Molecular cytogenetic analysis of ribosomal RNA genes by FISH has been performed in many plants, yet studies spanning an extensive polyploid series and encompassing most of the species within a genus are quite rare. In the strawberry genus Fragaria, prior to our initial report on F. vesca subsp. vesca 'Hawaii 4' in Shulaev et al. , only one unspecified accession of one diploid species, F. vesca, had been studied . In the present investigation, nine of the 12 known diploid species as well as several polyploid taxa were studied for genomic distribution of rDNA sites. In total, we extended chromosomal localization of both 5S and 25S rDNA clusters via FISH to 33 accessions representing 25 taxa (species and subspecies), covering ploidy levels from diploid to decaploid.
In all the Fragaria species and subspecies examined here, the 5S rDNA signals were displayed in proximal regions of chromosome short arms, while the 25S rDNA signals were on terminal chromosomal regions, indicating that the chromosomal positions of rDNA sites are highly conserved across Fragaria. Yet, as detailed below, marked variations in signal intensity were observed among diploid taxa, and intriguing instances and patterns of signal loss were observed, both within and between ploidy levels.
Typical and exceptional distribution patterns of 5S and 25S rDNA sites among diploid strawberry species
Among and within diploid taxa, the observation of two 5S and six 25S rDNA sites by FISH analysis (Figure 1) was generally consistent with previous findings in F. vesca [27, 28]. Thus, at the diploid level, one copy of the basic (x = 7) Fragaria genome is typified by the detectable presence of one 5S locus and three 25S loci, distributed such that one chromosome is "double-marked" by a proximal 5S rDNA locus and a terminal 25S rDNA locus, and two chromosomes are "single-marked" by terminal 25S rDNA loci (Figure 2A). By reasonable inference from its typicality, the described pattern may represent that of the ancestral Fragaria genome.
Among diploids, the greatest departures from the typical pattern were two distinct instances of rDNA site loss, each involving a different locus. First, absence of the 25S rDNA signals from the S1 pair left both accessions of F. nilgerrensis with only four 25S signals (e.g., CFRA 1358 in Figure 2F). This divergence of F. nilgerrensis from the other diploid species contributes to its status as a well-differentiated evolutionary unit, as suggested by the phylogenetic studies of Rousseau-Gueutin et al.  and Harrison et al. , and also by the sterility of the hybrids resulting from its crosses with other Fragaria species . Second, F. vesca 'Yellow Wonder' also displayed only four 25S signals, but the site losses were from the S2 pair (Figure 2C), also making 'Yellow Wonder' distinct among diploid taxa, including the other studied genotypes of F. vesca subsp. vesca. Thus, diminution of 25S rDNA site number from six to four has occurred at least twice among diploid Fragaria.
Polymorphisms were also detected in size and intensity of 25S rDNA FISH signals, which we described as "major" (large and strong) and "minor" (small and weak) signals, between different loci in the various diploid taxa. Yet of the three typically marked chromosome pairs, no one pair had consistently the brightest or least bright 25S signals across the diploid genotypes. Thus, in contrast to the conservation of a typical pattern of rDNA site distribution, the allocation pattern of 25S rDNA signal intensities among chromosome pairs varied among diploid genotypes to an extent that precluded typification. Furthermore, polymorphism in signal intensity and/or presence versus absence was observed between homologous 25S rDNA sites in many diploid genotypes. Signal absences from one but not both members of a typically marked pair occurred in three cases: F. nipponica (Figure 2E), F. iinumae (Figure 2D), and a genotype ('Pawtuckaway') of F. vesca subsp. americana (Figure 2B). The latter pattern of variability between homologs was consistently seen on separately prepared slides suggesting that this variability was not due to experimental artifact. Based upon available data, it would be premature to speculate about the possibility that unbalanced signal intensity between homologs is a precursor to eventual locus loss, culminating in a diminution of 25S site number from the typical six to four. However, all the variations summarized above indicate that the 25S rDNA arrays in diploid Fragaria exist in a dynamic state.
Characterization of 25S rDNA arrays in diploid strawberries: conservation of site number and chromosomal position vs. dynamics of copy number
As FISH is considered to be a semi-quantitative technique , it is reasonable to expect that size and intensity of hybridization signals is an indicator of targeted sequence copy number. Thus, the polymorphism of 25S rDNA signal intensities revealed among diploid Fragaria may imply different repeat copy numbers among different rDNA sites. Loss of 25S rDNA signal(s) could be due to a complete elimination of entire sites, or perhaps only to the loss of most copies of 25S rDNA repeats at the respective site(s), resulting in a diminished signal or leaving too few repeats to be detected by FISH. Such outcomes could be attributable to spontaneous deletion of an rDNA-containing fragment from the short arm of the chromosome(s) , or to unequal crossing over, which could lead to a loss (and/or gain) of repeats from different sites .
