Genome conflict in the gramineae

Authors

  • Neil Jones,

    Corresponding author
    1. Institute of Biological Sciences, The University of Wales Aberystwyth, Ceredigion, SY23 3DD and
    2. Aberystwyth Cell Genetics Group, UK;
    Search for more papers by this author
  • Izolda Pašakinskienė

    1. Lithuanian Institute of Agriculture, 58344 Dotnuva-Akademija, Kedainiai, and
    2. Šiauliai University, Višinskio 19, 77156 Šiauliai, Lithuania
    Search for more papers by this author

Author for correspondence: Neil Jones Tel: +44 1970 622230 Fax: +44 1970 622307 Email: rnj@aber.ac.uk

Abstract

Contents

  • Summary 1

  • I. Introduction 1
  • II. Intragenomic conflict 3
  • III. Intergenomic conflict 11
  • IV. Conclusions 15
  • Dedication 16

  • References 16

Summary

The genomes of grasses and cereals include a diverse and large collection of selfish genetic elements, many of which are fossil relics of ancient origin. Some of these elements are active and, because of their selfish nature and the way in which they exist to perpetuate themselves, they cause a conflict for genomes both within and between species in hybrids and allopolyploids. The conflict arises from how the various elements may undergo ‘drive’, through transposition, centromere and neocentromere drive, and in mitotic and meiotic drive processes in supernumerary B chromosomes. Experimental and newly formed hybrids and polyploids, where new combinations of genomes are brought together for the first time, find themselves sharing a common nuclear and cytoplasmic environment, and they can respond with varying degrees of instability to adjust to their new partnerships. B chromosomes are harmful to fertility and to the physiology of the cells and plants that carry them. In this review we take a broad view of genome conflict, drawing together aspects arising from a range of genetic elements that have not hitherto been considered in their entirety, and we find some common themes linking these various elements in their activities.

I. Introduction

There are compelling reasons for research interest in the Poaceae. Several well known species of the > 10 000 that comprise this family have been domesticated as crop plants, and collectively they contribute ≈ 60% of the world's food production (Keller & Feuillet, 2000). Their economic significance is matched by their occupancy of > 25% of the Earth's land area (Shantz, 1954). The so-called grass cereals attract the major share of research resources, but the forage grasses also have significant economic impact and are attracting increasing attention in the post-BSE (mad cow disease) era.

This review is concerned with a particular aspect of grass biology, namely that of genome conflict arising from a variety of selfish genetics elements, including entire selfish chromosomes, which are a major component of grass genomes (section II, 3). Conflict arises from the dynamic interactions involving defence and attack scenarios between the defining species-specific sequences and those selfish elements using the host genome as an environment in which to propagate themselves. It is inevitable that much of the knowledge base drawn upon comes from crop species, and within these we bias our thoughts to those with which we are most familiar, namely the Lolium/Festuca complex of forage grasses. We also use a broad canvas, taking us into new research on centromeres, and into natural populations of some species where there are selfish elements that are not usually included in such discussions, but that contribute to genome diversity and basic knowledge of plant genetic resources. If any justification is needed on grounds of utility, then centromeres are of importance in relation to chromosome stability in hybrids and allopolyploids (section III) and supernumerary B chromosomes in the regulation of recombination and diploidization in polyploids (section III, 6).

Grasses have great diversity in their genome organization, including variation in genome size and chromosome number. Genome sizes for just over 400 species (4% of the family) are available on the RBG Kew Plant DNA C-values database (http://www.rbgkew.org.uk/cval/homepage.html), and a small selection of these data are summarized in Table 1 for the species referred to in this review, along with as a few others to indicate the range of variability. Other authors have recently and usefully reviewed genomics in the Poaceae (Bennetzen & Kellog, 1997; Bennetzen et al., 1998; Draper et al., 2001; Feuillet & Keller, 2002). The time scales of grass evolution and the evolution of chromosome number and genome size have been covered in depth by Gaut (2002). Here we deal only briefly with a number of general features of the grass family genome, as a basis for this review.

Table 1.  Genome sizes, chromosome numbers and ploidy levels in a selection of species of the Poaceae, representing mainly cereals and forage grasses
SpeciesChromosome no.1C (pg)Mbp
  1. 1C values from the RBG Kew Plant DNA C-values database (http://www.rbgkew.org.uk/cval/homepage.html). Mbp values were converted from 1C DNA values on the basis that 0.1 pg = 82 Mbp (Bennett et al., 2000). Species marked in bold indicate that genetic maps exist (Feuillet & Keller, 2002; Gaut, 2002; Alm et al., 2003).

Oropetium thomaeum2n = 2x = 18 0.25   205
Brachypodium sylvaticum2n = 4x = 28 0.48   394
Oryza sativa2n = 2x = 12 0.50   410
Sorghum bicolor2n = 4x = 40 1.68 1 378
Lolium perenne2n = 2x = 14 2.08 1 706
Lolium rigidum2n = 2x = 14 2.18 1 788
Festuca pratensis2n = 2x = 14 2.23 1 829
Zea diploperennis2n = 2x = 20 2.65 2 173
Zea mays2n = 2x = 20 2.73 2 239
Lolium remotum2n = 2x = 14 3.03 2 485
Lolium temulentum2n = 2x = 14 3.13 2 567
Festuca mairei2n = 4x = 28 3.95 3 239
Lolium multiflorum2n = 2x = 14 4.10 3 362
Festuca arundinacea2n = 4x = 28 4.28 3 510
Aegilops cylindrica2n = 4x = 28 4.65 3 813
Aegilops speltoides2n = 2x = 14 5.15 4 223
Hordeum bulbosum2n = 2x = 14 5.50 4 510
Hordeum vulgare2n = 2x = 14 5.55 4 551
Festuca pratensis2n = 4x = 28 5.63 4 617
Festuca arundinacea2n = 6x = 42 6.05 4 961
Triticum monococcum2n = 2x = 14 6.23 5 109
Aegilops mutica2n = 2x = 14 6.30 5 166
Aegilops sharonensis2n = 2x = 14 7.05 5 781
Festuca gigantea2n = 8x = 56 7.23 5 929
Secale cereale2n = 2x = 14 8.28 6 790
Hordeum bulbosum2n = 4x = 2811.05 9 061
Avena sativa2n = 6x = 4213.2310 849
Aegilops crassa2n = 6x = 4215.7012 874
Triticum aestivum2n = 6x = 4217.3314 211
Triticale2n = 8x = 5625.9821 304

1. Genome size

Genome sizes range from 205 Mbp in Oropetium thomaeum to > 21 304 Mbp in Triticale. There is a 35-fold difference between wheat and rice; an eightfold difference between wheat and perennial ryegrass (Lolium perenne); and a fourfold difference between Festuca arundinacea and wheat (Table 1). This massive range in DNA values, which appears to be unrelated to the basic set of genes required for growth and development, is known as the C-value paradox (Thomas, 1971), where 1C is the DNA amount in an unreplicated set of chromosomes. This paradox has now been largely resolved by the discovery that repetitive DNA is the main cause of variation in genome size, although the story is not simple in that genome size is not directly proportional to the content of repetitive DNA. There is a fourfold difference in DNA amount between sorghum and maize, for example, but retrotransposons account for only half of this (SanMiguel et al., 1998). Brachypodium distachyon (2n = 2x = 10) is now being proposed as a new model species for the grasses on account of its small genome size (< 175 Mpb) and a number of other attractive features which are held to make it a competitive model with the larger genome of rice (Draper et al., 2001; Hasterok et al., 2004).

2. Chromosome number

Chromosome numbers vary over a wide range, and polyploidy features strongly in the evolution of the family (reviewed in Gaut, 2002). It is interesting to note some species in the database where there is data is for both diploid and polyploid forms of the same species, as in Festuca pratensis 2x = 1829 Mbp/4x = 4 617 Mbp (Table 1). Polyploidy has played a major role in the evolution and modelling of grass genomes, with present-day polyploids comprising ≈ 44% of species (DeWet, 1986). Extant diploids such as maize contain duplicated chromosome segments that originate from ancient polyploid events, but the mechanisms leading to diploidization of polyploids are not well understood, nor are the events that lead to changes in chromosome number in diploids (Gaut, 2002). The one certainty is that there has been much activity at the chromosome level during the 50–80 million yr of the history of the grass genome, and that this activity has played a significant part in grass genome organization. We should note the extensive chromosome number polymorphism caused by the presence of supernumerary B chromosomes in grasses (section II, 3). The fact that these optional extras have been bred out of modern cultivars is not sufficient reason to overlook them. In rye, for example, individual plants in natural or seminatural populations may have between 0 and four B chromosomes (sometimes six) in addition to the 14 A chromosomes of the basic set. This means the populations are polymorphic for nuclear DNA amounts, with each B adding the equivalent of 10% of the 1C DNA value or 5% of the diploid genome; the same is true for natural/primitive populations of maize and many forage grasses. It is estimated that a single B chromosome in rye has 800 Mbp, more than four times the entire genome size of Arabidopsis or Brachypodium distachyon, and about twice that of rice. B chromosomes are a variable component of the genomes of numerous species of Poaceae, and where known they should be included in chromosome number citations (e.g. rye is 2n = 2x = 14 + Bs).

