Meiosis Drives Extraordinary Genome Plasticity in the Haploid Fungal Plant Pathogen Mycosphaerella graminicola Authors: Alexander H. J. Wittenberg, Theo A. J. van der Lee, Sarrah Ben M'Barek, Sarah B. Ware, Stephen B. Goodwin, Andrzej Kilian, Richard G. F. Visser, Gert H. J. Kema, Henk J. Schouten Published: 2009, PLOS ONE DOI: 10.1371/journal.pone.0005863 ABSTRACT Meiosis in the haploid plant-pathogenic fungus Mycosphaerella graminicola results in eight ascospores due to a mitotic division following the two meiotic divisions. The transient diploid phase allows for recombination among homologous chromosomes. However, some chromosomes of M. graminicola lack homologs and do not pair during meiosis. Because these chromosomes are not present universally in the genome of the organism they can be considered to be dispensable. To analyze the meiotic transmission of unequal chromosome numbers, two segregating populations were generated by crossing genetically unrelated parent isolates originating from Algeria and The Netherlands that had pathogenicity towards durum or bread wheat, respectively. Detailed genetic analyses of these progenies using high-density mapping (1793 DArT, 258 AFLP and 25 SSR markers) and graphical genotyping revealed that M. graminicola has up to eight dispensable chromosomes, the highest number reported in filamentous fungi. These chromosomes vary from 0.39 to 0.77 Mb in size, and represent up to 38% of the chromosomal complement. Chromosome numbers among progeny isolates varied widely, with some progeny missing up to three chromosomes, while other strains were disomic for one or more chromosomes. Between 15–20% of the progeny isolates lacked one or more chromosomes that were present in both parents. The two high-density maps showed no recombination of dispensable chromosomes and hence, their meiotic processing may require distributive disjunction, a phenomenon that is rarely observed in fungi. The maps also enabled the identification of individual twin isolates from a single ascus that shared the same missing or doubled chromosomes indicating that the chromosomal polymorphisms were mitotically stable and originated from nondisjunction during the second division and, less frequently, during the first division of fungal meiosis. High genome plasticity could be among the strategies enabling this versatile pathogen to quickly overcome adverse biotic and abiotic conditions in wheat fields. INTRODUCTION Fungi provide attractive model systems to analyze processes that occur during meiosis. Many fungi are haploid, which greatly simplifies genetic studies. Furthermore, complete recovery of the meiotic products, or tetrads, is possible in ascomycete fungi, and these tetrads can be analyzed for the segregation of genetic markers. Tetrad analyses of Aspergillus nidulans and Neurospora crassa have been instrumental in answering fundamental questions concerning meiosis [1] – [3] . Here we describe genetic studies in another filamentous ascomycete, Mycosphaerella graminicola (asexual stage: Septoria tritici ). This fungus causes septoria tritici blotch (STB) of wheat, a disease characterized by necrotic blotches on the foliage. These blotches contain asexual (pycnidia) and sexual (pseudothecia) fructifications. M. graminicola represents an intriguing model for fundamental genetic studies of plant-pathogenic fungi. Field isolates of this pathogen usually have 18–21 chromosomes, the highest number reported among ascomycetes. Furthermore, these chromosomes have an extraordinary size range, varying from 0.39 to 6.09 Mb [4] . Genome plasticity - comprising processes such as inversions, deletions, insertions and translocations that translate into chromosome length polymorphisms (CLPs) as well as chromosome number polymorphisms (CNPs) - results in a genome size that varies between 32 and 40 Mb, similar to other filamentous ascomycetes [4] – [9] . M. graminicola has an active sexual cycle under natural conditions, which is an important driver of STB epidemics and results in high genetic diversity of populations in the field [10] – [12] . Analyses of a cross between two M. graminicola strains that originated from bread wheat fields in The Netherlands resulted in the first genetic linkage map of a Mycosphaerella species [13] , [14] . Although this map was a major milestone, the anonymous AFLP and RAPD markers complicated integration of genetic data