Variation in crossover interference levels onindividual chromosomes from human males

Jie Lian1,2,{, Yimeng Yin1,2,{, Maria Oliver-Bonet3, Thomas Liehr4, Evelyn Ko3, Paul Turek5,6,7,

Fei Sun1,2 and Renee H. Martin3,�

1Hefei National Laboratory for Physical Sciences at Microscale and 2School of Life Sciences, University of Science

and Technology of China, Hefei, Anhui 230026, China, 3Department of Medical Genetics, University of Calgary,

Calgary, Canada T2 N 4N1, 4Institute of Human Genetics and Anthropology, 07743 Jena, Germany, 5Department of

Urology, 6Department of Obstetrics and Gynecology and 7Department of Reproductive Sciences, University of

California San Francisco, San Francisco, CA 94143-1695, USA

Received April 7, 2008; Revised and Accepted May 20, 2008

Crossovers (COs) generated by homologous recombination ensure the proper segregation of chromosomesduring meiosis. COs exhibit interference, which leads to widely spaced COs along chromosomes. Strongpositive CO interference has been found in humans. However, little is known about the extent of humanCO interference. In this study, variations in CO interference over the entire human genome and among indi-viduals were analyzed by immunofluorescence combined with fluorescence in situ hybridization of testicularbiopsies from 10 control men. These methods allow for direct identification of the frequency and location ofCOs in specific chromosomes of pachytene cells. The strength of CO interference was estimated by fitting thefrequency distribution of inter-CO distances to the gamma model. Positive interference among CO onchromosomes was observed in these men, and the strength of inter-arm interference was significantly stron-ger than that for intra-arm CO. In addition, interference was observed to act across the centromere.Significant inter-individual and inter-chromosomal variations in the levels of interference were found, withsmaller chromosomes exhibiting stronger interference. Discontinuous chromosome regions (gaps) andunsynapsed chromosome regions (splits) in chromosome 9 had both cis and trans effects on CO interferencelevels. This is the first report that the interference level varies significantly across the whole genome and that,at least in the human male, anomalies in chromosome synapsis play an important role in altering CO inter-ference levels.

INTRODUCTION

Crossover (CO) formation is a crucial step in the establishmentof a physical connection between homologous chromosomesduring prophase I of meiosis, which facilitates proper chromo-some segregation. Improper segregation of homologs (i.e. non-disjunction) results in aneuploidy, the leading cause of fetalloss and birth defects in humans (1). In humans, positive inter-ference affects the distribution of COs along chromosomes(2–5). Positive CO interference, by definition, means thatthe occurrence of one CO discourages the formation of otherCOs in its vicinity (6). An obvious consequence of positive

interference would be that COs are more evenly spaced thanwould be expected with random placement across chromo-somes, a situation which reduces the risk of non-disjunction(7). Yeast mutants for which interference is absent show ahigh incidence of non-disjunction of homologous chromo-somes at the first meiotic division, indicating that interferenceis crucial to CO control and homolog segregation (8,9).Despite the clinical importance of CO interference, the mech-anisms of interference are still poorly understood (10).

CO interference has been examined by severalmathematical models. Among these, the gamma model hasbeen most commonly used for the analysis of interference

†The authors wish it to be known that, in their opinion, the first two authors should be regarded as joint First Authors.

�To whom correspondence should be addressed at: Department of Medical Genetics, University of Calgary, 3330 Hospital Dr, NW, Calgary, AB,Canada T2 N 1N4. Tel: þ1 4032207520; Fax: þ1 4032107931; Email: rhmartin@ucalgary.ca or feisun@ustc.edu.cn

# The Author 2008. Published by Oxford University Press. All rights reserved.For Permissions, please email: journals.permissions@oxfordjournals.org

Human Molecular Genetics, 2008, Vol. 17, No. 17 2583–2594doi:10.1093/hmg/ddn158Advance Access published on May 23, 2008

levels (2,11–13). The strength of interference among COs canbe calculated by the interference parameter y, which is esti-mated by fitting the frequency distribution of inter-CO dis-tances to the gamma model. When y ¼ 1, distribution ofinter-focus distances is exponential, representing random dis-tribution of COs along chromosomes and no interferencebetween CO foci. When y . 1, a positive interference is indi-cated, and as y increases, the interference level is stronger.Employing the gamma model, variations in interferencelevels across chromosomes and among individuals werefound in two systematic studies of human CO interference,on the basis of the Center d’Etude du PolymorphismeHumain (CEPH)-family genotyping data (2,3). Earlierstudies of human pedigree data also showed the variation ofCO interference levels. For example, smaller chromosomesexhibited stronger interference than larger chromosomes(14), and the level of interference on the p-arm tended to belower than that on the q-arm of chromosome 1 (15). Meioticrecombination and CO data in these early studies originatedfrom the construction of linkage maps, using geneticmarkers rather than precisely localized genetic exchanges, soCO interference was necessarily analyzed in chromosomalsegments and not at specific locations. There was no directevidence in these studies that interference showed heterogen-eity across the entire human genome.

