Abstract We examined the distribution of meiotic epitopes for the Dmc1 protein of lilies in a normal dip-loid, a triploid, and in a diploid species-hybrid. The trip-loid has an extra chromosome set; all three sets align, butonly two of the three axes intimately pair at a given loca-tion. Our findings with the triploid support the idea thatretention of the foci until the pachytene stage requires asuccessful homology check and synaptonemal complex(SC) initiation; the number of foci in the triploid dimin-ishes by approximately 30% from early zygotene topachytene, and the triploid pachytene values are similarto the pachytene values of the diploid. The species-hybrid lacks chromosome homology, has reduced SCformation and few reciprocal genetic exchanges. In thisspecies-hybrid the number of foci at early zygotene issimilar to that in the normal diploid but is dramaticallyreduced by mid-zygotene. The extent to which the num-ber of Dmc1 foci is reduced is similar to the extent thatSC formation is reduced. In contrast the extent of the reduction in reciprocal genetic exchange in the species-hybrid is much greater than the reduction in the numberof foci. We conclude that Dmc1 protein is involved in

homology checking, but the impact of failure to find homology affects SC formation and reciprocal geneticexchange differentially.

Introduction

Meiosis is a specialized type of nuclear division duringwhich the chromosome number is reduced in anticipa-tion of fertilization. The reduction in chromosome num-ber is accomplished by having one round of chromosomereplication be followed by two rounds of chromosomedivision. In the first meiotic division homologous chro-mosomes are separated; in the second meiotic divisionsister chromatid centromeres are separated. Ensuring thathomologous chromosomes are separated requires thatspecialized events occur in both premeiotic S-phase andin the first meiotic prophase.

In a standard meiotic prophase I homologous chromo-somes come together, pair up precisely along their axiallength, and form an elaborate tripartite structure knownas the synaptonemal complex (SC). While homologousSC formation is the rule, an SC can form between non-homologous chromosome segments in a variety of circumstances [e.g. species-hybrids (Toledo et al. 1979),synaptic accommodation (Moses and Poorman 1981),haploids (Loidl et al. 1991), meiotic mutants (Yoshida et al. 1998)].

A pathway for repairing double-strand breaks is alsopresent during prophase I. Double-strand breaks in theDNA are created, then processed to expose re-sected sin-gle-stranded DNA. The 3′ ends of such single strands thenbecome associated with RecA-like proteins. D-loop for-mation proceeds between the single-stranded DNA (asso-ciated with RecA-like proteins) and the double-strandedDNA of the homologous chromosome, and a double Holliday junction can be formed. Resolution of the Holli-day junction in a particular manner can result in reciprocalgenetic exchange (reviewed in Davis and Smith 2001).

One RecA-like protein found in eukaryotes is theRad51 protein. In vitro Rad51 can form a nucleoprotein

Edited by: P. Moens

We would like to dedicate this article to the memory of Dr. HerbertStern. He provided us (Y. Hotta and C. Hasenkampf) with an exem-plary role model – both as a rigorous, successful scientist and as akind and witty person of the highest integrity. We miss him dearly.

S. George · P. Behl · R. deGuzman · M. Lee · S. RusyniakC. Hasenkampf (✉ )Division of Life Science, University of Toronto at Scarborough,1265 Military Trail, Scarborough, Ontario M1C 1A4, Canadae-mail: hasenkampf@utsc.utoronto.ca

K. Hiratsuka · H. TakaseDepartment of Molecular Biology, Graduate School of Biological Sciences, Nara Institute of Science and Technology, 8916-5 Takayama, Ikoma, Nara 630-0101, Japan

Y. HottaDepartment of Nutrition and Health, Niigata University of Health and Welfare, 1398 Shimami, Niigata, Niigata 950-3198, Japan

Chromosoma (2002) 111:96–105DOI 10.1007/s00412-002-0193-5

O R I G I N A L A RT I C L E

Susan George · Pearl Behl · Rhoda deGuzmanMarian Lee · Stefan Rusyniak · Yasuo HottaKazuyuki Hiratsuka · Hisabumi TakaseClare Hasenkampf

Dmc1 fluorescent foci in prophase I nuclei of diploid, triploid and hybrid liliesReceived: 7 November 2001 / Revised: 7 March 2002 / Accepted: 11 March 2002 / Published online: 15 May 2002© Springer-Verlag 2002

97

filament with single-stranded DNA (Ogawa et al. 1993).Another RecA-like protein, Dmc1, has been found inmany eukaryotes. In contrast to RecA protein and otherRecA-like proteins, conditions under which Dmc1 formsa nucleoprotein filament with single-stranded DNA invitro have not yet been found. Instead, in vitro, Dmc1forms octameric rings or stacks of rings (Masson et al.1999; Passy et al. 1999). Both Rad51 and Dmc1 canbind, in vitro, to single-stranded DNA but neither proteindoes so as robustly as RecA protein, and it is likely thataccessory proteins would be needed for Rad51-mediatedstrand invasion in vivo (Gupta et al. 1999, 2001).

Proper functioning of Rad51 protein is important tohomologous recombination both in a mitotic cycle (forrepair of DNA damage) and during meiosis. Use of anti-bodies directed against Rad51 protein has revealed thatdistinct foci of Rad51 form in response to DNA damageduring a mitotic cycle (Gasior et al. 2001) and duringmeiotic prophase I (Bishop 1994). In yeast, null mutantsof Rad51 have increased sensitivity to radiation, andhave reduced levels of recombination during meiosis(Shinohara et al. 1992). In mice, loss of Rad51 functionresults in embryonic lethality (Lim and Hasty 1996).