sets. In addition, the number of markers was limited and the map resolution was too low to assess the complications anticipated during meiosis due to the CLPs and CNPs commonly observed among M. graminicola isolates [9] . The exact origin and maintenance of CNPs and CLPs are not known. A likely hypothesis is that they can be generated or lost during meiosis. Recombination between chromosomes that differ in length could give rise to derivatives with CLPs [15] . Nondisjunction during meiosis I or II would generate CNPs. To test these hypotheses, we used the recently developed Diversity Arrays Technology (DArT) for the first time on a haploid fungal genome [16] – [20] . The parallel genotyping of progeny isolates using several thousands of DNA fragments spotted on a microarray and subsequent analysis resulted in one of the most dense genetic linkage maps currently available for a fungus. This enabled high-resolution genetic linkage analyses to study the meiotic processing of CNPs and CLPs as well as the generation of new genome plasticity in M. graminicola . We frequently observed the loss of one or more chromosomes, disomy and translocations. This extraordinary genome plasticity helps to explain the high genetic diversity observed within natural populations of this fungus and most likely facilitates rapid adaptation to changing environments. MATERIALS AND METHODS We used three isolates of M. graminicola : IPO323 and IPO94269 were isolated from bread wheat in the Netherlands and IPO95052 was isolated from durum wheat in Algeria. Isolate IPO323 was crossed to both IPO94269 and IPO95052 using a previously developed in planta protocol [10] , resulting in 68 and 148 progeny, respectively. All progeny isolates were collected and analyzed individually. DNA of parents and progeny was isolated using the Wizard Genomic DNA purification kit (Promega Madison, WI), starting with approximately 10 mg of lyophilized spores. Tables S1 and S2 provide an overview of the progeny isolates used in this study. DART PROCEDURE Generation of genomic representations, library construction, target preparation and image analysis were essentially performed as described previously [17] , [18] , with the modifications described by Wittenberg et al. [16] . The adapter and primer oligonucleotide sequences used in this study are listed in Table S3 . For details see Text S1 . NOMENCLATURE OF MARKERS AFLP markers were designated by the primer combination used for the amplification and the approximate length of the generated fragment [14] . For both AFLP and DArT markers the prefix A or B indicated the phase of the marker; those originating from parent IPO323 had the prefix A while markers from parent IPO95052 were indicated by the prefix B. DArT markers identified in cross IPO323×IPO94269 originating from isolate IPO95052 could be assigned the prefix A or B, as IPO94269 was not used for the library construction. Markers segregating in both populations received the prefix C. In addition, DArT markers were designated by the enzyme combination used for complexity reduction (BamHI, MseI and RsaI: BMR or HindIII, MseI, RsaI: HMR), the 384-well plate number and the position of the fragment in that plate (i.e., AHMR_04I09). Recently, 23 SSR loci were identified in M. graminicola , 21 of which could be positioned on the existing linkage map along with two previously published SSR loci [21] , [22] . The newly generated DArT markers were used to integrate the new IPO323×IPO94269 map with the existing map of that population [14] . Moreover, six of these SSRs also differentiated the parents of the second mapping population. To enable the mapping of these SSRs in the IPO323×IPO95052 progeny, amplification reactions were performed as described by Goodwin et al. [21] . SELECTION OF UNIQUE SEGREGATION PATTERNS AND MERGING OF TWIN ISOLATES The binary scores of polymorphic markers were converted to the correct allelic phase based on the scores of the parents. A Perl script was written that grouped loci with identical segregation patterns after disregarding unknown scores. The marker with the highest call rate (percentage of scored individuals) was selected as a representative for each group. The script also calculated the call rate for each individual genotype and the global call rate for the whole dataset. Individual genotypes were incorporated into the scoring table when at least 95% of the grouped