3. Phylogeny and comparative mapping

There is a long history of efforts to construct the phylogeny of the Poaceae, and the advent of molecular analysis has recently contributed greatly to our knowledge of the evolutionary history of this ancient family. The chronology of the story is given by Gaut (2002). It is now recognized that there is a common basis to the linkage maps for genes and other markers in grasses and cereals, and well developed genetic and physical maps have been produced and compared. Notwithstanding the wide range of variation in nuclear DNA content, ploidy levels and chromosome number, comparative mapping in the economically important species has led to the view that the colinearity of genes and markers is well conserved between species, and that the grass family as a whole can viewed as a single genetic system built from ≈ 30 rice linkage blocks (Devos & Gale, 2000) and rearranged in various ways in different species. Genomics and comparative mapping are well reviewed in a number of recent publications (Keller & Feuillet, 2000; Feuillet & Keller, 2002; Gaut, 2002; Alm et al., 2003). The flow of genomics information over the past decade is now accelerating rapidly, and includes the growing body of sequence data from large bacterial artificial chromosome (BAC) inserts as well as the wealth of data from the rice model genome initiative (Table 2). In maize there is new and surprising evidence from sequencing a 100 kb stretch of a 230 BAC contig from the bz region that microcolinearity can also be compromised by intraspecific variation between lines. Retrotransposon clusters in the region differ in their patterns of interspersion between lines, and the density and content of genes also varies (Fu & Dooner, 2002).

Table 2.  Selected Internet sources for grass genomics and comparative mapping
ResourceWeb address
IGER Forage Grass Genome Databasehttp://ukcrop.net/perl/ace/search/FoggDB
Rice Genome Research Programhttp://rgp.dna.affrc.go.jp
Maize Genetics and Genomics Databasehttp://www.maizegdb.org
US Wheat Genome Projecthttp://wheat.pw.usda.gov/NSF
GrainGenes Database for Triticeae and Avenahttp://wheat.pw.usda.gov/index.shtml
International Triticeae EST Cooperativehttp://wheat.pw.usda.gov/genome

II. Intragenomic conflict

The nuclear genome is the complete complement of genetic material in a haploid chromosome set, including all sequences, whether genes or otherwise. But with very few exceptions we do not know what the genome is, other than the estimated number of base pairs or what the limits are to its variability in size and quality between individuals within a species. The major components can be broadly classified as protein-coding genes, rDNA repeats, and various repetitive elements including tandem (satellite DNA, mini- and microsatellites and telomeres) and dispersed repeat sequences [retrotransposons, transposons, miniature inverted-repeat transposable elements (MITES) and unstructured fragments]. Some of these components are visible by microscopy and special techniques including nucleolus organizer regions (NORs), telomeres, chromomeres, heterochromatin (including knobs in maize), Giemsa c-bands and parts identified by fluorescence in situ hybridization (FISH) and genomic in situ hybridization (GISH) probes.

The current state of knowledge on sequence organization of grass genomes is documented in number of recent reviews (Vershinin et al., 1995; SanMiguel et al., 1996; Bennetzen & Kellog, 1997; Bennetzen et al., 1998; Schmidt & Heslop-Harrison, 1998; Fedoroff, 2000; Langdon et al., 2000a; Wendel & Wessler, 2000; Walbot & Petrov, 2001; Cuadrado & Jouve, 2002; Feschotte et al., 2002; Feuillet & Keller, 2002; Gaut, 2002; Langdon et al., 2003; Schwarzacher, 2003), and as indicated in Table 2 bioinformatics information is accumulating rapidly.

There is general agreement that repetitive DNA is selfish. Selfishness means that transposable elements, and other sequences (satellite DNA), exist through their capacity to multiply within a genome without offering any obvious selective advantage to their host. Self-perpetuation is their sole function (see also B chromosomes, section II, 3). They have accumulated by transposition or retrotransposition; or by replication slippage, unequal crossing over and gene conversion in the case of satellite DNA. Transposons base their selfishness on being able to copy themselves into new locations, rather than through allelic competition, as McClintock first showed for the Ac element in maize half a century ago (Jones, 2004).

Intragenomic conflict arises because of the large number of repetitive elements, and the negative effects they have on their hosts. There has to be a limit to how much size inflation a genome can tolerate, as well as the cost of mutation when selfish elements land in near to genes and cause mutations or influence gene expression (Lippman et al., 2004). By virtue of their repetitive nature they may cause nonhomologous recombination and affect the integrity of meiosis. They also can cause deletions, duplications, inversions and translocations, and harm the physiology of the host cell (Burt & Trivers, 2005).

To balance the story, we should also be mindful of new thinking on how repetitive elements have the potential for functionality as controlling elements (Kashkush et al., 2003) or as components of centromeres (section II, 2).

1. Transposable elements

Size inflation by repetitive DNA in the grass genomes accounts for much of the C-value paradox and is largely caused by historical bursts of retrotransposon activity, as in maize, the genome of which is thought to have doubled its size within the past 1–3 million yr (SanMiguel et al., 1998). The interspersed repetitive DNA in plants is composed of a diversity of transposable elements, with MITES and long-terminal repeat (LTR) retrotransposons making the largest contribution. MITES are < 600 bp, and are thought to have originated from DNA transposons. They are preferentially located in the vicinity of genes, and rice contains up to 100 000 of them. How many there are in the forage grasses is not known. LTR retrotransposons, with long-terminal repeats in the same orientation, are the single largest component of most plant genomes. They occur in intergenic regions, are often nested within each other and may also be found near to genes. These elements make up as much 15% of the 430 Mb genome in rice, 50–80% of the 2800 Mb genome of maize, and 70% of the 4800 Mb genome of barley (SanMiguel et al., 1996; Feschotte et al., 2002), and we may assume that they are present in comparably high numbers in forage grasses. There is specific information on retro-elements which target centromeres in grasses and cereals (Langdon et al., 2000a), and other sources dealing with DNA transposons, satellite DNAs, and various repetitive fractions representing a variety of degenerate or truncated elements (Schmidt & Heslop-Harrison, 1998). Collectively these interspersed repetitive DNA elements constitute distinctive differences between genomes in hybrids and introgression lines, as we know from the way GISH can be used to discriminate between them. Variation in the LTR regions of retrotransposons can also used to construct phylogenetic trees in grasses, indicating that their organization varies between species (Bennetzen & Kellogg, 1997). Active retrotransposons comprise only a tiny part of their total contribution to genomes. They are normally transcriptionally silent, but some can be reactivated under certain stress conditions (Kashkush et al., 2003) and in certain genetic backgrounds. LTR retrotransposons have a long half-life and are difficult to eliminate. One way in which their number can be reduced is by crossing over in the LTR regions, when the whole element forms a loop and brings its LTRs into register with each other. This is known to occur because of the number of LTR regions found in genomes that lack their internal domain sequences, but the process is highly inefficient. Recent work highlights the ongoing activity of the En/Spm DNA transposons in the cereal grass Aegilops speltoides, where a plethora of chromosomal repatterning events are leading to new and fertile genomic forms in a small natural population in the Haifa bay area of Israel. The transposons are clustered at hot spots for chromosome rearrangements and are activated during male gametogenesis (Raskina et al., 2004).

Superimposed on the various kinds of DNA sequences in genomes are the ‘epigenetic histone codes’ which have a role in the dynamics of chromatin behaviour, the regulation of gene expression and the suppression of transposon mobility in higher plants (Avramova, 2002; Loidl, 2004). It has been shown, for instance, that the methylation of histone H3 in the euchromatin of plant chromosomes is related to genome size. In most species with small genomes (1C < 500 Mbp), which includes Arabidopsis thaliana, strong methylation of H3 is restricted to the localized blocks of constitutive heterochromatin in the chromocentres. In large-genome species it is uniformly distributed and covers euchromatin as well as heterochromatin (Houben et al., 2003). By using fluorescent antibodies, the presence of these epigenetic modifications has been revealed at many sites in the euchromatin where they were previously not visible by classic staining methods. New work using microarrays in Arabidopsis shows that transposable elements, and a variety of other tandem repeats, are embedded in the heterochromatin and are under the control of an ATPase DDM1 which modulates methylation and chromatin remodelling (Lippman et al., 2004). The authors further show that transposable elements can regulate genes epigenetically when they insert in them or very nearby. Epigenetic histone modifications therefore play a key role in silencing retro-elements, as well as in regulating gene expression. These epigenetic codes could come into conflict in polyploids, hybrids and introgression lines, where alien chromosomes may come together for the first time necessitating a resolution to their genetic and epigenetic differences, as discussed in section III.