Dmc1 generally appears to be meiosis specific (Massonand West 2001). In Dmc1 knockout mice both male and fe-male sterility is observed (Yoshida et al. 1998). In Saccha-romyces cerevisiae, Dmc1 null mutants have a reduction inrecombination (Bishop et al. 1992). Use of anti-Dmc1 anti-bodies reveals that Dmc1 protein, like Rad51 protein, oc-curs in distinct foci during prophase I (Bishop 1994). Therelationship of Rad51 foci and Dmc1 foci has been mostextensively studied in S. cerevisiae and mammals. Bishop(1994) found that in an SK1 strain of S. cerevisiae, Dmc1and Rad51 foci co-localize in wild-type individuals, but foci for either can form in the absence of the other [i.e.Rad51 foci can form in yeast Dmc1 null mutants (Bishop1994); Dmc1 foci can form in Rad51 null mutants, al-though the foci are less bright than in wild type (Shinoharaet al. 2000)].

Analysis of mutants suggests that Rad51 protein andDmc1 protein are also important in stabilizing the associ-ation of homologous chromosomes prior to the formationof SCs. Rockmill et al. (1995) have suggested that Rad51and Dmc1 are involved in a chromosome homologycheck. If homology is present, then somehow a stable ax-ial association can form between homologous chromo-somes. They further propose that this axial associationexpedites subsequent SC formation. In fact Rad51 orDmc1 null mutants do have delayed and reduced amountsof SC. Further it is likely that some of the SC that formsis nonhomologous. Rad51 knockout mice die, so theirmeiotic phenotype cannot be assessed. Knockout Dmc1mice are sterile. Cytological analysis of these mice re-veals that the homologous chromosomes generally fail toform SC; of the small amount of SC that is present, someis likely nonhomologous (Yoshida et al. 1998). Andersonet al. (1997) (in lily) and Tarsounas et al. (1999) (in mice)have found Rad51 and Dmc1 to be associated with com-ponents of the SC both before and after mature SC has

formed, and postulate that these proteins aid in the forma-tion of axial associations. Thus in addition to their rolesin the double-strand break pathway, Dmc1 and Rad51proteins appear to be important for the formation of axialassociations between homologous chromosomes, and inexpediting timely SC formation.

Strategies for studying Dmc1 function to date havebeen largely 'trans' type analyses. The function and distri-bution of both wild-type and mutant Dmc1 protein havebeen examined. We elected to use an alternate approach.We sought to elucidate further the function of the wild-type Dmc1 protein in normal lily diploid meiosis andcompare it with events in lilies with abnormal chromo-some situations. We decided to study the distribution of Dmc1 epitopes in a triploid lily with an extra set ofchromosomes, and in a diploid species-hybrid known asBlack Beauty. Black Beauty diploid hybrids have a veryreduced level of chiasma formation (Emsweller and Uhring 1966). The problems that arise during meiosis inthis diploid hybrid are likely due to lack of chromosomehomology, and not genic imbalances between the two pa-rental genotypes. This is inferred from the fact that if thechromosome number of the diploid hybrid is doubled, thesynthetic allotetraploid undergoes a normal meiosis witha high degree of reciprocal genetic exchange (Toledo et al. 1979). Here we report differences in the number ofDmc1 foci seen in our control diploid, in the triploid, andin the species-hybrid. We discuss these findings relativeto the function of the Dmc1 protein.

Materials and methods

The lilies used in this study were either Lilium tigrinum (diploid),L. tigrinum (triploid), or the cultivar Black Beauty, which is a spe-cies-hybrid that results from a cross of Lilium henryi with Liliumspeciosum. The Tigrinum lilies were purchased from Van Hof andBlokker bulb distributors (Mississisauga, Ontario), and the BlackBeauty bulbs were purchased from Dahlstrom and Watt BulbFarms (Smith River, Calif.). All plants were grown in a green-house under conditions of natural lighting, with a day/night tem-perature regimen of 25°C/15°C.

Squash preparations for examining chromosome morphologywere prepared by squeezing the contents of individual buds into afixative consisting of three parts 95% ethanol, and 1 part glacialacetic acid for 30 to 120 min. The meiotic cells came out of theanther as a packed filament of connected cells. Portions of the fila-ment were removed from the fixative and placed on a clean slidewith iron aceto-carmine. The material was stained for 10 min as itwas uniformly distributed on the slide. A cover glass was thenadded, and pressure was applied to produce the squash prepara-tion. The slides were dipped in liquid nitrogen, and the cover glasswas removed. Preparations were then dehydrated in a graded etha-nol series consisting of 70% ethanol, two changes in 95% ethanol,and two changes of 100% ethanol. Slides were placed in histoclear(DiaMed) for 3 min, then drained and mounted with a drop of Per-mount and a cover glass. Nuclei in prophase I and metaphase Iwere examined using a Zeiss axiophot microscope equipped witha 63× planapo lens. The criteria for determining the substages ofprophase I were a combination of the criteria of Erickson (1948)and Holm (1977), as indicated in Hasenkampf et al. (1992). Thetriploid and species-hybrid each had the potential for incompletesynapsis. Therefore the shape and location of the nucleoli, and theoverall extent of chromosome condensation were particularly im-portant criteria.