Any rDNA site is a stretch of DNA with sequence homology to other rDNA sites on other chromosomes . Thus, physical association of rDNA clusters between both homologous and nonhomologous sites is possible, supported by the fact that rDNA-bearing chromosomes appear to be non-randomly associated with each other at mitotic metaphase . It is widely accepted that association of genes with highly repetitive sequences would increase the opportunity for unequal exchanges. The distal chromosomal position of rDNA sites in strawberry may facilitate this association, which could lead to unequal exchanges and rDNA repeat duplications/deletions, and therefore changes (both increases and decreases) in copy number between both homologous and nonhomologous sites. Size polymorphism of the hybridization signals, among and within homologous sites could be explained by such events. On the other hand, somatic exchanges taking place between different sites could make homogenization and changes in rDNA copy number occur especially quickly . Therefore, it is reasonable to suggest that the rDNA repeats in strawberry are in a highly dynamic state because of their terminal positions and potentially high degree of association between sites. However, the high conservation in rDNA site number and chromosomal location, despite their apparent high dynamics of copy number among sites, may indicate conserved genome organization among strawberry species, at least for chromosomal segments involving rDNA sites.
When more than one pair of 18S-25S rDNA sites is present in the genome, some sites may be inactive . When sites are active, a secondary constriction is typically visible. Once genes are inactivated (silenced), the constriction disappears, even though the sequence is still present and detectable by FISH . In strawberry, satellites have been reported on a small pair of chromosomes in five diploid species in previous karyotype studies [23, 24]. In our work, one or two satellites and secondary constrictions are sometimes visible in metaphase chromosome preparations in some diploids (Figure 1), and were always associated with a 25S rDNA FISH signal, irrespective of signal size. In reports on some plants, active rDNA clusters that can form nucleolar organization regions (NORs) and produce large and intensive FISH signals have been described as major loci, while the ones without transcription activity but are still detectable by FISH as weak signals are described as minor loci (e.g., Hordeum) . So far, whether all or only some of the rDNA sites in strawberry are actively transcribed and form NORs awaits resolution by further work. The terms "major/minor sites" applied in the Results section are not intended to refer to the transcriptional activity of rDNA sites, but only to relative signal brightness.
Proportional increase of rDNA site number in lower (3x - 6x) polyploid strawberries
The triploid cytotype of the hybrid F. ×bifera was previously inferred to possess one copy of the F. vesca genome and two copies of the F. viridis genome . Three 5S and nine 25S rDNA sites observed in this triploid cytotype constitute simple multiples of both rDNA site numbers in its two diploid progenitors. The one larger and stronger 5S rDNA site is probably from F. vesca, while the two smaller ones are from F. viridis, in consideration of the copy number of subgenomes provided by the two donors.
In three tetraploids, multiples of 5S and 25S rDNA site numbers (Figure 3B-3D) are increased in proportion to the increase in whole genome copy number (i.e., in comparison to the typical diploids, the tetraploids had twice as many chromosomes and twice as many detected 5S and 25S rDNA sites). To date, the alternate possibilities of auto- or allo- polyploidy origin have not been resolved in these tetraploids . Various diploid species occurring in respectively overlapping geographical areas have been proposed as putative ancestors. Due to the highly conserved genomic distribution pattern of rDNA sites among diploids, no specific species were identifiable as putative ancestors of the tetraploids based on rDNA-FISH data. However, variable allocation of 25S rDNA signal intensities among loci between these tetraploids at least implies that the diploid ancestry of F. corymbosa may be distinct from those of either F. gracilis or F. tibetica. Lundberg et al.  suggested that F. corymbosa was an allotetraploid. The size polymorphism of 5S rDNA signals in F. corymbosa observed in our work is possibly supportive to this hypothesis.
In the hexaploid species F. moschata, six 5S and eighteen 25S rDNA sites would be expected as three times those in most diploids. In fact, six 5S rDNA sites appeared as distinct FISH signals, but one or two fewer signals for 25S rDNA sites were shown in, respectively, F. moschata genotypes CFRA 376 and CFRA 157 (Figure 3E and 3F). F. moschata was shown to be an allopolyploid and its subgenome donors include F. vesca and F. viridis, which were suggested by DNA molecular studies [13, 39, 40]. In F. vesca, some genotypes (e.g., 'Yellow Wonder' and 'Pawtuckaway') have fewer 25S rDNA sites than six. Thus, the 25S rDNA site number less than eighteen in a hexaploid would not be surprising, if any F. vesca genotype(s) having less than six 25S rDNA sites were involved in its origin. Alternatively, one or two rDNA sites could have been diminished or lost by loss of most or all of its repeats during or after the arising of the hexaploid. When Rousseau-Gueutin et al.  studied two low-copy gene sequences for the construction of phylogenetic trees of Fragaria species, they linked F. moschata to both F. vesca and F. viridis in the tree based on the sequence analysis of the gene GBSSI-2, but detected no affinity between F. moschata and F. viridis in the tree from the DHAR gene, for which physical elimination of one homoeologous copy of this gene was proposed to be a possible reason.