2. Centromeres

In monocentric chromosomes, the centromere is recognized cytologically as the primary constriction in each chromosome of the complement. It provides for the movement and orderly segregation of chromatids into daughter nuclei at mitosis and meiosis II, and the movement and segregation of half-bivalents during the first meiotic division. In terms of structure we recognize two domains: the central kinetochore domain where spindle fibre attachment occurs; and the heterochromatic paracentric cohesion domains on either side of the core region, which hold chromatids together from the end of the DNA synthesis phase until anaphase of mitosis and of meiosis I (Fig. 1a). There are also protein complexes (cohesins) that hold sister chromatids together throughout the whole of their length, from the end of the S-phase until anaphase (Heyting & van Heemst, 2000; Kitajima et al., 2004), but these are not part of the centromere. At anaphase this cohesin binding of chromatids is lost before that at the centromere cohesion domain itself. The kinetochore region is the part of greatest interest, not only because of its interaction with the spindle fibres and its essential role in the cell cycles, but also for its enigmatic structural properties and the controversial aspect of its involvement with selfish genetic elements and genome conflict.

Figure 1.

Structure of a representative plant centromere at (a) cytological and (b) molecular levels.

The kinetochore domain in higher eukaryotes is mainly comprised of clusters of long tandem arrays of satellite DNA interspersed with retrotransposons; collectively these satellites may add up to several megabases of sequence. They have the property of rapid sequence divergence, which has resulted in species-specific sequence profiles (Murata, 2002), but with some level of similarity between related species such as maize and rice. The number of repeats may also vary between the individual chromosomes within a species (Jiang et al., 2003).

In A. thaliana, for example, this CEN–DNA core is comprised of about 20 000 copies of a 178 bp tandem repeat interspersed with parts of Athila retro-elements, making ≈ 3 Mbp in total. Furthermore, this repeat region varies in sequence organization between different population ecotypes (Hall et al., 2003), an aspect that has implications for genome conflict (see below).

Addition lines of maize chromosomes added to oat (Avena sativa) have recently been used to study seven of the 10 maize centromeres individually, by DNA fibre fluorescence. The core regions were found to consist mainly of long arrays of CEN–DNA intermingled with clusters of centromere-specific retrotransposons (CRM), with the amount of these two kinds of sequences varying from ≈ 300 to more than 2800 kb between different centromeres (Jin et al., 2004).

In addition to its G/C-rich satellite repeats (AGGGAG)n, each barley centromere is estimated to contain about 200 copies of a complete gypsy-like element, cereba (≈ 7 kb), which collectively make up about 1.4 Mbp of this retrotransposon (Houben & Schubert, 2003). This class of Ty3/gypsy-like retrotransposons (centromere retrotransposons, CR), including Sau3A9, CCS1, RIRE7 and cereba, is conserved in all cereals and probably all grasses, and is exclusively localized to the core CEN–DNA regions. This contrasts with other retrotransposons, such as Ty1/copia-like elements, which may also occur in the centromere as well as in other parts of the chromosome arms. Given the high degree of conservation of the centromeric Ty3/gypsy-like elements, it is considered likely that their insertion in the centromeres of cereals and grasses (including the supernumerary B chromosomes of rye and maize; Miller et al., 1998) took place some 60 million yr ago, and coincident with the evolutionary divergence of these species, and that they were subsequently amplified during centromere evolution and are no longer intact or transpositionally active (Miller et al., 1998; Presting et al., 1998). The highly conserved nature of these retrotransposons is taken to mean that they have some functional significance in the centromere, despite the fact that they are specific to only the cereals and grasses. The nature of this function has yet to be understood, but the fact that they form nucleosomes with the same centromere-specific form of histone H3 (Cen-H3) as the CEN satellite DNA indicates that they may have been co-opted for some function concerned with chromosome segregation (Zhong et al., 2002). A recent analysis of published sequences (Langdon et al., 2000a) suggests that a single ancestral family, named crwydryn, targeted the centromeres and gave rise through active transposition to a variety of ‘universal cereale centromeric sequences’ in the evolutionary history of the Poaceae. FISH analysis indicates that crwydryn elements in rye are distributed throughout the centromeres (Langdon et al., 2000a); in maize the CR elements are found on all 10 chromosomes and tend to be inserted into the 156 bp satellite DNA repeats as large clusters, leaving long tandem arrays of centromeric satellite DNA uninterrupted to interact with CenH3 histone (Mroczek & Dawe, 2003). The DNA satellite sequences of the cereals, however, in contrast to the CR elements, are species-specific (for references see Hudakova et al., 2001).

The CEN–DNA is highly variable both in quality and quantity, and has no centromere-specific sequence information. This is in contrast to the highly conserved Cen-H3 histone protein with which it associates to form the nucleosomes, and which is conserved in the sense that it is found at the centromeres of all eukaryotes. The CEN–DNA is involved in a specialized chromatin structure in which the normal H3 histone is partly replaced by a Cen-H3 variant that shares a common core with H3, but differs in its amino-terminal tails and in some internal loop properties. The Cen-H3 tail also varies in length and sequence composition between different species (Jiang et al., 2003). Despite the fact that Cen-H3, like other histones, is one of the most highly conserved of all eukaryote proteins, it is thought to coevolve with the repetitive elements of the CEN–DNA of individual species by altering its DNA-binding preferences (Malik & Henikoff, 2002), thus making the centromere a paradoxical as well as highly dynamic structure. In the chromatid cohesion domains DNA is complexed with the normal H3 histone, although the H3 in these regions does undergo post-translational modifications during nuclear divisions (Houben et al., 2003). In rice the centromeres are small enough in terms of satellite DNA (≈ 750 kb which binds to Cen-H3) to permit full sequencing of the centromere core of chromosome eight (Cen8). Unexpectedly, because of its heterochromatic nature, Cen8 was found to contain a number of centromere-unique active genes embedded within it (Nagaki et al., 2004; Wu et al., 2004). It is suggested that gene activity is compatible with centromere function as the transcription of these genes occurs when centromeres are not otherwise active in terms of chromosome segregation and movement. Earlier suggestions that genes also occur within the centromere of Arabidposis (Copenhaver et al., 1999) were later contested (Nagaki et al., 2004).

The Cen-H3–DNA complex is constitutive. It is present throughout the cell cycle (Zhong et al., 2002) and can be visualized in interphase nuclei with the appropriate FISH/antibody probes and by immunoprecipitation. There is also an outer kinetochore domain of transient proteins, which form attachment sites for the spindle fibres and which are present only during chromosome segregation (Yu et al., 2000; Houben et al., 2003; Jiang et al., 2003). This coming together of the variable CEN–DNA with the relatively conserved Cen-H3 kinetochore proteins has led to the idea that centromere function must be epigenetically controlled, with the constitutive elements of the inner kinetochore proteins being stably inherited, but by what mechanism is unknown.

The chromatid cohesion domains (Fig. 1a) have both constitutive and transient properties and are packaged into heterochromatin. A point of confusion is that a number of authors use the term ‘centromere’ with reference only to the Cen-H3–DNA nucleosome complex, which is the component of much interest at the present time, while others refer to the centromere region which includes the kinetochore domain and the pericentromeric heterochromatin, and this needs to be borne in mind when reading the literature. The centromere is best defined as the domain which directs the formation of the kinetochore. The boundaries of the kinetochore domain are defined by the limits of the CEN–DNA region (Jiang et al., 2003). Fig. 1b shows a representative plant centromere in terms of its molecular organization, based on a synthesis of the literature. Some recent reviews give a more extensive reference list (Sullivan, 2002; Houben & Schubert, 2003; Jiang et al., 2003).

It is also a matter of some puzzlement that the CEN–DNA satellites are present in vast excess of what is needed for functionality, and that their number can vary between the individual chromosomes within a species. The centromere of the B chromosome (B) of maize has proved to be a useful experimental tool for studying centromere function in relation to size: its 9 Mb of repeats can be split into variously sized pieces by utilizing centromere misdivision at anaphase I of meiosis when the B is present as a univalent. As many as 25 functional derivatives have been found, some < 300 kb in size, and these size variants show a strong correlation between centromere size and meiotic transmission rates (Kasás & Birchler, 1998). Evidently it is not essential for the whole of the CEN–DNA to be present, and variously sized pieces of it can function perfectly well, although there is restriction in that a certain minimum size of satellite tandem array is needed.