98

For the immunocytochemical analysis floral buds at leptotene-early zygotene, mid-late zygotene, and early pachytene were used.One anther of a bud was used to prepare a squash preparation todetermine the meiotic stage. The remainder of the anthers wereused for immunocytochemistry if the bud was at the desired stage.For the immunocytochemistry the meiotic nuclei were prepared ina manner similar to the procedure used by Terasawa et al. (1995).The contents of the anthers were squeezed into a small tube con-taining 4% paraformaldehyde dissolved in phosphate-buffered saline (PBS) at pH 7.4 (PBS is 130 mM NaCl, 7 mM Na2HPO4,3 mM NaH2PO4). After fixation for 10 min at room temperature,the fixative was removed and the filaments were washed threetimes in a solution consisting of 0.6 M mannitol, 5 mM 2-N-morpholino ethane sulfonic acid, 50 mM EDTA, 0.15 mM sper-mine, pH 5.6. The filaments of cells were then incubated in theabove solution to which Y-23 pectolyase (Sigma) had been addedto a final concentration of 2%. After 5 min the solution was re-placed with 4% paraformaldehyde in PBS and the cells were fixedagain for 30 min. Next a small portion of a filament in 25 µm of fixative was placed on a clean slide, and 25 µm of 0.4% TritonX-100 was added. After 2 min a cover glass was added and the fil-ament was gently tapped into a monolayer to disperse the meioticcells evenly, and pressure was applied to the cover glass to squashopen the cells. Slides were then dipped in liquid nitrogen, and thecover glasses were removed. Slides were immediately transferredto a tray containing PBS and were washed three times, each washlasting 3 min. For immunocytochemistry, individual slides wereremoved from the tray, the PBS was blotted away, and 200 µm ofblocking solution was applied. The blocking solution consisted of0.5% blocking reagent (Roche), and 5% goat serum, both dis-solved in double-distilled water. After 30 min, the blocking solu-tion was shaken off the slide, and 200 µm of incubation buffer(0.5% blocking solution with 1% goat serum) containing the pri-mary antibody was added to the slide; slides were incubated for2 h at room temperature.

Three different primary immune sera were used. For the posi-tive controls the primary antibody solution was an immune serumdesignated 411, directed against lily histone H1 (Riggs 1994). Thenegative control experiments were done with the preimmune se-rum from the rabbit later used to generate immune serum 1. Twodifferent immune sera were used to detect Dmc1 epitopes. Thefirst immune serum was prepared using the entire lily Dmc1 pro-tein. Immune serum 1 reacts with purified lily Dmc1 protein, andwith the purified Dmc1 protein of Arabidopsis. It does not reactwith purified Arabidopsis Rad51 protein. Immune serum 1 reactswith only one band in our protein gels; however, we do not havepurified lily Rad51 protein and cannot absolutely eliminate thepossibility that immune serum 1 reacts with both lily Dmc1 andRad51 proteins. The other Dmc1 immune serum was kindly pro-vided by the laboratory of Dr. Tomoko Ogawa. The Ogawa im-mune serum was created using a synthetic peptide correspondingto the 18 amino acids at the N-terminus of the lily Dmc1 protein(the lily gene was originally called LIM15; Kobayashi et al.1993). This immune serum is specific for the Dmc1 protein anddoes not react with lily Rad51 protein (Terasawa et al. 1995; Anderson et al. 1997). All immune sera were diluted 1:750 for theimmunocytochemistry experiments.

Once the incubation in primary antibody was completed, slideswere washed four times (5 min per wash) in PBS with 0.1%Tween-20, followed by two washes in PBS, then one wash in in-cubation buffer. Slides were then drained briefly and 200 µm ofsecondary antibody was added. The secondary antibody was agoat, anti-rabbit IgG conjugated to fluorescein (Vector Laborato-ries) diluted 1:100 in incubation buffer. After a 2 h incubation atroom temperature the slides were washed in PBS four times, eachwash lasting 3 min. The washing buffer was dabbed off the slidesand a drop of aqueous Vectashield mounting media (Vector Labo-ratories) was added. Some slides were stained with 4′,6-diami-dino-2-phenylindole (DAPI) before mounting. Slides were ob-served using a 100× Zeiss planapochromat lens. The fluoresceinsignal was detected with the fluorescein filter set (Zeiss 487909);the DAPI signal was detected using Zeiss filter set 487902. Re-

sults from 20 nuclei were used for each stage for immune serum 1,and results from 10 nuclei for each stage were used for the im-mune serum 2 (Ogawa).

For the quantification of foci the fluorescein filter set wasused, and the nuclei were positioned in the center of the field ofview. The primary focal plane was selected. The focused imagewas captured using a Sony 3 CCD color video camera. The result-ing image was digitized and analyzed using Northern Eclipse Soft-ware (Empix, Missassauga, Ontario). The boundary of the nucleusin the digitized image was delimited on the image and the numberof discrete fluorescent foci within the delimited region was deter-mined. The color images were then converted to black and whitefor publication. Since the analyzed nuclei had been liberated fromthe cell and had been squashed into a flat preparation, most fociwere visible within the selected focal plane. However some outly-ing foci may not have been in the captured focal plane. Hence ourestimates of the number of foci are minimum estimates.

Immunoblots were prepared from samples of approximatelyfive to ten buds per stage; each bud contained thousands of meiot-ic cells. One anther per bud was used for cytological assessmentof the meiotic stage. The remaining anthers of each bud of the appropriate stage, were used for immunoblotting. The contents ofeach anther were squeezed into ice-cold White's plant tissue cul-ture medium (White 1963), pH 5.4, supplemented with proteaseinhibitors according to the manufacturer's instructions (Roche catalog no. 1873580). Once a collection for a particular samplewas complete, the White's medium was removed, and the meioticcells were homogenized in an equal volume of 2×DB (2% SDS,0.6% w/v β-mercaptoethanol, 20 mM TRIS, pH 8.6, 20% glyc-erol). The protein concentration of each sample was determined.Gel electrophoresis and immunoblotting were done as indicated inHasenkampf et al. (1992).

To confirm the protein concentration estimates and assess thequality of the preparations, 5 µg of each protein sample were sub-jected to SDS-polyacrylamide gel electrophoresis and stained withCoomassie Brilliant Blue. If the concentration of the core histoneswas the same in each protein extract, then additional 5 µg sampleswere used in gels to produce the immunoblots. While no directloading controls were present for the individual lanes of each gelused to produce the immunoblots, the protein concentrations ofeach sample were confirmed (as indicated above), and the immu-noblot series done for each type of lily was repeated at least threetimes for immune serum 1, and at least twice for immune serum 2(Ogawa). For each type of lily the replicates of the immunoblotsshowed the same trend.