In genotype CFRA 157, five of the six 5S rDNA sites are co-localized with 25S rDNA sites, which means loss of one 25S rDNA site (if the latter hypothesis discussed above is true) has occurred on one of the six "double-marked" chromosomes and the second one on a "single-marked" chromosome. These two chromosomes could not be homologs, suggesting that elimination of 25S rDNA sites could occur simultaneously on non-homologous chromosomes. Therefore, in species with a high site number, loss of rDNA copies could proceed very quickly simply because there are many opportunities for occurrence of this event. This speculation gains support from the cases of octo- and deca-ploid strawberries, in which eight 5S and twenty-four 25S rDNA sites, and ten 5S and thirty 25S rDNA sites, respectively, would be expected but much less are actually observed.
Remarkable rDNA site number reduction in octoploid strawberries
Each of the two wild octoploid species includes multiple subspecies. In this work, we examined three subspecies of F. chiloensis and of F. virginiana, respectively. Numbers of 5S and 25S rDNA sites are consistent among all these subspecies, and a strong reduction in rDNA site number for both kinds of rDNA was observed as compared with proportionate multiples of the typical diploid numbers. The strong reduction in rDNA site number might be attributable to two factors. First, more than a half of the rDNA sites inherited from the lower-ploidy ancestors might have failed to participate in associations occurring among other rDNA-bearing chromosomes, where homogenization through unequal crossing-over and gene conversion could have maintained homology of their DNA sequences and therefore their transcriptional function. Thus, susceptibility to loss of some rDNA arrays may be due simply to the high initial number of rDNA sites. Though rDNA site number is expected to be correlated with genome size, perhaps only a restricted number of rDNA sites under a certain threshold could be associated together and be maintained in homology, and this low number might be sufficient to support normal cellular activity. Thus, extra sites beyond that restricted or necessary number would suffer accumulation of mutations from lack of homogenization forces, and be subject to elimination.
A second possibility is nucleolar dominance, as originally described by Navashin  in some species of Crepis. This phenomenon occasionally occurs in natural allopolyploids as well as synthetic interspecific hybrids , which show exclusive transcriptional activity of genes encoding 18S, 5.8S, and 25S rRNA that belong to one of the parental genomes with concurrent lack of expression of rDNA from the other parental genome . However, how fast one of the parental rDNA repeat types may be removed from the hybrid genome and by what evolutionary forces is currently unresolved. If nucleolar dominance also occurs in strawberry, loss of rDNA sites could be explained as occurring after loss of the transcriptional activity of these sites.
As an explanation for rDNA site loss, the hypothesis based on nucleolar dominance could be integrated with that of high initial copy number if association of rDNA sites is demonstrated to occur only between those derived from a subset of the diploid ancestors. Although genome composition of octoploid strawberries has not been fully elucidated , putative subgenome donors including ancestors of F. vesca and F. iinumae have been strongly supported by phylogenetic analysis on DNA sequence data of multiple nuclear genes [1, 13], whereas in the phylogenetic tree constructed on nuclear ITS sequences , F. iinumae is not clustered with any octoploids but is a sister species to all the others. As nuclear ITS is included in the 45S rDNA unit, it is reasonable to speculate that rDNA site loss could be subgenome-specific and that rDNA repeats from the ancestral F. iinumae subgenome(s) were lost during or after the establishment of the ancestral octoploid(s).
Davis et al.  noticed the evident diminution of genome size by 12% to 16% in two octoploid cultivars as compared with an expectation of four times the size of a diploid genome. This diminished size could be due to events, such as losses of DNA segments, that occurred during or after the origination of the octoploids . The marked loss of rDNA sites in octoploids could have been a part of a broader, generalized loss of DNA segments, or could have occurred via a separate and distinct mechanism.
Origins of higher ploidy strawberries
Identical site numbers of both 5S and 25S rDNA among all the octoploid subspecies of F. chiloensis and F. virginiana suggest that the wild octoploid species are closely related and very likely share a common genome composition or even have a common ancestor, as proposed by phylogenetic analysis based on different DNA sequences [13, 30, 45]. Moreover, conservation of 5S and 25S rDNA site numbers among the octoploid species and subspecies suggests that site loss to observed levels may have been an early event, preceding the divergence of the various octoploid taxa from a common octoploid ancestor. The origin of F. moschata may have followed an independent and perhaps more recent evolutionary pathway as compared with the octoploid lineage(s), given that F. moschata still "maintains" most of its rDNA sites.