(i) Centromere drive  The centromere is a zone of genome conflict. There is an active discussion currently taking place about how this conflict involves the ever-changing nature of the CEN–DNA and the highly conserved Cen-H3 histone which preserves the centromere-specific histones. The argument is that the highly repetitive satellite CEN–DNA, by its very nature, undergoes rapid evolutionary change for the number of repeat units and their sequence divergence, and this property results in variation in a centromere's capacity to capture spindle fibre-binding proteins and to engage a race to the poles. The theory is that during female gametogenesis, because only one functional megaspore is produced, the first centromere to reach the pole that will eventually form the egg cell will have a competitive advantage and could spread throughout a population by ‘selfish centromere drive’. The counteracting force is considered to be the way the Cen-H3 can continuously respond to these variations in the CEN–DNA to preserve centromere organization and stability of function. There are parallels here with host–parasite interaction scenarios and what is described as an ‘arms race’ (Henikoff & Malik, 2002; Malik & Henikoff, 2002; Jiang et al., 2003) between driving centromeres and genome ‘fight-back’ by the genes that code for Cen-H3, to prevent aggressive centromeres reaching fixation in a population and impairing the function and the correctness of chromosome segregation.

An overview of centromere drive is given by Burt & Trivers (2005), and while there is evidence to support the theory of drive in animals (Palestis et al., 2004), it is lacking in plants as far as the normal genome is concerned. As mentioned above, data on Arabidopsis show that variation in the CEN–DNA repeats can take on patterns specific to different population ecotypes (Hall et al., 2003). Forty-one different ecotypes were analysed, and the survey revealed that the consensus sequences from each ecotype varied significantly from the Arabidopsis consensus. This suggests that base substitutions in the satellites of the centromere had spread through a 5 million yr old genome at a faster rate than for any other single copy sequence. The Arabidopsis satellites are dynamic, and it seems that equilibria are established which maintain particular highly conserved satellite domains against the forces that drive sequence change. Herein lies the conflict, but other than invoking drive, we cannot do more than speculate on the mechanism involved. We can also remind ourselves that there is variation in centromere organization not only between different species, but also between chromosomes within a species (Jiang et al., 2003). It is too fanciful to speak of centromeres having different strengths, but this idea does encapsulate the view of centromere drive.

Having introduced a note of caution, we can now point to two situations where there is solid evidence for centromere selfishness, which drives not only the centromeres but also the chromosomes attached to them, thereby creating genome conflicts. We first deal with neocentromeres, then in the following section with supernumerary B chromosomes.

(ii) Neocentromeres  Neocentromeres are places in the chromosome other than the true centromere where additional spindle attachments can be developed, and which can facilitate poleward movement of chromosome arms in advance of the centromere proper – in other words, they are selfish regions of chromosomes and another cause of genome conflict. The best known example is in maize, occurring in the presence of the so-called abnormal-10 chromosome. There are two forms of maize chromosome 10: the normal form (k10) and the abnormal form (K10), which occurs naturally in a number of races from Latin America and the southern USA. Abnormal K10 differs from normal k10 in having a heterochromatic segment attached to the end of its long arm. In the presence of K10, neocentric activity develops at the sites of the heterochromatic knobs on several chromosome arms during meiosis and drives them to the poles ahead of the normal centromeres (Rhoades, 1952). Knobbed sites at which one homologue lacks a knob (heterozygote), or has a smaller one, in the presence of K10 will be preferentially segregated into the basal megaspore during female meiosis and will contribute to a process where knobs are selected for increasing size. This happens because the four products of megasporogenesis are in a row, with the basal nucleus at the bottom of the row becoming the egg, and the orientation of knobbed chromatids towards the outer pole during meiosis takes the knobbed chromosome into the functional megaspore (Fig. 2). The transmission rate of the K10 knob through the female track is about 70%, depending on the background genotype and environment. In addition to abnormal-10, knobs are known on all 10 of the maize chromosomes and can occupy some 22 different sites; although they are highly polymorphic in populations. In plants that are homozygous for normal chromosome 10, knobs are quiescent and follow the normal mendelian segregation pattern.

Figure 2.

Involvement of the neocentromere of maize abnormal chromosome 10 (K10) in preferential segregation of the knobbed region into the functional megaspore during female meiosis. In plants heterozygous for K10/k10 there is often a cross-over between the centromere and the knob, and during the first division of meiosis each pole will receive a chromatid with and without a knob, with the knobbed chromatids reaching the pole first because of their neocentric activity. In maize the interphase between meiosis I and II is of short duration, with very little movement of chromosomes in the interphase nucleus; with the knobbed chromatids lying nearer to the pole than the knobless ones, their anaphase I relationship is maintained at anaphase II and the knobbed chromatids pass preferentially into the outer poles of the linear tetrad. The egg then develops from the basal cell at the chalazal end of the linear tetrad, and for this reason the knobbed chromatid is transmitted at greater than mendelian expectations. This process cannot happen through the male side because of the different way in which the tetrad of microspores is formed. (Based on John & Lewis, 1963.)

Recent studies have centred around the nature of the neocentromeres and the molecular basis of their action and genetic control (Birchler et al., 2003; Burt & Trivers, 2005). It seems that neocentromeres are active from prometaphase through to anaphase; they move 50% more quickly than normal centromeres; they interact laterally with kinetochore fibres rather than in the normal end-on fashion; and they lack two major kinetochore proteins (CENPC and MAD2) found in normal centromeres. The knob on abnormal-10 contains ≈ 1 million copies of a 180 bp repeat which is present in most knobs, and recent studies have thrown up a much more complex picture of abnormal-10 than hitherto understood. It now appears that another repeat, the 350 bp TR-1, coexists with the 180 bp repeat in many of the knobs and also at three prominent chromomeres proximal to the knob itself, and that these chromomeres also exhibit independent neocentric activity. In other words, there are two quite separate neocentromere systems working in abnormal-10, both of which have their own trans-acting and closely linked protein-coding genes regulating their activity (Hiatt et al., 2002). The movement of neocentromeres is a matter of some speculation, as they do not seem to be dependent on normal spindle dynamics in the same way as true centromeres. It is proposed that ‘the mechanism of their motility involves microtubule-based motors’, and that their repeats recruit binding proteins as agents to promote the selfish transmission of their repeats (Hiatt et al., 2002).

The genome conflict perspective cannot be more aptly phrased than that ‘neocentromere drive provides a plausible mechanism for the evolution and maintenance of repeat arrays that occur in interstitial positions’ in chromosomes (Hiatt et al., 2002). As Birchler et al. (2003) point out, ‘as an example of meiotic drive, one must wonder why Ab10 and the knobs it affects have not swept to fixation in maize and teosinte populations’. The knobs do not drive to fixation; the opposing force, and ‘cost’, in this particular genome conflict zone involves the deleterious effect of knobs on the function of male gametes (Buckler et al., 1999), leading to evolutionarily stable polymorphisms and yet another class of selfish elements contributing to the evolution of the maize genome. It is also clear that excessive expansion of the genome by repeat sequences, which can recruit and exploit proteins to give them an advantage in meiotic drive, would eventually compromise the meiotic process itself, and this in turn would have another balancing outcome.

Heterochromatic knob polymorphisms are found in other genera and species, although none is so well known and well characterized as in maize. Examples include Vicia, Scilla, Anemone, rye and Trillium; and neocentromeres are known in Elymus, Bromus, Phalaris, Festuca, Pennisetum and rye; in Pennisetum and Festuca the neocentromeres form at knobs (Burt & Trivers, 2004). The molecular characterization and genetic basis of these other knob polymorphisms remain mysterious, but nonetheless they represent another component in the story of centromere-related genome conflict.

3. B chromosomes

The ideal strategy for selfish genetic elements to exploit drive is to escape the regular genome and establish themselves as autonomously replicating supernumerary chromosomes, free of genes and independent of the normal genome in their inheritance. This is what a B chromosome is: it is a supernumerary which does not associate with the normal A chromosome set at meiosis; which may be present or absent among individuals of a population (dispensable); and which has its own nonmendelian transmission process. B chromosomes (Bs) have been described in detail by Jones & Rees (1982), and more recently reviewed and updated for plants (Jones, 1995; Puertas, 2002; Jones & Houben, 2003; Camacho, 2004). Bs are known in > 1000 species of flowering plants (Jones, 1995) and are especially common in the Composite, Liliaceae and Gramineae. Rye is a useful model for describing the basic system, and from this we can extend the discussion to some other cereals and grasses such as Lolium and Festuca, wheat and maize.