Results

Bright-field microscopy

A light microscopy study was performed to access the extent of pairing and reciprocal genetic exchange inthe normal diploid tiger lily, in the triploid tiger lily, andin the species-hybrid Black Beauty. All three types of lilies had a high degree of pairing during pachytene(Fig. 1A, C, E). Metaphase I analysis was done for eachtype of lily to determine the extent of reciprocal geneticexchange. For the metaphase I analysis 30 nuclei (fromthree different plants) were examined for each type ofplant. For each nucleus the numbers of univalents, biva-lents, and trivalents were recorded.

The diploid tiger lily served as the positive control,being expected to have normal homologous chromosomepairing, synapsis, and reciprocal genetic exchange. In thenormal diploid, two by two alignment was followed bymore intimate pairing such that each pair of chromo-

99

somes appeared as a single uninterrupted bivalent. Asexpected, the normal diploid formed 12 bivalents (e.g.Fig. 1B) in all nuclei examined, indicating successfulcompletion of all events necessary to produce reciprocalgenetic exchange. While a precise count of the numberof reciprocal exchanges was not undertaken, two or moreexchanges per bivalent were seen in all 30 cells exam-ined, and the chiasma frequency is likely very similar tothe reported value of 54.8±6.0 per nucleus (Stack et al.1989).

In the triploid all three homologs regularly alignedthree by three, but the more intimate synapsis occurredtwo by two. Thus each set of (three) aligned chromo-somes had chromosome regions where one chromosomeaxes was not synapsed (Fig. 1C, arrowheads). This resultis similar to that of Loidl and Jones (1986) in an Alliumautotriploid, and that of Moens (1968) in a triploid tigerlily. The lily triploid is capable of forming a wide varietyof types of chromosome configuration at metaphase I,depending on the number of reciprocal genetic exchang-es, and which two (of the three) chromosomes partici-pate in each exchange. The most common situationfound was 11 trivalents, 1 bivalent, and 1 univalent(Fig. 1D). This configuration was seen in 16 of the 30nuclei examined. In 11 of the 30 nuclei all 12 sets of

three homologs were associated as 12 trivalents. The re-maining 3 nuclei had 10 trivalents, 2 bivalents, and 2univalents. Thus in general there was a high degree oftrivalent formation, indicating that there was a high de-gree of reciprocal genetic exchange, and that most oftenall paired chromosome arms participated in a genetic ex-change. Therefore the presence of an extra set of chro-mosomes, and the occurrence of unsaturated SC forma-tion (one-third of the total axial length was aligned butnot synapsed) did not block the ability of the nuclei toaccomplish high levels of reciprocal genetic exchange.

Black Beauty is a diploid species-hybrid with 24chromosomes, 12 each from two different but relatedspecies. Thus the chromosomes are homeologous, notstrictly homologous. The homeologous chromosomes accomplish pairing to a high degree (Fig. 1E); unpairedregions were only occasionally seen. Previous ultrastruc-tural analyses of diploid Black Beauty indicated that itforms an SC (Toledo et al. 1979, and Hasenkampf andShull, unpublished results). Even though the appearanceof the SC is normal, the extent of SC formation in diploid Black Beauty is reduced. Toledo et al. (1979) estimate that SC formation is 64% complete in diploidBlack Beauty, and Hasenkampf and Shull (unpublishedresults) estimate that SC formation only includes 37%

Fig. 1A–F Pachytene andmetaphase I chromosome con-figurations. A, B Micrographsof pachytene and metaphase Inuclei of the diploid control. C, D Micrographs of the trip-loid, Lilium tigrinum. The arrowheads in C point to regions where all three homo-logs are aligned and have axialassociations. In D, 11 triva-lents, 1 bivalent (II), and oneunivalent (I) can be seen. E, F Micrographs of the spe-cies-hybrid. The nucleus in Fhas 16 univalents, and 4 biva-lents (II). Bar in E represents10 µm for A, C, and E; bar inF represents 20 µm for B, D, F

100

of the paired chromosome length. Previous reports indi-cated that Black Beauty diploids are largely achiasmatic(Emsweller and Uhring 1966; Toledo et al. 1979). Wefound a similar result in our study. In the metaphase I nuclei examined most chromosomes occurred as univa-lents. Occasional bivalents were observed (Fig. 1F, biva-lents indicated by roman numeral II). Three nuclei hadno bivalents; eight nuclei had two bivalents; eight nucleihad three bivalents; five nuclei had four bivalents; fivenuclei had five bivalents; and one nucleus had six. Thusof the 12 potential bivalents that could form if reciprocalgenetic exchange occurred normally, most nuclei had on-ly two or three bivalents.

In summary, the diploid control exhibited normalpairing and reciprocal genetic exchange. The triploid lilyhad three by three alignment of the chromosomes, withpairing partner switches, and a high degree of reciprocalgenetic exchange. Trivalent formation was the norm atmetaphase I indicating that any two of the three homo-logs could participate in the exchange. The diploid spe-cies-hybrid had largely normal chromosome pairing, butgreatly reduced levels of reciprocal genetic exchange.

Immunoblots

Protein extracts for immunoblotting were prepared fromlily anthers at a variety of premeiotic and meiotic stagesfor each of the three lily varieties. The stages examinedwere the premeiotic interphase, leptotene-early zygotene,mid-zygotene, pachytene (early, mid- and late) and dip-lotene-diakinesis. Results of immunoblotting are shownin Fig. 2A, B. Immune serum 1 and immune serum 2

gave similar results, except that immune serum 1 regu-larly gave the stronger signal.

For the diploid control and the triploid lily the resultswere similar. That is, the strongest immunostaining wasobserved in the leptotene-early zygotene samples, it de-clined in mid-zygotene and was gone (or reduced) in thepachytene extracts. Results obtained with the species-hybrid Black Beauty were different. Immunostainingwas only detectable in the leptotene-early zygotene sam-ples. Immune serum 1 had a modest amount of stainingat this early stage, but immune serum 2 had only a barelyperceptible band. This was the first indication that theDmc1 protein might be behaving differentially in an un-usual chromosomal setting; further differences were seenin the immunocytochemistry results.