Recently, CO positions in meiotic chromosomes have beenidentified precisely by immunofluorescence methods, usingantibodies against mut L homolog 1 (MLH1; marks thelocation of COs) and synaptonemal complexes (SCs; protein-aceous structures linking homologous chromosomes duringmeiotic prophase I) (16–18). Results from the application ofthese techniques suggested that there is a variation in COinterference (as measured by inter-CO distances) in humanmales (19,20). In the human male, chromosome synapsisanomalies such as discontinuous regions (gaps), and unsy-napsed regions (splits) in the SC (21,22) significantly alteredthe distribution of COs on SCs (23). These studies indicatethat incompletely formed SCs may affect the levels of COinterference in the human male.

In the present study, the distribution of meiotic COs in eachchromosome was identified by immunofluorescence methodscombined with centromere-specific multicolor fluorescencein situ hybridization (cenM-FISH) on testicular tissue from10 control men with normal spermatogenesis. The strengthof CO interference was measured by the interference par-ameter y of the gamma model, and then interference hetero-geneity across the entire human genome and amongindividuals was examined. In addition, the effects of SC for-mation on CO interference were determined by analyzinginterference strength in chromosome 9 (the SC most fre-quently observed to have gaps/splits in pachytene stagemeiotic cells).

RESULTS

An example of a pachytene SC spread, with identification ofindividual autosomes and cenM-FISH signals in the samecell, is shown in Figure 1. MLH1 foci in about 100 pachytene-stage spermatocytes were analyzed for each of the 10 men in

order to determine the distribution of COs. The observedinter-CO distances (expressed as percentage of the lengthof the SC) were fitted to the gamma distribution. TheKolmogorov–Smirnov test was used to examine thegoodness-of-fit for the observed frequency distribution ofinter-CO distances to the gamma distribution. The resultsshowed that the gamma model gives a very good fit tointer-CO distances along SC in most cases, except that somefits were less perfect because of the limited range of observable

Figure 1. Human pachytene spermatocyte with SCs shown in red, centromeresin blue and MLH1 foci in yellow (upper). Gaps in SC1 and SC9 are markedwith a ‘g’. Subsequent cenM-FISH analysis permits identification of individualchromosomes so that recombination (MLH1) foci can be analyzed for each SC(lower).

2584 Human Molecular Genetics, 2008, Vol. 17, No. 17

inter-CO distances or local chromatin properties (12). Theinterference parameter y, a measure of the strength of interfer-ence among COs, was then calculated from the gamma distri-bution. No interference parameter was estimated forchromosome 21, since it rarely has more than one CO.

Characterization of CO interference

Examples of the estimate of y for meiotic chromosomes areshown in Figure 2. The expected frequency distribution ofinter-CO distances (the solid curves), obtained from thegamma model, fits well to the observed inter-CO distances(the bars). Furthermore, the fit of inter-CO distances to thegamma distribution (indicating CO interference) was signifi-cantly better than the fit to the exponential distribution (corre-sponding to random distribution of COs) (Fig. 2). Allestimated interference parameters (y) were significantlygreater than 1 (P , 0.0001, likelihood-ratio tests), providingstrong evidence in favor of genome-wide positive CO interfer-ence in human males.

In an earlier study, the centromere was assumed to play arole in recombination that acted as a barrier to interference(24), but recent evidence from several species suggests thatCO interference acts across the centromere (2,25–27). Toverify the role of the centromere in CO interference inhuman males, the relationship between the distance from thecentromere to the nearest CO on each arm (of chromosomeswith at least one CO on each arm) was examined. The dis-tances were expressed as percentage of SC length. Pearsoncorrelation analysis showed a significant negative correlationbetween them (r ¼ 20.3748, P , 0.0001) (Fig. 3), suggestingthat when a CO occurred near the centromere in one arm, thenearest CO in the other arm tended to be farther away from thecentromere. Thus, CO interference does continue uninter-rupted across the centromere in human male meiosis.

Intra-chromosomal variation in CO interference

As described earlier, CO interference acts across the centro-mere, indicating that COs are not formed independently inthe two arms. Using pooled data for each chromosome from

Figure 2. Examples of frequency distributions of inter-CO distances on two long chromosomes (1 and 2) and two short chromosomes (19 and 20) from the sameindividual. The bars represent the observed relative frequencies of inter-CO distances. The solid curves correspond to the fitted distributions under the gammamodel, which shows a reasonable fit to these data (Kolmogorov–Smirnov Test). The estimates of the interference parameter y, u (SE), are obtained by fitting theobserved inter-CO distances to the gamma distribution. The dashed curves correspond to the fitted distributions under the exponential distributions which rep-resents no interference models (i.e. y ¼ 1). The fit of inter-CO distances to the gamma distribution is far better than the fit to the exponential distribution, demon-strating that the distribution of COs on chromosomes is not random, on the contrary, shows positive interference in humans.

Human Molecular Genetics, 2008, Vol. 17, No. 17 2585

the 10 individuals, the y of inter-CO distances on the p- andq-arms of a chromosome was estimated to evaluate intra-arminterference heterogeneity. Using SCs 1–7 that had at leasttwo COs on each arm (since SCs 8–22 rarely have morethan one CO on the p-arm), no significant difference in theinter-CO y was observed between the p- and q-arms (P ¼0.84, one-way ANOVA) (Fig. 4A). This is consistent with pre-vious results (3), and suggests that CO interference is constantalong the chromosome arms.