(i) B-drive in rye (Secale cereale)  Early cytogeneticists were confused by the chromosome number of rye: it seemed to vary from one plant to another. Gotoh (1924) finally solved this puzzle when he realized that rye has a diploid chromosome number of 2n = 2x = 14 (the basic A chromosome complement) plus a variable number of additional chromosomes (B chromosomes) which are optional extras, and which may be present or absent from individuals of a population. Further studies over 60 yr have amassed a wealth of knowledge on the occurrence of the Bs in rye populations, their phenotypic effects, mode of inheritance and molecular organization (Jones & Puertas, 1993; Jones, 1995; Jones & Houben, 2003), leading to the conclusion that they are selfish/parasitic chromosomes maintained by mitotic drive (Jones, 1985, 1991). The essential features are: the Bs are found in virtually all wild and semi-wild populations wherever rye is grown, as well as in some cultivars; the number of Bs is constant within an individual but varies from zero to six in different plants of a population, with most plants having 0, two or four Bs; at meiosis they pair only among themselves, in various combinations, and never with the A chromosomes, and meiosis is described as irregular; in experimental crosses they show numerical accumulation caused by mitotic drive in the gametophytes (see below); they are harmful for the vigour and fertility of host plants, especially in high numbers; and the Bs have no known genes, although they do have some genetic organization which facilitates or modulates their drive. The population equilibrium frequency of B numbers is a balance between the forces of drive and negative effects on reproductive fitness, and this may vary from one population to another.

The drive process is centromere-based and takes place in both male and female gametophytes, although rye is exceptional in this respect, and in other grasses there is drive only through the male side. At the first pollen mitosis the rye Bs do not separate at anaphase (Fig. 3), and as the A chromosome chromatids reach the poles the Bs remain undivided at the equator of the spindle and their separation is delayed long enough for them to them be included in an unreduced number in the generative nucleus. The process is described at the cytological level, and may depend on the fact that the spindle is asymmetrical, placing undivided Bs closer to the generative rather than the tube nucleus. This directed nondisjunction is highly effective and delivers the unreduced number of Bs into the generative nucleus 86% of the time, regardless of the genotype or population concerned. The second pollen mitosis is normal, so the sperm come to carry the unreduced B number. As this happens through both male and female tracks, the build-up of B chromosomes in the population is inevitable. Selection for high seed set eliminates Bs in modern cultivars as they adversely affect fertility.

Figure 3.

Genetic organization of the nondisjunction properties of the rye B chromosome and photograph showing nondisjunction of chromatids of a single B at first pollen grain mitosis (mitotic drive). Nondisjunction depends on a genetic element located in the distal region of the long arm, and when this region is deleted the Bs disjoin normally. Deleted Bs do undergo nondisjunction when a standard B is present in the same pollen grain nucleus, indicating that the controlling element makes a product that is trans-acting and is recognized by receptors in the pericentric sticking sites, which delay separation of the B chromatids. The spindle is asymmetrical at the first pollen grain mitosis, and the conjoined B chromatids lie on the equator at the blunt end, nearest to the pole which will form the generative nucleus – there is directed nondisjunction into the nucleus that will form the male gametes. One theory to explain the genetic control is that the D1100 (green) and E3900 (pink) repetitive elements in the heterochromatic B-specific domain of the long arm recruit proteins (IIII) which interact with the pericentric receptors, causing them to stick together transiently (Langdon et al., 2000b).

We have some knowledge of the mechanism of nondisjunction at pollen mitosis. As seen in Fig. 3, there are sticking sites on either side of the centromere of the B, probably in the pericentromeric domains (but this is not certain), and it is this region of delayed separation that holds the B-chromatids together longer than the As. Does this mean that the opportunistic Bs have a different centromere organization to that of the As, at least in the pericentromeric region? Furthermore, it is known that there is a genetic element (whatever that means) in the distal part of the long arm of the rye B that produces a trans-acting product (protein?) which is essential for the sticking action of the sensitive receptors on either side of the centromere region. This is known because, in Bs that are deficient for the distal part of the long arm of the B, nondisjunction fails unless there is another B in the same cell which carries this distal segment (Fig. 3). We lack such detail for the female side of the process, but in any event there is a powerful centromere-based mitotic drive process which propagates these parasitic chromosomes in natural populations. A recent study has concentrated on the sequence organization of the distal end of the long arm of the B. In this region there are two B-specific sequences that have a complex organization (Langdon et al., 2000b), but it has not been possible thus far to identify genes which could code for proteins to interact with the pericentric receptors. The rest of the rye B has repetitive DNA similar to that of the A chromosomes. In addition, it also has a highly conserved structure and, at the cytological level, it has a similar form in all the many different populations where it occurs. It seems that it is a highly optimized selfish element for the function for which it has evolved – its own replication. One highly speculative possibility is that the B-specific heterochromatin segment at the end of the long arm in some way recruits proteins, by an epigenetic mechanism, which interact with the pericentromeric receptors to facilitate nondisjunction (Fig. 3). In this respect there are similarities with the neocentromeres of maize, although the information for rye is much less specific than for maize.

How do host plants deal with such a genome conflict perpetrated by these invasive and selfish genetic elements, if they do so at all? One possible way would be to restrict the capacity for meiotic pairing of the Bs, and in this way to cause their elimination as univalents. This would reduce their capacity for drive. In fact there is variation in B chromosome transmission rate in different experimental crosses, and selections can be made for both high (H) and low (L) transmission rate lines (Puertas, 2002). In H lines the Bs form bivalents in ≈ 90% of pollen mother cells, whereas in L lines they do so in only 20% of pollen mother cells. Genetic analysis of such lines indicates that the ‘genes’ for variation in transmission rate are located in the Bs rather than the As, and that such genes may actually be the sites for chiasma formation, or at least homologous arm-binding sites, in the Bs themselves (Puertas et al., 1998). Variation in transmission is seen as a property of the Bs, and as parasites they are evolving ways to ameliorate their own worst excesses, and to maintain some equilibrium with their host genomes (see also iii below). The Bs in rye are a highly aggressive and autonomous class of selfish genetic elements, which accounts for the way they have spread themselves throughout virtually every wild or semi-wild rye population in the world (Jones & Puertas, 1993).

(ii) B-drive in Festuca pratensis and other Gramineae  In F. pratensis B chromosomes occur in wild and semi-wild populations over the whole range of the species distribution in Sweden, as well as in the UK and other parts of Europe. The Swedish polymorphism was studied by Bosemark (1956) and again by Trueman et al. (1989); in the intervening 30 yr period little had changed, other than a slight increase in the average B frequency in the 50 populations surveyed. There are close parallels between the B-system of F. pratensis and that of rye. In F. pratensis directed nondisjunction, and thus mitotic drive, occurs in the same way (but only through the male track) and appears to involve a similar system of genetic control with a trans-acting element as its basis (Bosemark, 1956), although our knowledge is less complete in F. pratensis than in rye. Notwithstanding this knowledge gap, we can be sure that a genome conflict exists, and that some kind of mutual antagonism is dealt with between a selfish chromosome and its host genome within which it necessarily resides.

A similar story can be told for more than 260 species of the Gramineae, including Lolium perenne and several other Lolium and Festuca species, as well as species of Aegilops and Avena (Jones & Rees, 1982; Jones, 1995). Genome conflict caused by B chromosomes is a widespread and endemic feature of the Gramineae, although the rye and the F. pratensis model has some variations. In Aegilops mutica and A. speltoides, for example, which are unusual, Bs are eliminated from the roots and found only in the aerial parts of the plants, but they still undergo drive through directed nondisjunction at the first pollen mitosis. Bs in Phleum nodosum undergo preferential meiotic segregation in egg mother cells: this is exceptional for the Gramineae, and constitutes an example of meiotic rather than mitotic drive.

(iii) The B-polymorphism in maize  Maize is a model organism par excellence for cytogenetics of the Gramineae, and also has widespread B chromosome polymorphisms in primitive strains of Indian corn in the southern USA and in South America (McClintock et al., 1981). There is a wealth of knowledge on the maize B chromosome (Carlson, 1986), but the key factors for B-drive are nondisjunction at the second pollen mitosis; selective fertilization of the egg cell by the B-carrying sperm; and suppression of meiotic elimination of unpaired Bs (Puertas, 2002). The genome conflict caused by Bs in maize is less severe than that in rye in terms of adverse phenotypic effects, but nonetheless they are parasitic and have the capacity to drive themselves selfishly in natural populations. Genetic analysis of high- and low-transmission lines from natural populations has uncovered some of the complex attack and defence interactions between B and A chromosomes. It has been known for some time that nondisjunction is controlled by the Bs themselves, although this process would not contribute to drive unless, as is the case, there is preferential fertilization of the egg by the B-carrying sperm (Jones & Rees, 1982). Control over preferential fertilization is therefore a critical factor in modulating the B-drive and, according to experiments by Puertas and colleagues (González et al., 2003), this control resides in a gene in the A chromosome genome. In high-transmission lines the Bs take advantage of the mBtH allele to facilitate their attack, but in low-transmission lines the mBtL allele mounts the defence by causing random fertilization by 0B and +B sperm, thereby neutralizing the attack by preventing preferential fertilization. The third component of the story involves the pairing behaviour of Bs at meiosis. Another gene (fBt) in one of the A chromosomes has a dominant allele (fBtL) for low transmission, which causes meiotic loss of unpaired B chromosomes. The recessive allele fBtH has the opposite effect. This analysis of B–A interactions provides us with the best example we have in plants of intragenomic conflict caused by supernumerary B chromosomes. The genes concerned represent ‘a polymorphic system of attack and defense between B and A chromosomes’ (González et al., 2003). The situation is represented in the model in Fig. 4.