Immunocytochemistry

Fluorescent immunocytochemistry was undertaken usingsquash preparations of chromosomes at early zygotene,mid-zygotene and early pachytene. Positive controlslides were treated with anti-histone H1 immune serumto ensure that squash preparations for all three lily types,at all stages, were immunoreactive. As expected, uni-form staining of the chromatin was observed for all threevarieties at these three stages (data not shown).

Immunostaining was seen for all three lily varieties,at all three stages examined, with both Dmc1 immunesera (Immune serum 1, Fig.3, Immune serum 2, Fig.4).With each immune serum, two patterns of staining wereseen. There was a uniform diffuse staining of the chro-matin (also seen in Franklin et al. 1999). This diffusestaining was not seen in the negative control slides treat-ed with preimmune serum and thus does not representbackground staining. In addition to the diffuse staining,many bright discrete foci were seen along the axes of thechromosomes similar to foci reported in other immuno-cytochemical studies of Dmc1 and Rad51 proteins (e.g.Bishop et al. 1992; Bishop 1994; Terasawa et al. 1995;Dresser et al. 1997; Tarsounas et al. 1999). To confirmthe position of the foci on the chromosomes, prepara-tions were stained with DAPI and the images were com-pared using DAPI and fluorescein filters (data not

Fig. 2A, B Immunoblot profiles in diploid, triploid and species-hybrid lilies. Five micrograms of nuclear proteins were loaded ineach lane. The blots in A were probed with immune serum 1. Thisimmune serum was raised against the entire Dmc1 protein. Theblots in B were probed with immune serum 2 (Ogawa), which isspecific for Dmc1 protein. A, B Lane 1 premeiotic S-phase, lane 2leptotene-early zygotene, lane 3 mid- to late zygotene, lane 4pachytene (early, mid- and late), lane 5 post-pachytene stagesranging from diakinesis through the second meiotic division. Thetop molecular weight marker (lane M) represents 32.5 kDa, andthe lower marker represents 25 kDa

101

shown). Corresponding images show that the foci arealong the chromosomes as observed in numerous otherstudies (e.g. Terasawa et al. 1995; Yoshida et al. 1998;Tarsounas et al. 1999).

Foci in the normal diploid lily

In addition to the qualitative assessment, we also quanti-fied the number of foci seen per nucleus for both immunesera. To assess our ability to detect foci we compared theestimate we obtained with our diploid control, L. tigri-num, with that obtained by Terasawa et al. (1995), usingdiploid Lilium longiflorum. They report the values forDmc1 foci at zygotene as being between 800 and 1000.Using the same Dmc1 immune serum(our immune serum2) we obtained a mean value of 656 foci (SD = 51). Oursomewhat lower zygotene value may be due to species-specific differences between L. longiflorum and L. tigri-num, or may be due to a failure, on our part, to detect focioutside the main focal plane. Using our other immune se-rum on the early zygotene nuclei of our diploid control,we observed a mean value of 1015 foci per nucleus (SD = 162).

We compared the number of Dmc1 foci seen at earlyzygotene, mid-zygotene and pachytene in our diploidplants using the two immune serum. The results of ourcomparisons are illustrated in Fig. 5A. The bar graphsrepresent the mean number of foci observed for the dip-loid at each of the three stages examined, for each im-mune serum. As can be seen in Fig. 5A, the two immuneserum generally revealed the same trend – a reduction innumber of foci from early zygotene to early pachytene.However, there was one notable difference between thetwo sera. Immune serum 1 always detected more focithan did immune serum 2. Immune serum 2 was raisedagainst the 18 amino acids at the N-terminus of the lilyDmc1 protein. Immune serum 1 was raised against theentire Dmc1 protein. Thus immune serum 1 likely is de-tecting a greater range of Dmc1 epitopes. Hence immuneserum 1 is likely the more sensitive immune serum, capable of recognizing less bright collections of Dmc1proteins. The difference in number of foci detected withthe two sera was most pronounced in the early zygotenenuclei, then decreased such that by early pachytene thenumber of foci detected with both serum was very simi-lar. We interpret this to mean that early in the process ofpairing (i.e. in early zygotene) there is a set of bright focidetectable with both immune sera, and a set of lessbright foci that are only detectable with immune serum1. Then as the pairing process progresses, the less in-tense foci either disappear or merge.

Foci in the diploid control compared with foci in the triploid, and in the species-hybrid

The triploid lily (Fig. 5B), like the diploid, shows a re-duction in number of foci from early zygotene to early

Fig. 3A–I Foci in lily nuclei reacted with immune serum 1. A, D, G Micrographs of the diploid control, L. tigrinum. B, E, HMicrographs of the triploid, L. tigrinum. C, F, I Micrographs ofthe species-hybrid, Black Beauty. A, B, C Early zygotene nuclei.D, E, F Mid-zygotene nuclei. G, H, I Early pachytene. Bar rep-resents 10 µm for all panels

Fig. 4A–I Foci in lily nuclei reacted with immune serum 2. A, D, G Micrographs of the diploid control, L. tigrinum. B, E, HMicrographs of the triploid, L. tigrinum. C, F, I Micrographs ofthe species-hybrid, Black Beauty. A, B, C Early zygotene nuclei.D, E, F Mid-zygotene nuclei. G, H, I Early pachytene. Bar rep-resents 10 µm for all panels

102

pachytene. The triploid has 33% more chromosome axisthan the diploid, and at early and mid-zygotene hassomewhat more foci than the diploid. Nonetheless byearly pachytene the triploid has approximately the samenumber of foci as the diploid. Thus the number of brightfoci detectable at early pachytene is related to the synap-sed chromosome length, rather than the total chromo-some length present.

Lastly we examined the foci pattern in the diploidspecies-hybrid Black Beauty (Fig. 5C). At early zygo-tene the diploid species-hybrid has approximately 80%as many foci as the normal diploid at the same stage.However, in mid-zygotene there is a dramatic reduction

in foci number, such that by early pachytene the species-hybrid has only 43% as many foci as the diploid control.