The presence of four 5S and fifteen 25S rDNA sites suggests that the decaploid cytotype for CFRA 110 of the F. virginiana subsp. platypetala originated from doubling of an interspecific hybrid between an octoploid and a diploid species, which are probably F. virginiana and an American subspecies of F. vesca, respectively, due to the North American geographic collection site of CFRA 110. Thus, this decaploid cytotype could be comprised of two 5S and five 25S rDNA sites from the F. vesca progenitor, and two 5S and ten 25S rDNA sites from the octoploid progenitor. Contrastingly, the site numbers of six for 5S and twelve for 25S rDNA in decaploid F. iturupensis do not fit an origination model involving a simple combination of an octoploid and a diploid. In such an origin, the decaploid would be expected to have fewer (four) 5S sites and more (15 or 16) 25S sites, as seen in decaploid CFRA 110. Instead, the rDNA site numbers in F. iturupensis imply a different and probably more complex origin of this decaploid species as compared with CFRA 110.
Identification of individual chromosomes by rDNA markers
Due to the extremely small size (i.e., 0.61-1.85 μm for strawberry)  and morphologically minimal differentiation of Fragaria chromosomes, chromosome-specific markers provided by FISH are needed for the identification of individual chromosomes, and are critical for tracking homo- or homoeo-logous chromosomes among species and in polyploids.
A previous rDNA-FISH technique performed on F. vesca enabled the construction of a karyotype with three pairs of marked chromosomes in diploid strawberry . Confirmatory results were obtained in our lab recently on F. vesca 'Hawaii 4' . For most accessions of F. vesca, our data presented here are congruent with the two previous studies. Among the three pairs of chromosomes with rDNA markers, one pair is double marked by 25S and 5S rDNA, and the other two single marked pairs can be distinguished by their different size and/or signal intensities of 25S sites they bear. The other eight chromosomes in a diploid complement are still challenging for even matching of homologs, except for the largest pair, which can be grouped by its size in most cells. To date, various genomic resources for F. vesca including a fosmid library , mapped and annotated fosmid clones , a BAC library , and the draft 'Hawaii 4' genome  have been established. Assembly has been anchored to the genetic linkage map into seven pseudochromosomes. These all give good opportunities for developing new, chromosome-specific probes that can expand the scope of karyotypic resolution in Fragaria.
We established several fosmid clones for identification of more individual chromosomes besides the rDNA marked ones in F. vesca (unpublished data). A comprehensive molecular karyotype in strawberry could be constructed. By finding sequence homology between 25S and 5S rDNA probes used for chromosome identification and scaffolds mapped to the pseudochromosomes of the 'Hawaii 4' linkage map, the chromosome pair double marked by 25S and 5S rDNA sites and the other small chromosome pair single marked by 25S rDNA signals correspond to pseudochromosome VII and pseudochromosome VI, respectively . As additional probes are developed, the seven linkage groups or pseudochromosomes will be assigned to each specific chromosome so that any sequences of interest defined in a certain pseudochromosome could be verified cytologically along real chromosomes with the assistance of chromosome-specific landmarks in the comprehensive molecular karyotype.
rDNA site numbers in the Rosaceae family
Physical mapping of rDNA sites by FISH has been performed in several genera of the Rosaceae family, but very few species have been examined within each of these genera. In overview, the number of detected 25S rDNA sites in diploid rosaceous species has been either two (in Rosa and Rubus parvifolius) [49–52], four , or six (in Prunus) [53–55]. Apple (Malus x domestica), in which a relatively recent genome-wide duplication is indicated , has a distinctively high chromosome number of 2n = 34 in Rosaceae, and eight 25S rDNA sites were observed in it . For 5S rDNA, a constant number of two for diploids has been observed in most of these genera except for Prunus [53, 54] and Rosa [50, 52], in which four sites have been found. In terms of chromosome position, all the rosaceous species including Fragaria exhibit a similar distribution pattern, in which 25S rDNA repeats are clustered at terminal regions while 5S rDNA sites are in interstitial and proximal regions of chromosomes. However, "double-marking" by both 5S and 25S rDNA as seen in Fragaria was only reported for diploid Prunus subhirtella  and a pentaploid Rosa canina . Comparative molecular cytogenetic analysis in the Rosaceae will benefit from the development of additional probes targeting conserved sequence sites at multiple chromosomal locations.