Figure 4.

Model showing coevolution of genes for B chromosome attack and defence in a native population of maize. (Based on González et al., 2003.)

Two other aspects of the maize B are of interest in relation to centromeres. First, it has B-specific sequences that are a major component of the B centromere, making it distinctive within the maize genome, but which are also represented in the centromere of chromosome 4 (Page et al., 2001) and in the TR-1 repetitive elements of the heterochromatin knobs. There is some indication from comparative sequence analysis that knobs and Bs share some common ancestry (Hsu et al., 2003) and thereby some relatedness as selfish genetics elements causing genome conflict. Second, the Bs interact with the knobbed A chromosomes in pollen mitosis, causing their elimination (Jones & Rees, 1982), and there is a negative correlation between the numbers of knobs and the number of B chromosomes in natural populations (Longley, 1938). It seems that one kind of selfish element is in conflict with another, as well as with the host genome.

III. Intergenomic conflict

When genomes of different species are combined together within the cellular and nuclear environment of one of their parents, either as interspecific or as intergeneric hybrids, they may express various states of intergenomic conflict. The theoretical basis of this level of genomic conflict in the hybrids anticipates antagonisms in terms of several components, including nucleocytoplasmic incompatibilities. Chromosome pairing at meiosis will be compromised. There may be differential patterns or states of chromatin organization in terms of euchromatin and heterochromatin, and associated epigenetic effects on patterns of gene expression. Promoters may come into conflict and transposons may respond in new ways to a changed nuclear environment. Transposon activation can result in gene mutation, chromosome breakage and changes to gene expression, where they insert near to or inside genes. The whole story of how DNA is embedded chromatin and how the histone codes operate is suddenly altered, and must affect the way in which genomes interact and accommodate to sharing a nucleus. We can expect to see silencing, activation and other unexpected outcomes for sequence organization as two alien genomes come to terms with each other. As seen below, these conflicts are not necessarily related to the size difference between the genomes concerned. In addition to these considerations, we can anticipate problems of spatial separation of genomes and conflict over nuclear territories. Centromeres may be involved in determining some aspects of the likely conflicts, not least because their activities may be under separate genetic control, and those from one genome may suffer suppression and silencing by the other. The genes that code for the Cen-H3 histones may experience dominance or interaction between them, and this could lead to chromosome instabilities, even at a tissue-specific level, with whole genomes or part of the genome being eliminated as the response. Where genomes do accommodate each other, there is no rule to say that the balance of their chromatin contributions will remain the same as it was at the time of hybridization. Intragenomic recombination could remodel a hybrid, as explained below. B chromosomes in hybrids may also have unpredictable effects because of the novelty of their situation.

Navashin (1934) provided an early case study of genome conflict with his account of nucleolar dominance (amphiplasty) in Crepis hybrids. The insight was to cross the F1 back to the parent species, demonstrating that suppression of nucleolus formation was reversible and that the change was one of the state of expression of the genetic information for assembly of the nucleolus rather than its elimination – what we now call ‘epigenetic’. It is axiomatic that the nucleolus is a highly visible marker for gene expression, but also true that many other nonvisible effects were probably occurring at the same time. In the 70 yr since Navashin announced his discovery, we have still not arrived at a definitive explanation for the phenomenon of nucleolar dominance, other than to say that it is an epigenetic process and involves the silencing of genes coding for rRNA in one of the parent species. The matter has recently been reviewed in detail by Pikard, (2000) and Viegas et al. (2002), and various theories and imaginative schemes to account for the effect have been elaborated.

Changes at the whole genome/whole chromosome level can be followed visually using the GISH technique. This differential ‘painting’ of species chromosomes provides us with a rich cytological resource for studying genome conflict in hybrids. There are obvious practical drivers to such studies, as well as many fundamental biological questions to be answered.

1. Chromosome elimination in hybrids

Chromosome elimination of one of the genomes in hybrids from crosses between different species, or even different genera, is well known as a means of producing doubled haploids for plant breeding. More than three decades have passed since the phenomenon was first correctly interpreted in Hordeum vulgare × Hordeum bulbosum (Kasha & Kao, 1970), and over this period many other interspecies and intergeneric crosses have been made, both to investigate the mechanisms involved and to explore the potential of the process for production of pure-breeding homozygous lines.

There is a vast amount of literature on this subject, and in assessing it we highlight some key aspects which we believe we can now make sense of in terms of centromere organization and function, discussed earlier (section II, 2).

(i) Hybrids between a given pair of species  Hybrids between a given pair of species vary in their capacity for genome elimination. In some crosses stable F1 plants can be established, while in other genotypes elimination of one of the parental genomes can occur. When elimination does occur, the extent to which it happens also varies so that some plants can remain as mosaics for long periods (e.g. H. vulgare ×H. bulbosum; Thomas & Pickering, 1983; Riera-Lizarazu et al., 1996). Genotype differences in this context may be a manifestation of variation in centromere organization (section II, 2).

(ii) Eliminated chromosomes  Eliminated chromosomes often fail to congress onto metaphase plate or to reach the anaphase poles during the early divisions of the zygote when the elimination is taking place, as in H. vulgare × H. bulbosum (Bennett et al., 1976). As these authors noted, it could be a question of lack of efficiency in attachment to spindle microtubules. Later studies in wheat × maize crosses demonstrated that all the much smaller maize genome chromosomes were lost during the first three cell divisions in most embryos. Furthermore, the maize centromeres were seen to be either tiny or invisible, and without affinity for spindle attachment. The maize NORs were also suppressed (Laurie & Bennett, 1989). The story is the same in wheat × sorghum crosses (Laurie & Bennett, 1988). In barley × maize hybrids, however, the maize chromosomes in the zygote bucked the trend and had well defined centromeres – even so, they were still eliminated (Laurie & Bennett, 1988).

(iii) Spatial separation in stable F1 hybrids  In stable F1 hybrids the genomes may show a degree of spatial separation, as in H. vulgare cv. Tuleen 346 × H. bulbosum. GISH observations in this material showed that parental genomes tend to remain separated throughout the whole of the cell cycle, with the H. vulgare chromosomes occupying a more central domain of the nucleus than those of H. bulbosum. The authors also noted that the H. bulbosum chromosomes tended to lag at mitosis and suffered some elimination (Anamthawat-Jónsson et al., 1993). They further commented that the distinctive behaviour of the two parents in the hybrid suggested that centromere activity of the parents must be under separate genetic control, and noted differences in the organization of satellite DNA in their pericentromeric and centromeric regions. Later studies on the same hybrid (Schwarzacher et al., 1992), using EM serial-thin section reconstructions, confirmed genome separation as well as the near-identical size of the two genomes. What the authors describe as centromere-associated structures of H. vulgare were larger than those of the more peripheral H. bulbosum chromosomes which have ‘weaker’ (smaller) centromeres. Genome separation was also found for F1 hybrids between H. vulgare and Secale africanum, with the Hordeum chromosomes nearer to the centre of mitosis than in Secale (Finch et al., 1981; Leitch et al., 1991); and also for Hordeum chilense × S. africanum where the two genomes occupy distinct nuclear domains (Leitch et al., 1990).

(iv) Tissue-specific elimination in an F1 hybrid  A confusing occurrence of tissue-specific elimination in the F1 hybrid Hordeum marinum × H. vulgare cv. Tuleen 346 complicates the centromere story. In this case, elimination of the Tuleen 346 genome occurs in the endosperm, while in the embryo it is the H. marinum genome that is lost; in other words there is ‘alternative elimination’ of parental genomes in different tissues (Finch & Bennett, 1983; Finch, 1983). The authors observed that the eliminated chromosomes had smaller centromeres and tended to occupy more peripheral positions. They suggest tissue-specific suppression of genes for centromere function, which we now have reason to believe are the genes coding for the Cen-H3 histones (section II, 2). Interestingly, the eliminated chromosomes also showed suppression of the NORs.