As can be seen in Fig. 3F, I and Fig. 4F, I, the BlackBeauty nuclei at mid-zygotene and early pachytene havea smaller volume than either the diploid or triploid nucleiat the same stage. It is unclear why the paraformalde-hyde-fixed Black Beauty nuclei at these stages resistedsquashing, but this phenomenon was seen for most of theBlack Beauty nuclei examined. Nonetheless we do notthink the smaller nuclear volume significantly affectedour ability to count the foci; we think the observed re-ductions were real. This view is supported by the BlackBeauty immunoblot results (Fig. 2, right-most blots).The Dmc1 protein is only detectable in the leptotene-early zygotene protein extracts of the Black Beauty spe-cies-hybrid and was difficult to detect at all with immuneserum 2. Thus we conclude that the species-hybrid is un-able to retain a normal level of Dmc1 foci beyond earlyzygotene.

Discussion

We have examined three different varieties of lilies withthree different meiotic 'conditions'. We first examinedthe temporal distribution of Dmc1 epitopes in our con-trol diploid using immunoblotting. We found the highestlevel of immunostaining in the leptotene-early zygoteneextracts. Staining declined during zygotene, and wasusually absent in the pachytene extracts. Our immuno-blot results are similar to those obtained by Terasawa et al. (1995) except that they observed a low level ofstaining in their pachytene extracts. This difference maybe due to the fact that our pachytene extracts are a mix-ture of early, mid- and late pachytene, and their extractmay have been of only early pachytene. This interpreta-tion is consistent with the results from our immunocyto-chemistry. We selected very early pachytene nuclei forthe immunocytochemical analysis and we did observeimmunostaining in the early pachytene nuclei.

Dmc1 foci and synapsis

The triploid tiger lily is able to undergo chromosomealignment, intimate pairing and reciprocal genetic ex-change; its only observed cytological difference from thecontrol was that approximately one-third of its axiallength was excluded from intimate association and recip-rocal genetic exchange owing to the lack of an availablepairing partner. Immunoblots done with extracts from thetriploid tiger lily gave the same trend as the diploid. Thegreatest signal was seen in extracts from leptotene-earlyzygotene, then it declined toward pachytene. We thencompared the numbers of chromosomal foci seen in ourcontrol diploid with that seen in the triploid nuclei. Atearly and mid-zygotene the triploid had somewhat morefoci than the diploid control. But by early pachytene the

Fig. 5 Distribution of foci at different prophase I stages. Themean numbers of foci for each stage, for each type of lily sample,are presented as a series of bar graphs. Solid bars values obtainedusing immune serum 1; hatched bars values obtained with im-mune serum 2; eZ early zygotene, mZ mid-zygotene, P earlypachytene. A Diploid L. tigrinum. For the nuclei reacted with im-mune serum 1 the standard deviations for the eZ, mZ and P sam-ples were 162, 161, and 79, respectively. Paired t-tests were done,and the observed reductions in mean number of foci from eZ tomZ, and mZ to P were statistically significant (P<0.005 andP<0.001, respectively). For immune serum 2 the standard devia-tions for the eZ, mZ and P samples were 51, 52 and 119, respec-tively. The observed, small differences in the stage-specific meanvalues were not statistically significant. B Triploid L. tigrinum.For the nuclei reacted with immune serum 1 the standard devia-tions for the eZ, mZ and P samples were 79, 102 and 166, respec-tively. All of these stage-specific differences in foci number werestatistically significant (P<0.001). For immune serum 2 the stan-dard deviations for the eZ, mZ and P samples were 109, 113 and78, respectively. The difference in number of foci from early zy-gotene to mid-zygotene was not statistically significant; the reduc-tion from mid-zygotene to early pachytene was (P<0.001). C Dip-loid species-hybrid, Black Beauty. For the nuclei reacted with im-mune serum 1 the standard deviations for the eZ, mZ and P sam-ples were 127, 96, and 71, respectively. All of the stage-specificdifferences in mean number of foci were statistically significant(P<0.001). For immune serum 2 the standard deviations for theeZ, mZ and P samples were 105, 42, and 52, respectively. The re-duction in number of foci from eZ to mZ was statistically signifi-cant (P<0.001), whereas the small reduction from mid-zygotene toearly pachytene was not

103

number of foci in the diploid and triploid was very simi-lar. Thus by the end of zygotene, the number of Dmc1foci present is proportional to the amount of intimatelypaired chromosome axis (which is the same in the dip-loid and triploid), and is not proportional to the totalamount of chromosome axis (which is 33% higher in thetriploid).

The diploid species-hybrid, with homeologous chro-mosomes, has relatively normal chromosome alignment,an intermediate reduction in SC formation, and a largereduction in reciprocal genetic exchange, presumablydue to a failure at the chromosome homology checkingstage. Our results with the species-hybrid suggest thatDmc1 foci are only stabilized if chromosome homologyis found at that site. In the species-hybrid the number offoci seen at early zygotene is similar to the number seenin the normal diploid control, but the total number of foci is dramatically reduced in the mid-zygotene nuclei.By early pachytene the number of foci in the hybrid isreduced to 45% of its early zygotene level, (and the hybrid has only 43% as many foci as the diploid controlat the same stage). A comparable result was seen withour immunoblots in which immunostaining was onlyseen in the early zygotene extract. It appears that re-duced chromosome homology leads to reduced numbersof stabilized Dmc1 foci.

Rockmill et al. (1995) suggested that both Dmc1 pro-tein and chromosome homology are required for axialassociations. Axial associations, in turn, are important tothe timely initiation of SC. Ultrastructural studies usingRad51 and/or Dmc1 antibodies support this conclusion(Anderson et al. 1997; Tarsounas et al. 1999). Our resultsare in agreement with this conclusion. The reduction inthe number of Dmc1 foci in our species-hybrid is similarto the reduction in SC formation (43%–67% reductionsin SC formation; Toledo et al. 1979, and Hasenkampfand Shull, unpublished results). A successful homologycheck appears to be needed to stabilize the Dmc1 foci ata site, and to expedite the timely initiation of SC.

