bin 3.06 on the long arm of chromosome 3.

Our exploration of the maize genome for candidate genes in the dy mapping interval, in particular, genes that if mutated could result in one or more of the dy-associated phenotypes, is summarized in Fig. 7. The genomic region of the dy mapping interval (umc1730 – bnlg1047a) spans 7.7 Mbp and includes 154 gene models, some of which are named, whereas others are unclassified and predicted to encode proteins with unknown function. For this last group, we carried out BLAST homology searches using other databases (www.plantgdb.org; GenBank) to identify proteins or protein domains for our assessments.

Omitting consideration of transposable-element genes, we identified about 20 genes that matched our search criteria. Among these, we found one very compelling candidate gene, a maize SUN-domain protein gene, ZmSUN3 (Murphy et al., 2010) (Fig. 7C,D). SUN-domain proteins are localized to the inner nuclear membrane of the NE, where they function as part of a physical structure that tethers the cytoplasm to the nucleoplasm, including specific roles in meiotic telomere dynamics during meiotic prophase in animals and fungi (Hiraoka and Dernburg, 2009; Starr and Fridolfsson, 2010). The ZmSUN3 gene in the maize dy mapping region is therefore considered a compelling candidate gene for further examination. It belongs to the recently identified PM3-type class of SUN-domain proteins, characterized by having a SUN domain in the middle of the protein, three transmembrane domains, a coiled coil domain, and a highly conserved PAD region of unknown function (diagrammed in Fig. 7D). The ESTs that map to this region (Fig. 7C) and gene structure models (Fig. 7D) show the gene intron-exon structure of ZmSUN3 and the resulting predicted protein and its domains.

Our cloning and sequencing of multiple genomic PCR products and cDNA clones from various genetic sources did not reveal any obvious dy-specific genetic lesion in the SUN3 gene. In particular, we observed several SNPs and a few genotype-specific amino acid variants but did not detect any frameshift, nonsense, or mobile-element insertion in the ZmSUN3 gene in the dy genetic background. The dy mutation has been propagated by self-fertilization for many decades and is presumed to be in an isogenic, albeit undefined, genetic background. We did notice, however, that the dy gene produced two different transcripts, the canonical ZmSUN3 (Murphy et al., 2010) and a less abundant splice variant here designated ZmSUN3-sv1 (Fig. 7C, transcripts labeled ‘sv1’). The splice variant results from intron 3 retention, and the resulting mRNA would encode a shorter protein that lacks the two C-terminal transmembrane domains. Accumulation of this truncated SUN3 protein in the meiotic nucleus could disrupt the targets, folding, or interactions of SUN3 with its normal partners. Such a scenario would be compatible with the dy phenotypes observed, especially the precocious telomere detachment and the NE distortion. In this scenario (Fig. 7F), accumulation of aberrant SUN3 protein could disrupt its meiotic function. To investigate this possibility further, we amplified SUN3 cDNA from normal and dy male flowers at meiosis stage and found that the splice variant was the most abundant form in dy (Fig. 7E). We also observed the SUN3-sv1 transcript in non-meiotic tissues and other genotypes, but to a lesser relative degree (not shown), consistent with random EST reads performed by others (Fig. 7C).

The presence of the variant transcript and its increased proportion in dy/dy anthers, we consider a possible scenario in which dy is a gain-of-function, antimorphic mutation resulting from a splice variant that accumulates to detrimental levels in the homozygous (dy/dy) or hemizygous (dy/-) plants, but not in heterozygous (+/−) or wild-type (+/+) plants. We suggest, therefore, that the splice variant of SUN3 may be the basis for the mutant phenotypes observed.

Finally, we recognize that other genes besides SUN3 that reside in the linkage deserve careful consideration. These other candidate genes are listed in (Fig. 8) alongside a diagrammatic summary of a hypothetical meiotic SUN LINC complex. We considered genetic disruption of processes that involve gene products in any of three cellular domains: the cytoplasm, the nuclear envelope, and the nucleoplasm. In the cytoplasm, candidate genes would include components of the motility system, such as tubulin or actin; molecular motors; or connector proteins that link these to the nuclear envelope. In the nuclear envelope, candidate genes would include KASH and SUN domain proteins. In the nucleoplasm, candidate genes are more numerous, reflecting the abundance of structural and enzymatic gene products required for meiosis. Finally, we considered candidate genes for regulatory control or any of these processes in a meiotic SUN LINC complex pathway, such as check point or signal transduction genes. Overall, considering all of these candidates, we find that SUN3 is the most compelling candidate and favor the hypothesis that the mutant phenotypes of dy are caused by an allele of SUN3 that imparts a disruptive gain-of-function to the meiotic nuclear envelope.