Subsequently, inter- and intra-arm interference y were com-pared to further assess the heterogeneity of intra-chromosomalinterference. Inter-arm CO interference y values were esti-mated by fitting trans-centromere inter-CO distances to thegamma model, from SCs with at least one CO on each arm.For evaluating intra-arm CO interference, the analysis wasrestricted to chromosomes with at least two COs in the samearm, using pooled inter-CO distances across the p- andq-arms for each chromosome. The fitted curves for the esti-mated y from inter- and intra-arm are shown in Figure 4B.A significantly higher interference level was found for inter-arm COs than for intra-arm COs (P , 0.0001, one-wayANOVA), providing strong evidence in favor of non-constantinterference across the whole chromosome. Furthermore, shortchromosomes exhibited higher interference levels for inter-arm COs when compared with longer chromosomes,whereas intra-arm interference strength was at similar levelsto those of the longer chromosomes (Fig. 4B). Therefore,the overall interference strength for short chromosomes wasstronger than for longer chromosomes.

Inter-chromosomal and inter-individual variationin CO interference

The y for each chromosome was estimated by pooled datafor the frequency distribution of inter-CO distances from all

individuals. A significant difference in the levels of interfer-ence over the entire human genome was observed, withsmaller chromosomes exhibiting higher levels of interferencethan larger chromosomes, with the exception of chromosome22 (Fig. 5). Furthermore, the heterogeneity of inter-chromosomal interference levels (Fig. 5) is remarkably paral-lel to that of inter-arm interference levels (Fig. 4), suggestingthat the difference in inter-arm interference levels may be themain contributor to variation in inter-chromosomal interfer-ence levels.

Inter-individual variation in interference was assessed byestimating y for each chromosome for each donor. Significantdifferences in y among individuals were found for chromo-somes 1, 3, 5–8, 13 and 15–20 (Fig. 6). In addition, thedegree of variation in interference levels was lower in thelarger chromosomes than in smaller chromosomes (P , 0.05for chromosomes 1, 3, 5–8, 15 and 17, whereas P , 0.0001for chromosomes 13, 16, 18, 19 and 20).

The role of the mature SC in CO interference

To evaluate the relationship between abnormal synapsis andCO interference, the effect of gaps/splits on the interferencelevels of chromosome 9 and other chromosomes was exam-ined. The inter-arm y for normally synapsed, gapped andsplit SC 9 is shown in Table 1. The interference levels forgapped and split SC 9 were significantly elevated comparedto normally synapsed chromosomes (for gapped SC 9,P , 0.0001; for split SC 9, P , 0.05), indicating that gaps/splits have a cis effect on chromosome 9’s interferencepattern. Furthermore, y values in Table 2 indicate that gaps/splits also exerted a trans effect on the interference patternof other chromosomes. SCs 13, 14, 16 and 18 in cells withsplit SC 9 and SC 20 in cells with gapped SC 9 were both sig-nificantly increased compared to normally synapsed chromo-somes, and SC 19 showed a significant reduction in the COinterference level in cells with a gapped SC 9.

To examine a possible trans effect among chromosomes,correlation analysis between the y for each chromosomefrom the 10 individuals was performed. Interestingly, signifi-cant positive correlations between the y of all the chromo-somes except chromosomes 2, 4, 6 and 8 were found(Fig. 7, P , 0.05, Pearson’s correlation analysis).Furthermore, this correlation always occurred among chromo-somes with similar SC length. For example, among longerSCs, y for chromosome 1 was positively correlated withchromosomes 3 and 5, and y for chromosome 3 was positivelycorrelated with chromosomes 5 and 12. Among shorter chromo-somes, chromosome 16 was positively correlated with chromo-somes 15 and 17, and chromosome 19 was positively correlatedwith chromosomes 17 and 20 (Fig. 7).

DISCUSSION

Our results provide the first direct evidence for significantvariation in the levels of CO interference across chromosomesand among human males. Variation in the inter-arm interfer-ence levels is by far the largest contributor to variation ininter-chromosomal CO interference. The heterogeneity of

Figure 3. The relationship between the two distances from centromere to thenearest COs on the p [d(p)] and q [d(q)] arms of those chromosomes that haveat least one CO on each arm. The distances were expressed as percentageof the SC length. A significant negative correlation was found between thetwo distances, indicating that interference acts across the centromere(r ¼ 20.3748, P , 0.0001).

2586 Human Molecular Genetics, 2008, Vol. 17, No. 17

inter-chromosomal interference also demonstrated a negativecorrelation between SC length and the global (chromosomal)level of interference, indicating that the strength of CO inter-ference might be modulated by the total length of the SC(28,29). In other words, if CO interference acts over a rela-tively fixed length of SC, short SCs will have fewer COsand higher levels of interference than longer SCs.