Dmc1 foci and reciprocal genetic exchange?

In the three types of lily we have examined, there is avery good correspondence between the extent of stabi-lized Dmc1 foci and the extent to which intimate pairingoccurs. Thus one consequence of the homology checkappears to be the timely formation of SC. In S. cerevisiaeRad51 and Dmc1 also both appear to be important forachieving normal levels of reciprocal genetic exchange(Bishop et al. 1992; Shinohara et al. 1992), and Dmc1seems to have a role in ensuring that recombination will occur between nonsister homologous chromatids(Schwacha and Kleckner 1997). Given the need for theDmc1 protein for recombination, a successful homologycheck also is likely required to allow the other DNA interactions that give rise to reciprocal genetic exchangebetween homologous chromosomes. A successful Dmc1-mediated homology check might trigger chromatin

changes that would promote the ability of Rad51 to com-plete strand exchange between homologous chromo-somes (Masson and West 2001). In this view, the pri-mary role of Dmc1 is chromosome homology checking.Rad51 might be blocked from engaging in interhomologinteractions, until chromosome homology checking has occurred. If homology checking is successful in aspecific chromosome region, then perhaps chromatinchanges would allow Rad51 to bind with accessory proteins that could enhance the in vivo ability of Rad51to accomplish strand exchange (Dresser et al. 1997).

We propose that in the lily the locations where DNAinteractions between homologs give rise to reciprocal ge-netic exchange are a subset of the sites (or might even bedifferent sites) than the ones that give rise to the visiblefoci within which Dmc1 is functioning in homologychecking and in the initiation of SC formation. In lilies(diploid, triploid or species-hybrid) the number of ob-served Dmc1 foci, even at early pachytene, is at least 10×higher than the number of reciprocal genetic exchanges.

In the species-hybrid, the reduction in number of sta-bilized Dmc1 foci (45%) correlates well, and directly,with the observed reduction in SC formation. We assumethat the extent of stabilized foci represents the extent towhich the homeologous chromosomes of the two paren-tal species share regions of chromosome homology, and'pass' the homology check. Based on this assumption onemight then expect Black Beauty to have 45% as manychiasmata as the diploid control. However, chiasmataformation in the hybrid is much more drastically re-duced: Black Beauty has only ~10% of the level of re-ciprocal exchange seen in our normal, diploid control.This result suggests that a further, more stringent, levelof chromosome homology is required to allow reciprocalgenetic exchange, than is needed for timely SC initia-tion. The higher stringency for homology might be en-forced at the same site, but at a later step, downstream ofthe role of Dmc1/Rad51 in the double-strand pathway.

Alternatively Dmc1-mediated homology checks mightneed to be successful at several sites of a particular chro-mosome region (perhaps a chromosome arm) before pro-posed chromatin changes could occur that would promotethe ability of Rad51 to complete strand exchange betweenhomologous chromosomes. The higher homology de-mands seen for reciprocal genetic exchange might be accomplished quantitatively by requiring that several ad-jacent chromosome regions all pass the homology check.We favor this quantitative multi-locational check of ho-mology because demanding homology at more than onesite within a chromosome region likely more accuratelymeasures chromosome homology than would one highstringency test of DNA sequence complementarity at asingle site.

The pachytene checkpoint

Studies of certain types of meiotic mutants have pointedto the existence of a pachytene checkpoint (reviewed in

104

Roeder and Bailis 2000). It is proposed that the pachy-tene checkpoint is activated as recombination is initiated;once activated, nuclei cannot pass the checkpoint untilrecombination is completed, and the block is deactivat-ed. The Black Beauty hybrid proceeds from pachytene tometaphase I, despite the fact that it has very few recip-rocal genetic exchanges. If the pachytene checkpoint model is correct, then either there is no pachytene check-point in the lily, or recombination is not initiated in re-gions that fail in homology checking, or the completionof even a few reciprocal exchanges is enough to allowthe nucleus to pass the checkpoint.

In conclusion, we interpret our results to indicate thatthe primary role of Dmc1 is homology checking. In thelily, if the homology check is successful, then SC forma-tion can be initiated locally, and if several successful homology checks occur, regional chromatin events canbe triggered that would allow Rad51 to interact with theappropriate accessory factors and promote strand ex-change between homologous chromosomes. Thus ourstudies support the widely held view that Dmc1 func-tions in homology checking. In lilies it is likely that suc-cessful homology checks promote both SC initiation andreciprocal exchange, but that these post-homologychecking events are separable and can be regulated dif-ferentially.

Acknowledgements The authors wish to thank Dr. TomokoOgawa for the generous gift of Dmc1 immune serum. They alsowish to thank Dr. Dan Riggs for assistance with the immunoblot-ting portion of this study. This project was supported by a Re-search grant to C.A.H. from the Natural Science and EngineeringResearch Council of Canada, a University of Toronto fellowshipto S.G., and a research grant to K.H. from the Japanese Society forPromotion of Science Research for the Future Program (JSPS-RFTF9716001). The experiments reported here comply with thecurrent laws of the country in which the experiments were done.