In the current study, we have undertaken a detailed 3D cytological analysis of dy and discovered several new phenotypes that place it in a pathway integrating NE functions with meiotic chromosome segregation. Collective, our newly discovered dy phenotypes – (1) incomplete telomere clustering in early prophase, (2) defects in progression of synapsis as revealed by ASY1 immunostaining, (3) marked nuclear morphology distortion using nuclear shape analysis, and (4) anaphase bridges – show that dy is pleiotropic and may involve direct or indirect disruption of a LINC complex-like pathway that operates during meiotic prophase in maize.

Our immunofluorescence staining of ASY1 provided clear evidence of synaptic irregularities. The ASY1 antibody was raised against a recombinant protein made from an ASY1 cDNA of Arabidopsis (Armstrong et al., 2002) and shown to detect chromosome axis in meiosis of other plant species including maize (Golubovskaya et al., 2006). This antiserum has previously been used successfully to characterize defects in synapsis in other maize mutants (Golubovskaya et al., 2011). The ASY1 protein associates with the chromosomal axis of unsynapsed meiotic chromosome fibers and is thought to be removed from the chromosomes as they synapse (Armstrong et al., 2002; Sanchez-Moran et al., 2008; Sanchez-Moran et al., 2007). This pattern is seen for wild-type maize in this study (Fig. 2B,E) but not for dy, where persistent ASY1 staining is believed to indicate incomplete synapsis leading to asynapsis. This view differs slightly from previously published descriptions of the mutant, in which the homologs synapse and exhibit a normal pachytene but fail to maintain chiasmata (Maguire et al., 1991), unless complete synapsis in dy is followed by partial desynapsis and ASY1 reloading.

Another possible explanation for the ASY1 staining pattern in dy is that the processing or resolution of meiotic chromosome tangles or interlocks is affected by the mutation. Disruption of interlock resolution could leave physical barriers to the formation of complete synaptonemal complexes, such as those described by (Maguire et al., 1991). Exactly how the completion of synapsis is monitored by the cell in maize meiosis remains unknown, but incomplete synapsis could trigger checkpoints or affect feedback loops. In addition, we note that any of these defects – asynapsis, desynapsis, or interlock resolution – could occur simultaneously and to different degrees in individual dy nuclei. Regardless of the mechanism, our findings of the effect of dy on nuclear shape distortion establish a functional link between meiotic chromosome metabolism inside the nucleus and organelle integrity and likely cytoplasmic forces operating on or outside of the nucleus. Other mutants of maize, such as pam1 (Golubovskaya et al., 2002; Golubovskaya and Mashnenkov, 1977), also link telomere positioning to meiotic chromosome segregation, suggesting that the Pam1 gene also functions in the NE-telomere-synapsis pathway described here for dy. Indirect evidence that is consistent with our distortion-detachment model (Fig. 7F) comes from a recent study on Arabidopsis SUN proteins, shown to play a role in somatic nuclear morphology in part through a SUN-KASH, LINC complex-type mechanism (Zhou et al., 2012).

Maize has historically been an excellent model system for forward genetic studies. Over 50 meiotic mutants of maize have been isolated (representing at least 35 different genetic loci), but only 14 have been genetically mapped, and still fewer (n = 6) have actually been cloned and characterized (Cande et al., 2009; Golubovskaya et al., 2003). We have mapped another historical maize meiotic mutant, desynaptic (dy), to a region on chromosome 3L, bin 3.06, using two B-A translocation mapping stocks and an SSR-based BSA mapping strategy. The requirement for cytological verification in progeny F3 families (Fig. 5) limited the mapping resolution in our study to a 7.7 Mbp region (Figs 6, 7).

Maize dy has previously been defined as a recombination modifier gene (Ji et al., 1999). Relatedly, genetic control of recombination frequency in maize was subjected to QTL mapping by two groups, and in each study, the biggest effect QTL mapped to regions immediately adjacent to our dy mapping interval (Dole and Weber, 2007; Esch et al., 2007). These map positions were estimated to be between 166.0 and 168.5 MB (Esch et al., 2007) and at 181.1 MB (Dole and Weber, 2007) on the physical map of maize, which map just beyond our proximal and distal boundaries, respectively. Interestingly, the QTL with the greatest effect on recombination in Arabidopsis mapped near the end of chromosome 1, a region including At1G71360, predicted to encode a PM3-type SUN-domain.

A key to positional map-based cloning is the availability of other alleles for validation of candidate genes. To date, only one other mutation has been shown to be similar to dy, namely desynaptic1 (dsy1), for which allelism test crosses were carried out between dsy1 and dy (Golubovskaya and Mashnenkov, 1997; Maguire and Jackson, 1998). These authors described the progeny as having normal-looking pachytene, metaphase I, anaphase I, and quartet stage cells, yet concluded that they were allelic on the basis of a variant diakinesis phenotype. Some alleles might possibly partially complement each other, or the progeny might be nonallelic but, in double heterozygous form, reveal a subtle diakinesis phenotype. Indeed, we favor requiring a stricter definition of allelism, in which primary phenotypes, such as univalents and micronuclei, would be present and conspicuous in the male meiocytes of test-cross progeny. In this regard, we would expect dsy1 mapping to reveal a locus unlinked to dy on 3L.