Based on the comparison of inter- and intra-arm CO dis-tances in pachytene cells from two men, Codina-Pascualet al. (19) reported that the level of trans-centromere interfer-ence was higher than that of intra-arm interference. In thesame paper, they also proposed that the first CO in nonacro-centric chromosomes is formed in subtelomeric regions ofeach arm, and exerts interference on the next CO to be estab-lished along the SC arm towards the centromere. They furtherproposed that when the interference from both sides of thecentromere coincides, it would result in a higher trans-centromere interference level than that for intra-arm.However, this hypothesis may not completely explain theheterogeneity seen in intra-chromosomal interference. First,although subtelomeric regions are hot regions for recombi-nation (5,19), it is not yet certain that CO formation is actuallyinitiated in subtelomeric regions and proceeds proximallytoward the centromere. Second, since pericentromericregions are usually heterochromatic and characterized byrepetitive DNA, reduced recombination frequencies, fewprotein-coding regions and gene silencing (30), the inhibitoryeffect of pericentromeric heterochromatin on recombinationcannot be excluded. Therefore, a complete explanation ofthe heterogeneity seen in intrachromosomal interference hasnot yet been elucidated.

Our results suggest that the mature SC plays an importantrole in CO interference in the human male. The SC is a tripar-tite structure with two parallel lateral elements [correspondingto the protein, synaptonemal complex protein 3 (SCP3)] and acentral element held together by transverse filaments (cor-responding to the protein SCP1) (31). It is not clear whichof the SC components play a major role in regulating CO

interference levels. In this study, interference levels for inter-arm COs were found to be significantly increased in bothgapped and split SC 9 compared with controls (Table 1). Inaddition, there was no significant difference in the interferencelevels between SCs with gaps and those with splits (Table 1;P ¼ 0.71). The effects of gaps and splits on interferencelevels on chromosomes exhibited both a cis and a transeffects.

To date, the molecular mechanisms of interference actionand variations are poorly understood. Over the past fewyears, several models have been proposed to explain CO inter-ference. One of the main models has been the ‘counting’model, in which recombination precursors mature to COs insuch a way that a fixed number of non-CO events occurs

Figure 4. Analysis of intra-chromosomal interference heterogeneity. (A) Estimates of y (+SE) for p- and q-arms of chromosomes 1–7 exhibiting at least twoCOs on each arm. No significant difference in y between the two arms. (B) Estimates of y for inter- (solid curves) and intra-arm (dashed curves), with data pooledacross individuals for each chromosome. The circles signify those chromosomes that have no y available (chromosomes 13, 14, 15, 21 and 22 for inter-arm;chromosome 21 for intra-arm), since they rarely have a CO in the p-arm or more than one CO in the whole chromosome, respectively. The inter-arm y wassignificantly higher than that for intra-arm, and short chromosomes exhibited more significant variation in y than did long chromosomes.

Figure 5. Estimates of y (+SE) for each chromosome, with data pooledacross individuals. Compared to long chromosomes, short chromosomesexhibited a higher level of CO interference. There is no y for chromosome21 because it rarely has more than one MLH1 focus.

Human Molecular Genetics, 2008, Vol. 17, No. 17 2587

Figure 6. Inter-individual variation in y (+SE) for each chromosome. For each plot, the horizontal dashed line corresponds to the mean y from 10 individuals.The P-values exhibit the significance of variance among individuals for each chromosome (one-way ANOVA). Note that there are no y for chromosomes 21 and22 due to insufficient data from each individual. Significant variations in y among individuals were found on chromosomes 1, 3, 5–8, 13 and 15–20.

2588 Human Molecular Genetics, 2008, Vol. 17, No. 17

Figure 6. Continued

Human Molecular Genetics, 2008, Vol. 17, No. 17 2589

between any two adjacent COs (32). As a consequence, thedistribution of CO precursors is obtained by counting earlyrecombination nodules and interference strength is assumedto be constant along the chromosomes (33). However, aphenomenon called ‘CO homeostasis’ [where both CO fre-quency and CO interference remain at wild-type levels whenthe numbers of double-strand breaks (DSBs) are reduced]has been demonstrated in yeast (34). These findings indicatethat CO frequency does not decrease, as expected by thecounting model, in order to maintain a fixed number ofnon-CO events. The data from the current study, as wellas those observed in Arabidopsis thaliana Chromosome 4(35), have shown that CO interference acts variably acrosschromosomes.

More recently, another interference model, known as themechanical stress model (36,37), proposes that the expansionof chromatin against some constraining elements brings

Figure 6. Continued

Table 1. Estimated interference parameters (y) in normally synapsed, gappedand split SC 9

Chromosome type Estimated inter-arm y

No. of intervals y SE

Normal 502 14.6 0.9Gapped 229 21.2� 2.0Split 33 23.2�� 5.7

�P , 0.0001, ��P , 0.05 compared to normally synapsed chromosomes.

Table 2. Estimated interference parameters (y) of affected SCs in cells withnormally synapsed, gapped and split SC 9

SC Chromosome type No. of intervals y SE

13 Normal 493 11.1 0.7Gapped 214 13.0 1.2Split 30 23.3� 6.0

14 Normal 487 11.0 0.7Gapped 207 13.6 1.3Split 29 21.2�� 5.5

16 Normal 547 11.2 0.7Gapped 232 11.8 1.1Split 34 24.5� 5.9

18 Normal 420 16.3 1.1Gapped 187 13.1 1.3Split 30 27.2�� 7.0

19 Normal 491 16.9 1.1Gapped 208 11.2�� 1.1Split 26 11.2 3.1

20 Normal 467 18.6 1.2Gapped 193 30.3� 3.1Split 31 12.1 3.0

�P , 0.001,��P , 0.05 compared to normally synapsed chromosomes.