markers could be scored. In M. graminicola , twin progeny isolates arise from the mitotic division that follows meiosis II in the ascus, resulting in four pairs of genetically identical ascospores. Although the random-ascospore progenies that resulted from the crossing protocol minimized the isolation of twin isolates, the large number of markers identified identical progeny efficiently. These were used to calculate the reproducibility of the different marker types and were merged before the mapping analyses. CONSTRUCTION AND COMPARISON OF THE LINKAGE MAPS The genetic linkage maps of the individual crosses as well as the bridge map were constructed with the software package JoinMap 3.0 [23] . A detailed description of the mapping process for the individual maps is given in Text S1 . The use of IPO323 in both crosses enabled the efficient generation of an integrated bridge map of the M. graminicola genome. The bridge map was used to compare the order of the loci in the constructed IPO323×IPO94269 and IPO323×IPO95052 maps. We used MapChart 2.2 [24] for the graphical representation of the genetic linkage maps. EVALUATION OF LOSS OR GAIN OF CHROMOSOMES We used graphical genotyping to compare the marker scores (A or B) and the phase (A or B) of the markers, which enabled us to identify whether each marker was present or absent in a particular progeny isolate. In cases where a linkage group (LG) was constructed from both marker types and a specific progeny isolate lacked all of these markers, we concluded that the isolate missed that LG. In cases where a LG was constructed from both marker types and a specific progeny isolate was scored present for all markers, we concluded that the isolate had an extra copy, derived from the other parent, of that particular LG. Hence, chromosome polymorphisms in progeny isolates were determined in silico if A and B markers that were assigned to a specific LG were always absent or present in a particular progeny. PCR VERIFICATION OF LOSS AND GAIN OF CHROMOSOMES DNA samples of the parental isolates (IPO323, IPO94269 and IPO95052), progeny isolates that showed absence of specific LGs by graphical genotyping ( Table S4 ) and two control progeny without these aberrations were used as templates in the PCR reactions. PCR was performed using SSR markers and specific primer pairs developed from the sequenced DArT markers located on the missing linkage groups ( Tables S5 and S6 ). To assure that absence of an amplicon was not caused by PCR failure, a positive PCR control was included that should be present in all parents and progeny that were tested. The SSR marker loci ac-0007 (LG8) and gga-0001 (LG12) were amplified in combination with the PCR control SSR locus ag-0003 (LG2). For the amplicons derived from the DArT marker sequences, the DArT fragments CABMR_07D07 (129 bp; LG1) or AHMR_08O09 (728 bp; LG15) served as positive PCR controls. PCR reactions were performed in a total volume of 20 µl containing 20 ng of genomic DNA, 1×PCR buffer (Roche), 1 µl of each of the forward and reverse primers used as a control (2 µM), 2 µl of each forward and reverse primer (2 µM), 0.8 µl of dNTPs (5 mM) and 0.2 µl of Taq DNA polymerase (5 U/µl). Amplification conditions were as follows: 94°C for 2 min, 12 cycles of 94°C for 30 sec, 66°C for 30 sec minus 1°C per cycle, 72°C for 30 sec; 27 cycles of 94°C for 30 sec, 53°C for 30 sec, 72°C for 30 sec; 72°C for 7 min, followed by a cooling-down step to 10°C. The SSR amplicons were separated on 6% non-denaturating acrylamide gels using a Mega-Gel Dual High-Throughput Vertical Electrophoresis Unit (CBS Scientific, Del Mar, California, USA). Amplicons based on the DArT sequences were separated on 2.5% agarose gels. RESULTS Among the 68 progeny isolates from the M. graminicola IPO323×IPO94269 cross, 1042 new DArT markers were obtained. The DArT markers were added to the first genetic linkage map of M. graminicola [14] , consisting of 271 AFLP markers, 57 RAPD markers and two markers for the biological traits avirulence ( Avr ) and mating type ( mat ). Twenty-five SSR markers also were added to the combined linkage map ( Table S7 ) [21] . For the 148 progeny isolates of the M. graminicola IPO323×IPO95052 cross, 1154 DArT markers were obtained that were combined with six SSR markers and the markers for the two biological traits ( Table S8 ). After analysis of the marker data, 31 twins were detected in