(v) The centromere underlies chromosome instabilities  The most compelling evidence we have to date that the centromere underlies chromosome instabilities in hybrids comes from addition lines of individual maize chromosomes added to oat, using DNA fibre-fluorescence in situ (Jin et al., 2004). In this material the CEN–DNA is intermingled with CRM, and collectively these components make up a range of sizes varying between ≈ 300 and 2800 kb for individual maize centromeres. Evidently not all CEN–DNA or CRM is associated with Cen-H3. The point of most interest here is that, in addition lines with two different genes coding for the Cen-H3 histone, one from maize and one from oat, the oat gene is dominant and the oat Cen-H3 is the form of histone that becomes incorporated into the maize centromeres. The maize Cen-H3 gene is silenced in the oat background, and the oat Cen-H3 functions to organize the kinetochore on the maize chromosomes. The fibre-FISH patterns therefore remain identical, whether a maize centromere is within its natural maize environment or in the oat background. We also know that in oat × maize hybrids it is the maize chromosomes that are eliminated, and the maize centromeres that are impaired in the hybrid condition; we still have to make sense of these apparent contradictions. All we can say at the moment is that in the oat × maize crosses the elimination of the maize chromosomes is more gradual than in crosses such as wheat × maize and barley × maize (Riera-Lizarazu et al., 1996), but nonetheless it does occur. In speaking about genome conflict, we can at least now say that the centromeres find themselves compromised in hybrids, and are subject to suppression and silencing together with the NOR rRNA genes.

In wheat × maize hybrids the sharing of a common nucleus by the two genomes, albeit transiently, has been found to cause variation in amplified fragment-length polymorphism (AFLP) in the wheat parent (Brazauskas et al., 2004). Actions are evidently taking place below the resolution of the light microscope, without obvious expression in the phenotype, and we might surmise that such actions are taking place in other kinds of hybrid situations and without our knowledge, but this aspect is unlikely to be centromere-related.

2. Synaptic adjustment

A large difference in genome size is not necessarily a factor in creating genome size conflict in a hybrid. The diploid hybrid ryegrass Lolium temulentum × L. perenne (2n = 2x = 14) has two sets of chromosomes which are structurally and genetically dissimilar, differing in DNA amounts by about 50%, yet they form a stable F1 hybrid with regular bivalents at meiosis, and have chiasma frequencies similar to those of the parents. The hybrid displays a remarkable capacity to resolve differences in terms of its synaptonemal complex irregularities and genome size differential, and to produce homologous bivalents with functional and morphological integrity (Jenkins & White, 1990). This story was recently revisited, using a moderately repetitive genomic clone from L. perenne which is 10 times more abundant in L. perenne than in L. temulentum (Jenkins et al., 2000). The sequence pLPBB2-123 was used as a FISH probe, but effectively acted as GISH in terms of its capacity to discriminate between the two genomes. At pachytene of meiosis the signal is evenly distributed along the length of the chromosomes, indicating that the sequence itself has little effect on the structural integrity or recombination efficiency of the hybrid, although it did highlight some structural irregularities (a foldback loop without signal emanating from a homologous bivalent) which could be assigned to the longer chromosomes of L. temulentum. Terminal overlaps of L. temulentum chromatin were also seen in some cells, so the structural integrity does show a modest level of compromise at this level of resolution, but function is not impaired. It would be profitable to make similar assays using other clones, and to link cytological effects with the presence of various fractions of repetitive DNA elements. If size is not a critical factor, then what is it that makes for conflict? No centromere studies are available for the species involved here, and we have no idea how their respective centromeres would adjust to whatever differences they may have.

3. Instabilities in allopolyploids

GISH is a powerful tool to visualize the nature of conflict in hybrids. It is known that F. arundinacea is a natural allohexaploid which originated as a hybrid between F. pratensis and Festuca glaucescens, and that it has the genome composition FpFpFgFgFgFg (Humphreys et al., 1995). In the past few years, two cases of ‘dramatic’ genome rearrangements taking place in somatic tissues were recorded in Lolium multiflorum ×F. arundinacea hybrids. First, in colchicine-doubled F1C0 octoploids of L. multiflorum × F. arundinacea (2n = 8x = 56), some genotypes restructured themselves as ‘novel diploids’ by processes of diploidization (2n = 14) and somatic recombination (Pašakinskienėet al., 1997). GISH showed them to be new genomic variants derived from F. pratensis, L. multiflorum and F. glaucescens, with genomes of the three species being represented in different proportions and as variable patterns, but in any case with F. pratensis chromatin making the genomic basis of the novel diploids. Second, similar events were found recently in a selected F2C1 hybrid from the same population of plants. The plant concerned was a hexaploid genotype F2 3–18 (2n = 6x = 42), and was characterized as a ‘super-recombinant’. GISH revealed its genome to have a high number of recombinant chromosomes, with some being constructs composed of chromatin from the three species, L. multiflorumF. pratensisF. glaucescens. Such a recovery of a functional tri-specific chromosome set is unique in the history of investigations of L. multiflorum × F. arundinacea hybrids. This genotype was observed for a number of years, and has high vigour complemented by a good level of fertility. Its instability showed in phenotypic segregation, as it was multiplied vegetatively. The initial hexaploid gave rise to a few somatic segregants which again preferred the diploid chromosome number of 2n = 2x = 14, for both Festuca and Lolium phenotypes (Pašakinskienė & Jones, 2004).

The instabilities described above in the L. multiflorum×F. arundinacea hybrids are genotype-specific, and the segregated plants which become diploid (2n = 2x = 14) are constructed de novo as a resolution of the tri-specific genomic conflict involving L. multiflorumF. pratensisF. glaucescens. In most cases the chromosomes of the novel diploid are very similar to the chromosomes of the pure F. pratensis diploid, but in other cases the F. pratensis genome has gained variously sized blocks of L. multiflorum chromatin, and the presence of F. glaucescens is not obvious in all the segregants. In any case, the reconstructed chromosomes are different from those existing in F. pratensis as a constituent genome of F. arundinacea, according to their GISH banding pattern (Pašakinskienėet al., 1998). We assume that the novel diploids could have resulted from concerted transposition, where at the some stage the entire newly made allopolyploid genome was a ferment of rearrangement of its constituent species-specific parts. It is clear that the centromeres have probably also played an important role in fixation of the novel diploids. We speculate that the centromeres must be ‘novel’ as well, built on the basis of the centromeric components of the parental species involved in the chromosome set of the hybrid genome.

The phenomenon of genome conflict arising through somatic recombination is not well documented in plants. An earlier case, based on evidence of somatic translocation during potato dihaploid induction from the cross Solanum tuberosum× Solanum phureja, was reported by Wilkinson et al. (1995); more recently Kondo et al. (1999) described reciprocal somatic translocations in F1 hybrids between Dendranthema japonica and Tanacetum vulgare, discriminated by GISH.

4. Rapid changes in sequence organization in polyploids

It is now apparent that changes in sequence organization and transcription may take place immediately when new allopolyploids are produced (Pikard, 2001). An example in the cereal grasses is to be seen in a newly synthesized wheat allopolyploid (Kashkush et al., 2002). The transcriptome response was investigated by analysing 3072 transcripts in a first-generation synthetic bivalent-forming allopolyploid and its two diploid progenitors Aegilops sharonensis and Triticum monococcum. cDNA–AFLP band patterns revealed that 60 out of the 3072 transcripts were altered in a reproducible way. Forty-eight transcripts disappeared and 12 were activated. The disappearance of transcripts was caused by gene silencing, or in some cases gene loss, as confirmed by sequencing. Silencing included genes for RNA, metabolism, disease resistance and cell-cycle regulation; these changes occurred via genetic and epigenetic alterations immediately after the polyploids were formed. Gene loss is an irreversible process, whereas silencing by methylation is epigenetic. The authors conclude that wide hybridization and/or chromosome doubling triggers a ‘genome shock’ as proposed by McClintock (1984). A more detailed and wider review of this aspect is presented by Liu & Wendel (2002).