From our initial screen for candidate genes, we identified a novel protein-coding gene predicted to encode a SUN domain-containing protein. Although we failed to detect a genetic lesion in this gene at the genomic DNA level, we have evidence for the existence of a significant amount of a splice variant predicted to encode a truncated protein lacking its two C-terminal transmembrane domains. Further experiments with qRT-PCR will allow for a more quantifiable detection of this transcript from different lines of maize. We hypothesize that, given enough available unspliced mRNA product, the translated protein is likely to be unable to carry out one or more essential cellular functions, including either telomere tethering to the NE or NE-cytoplasmic force transduction. The overall phenotypes observed with dy bear a striking similarity to those of MmSun1 knockout mice, in which meiotic telomere-NE detachment precedes meiotic failure and sterility (Ding et al., 2007).

If the aberrant transcript, Sun3-sv1, is causing the mutant phenotype, it might explain the variable expressivity observed with homozygous dy. In this case, ZmSUN3-sv1 RNA levels may have to accumulate to a critical threshold level to disrupt recombination and subsequent disjunction of homologs. This accumulation could produce a gain-of-function antimorphic mutation, for which similar alleles would be relatively rare. Alternatively, we have considered the possibility that SUN3 is a dying pseudogene, producing aberrant transcripts or even anti-sense transcripts that could contribute to gene silencing of ZmSUN3 or its closely related unlinked duplicate gene, ZmSUN4. Finally, it is possible that the dy mutation may be located not in the SUN3 gene but in another nearby gene, such as one of the candidates listed in Fig. 8. Among these other candidates are genes whose products could function in a LINC-complex-type pathway. These include genes encoding Actin-2, Beta tubulin, pleckstrin-homology domain, and MAR-binding filament-like proteins. These genes represent targets for reverse genetic analysis, in which meiosis-specific knockout or knockdowns would be expected to disrupt nuclear morphology, telomere positioning, synaptic progression, and eventual meiotic chromosome segregation. Further genetic and molecular analysis of Dy and its interacting partners is expected, therefore, to open new avenues for exploration of the intricate processes at work in meiotic cells.

Maize (Zea mays L., 2n = 2x = 20) inbred lines (A344+, KYS+ and dy) were grown at the Mission Road Research Facility (Tallahassee, FL) greenhouse and summer field plots. Male meiotic cells from meiosis-stage anthers were fixed with formaldehyde in 1× meiocyte buffer A (MBA) as previously described (Bass et al., 1997; Murphy et al., 2010) with either 4% paraformaldehyde, for 3D telomere FISH, or 0.5–1% paraformaldehyde and 0.05% Tween-20, for protein immunolocalization, for 30 min at room temperature. Anthers were then rinsed and stored in MBA for several weeks at 4°C.

Cells were prepared for 3D telomere FISH analysis essentially as described by Bass and colleagues (Bass et al., 1997). A fluorescein-conjugated telomere-specific oligonucleotide probe [5′-FITC-(CCCTAAA)4-3′] was diluted to 0.13 µM in the hybridization buffer. For immunostaining, fixed anthers were subjected to a 1 hour room-temperature treatment in a triton permeabilization buffer as previously described (Murphy et al., 2010). The acrylamide pads on the slides were then incubated in blocking buffer containing a 1∶250 dilution of primary rabbit polyclonal antibody, anti-ASY1 (a gift from F. Christopher Franklin, University of Birmingham, UK) then in a FITC-goat anti-rabbit secondary antibody at 1∶2000 in blocking buffer for detection. After staining, cells were counterstained with DAPI at 3 µg/ml in 1× PBS and mounted as previously described (Bass et al., 2003; Bass et al., 1997). Staging of meiotic nuclei was based on chromosome fiber appearance (Bass et al., 2003).

All images were collected with an IX-70 inverted wide-field epifluorescence microscope (Olympus Corporation, Center Valley, PA) with a single oil-immersion lens, 60× (NA 1.4 Plan Apo). Data were sampled in the X, Y, and Z dimensions with voxel dimensions of 0.07×0.07×0.3 µm³ and were collected for an area extending ∼3–4 µm beyond the nuclear border in Z, and at least 5 µm beyond the edge of the nucleus in X and Y. The resulting data sets were then subjected to 12 cycles of iterative 3D deconvolution before data analysis. Deconvolved images were adjusted by linear scaling for brightness and contrast with SoftWorx (Applied Precision, Issaquah, WA). Telomere position analysis was carried out with 3D modeling and measurement software (SoftWorx). The 3D nuclear shape analysis was carried out on a 0.9 µm projection (3 Z slices) from the middle of the nucleus. Nuclear edges were defined with 2D EditPolygon, and the resulting images were imported into ImageJ (Abramoff et al., 2004) for measurements and analysis. The nuclear circularity [4π(area/perimeter²)] and aspect ratios were determined on these polygon circles with tools available in ImageJ. One-way analysis of variance was used to test for differences between the samples analyzed. A Tukey’s post-hoc multiple comparison test was used to determine which of the means were statistically different from one another at the 95% confidence level.