the M. graminicola progenies ( Table S9 ). These twins result from the mitotic division that follows meiosis II in the ascus. The twin data enabled the dissection of mitotic or meiotic events that drive the generation of CLPs and CNPs. Eventually, the merged scoring tables comprised 60 individuals for the IPO323×IPO94269 cross and 125 individuals for the IPO323×IPO95052 cross ( Table S10 ). Because twins can be regarded as biological replicates, they also were used to evaluate the reproducibility of the marker scores for the different marker technologies. In our study, DArT and AFLP markers appeared to be more reproducible than the RAPD markers. Therefore, RAPD markers were excluded to improve the quality of the maps. Although the reproducibility for both DArT and AFLP was very high, the frequency of double crossovers in the final maps was much lower for DArT than for AFLP markers (0.24% compared to 0.96%), indicating the superior reliability of the DArT markers. CONSTRUCTION AND COMPARISON OF THE LINKAGE MAPS The combined genetic linkage maps contain 2078 markers comprising 1793 DArT, 258 AFLP, and 25 SSR DNA markers, plus the two markers that co-segregate with the biological traits Mat and Avr ( Table S11 ). The grouping and the order of the markers in the M. graminicola IPO323×IPO94269 cross were highly similar to those in the previous maps [14] , [21] . Compared to the previous map both new maps span a considerably larger part of the genome. In both crosses close to 99% of the segregating markers were reliably positioned, indicating that the current genetic linkage maps cover the complete genome. The new genetic linkage map of the IPO323×IPO94269 cross is 638 cM longer than the first linkage map, and spans 1854 cM with 1317 markers on 451 unique map positions, with an average distance of 4.1 cM between the markers ( Table S12 ). Nearly all markers (98.2%) were positioned on 24 LGs. Some of the smaller LGs that were observed in the first map merged with other LGs [14] : 10 LGs in the first map merged into five larger LGs, while six small new LGs were formed. For example, LGs 3 and 4 in the first map merged with LGs 22 and 17, respectively, in the new map. The order of the AFLP markers in the first and new map remained similar, although more AFLP markers were positioned in the latter (223 vs. 258 out of 271, representing 82.3% and 95.2%, respectively). The genetic linkage map of the M. graminicola isolate IPO323×IPO95052 cross spans 1946 cM and contains 1144 markers on 486 unique map positions on 23 LGs (comprising 98.5% of the generated markers), with an average distance of 4.0 cM between the markers ( Table S12 ). We also constructed a bridge map to compare the individual linkage maps using markers that segregated in both mapping populations. The resulting integrated map spans 1435 cM (∼75% of both individual maps) and contains 372 markers on 251 unique map positions. A total of 22 LGs from each of the individual crosses was aligned with the bridge map, and the marker order was similar to those on the two individual genetic maps ( Figure 1 and Figure S1 ). The 21 LGs in the bridge map is close to the estimated number of chromosomes based on electrophoretic and cytological karyotyping [4] , [14] and is identical to the number of chromosomes of the finished genome sequence ( http://genome.jgi-psf.org/Mycgr3/Mycgr3.home.html ) ( Table S13 ). Common markers are shown in bold and start with the prefix C, SSR markers are shown in blue and markers that are translocated in red. DArT markers were named according to phase of the marker (A = IPO323, B = IPO95052 or IPO94269), complexity reduction method used (BMR or HMR), and location in the spotting plate (e.g. BBMR_15L11). LG and AFLP nomenclature is according to Kema et al., 2002. Segregation distortion of the markers is indicated with * (P<0.05), ** (P<0.01), *** (P<0.005) or **** (P<0.001). TRANSLOCATIONS We identified eight DArT markers that were positioned very differently in the two maps, which is indicative of translocations. They represented five translocations between isolate IPO323 and either IPO94269 or IPO95052 and involved four inter-LG and one intra-LG translocations (CBBMR_14G17 in LG 6) ( Table S14 ). Another translocation between IPO323 and IPO94269 involved an SSR locus [21] that segregated in a diploid fashion in the isolate IPO323×IPO94269 cross (1∶1∶1∶1 ratio, χ 2  = 1.25, 0.25