2590 Human Molecular Genetics, 2008, Vol. 17, No. 17

Figure 7. The relationship between the interference parameters for each chromosome from 10 individuals. Significant positive correlations on y were foundbetween all chromosomes, indicating that there is inter-chromosomal effect on CO interference. A significant positive correlation between the y for chromosome1 and for chromosome 14, and for the distances from the MLH1 focus to the centromere in chromosome 21 was also found (Fig. 7H). These distances wereexpressed as percentage of the length of q-arms of chromosome 21.

Human Molecular Genetics, 2008, Vol. 17, No. 17 2591

about mechanical stress, which promotes the development ofCOs among DSBs. The occurrence of CO results in localstress relief at the affected site, which then spreads outwardsin both directions along the chromosome axes with decreasingstrength, thereby preventing the formation of a new CO nearbyas it travels. Moreover, SC nucleation and spreading would bepromoted by the stress effect as well. The model predicts that(i) a chromatin-constraining structure is required for generat-ing interference and conducting the interference signal;(ii) CO development events among nearby recombinationintermediates are not favored and (iii) the interference strengthwill decrease as the physical size of the SC increases in adistance-dependent manner. In the current study, significantvariation in inter-chromosomal interference was observed,lending support to some features of the stress model, such asthe distance-dependent manner of interference. It also pro-vides, as does the Codina-Pascual et al. (19) study, some evi-dence to support the prediction in the stress model that COinterference might be modulated by SC length. Evidencefrom Arabidopsis thaliana has also demonstrated that theinterference level between COs separated by a fixed geneticdistance is a function of physical distance (38). Further exper-iments are needed to support/refute these models or generatenew models for the characterization of CO interference andits variation.

In summary, our study provides the first detailed analysis ofheterogeneity in the level of CO interference over the entirehuman genome. Unambiguous evidence for significant inter-individual, inter- and intra-chromosomal variation in interfer-ence levels was observed in human males, with inter-arminterference levels providing the major contribution to inter-chromosomal CO interference. In addition, SC length modu-lates the interference strength at the genome-wide level,since (i) incomplete synapsis (gaps/splits in chromosome 9)had a cis and trans effects on the levels of CO interference,(ii) the interference levels of one chromosome were similarto levels of non-homologs with a similar SC length and (iii)the SC transverse filament protein SCP1 played a major rolein regulating interference levels among COs. The results inthis study could help to further understanding of current COinterference models and the role of the SC in CO interference.Finally and most importantly, a better understanding of COinterference may provide valuable insight into chromosomalnon-disjunction, which, when elucidated, may have far-reaching clinical benefits.

MATERIALS AND METHODS

Testicular sample collection

Testicular samples were obtained from non-chemotherapypatients undergoing orchiectomy for testicular or prostatecancer (n ¼ 3, age range: 47–81, Calgary, Canada) and vaso-vasostomy (n ¼ 7, age range: 35–61, University of California,CA, USA). These patients were ascertained for reasons unre-lated to meiotic defects or infertility. Testicular tissues werekept in phosphate buffered saline (pH 7.4) until use, and trans-ferred on ice to Calgary by air courier where appropriate. Wehave previously demonstrated that cold storage of testiculartissue does not affect recombination frequencies (39). Histo-

logical examination showed normal spermatogenesis in these10 donors (ages: 47–81 years). There was no significantdifference in the frequency of recombination or meioticabnormalities in cancer patients compared to men undergoingvasectomy reversals (20). Donors gave informed consent, andthis study received ethical approval from the institutionalreview boards at the University of Calgary and at theUniversity of California, CA, USA.

Fluorescence immunostaining and cenM-FISH

Slides with chromosome spreads were subjected to immuno-fluorescence staining as described previously (5). Primaryantibodies against the following proteins were used: SCP1(transverse filament proteins of the SC, 1:1000 dilution, agift from P. Moens, York University), SCP3 (lateral elementproteins of the SC, 1:250 dilution, a gift from T. Ashley,Yale University), MLH1 (marks the site of COs, 1:100dilution, Oncogene, San Diego, CA, USA) and CREST (Cal-cinosis, Raynaud’s phenomenon, Esophageal dysfunction,Sclerodactyly, Telangiectasia, marks the centromere; 1:100dilution, a gift from M. Fritzler, University of Calgary).These primary antibodies were detected using a cocktail ofsecondary antibodies (donkey antisera) conjugated with differ-ent fluorochromes: blue 1-amino-4-methylcoumarin-3 aceticacid (AMCA) and red Cy3 (1:100 dilution, Jackson Immuno-research, West Grove, PA, USA), green Alexa 488 and redAlexa 555 (1:125 dilution, Molecular Probes, Eugene, OR,USA). Primary and secondary antibodies were incubated over-night at 378C, and for 90 min at 378C, respectively. Slideswere examined on a Zeiss Axiophot epifluorescence micro-scope equipped with propidium iodide (PI), fluorescein iso-thiocynate (FITC) and 40,6-diamidino-2-phenylindole (DAPI)filters and a cooled charged coupled device (CCD) camera.Three fluorescent images (red, green, and blue) of the SC,CO sites, and CREST locations, respectively, were capturedusing Applied Imaging Cytovision 3.1 software (AppliedImaging Corporation, Santa Clara, CA, USA). Spreads werelocalized using a gridded finder slide.