Approximately 1.5 g of leaf tissue was harvested, flash frozen in liquid nitrogen, and stored at –80°C. DNA was extracted with an aqueous extraction buffer as previously described with some modifications (Dellaporta et al., 1985). Initially, 9 mL of prewarmed (65°C) extraction buffer (50 mM Tris-HCl, pH 8, 10 mM EDTA-NaOH, pH 8, 100 mM NaCl, 1% w/v SDS, 1% w/v polyvinylpyrrolidone and 1% w/v polyvinylpolypyrrolidone) was added to the tissue powder, shaken vigorously for 20 seconds, and incubated at 65°C for 20 minutes. The final DNA pellet was resuspended in 400 µl of TER (10 mM Tris-HCl, pH 8, 0.02 mM EDTA, 40 µg/ml RNase A), and the DNA was re-precipitated by the addition of 40 µl of 3 M sodium acetate (pH 5.2) and 800 µl of ice-cold ethanol. The DNA was pelleted by centrifugation, washed with 70% ice-cold ethanol, air dried, and resuspended in 250 µl of TER buffer. The resulting DNA yield and quality were assessed by agarose gel electrophoretic analysis, and spectroscopic quantification of the DNA yield was performed with a Nano-drop (Thermo Scientific, ND-1000, Wilmington, DE).

A monohybrid mapping population was created from genetic crosses between homozygous maize dy/dy and inbred line KYS+/+, and the resulting F1 progeny seed were planted and selfed to produce a segregating population. F2 individuals were used for leaf DNA preps and self-fertilized for cytological scoring of F3 meiotic cells. DNA samples were isolated from mutant (dy/dy) and nonmutant (+/+ or +/dy) and used for bulked-segregant analysis (Michelmore et al., 1991) mapping (Sigma, Maize Mapping SSR primer set, M4193). Upon detection of two consecutive dy-linked markers (p-bnlg1160, p-bnlg197), additional nearby SSR markers were obtained for finer-resolution mapping (supplementary material Table S1). B-A translocation mapping (Beckett, 1978) was used to confirm the linkage analysis data. B-A translocation stocks (Tb-3La and Tb-3La-2s[6270]) were obtained from the Maize Genetics Cooperation Stock Center. Plants harboring B-A translocations were crossed as male parents to F3 homozygous dy plants. Sixteen progeny resulting from these crosses were planted, grown to meiosis stages, and scored cytologically for the dy phenotype including chromosome bridges, laggards and micronuclei.

Candidate genes were initially identified by inspection of the genome annotations and gene models from the reference genome of B73 (AGPv1 and AGPv2 at http://maizesequence.org and http://maizegdb.org). Specifically, the dy-linked genes within the dy mapping interval (umc1730 at Chr3:171,427,820 to bnlg1047a at Chr3:179,119,635) for B73 AGPv2 were examined, and candidates were tabulated if they or BLAST-based homology searches indicated relevance to dy phenotypes.

For genomic and cDNA molecular cloning of ZmSUN3, DNA from B73+, KYS+, A344+ and verified dy/dy gene-specific primers (supplementary material Table S2) were designed to span the entire open reading frame, including the 5′ and 3′ UTR regions. PCR products from the individual reactions were cloned (Topo-TA PCR 2.1, Life Technologies, Grand Island, NY), sequenced, and assembled into a gene structure by means of the contig-assembly software and sequence-analysis program Sequencher v 4.7 (Gene Codes Corporation, Ann Arbor, MI). For RT-PCR, total RNA was isolated from meiotic florets microdissected from pre-emerged tassels with the plant RNeasy kit (Qiagen, Valencia, CA). The total RNA was treated with 10 U of RNase-free DNase-I (Sigma-Aldrich, St Louis, MO) and then reverse transcribed into cDNA with SuperScript III (Life Technologies, Grand Island, NY). Gene-specific primers were designed for ZmSUN3 on the basis of the cDNA or gene model sequences from the maize genome browser (http://maizesequence.org). RT-PCR products were then resolved, analyzed, cloned and sequenced as described above.

We thank the members of the Bass laboratory, including A. Brown, E. S. Howe and D. Vera, for critical reading of the manuscript and insightful comments. We thank A. B. Thistle for editing the manuscript.