5. Genome balance in allopolyploids

GISH has enabled the discovery of changes to ‘genome balance’ in what are otherwise stable Lolium–Festuca allopolyploids. In the F8 population of a tetraploid hybrid ‘Prior’ between L. perenne and F. pratensis meiosis in the early generations, as shown by conventional staining with acetocarmine, was characterized as stable with a high level of bivalent formation. A GISH study, however, revealed that extensive recombination had taken place between homologues of the two genomes, and that the balance of chromatin was not equal (Canter et al., 1999). The substitution of Festuca-origin chromosomes by those of Lolium origin resulted in a mean of 17.9 Lolium and 9.7 Festuca chromosomes per genotype. In terms of chromatin amounts, this equates to a mean length of 62.1%Lolium and 37.9%Festuca. Clearly the genome conflict had been dealt with by a change in the genome balance, over the eight cycles of sexual reproduction, in favour of the dominant Lolium genome. In an earlier study with the hybrids L. multiflorum × F. pratensis (Zwierzykowski et al., 1998), it was shown that the proportion of the genomes occupied by the L. multiflorum chromatin ranged from 49.2 to 66.7%, and this likewise confirms the balance in favour of Lolium over Festuca. A new example of Lolium-dominant behaviour has recently been found in the Lithuanian variety Punia, made from a cross at the tetraploid level of F. pratensis× L. multiflorum (Pašakinskienė & Jones, 2003). The reasons for the dominance of Lolium over Festuca in this way are not understood. Various theories have been proposed including gametic competition, pollination effects or selection for vigour in the early stages of seedling growth, but no definitive answers have yet emerged. In the light of recent knowledge of centromere organization and function (section II, 2) we could conjecture that Lolium centromeres are more competitive than those of Festuca, and this may account for the predominance of Lolium chromatin in these hybrids. Does centromere drive operate here, at female meiosis, in a way we have hitherto not suspected, and could this account for the dominance of the Lolium chromatin over that of Festuca?

In this particular genomic conflict, where L. perenne or L. multiflorum competes against F. pratensis, the Lolium chromosomes are the ones that behave as selfish and tend to colonize the genomic space in these Lolium–Festuca allotetraploids. The phenotypic outcome of this dominant Lolium chromatin is also evident in the hybrid variety Punia, which clearly expresses characteristics of root cell growth and response to low temperatures closer to that of L. multiflorum than to that of F. pratensis (Šimkûnas & Pašakinskienė, 2003). An interesting question remains as to which Lolium chromosomes out of the whole set have dominant centromeres and become retained as multiples, and which ones of Festuca lose out in the competition. A similar story to that of Festulolium was earlier described by Anamthawat-Jónsson (1999). Using GISH she recovered a unique set of chromosomes in Triticum hybrids (2n = 4x = 28, AABB) × Leymus (2n = 4x = 28, NNXX). The allopolyploids stabilized over a number of years as a set of six pairs of Leymus mollis and 15 pairs of the wheat parent (Anamthawat-Jónsson, 1999). The unique composition probably resulted from the stable replacement of one pair of Leymus chromosomes by the addition of one pair from wheat, together with the selective elimination of eight pairs from Leymus. It is clear that the conflicts and chromosome instabilities in the allopolyploids have a higher degree of complexity and a larger variety of outcomes. There are more conflicts to resolve, and more ways in which resolutions can occur.

6. B chromosomes and hybrids

Mochizuki (1957, 1964) first discovered that the B chromosomes of A. mutica could suppress homoeologous pairing in the F1 hybrid Triticum aestivum × A. mutica which was nullisomic for the 5B chromosome (B of 5B not to be confused with B chromosome) of wheat. In other words, the Bs could substitute for the 5B pairing control locus of hexaploid wheat, which was a potentially exciting discovery. Not surprisingly, many other species and intergeneric crosses were subsequently made by other people to investigate this bizarre and inexplicable phenomenon, including the F1 of L. temulentum× L. perenne+ Bs. As noted (section III, 2), these two species differ by 50% in their nuclear DNA contents, but nevertheless some genotypes form high frequencies of effective bivalents at metaphase I of meiosis in the F1s (without Bs). At the tetraploid level in plants without Bs there is homoeologous association and many multivalents are formed, but in the presence of B chromosomes synapsis is restricted to homologous pairs and the allopolyploid is effectively diploidized (for a full review of B chromosomes in hybrids see Jenkins & Jones, 2004). Considering we are dealing here with an entirely experimental system, one which neither the B chromosomes nor the hybrids themselves have experienced before, the situation is puzzling, although such dramatic effects are not found in all cases where Bs are placed in hybrids. In terms of genome conflict, we have a novel state of affairs where a B chromosome from one of the parents of a hybrid ‘intervenes’ in the genome conflict, and somehow regulates the interplay of homologues of the two genomes in the F1. This story is left suspended, and the best we can do at present is to construct hypotheses in terms of physical effects of the Bs involving genome separation, or changes to centromere-clustering patterns (Naranjo & Corredor, 2004), or even genes on Bs that express themselves only in new and unfamiliar nuclear environments. There is also the possibility, in the case of the Aegilops Bs, that they may have been derived from the pairing control chromosome 5B, but there is no evidence to support this view. In any event it is odd that a selfish genetic element can be manipulated experimentally to have a beneficial effect on a conflicting scenario in certain allopolyploids.

IV. Conclusions

The idea that the genome is dynamic and constantly evolving is not new. It can be traced back to the inspirational ideas of Barbara McClintock (McClintock, 1984) and to much earlier times in terms of numerical and structural chromosome variation and evolution (Schubert & Houben, 2004). The new element in our thinking comes from new approaches to genome structure and organization made possible by technological advances in genetics, and in the use of fluorescent and immunofluorescent probes to track chromatin and epigenetic histone modifications. In using the resolving power of the genomics toolbox we are raising our level of awareness of the extent to which a variety of selfish genetic elements have participated in the evolution of the genome and its ongoing state of conflict, which continuously presents new opportunities for genome modification. An overview of the various components contributing to conflict and its resolution is summarized in Fig. 5.

Figure 5.

Model of the components of genome conflict.

The genomes of grasses and cereals are excessively large in relation to the number of genes identified in them, with some notable exceptions at the lower end of the size range. Much of this size inflation is caused by amplification and colonization by transposable elements and other repetitive elements, including those that find their home in the centromere. These elements are associated with the evolution and divergence of species and are largely, but not entirely, silent relics of evolution. Those that are active contribute to genome evolution, and to an active debate about their present status and significance. The idea that selfish DNA is relic rubbish is gaining new interpretations in the face of new ideas about the total integrity and coherence of the dynamic (rapidly evolving) genome (Han & Boeke, 2004; Han et al., 2004). In this respect we should also note some recent reports on how environment can modulate the activity of retrotransposons in some cereal and grass species (Kalendar et al., 2000; Ceccarelli et al., 2002).

Polyploidy plays an obvious part in genome expansion, either through doubling the chromosome number within a species (autopolyploidy), or through interspecific hybridization followed by chromosome doubling to restore fertility (allopolyploidy). Rediploidization can subsequently reduce the number back to the diploid level, as is now thought to have happened for example in Arabidopsis and maize – but with a greatly increased genome size and still with the opportunity for chromosomes to ‘grow’ by sequence amplification through replication slippage or unequal crossing over (meiotic, somatic sister chromatid exchange), and the active spread in the genome by retro-elements (Schubert & Houben, 2004). While the process of polyploidization is relatively well understood, the mechanism of rediploidization is not known. What is known, however, is that the tolerance for multiple sets of chromosomes is high, especially in apomictic grasses which avoid the conflict caused by multiple sets of homologous chromosomes at meiosis (e.g. Poa litorosa, 2n = 38x = 266; Hair & Beuzenberg, 1961). It is also known that there are upper and lower species-specific limits to the size of individual chromosomes which will permit their proper segregation (Schubert & Houben, 2004). We are also grappling with the complexities and instabilities associated with allopolyploids, and the role centromeres may play in genome balance and in the instabilities and reconstruction that take place rapidly in newly formed allopolyploids.

B chromosomes tend to be overlooked because they are not present in highly developed crops, but they are found in the wild relatives of cultivars and constitute a significant component of genome size variation. They are also a potential resource for future use in genetic manipulation, for example in genome restructuring by insertion of B segments into A chromosomes; development as vectors for transgenes if their inheritance could be stabilized; and as tools to regulate recombination in A chromosomes. Supernumerary chromosome segments (Parker et al., 1998; Wilby & Parker, 1998), which are not considered here, have mechanisms of meiotic drive, and add to genome size polymorphism and genome conflict in some plant populations. Meiotic drive may also enhance the transmission of certain genes over and above mendelian expectations (Parker et al., 1998) and, as with supernumerary segments, the basis of this drive process is not understood.

The centromeres are the most intriguing part of the genome in terms of their mysterious structure and dynamism, and offer great intellectual challenges in understanding their structural and functional biology. It remains confusing that a part of the genome with such an essential function should itself be a hotbed of molecular ferment and selfishness. This is not what we might have expected, but it is the emergence of the unexpected that makes research into genome organization and function such a compelling science.

Dedication

This review is dedicated to the memory of Rokas Pašakinskas who died in a tragic accident in October, 2003, at the tender and promising age of 18 years.

Ancillary