Each pachytene-stage nucleus used for analysis met thefollowing criteria: (i) the correct numbers of bivalents (22autosomes and 1 sex body) were present, (ii) the SCs werenot overlapped or bent back on themselves, allowing all focito be scored and (iii) background was fairly low, allowingthe SCs to be distinguished from background noise and fromeach other. MLH1 signals were scored if they were distinctand localized on an SC. SCs were classified as normallysynapsed if they were completely linear, without anyobvious bubbles, forks, loops or irregularities.

After the analysis of the captured immunofluorescenceimages, cenM-FISH was carried out on the same spermato-cytes. This technique allowed simultaneous identification ofeach autosome. Techniques developed by Nietzel et al. (40)and Oliver-Bonet et al. (41) were modified to make use ofthe microwave-decondensed/codenatured FISH technique(42). Cells were microwave decondensed for 5 s in dithiothrei-tol (DTT) and 30 s in 3,5-diiodosalicylic acid, lithium salt(LIS)/DTT at medium power (550 W). Hybridization buffer(10% dextran sulfate, 2 � standard sodium citrate (SSC),55% formamide) was pre-warmed to 508C, added to the

2592 Human Molecular Genetics, 2008, Vol. 17, No. 17

cenM-FISH probes and warmed at 508C until all probe wasdissolved. Probes were applied to the slide, a glass coverslip was sealed in place with rubber cement, the probes andcells were microwave codenatured for 80 s at 1100 W andthe slide was incubated in a humid chamber at 378C forapproximately 24 h. A posthybridization wash (0.4 � SSC1% NP-40, 708C) was carried out, streptavidin–Alexa 647(Molecular Probes) solution was applied under a plasticcover slip, and the slide was incubated at 378C for 40 min ina humid chamber. The slide was washed, with constant agita-tion, for 10 min in 4 � SSC, air dried and mounted in DAPI.Cells previously analyzed by antibody immunostaining wererelocated, and six fluorescent images (blue, aqua, green,gold, red and far red) were captured for each cell, usingApplied Imaging Cytovision 3.1 software (Applied ImagingCorporation, Santa Clara, CA, USA).

Calculating the inter-CO distances on SCs

After cenM-FISH identification of each chromosome, imagesof the corresponding SC spreads were analyzed for CO distri-bution and SC length measurement, using MicroMeasure 3.3(available from the MicroMeasure website, http://www.colos-tate.edu/Depts/Biology/MicroMeasure). MicroMeasure is animage analysis application that allows collection of data fora wide variety of chromosome parameters from electronicallycaptured images (43). The inter-CO distances for eachchromosome were calculated by measuring the intervalsbetween adjacent COs along the SCs, and were expressed asa percentage of the SC length.

Estimation of the interference parameter y

Gamma distribution is widely used for the analysis of dis-tances between events along a linear axis (12). The interfer-ence parameter y is estimated by fitting the observedfrequency distribution of inter-CO distances to the gamma dis-tribution by the maximum likelihood method, using matlab 7.0software (Mathworks Inc., MA, USA). When y ¼ 1, gammadistribution shows exponential distribution, indicating nointerference among foci. With y , 1, gamma distribution isexponentially shaped and asymptotic to both the vertical andhorizontal axes, indicating negative interference. When y .1, gamma distribution assumes a mounded, skewed shapeand represents positive interference. As y increases, thestrength of positive interference is stronger, spacing COsmore evenly across the chromosomes.

The goodness-of-fit for a gamma distribution and the sig-nificance of the y value were also assessed by the followingtests using matlab 7.0 software (Mathworks Inc., MA, USA).First, a Kolmogorov–Smirnov test was performed to seehow well the observed frequency distribution of inter-focusdistances fits to the gamma distribution. Second, thegoodness-of-fit for a gamma distribution was further assessedby maximum likelihood estimates of the shape parameters andcomputer simulations for 10 000 inter-focus distances fromeach interference parameter, and then the fitted distributioncurves were obtained under the gamma model. Finally,likelihood-ratio tests were used to evaluate whether thegamma distribution or the exponential distribution fits the

data significantly better, that is, to determine whether ydiffers significantly from 1.0.

Statistical analysis

One-way analysis of variance (ANOVA) and subsequent pair-wise t-tests using the pooled variance from the ANOVA wereused to examine the differences in interference levels amongindividuals. The Student’s t-test was applied to examine thedifference in interference levels between normally synapsedand gapped/split SC 9 and between SCs other than SC 9both in cells with normally synapsed SC 9 and in cells withgapped/split SC 9. The relationships between the two sets ofdistances from centromere to the nearest COs on the p- andq-arms and between the y for each chromosome were evalu-ated using Pearson’s correlation analysis.