References

Anderson LK, Offenberg HH, Verkuijlen WMHC, Heyting C(1997) RecA-like proteins are components of early meioticnodules in lily. Proc Natl Acad Sci U S A 94:6868–6873

Bishop D (1994) RecA homologs Dmc1 and Rad51 interact toform multiple nuclear complexes prior to meiotic chromosomesynapsis. Cell 79:1081– 1092

Bishop DK, Park D, Xi L, Kleckner N (1992) DMC1: a meiosis-specific yeast homolog of E. coli recA required for recombina-tion, synaptonemal complex formation, and cell cycle progres-sion. Cell 69:439–456

Davis L, Smith GR (2001) Meiotic recombination and chromo-some segregation in Schizosaccharomyces pombe. Proc NatlAcad Sci U S A 98:8395–8402

Dresser ME, Ewing DJ, Conrad MN, Domingues AM, Barstead R,Jiang H, Kodadek T (1997) DMC1 functions in a Saccharomy-ces cerevisiae meiotic pathway that is largely independent ofthe Rad51 pathway. Genetics 147:533–544

Emsweller SL, Uhring J (1966) Lilium X 'Black Beauty' – diploidand amphidiploid. Lily Yearbook 29:1–3

Erickson RO (1948) Cytological and growth-correlations in flowerbud and anther of Lilium longiflorum. Am J Bot 35:729–739

Franklin AE, McElver J, Sunjevaric I, Rothstien R, Bowen B,Cande WZ (1999) Three-dimensional microscopy of theRad51 recombination protein during meiotic prophase. PlantCell 11:809–824

Gasior SL, Olivares H, Ear U, Hari DM, Weichselbaum R, Bishop DK (2001) Assembly of RecA-like recombinases: dis-tinct roles for mediator proteins in mitosis and meiosis. ProcNatl Acad Sci U S A 98:8411–8418

Gupta RC, Folta-Stogniew E, Radding CM (1999) Human Rad51protein can form homologous joints in the absence of netstrand exchange. J Biol Chem 274:1248–1256

Gupta RC, Golub E, Bi B, Radding CM (2001) The synaptic activ-ity of HsDmc1, a human recombination protein specific tomeiosis. Proc Natl Acad Sci U S A 98:8433–8439

Hasenkampf CA, Qureshi M, Horsch A, Riggs CD (1992) Tempo-ral and spatial distribution of meiotin-1 in anthers of Liliumlongiflorum. Dev Genet 13:425–434

Holm PB (1977) The premeiotic DNA replication of euchromatinand heterochromatin in Lilium longiflorum. Carlsberg ResCommun 42:249–281

Kobayashi T, Kobayashi E, Sato S, Hotta Y, Miyajima N, Tanaka A, Tabata S (1993) Characterization of cDNAs in-duced in meiotic prophase in lily microsporocytes. DNA Res1:15–26

Lim DS, Hasty P (1996) A mutation in mouse Rad51 results in anearly embryonic lethal that is suppressed by a mutation in p53.Mol Cell Biol 16:7133–7143

Loidl J, Jones GH (1986) Synaptonemal complex spreading in Allium. I. Triploid A. sphaerocephalon. Chromosoma 93:420–428

Loidl J, Nairz K, Klein F (1991) Meiotic chromosome synapsis ina haploid yeast. Chromosoma 100:221–228

Masson J, West SC (2001) The Rad51 and Dmc1 recombinases: anon-identical twin relationship. Trends Biochem Sci 26:131–136

Masson J, Davies AA, Hajibagheri N, Benson F, Stasiak AZ,Stasiak A, West S (1999) The meiosis-specific recombinasehDmc1 forms ring structures and interacts with hRad51.EMBO J 18:6552–6560

Moens PB (1968) The fine structure of meiotic chromosome pair-ing in the triploid, Lilium tigrinum. Can J Genet Cytol 10:799–807

Moses MJ, Poorman PA (1981) Synaptonemal complex analysisof mouse rearrangements II. Synaptic adjustment in a tandemduplication. Chromosoma 81:519–535

Ogawa T, Yu A, Shinohara A, Egelman EH (1993) Similarity of the yeast Rad51 filament to the bacterial RecA filament.Science 259:1896–1899

Passy SI, Yu X, Li Z, Radding CM, Masson J-Y, West SC, Egelman E (1999) Human Dmc1 protein binds ssDNA as anoctameric ring. Proc Natl Acad Sci U S A 96:10684–10688

Riggs CD (1994) Purification and characterization of histone H1from meiotic cells of a monocotyledenous plant, Lilium longiflorum. Genome 37:736–741

Rockmill B, Sym M, Scherthan H, Roeder GS (1995) Roles for two RecA homologs in promoting meiotic chromosomesynapsis. Genes Dev 9:2684–2695

Roeder GS, Baillis JM (2000) The pachytene checkpoint. TrendsGenet 16:395–402

Schwacha A, Kleckner N (1997) Interhomolog bias during meioticrecombination: meiotic functions promote a highly differenti-ated interhomolog-only pathway. Cell 90:1123–1135

Shinohara A, Ogawa H, Ogawa T (1992) Rad51 protein involvedin repair and recombination in S. cerevisiae is a RecA-likeprotein. Cell 69:457–470

Shinohara M, Gasior SL, Bishop DK, Shinohara A (2000)Tid1/Rdh54 promotes co-localization of Rad51 and Dmc1 during meiotic recombination. Proc Natl Acad Sci USA97:10814–10819

Stack SM, Anderson LK, Sherman JD (1989) Chiasmata and recombination nodules in Lilium longiflorum. Genome32:486–498

105

Tarsounas M, Morita T, Pearlman R, Moens PB (1999) Rad51 andDmc1 form mixed complexes associated with mouse meioticchromosome cores and synaptonemal complexes. J Cell Biol147:207–219

Terasawa M, Shinohara A, Hotta Y, Ogawa H, Ogawa T (1995)Localization of RecA-like recombination proteins on chromo-somes of the lily at various meiotic stages. Genes Dev9:925–934

Toledo LA, Bennett MD, Stern H (1979) Cytological investigationsof the effect of colchicine on meiosis in Lilium hybrid cv.“Black Beauty” microsporocytes. Chromosoma 72:157–173

White PR (1963) The cultivation of animal and plant cells, 2ndedn. Ronald Press, New York

Yoshida K, Kondoh G, Matsuda Y, Habu T, Nishimune Y, Morita T (1998) The mouse RecA-like gene Dmc1 is requiredfor homologous chromosome synapsis during meiosis. MolCell 1:707–718