ACKNOWLEDGEMENTS

R.H.M. holds the Canada Research Chair in Genetics. F.S. andM.O.-B. are the recipients of a CIHR Strategic TrainingFellowship in Genetics, Child Development and Health.

Conflict of Interest statement. None declared.

FUNDING

Canadian Institutes of Health Research (MA7961 to R.H.M.);the National Basic Research Program of China (Project No.2007CB947400 to F.S.); the National Natural Science Foun-dation of China (No. 30770811 to F.S.).

REFERENCES

1. Page, S.L. and Hawley, R.S. (2003) Chromosome choreography: themeiotic ballet. Science, 301, 785–789.

2. Broman, K.W. and Weber, J.L. (2000) Characterization of humancrossover interference. Am. J. Hum. Genet., 66, 1911–1926.

3. Lin, S., Cheng, R. and Wright, F.A. (2001) Genetic crossover interferencein the human genome. Ann. Hum. Genet., 65, 79–93.

4. Schmitt, K., Lazzeroni, L., Foote, S., Vollrath, D., Fisher, E., Goradia, T.,Lange, K., Page, D. and Arnheim, N. (1994) Multipoint linkage map ofthe human pseudoautosomal region, based on single-sperm typing: dodouble crossovers occur during male meiosis? Am. J. Hum. Gen., 55,423–430.

5. Sun, F., Oliver-Bonet, M., Liehr, T., Starke, H., Ko, E., Rademaker, A.,Navarro, J., Benet, J. and Martin, R.H. (2004) Human male recombinationmaps for individual chromosomes. Am. J. Hum. Genet., 74, 521–531.

6. Muller, H.J. (1916) The mechanisms of crossing-over. Am. Nat., 50, 193–221.

7. Coop, G. and Przeworski, M. (2007) An evolutionary view of humanrecombination. Nat. Rev. Genet., 8, 23–34.

8. Chua, P.R. and Roeder, G.S. (1994) Tam1, a telomere-associated meioticprotein, functions in chromosome synapsis and crossover interference.Genes Dev., 11, 1786–1800.

9. Sym, M. and Roeder, G.S. (1994) Crossover interference is abolished inthe absence of a synaptonemal complex protein. Cell, 79, 283–292.

10. Hillers, K.J. (2004) Crossover interference. Curr. Biol., 14, R1036–R1037.

11. de Boer, E., Dietrich, A.J., Hoog, C., Stam, P. and Heyting, C. (2007)Meiotic interference among MLH1 foci requires neither an intact axialelement structure nor full synapsis. J. Cell Sci., 120, 731–736.

12. de Boer, E., Stam, P., Dietrich, A.J., Pastink, A. and Heyting, C. (2006)Two levels of interference in mouse meiotic recombination. Proc. NatlAcad. Sci. USA, 103, 9607–9612.

Human Molecular Genetics, 2008, Vol. 17, No. 17 2593

13. McPeek, M.S. and Speed, T.P. (1995) Modeling interference in geneticrecombination. Genetics, 139, 1031–1044.

14. Collins, A., Forabosco, P., Lawrence, S. and Morton, N.E. (1995) Anintegrated map of chromosome 9. Ann. Hum. Genet., 59, 393–402.

15. Forabosco, P., Collins, A. and Morton, N.E. (1995) Integration of genemaps: updating chromosome 1. Ann. Hum. Genet., 59, 291–305.

16. Anderson, L.K., Reeves, A., Webb, L.M. and Ashley, T. (1999)Distribution of crossing over on mouse synaptonemal complexes usingimmunofluorescent localization of MLH1 protein. Genetics, 151, 1569–1579.

17. Baker, S.M., Plug, A.W., Prolla, T.A., Bronner, C.E., Harris, A.C., Yao,X., Christie, D.M., Monell, C., Arnheim, N., Bradley, A. et al. (1996)Involvement of mouse Mlh1 in DNA mismatch repair and meioticcrossing over. Nat. Genet., 13, 336–342.

18. Roeder, G.S. (1997) Meiotic chromosomes: it takes two to tango. GenesDev., 11, 2600–2621.

19. Codina-Pascual, M., Campillo, M., Kraus, J., Speicher, M.R., Egozcue, J.,Navarro, J. and Benet, J. (2006) Crossover frequency and synaptonemalcomplex length: their variability and effects on human male meiosis. Mol.Hum. Reprod., 12, 123–133.

20. Sun, F., Oliver-Bonet, M., Liehr, T., Starke, H., Turek, P., Ko, E.,Rademaker, A. and Martin, R.H. (2006) Variation in MLH1 distribution inrecombination maps for individual chromosomes from human males.Hum. Mol. Genet., 15, 2376–2391.

21. Codina-Pascual, M., Oliver-Bonet, M., Navarro, J., Campillo, M., Garcia,F., Egozcue, S., Abad, C., Egozcue, J. and Benet, J. (2005) Synapsis andmeiotic recombination analyses: MLH1 focus in the XY pair as anindicator. Hum. Reprod., 20, 2133–2139.

22. Sun, F., Trpkov, K., Rademaker, A., Ko, E. and Martin, R.H. (2005)Variation in meiotic recombination frequencies among human males.Hum. Genet., 116, 172–178.

23. Sun, F., Oliver-Bonet, M., Liehr, T., Starke, H., Trpkov, K., Ko, E.,Rademaker, A. and Martin, R.H. (2005) Discontinuities and unsynapsedregions in meiotic chromosomes have a cis effect on meioticrecombination patterns in normal human males. Hum. Mol. Genet., 14,3013–3018.

24. Mather, K. (1938) Crossing-over. Biol. Rev., 13, 252–292.25. Colombo, P.C. and Jones, G.H. (1997) Chiasma interference is blind to

centromeres. Heredity, 79 (Pt 2), 214–227.26. Drouaud, J., Camilleri, C., Bourguignon, P.Y., Canaguier, A., Berard, A.,

Vezon, D., Giancola, S., Brunel, D., Colot, V., Prum, B. et al. (2006)Variation in crossing-over rates across chromosome 4 of Arabidopsisthaliana reveals the presence of meiotic recombination ‘hot spots’.Genome Res., 16, 106–114.

27. Lhuissier, G.G., Offenberg, H.H., Wittich, P.E., Vischer, N.O. andHeyting, C. (2007) The mismatch repair protein MLH1 marks a subset ofstrongly interfering crossovers in tomato. Plant Cell, 19, 862–876.

28. Kaback, D.B., Barber, D., Mahon, J., Lamb, J. and You, J. (1999)Chromosome size-dependent control of meiotic reciprocal recombination

in Saccaromyces cerevisiae: the role of crossover interference. Genetics,152, 1475–1486.

29. Kleckner, N., Storlazzi, A. and Zickler, D. (2003) Coordinate variation inmeiotic pachytene SC length and total crossover/chiasma frequency underconditions of constant DNA length. Trends Genet., 19, 623–628.

30. Dillon, N. (2004) Heterochromatin structure and function. Biol. Cell., 96,631–637.

31. Schmekel, K. and Daneholt, B. (1995) The central region of thesynaptonemal complex revealed in three dimensions. Trends Cell Biol., 5,239–242.

32. Foss, E., Lande, R., Stahl, F.W. and Steinberg, C.M. (1993) Chiasmainterference as a function of genetic distance. Genetics, 133, 681–691.

33. Stahl, F.W., Foss, H.M., Young, L.S., Borts, R.H., Abdullah, M.F. andCopenhaver, G.P. (2004) Does crossover interference count inSaccharomyces cerevisiae? Genetics, 168, 35–48.

34. Martini, E., Diaz, R.L., Hunter, N. and Keeney, S. (2006) Crossoverhomeostasis in yeast meiosis. Cell, 126, 285–295.

35. Drouaud, J., Mercier, R., Chelysheva, L., Berard, A., Falque, M., Martin,O., Zanni, V., Brunel, D. and Mezard, C. (2007) Sex-specific crossoverdistributions and variations in interference level along Arabidopsisthaliana chromosome 4. PLoS Genet, 3, e106.

36. Borner, G.V., Kleckner, N. and Hunter, N. (2004) Crossover/noncrossoverdifferentiation, synaptonemal complex formation, and regulatory surveillanceat the leptotene/zygotene transition of meiosis. Cell, 117, 29–45.

37. Kleckner, N., Zickler, D., Jones, G.H., Dekker, J., Padmore, R., Henle, J.and Hutchinson, J. (2004) A mechanical basis for chromosome function.Proc. Natl Acad. Sci., 101, 12592–12597.

38. Stahl, R.W., Foss, H.M., Young, L.S., Borts, R.H., Abdullah, M.F. andCopenhaver, G.P. (2004) Does crossover interference count inSaccaromyces cerevisiae? Genetics, 168, 35–48.

39. Sun, F., Trpkov, K., Rademaker, A., Ko, E., Barclay, L., Mikhaail-Philips,M. and Martin, R.H. (2004) The effect of cold storage on recombinationfrequencies in human male testicular cells. Cytogenet. Genome Res., 106,39–42.

40. Nietzel, A., Rocchi, M., Starke, H., Heller, A., Fiedler, W., Wlodarska, I.,Longarevic, I.F., Beensen, V., Claussen, U. and Liehr, T. (2001) A newmulticolor-FISH approach for the characterization of markerchromosomes: centromere-specific multicolour-FISH (cenM-FISH). Hum.Genet., 108, 199–204.

41. Oliver-Bonet, M., Liehr, T., Nietzel, A., Heller, A., Starke, H., Claussen,U., Codina-Pascual, M., Pujol, A., Abad, C., Egozcue, J. et al. (2003)Karyotyping of human synaptonemal complexes by cenM-FISH.Eur. J. Hum. Genet., 11, 879–883.

42. Ko, E., Rademaker, A. and Martin, R.H. (2001) Microwavedecondensation and codenaturation: a new methodology to maximizeFISH data from donors with very low concentrations of sperm. Cytogenet.Cell Genet., 95, 143–145.

43. Reeves, A. (2001) MicroMeasure: a new computer program for thecollection and analysis of cytogenetic data. Genome, 44, 439–443.

2594 Human Molecular Genetics, 2008, Vol. 17, No. 17