This is the introduction to my thesis.
I tried to illustrate what I find interesting about meiotic chromosome pairing, a problem that has perplexed biologists for a century.

Introduction to The Problem of Meiotic Chromosome Pairing

Background: the biology of meiosis

 The commingling of genetic information from two sources is a universal
process in biology.  By shuffling genes and genomes together, a lineage
can eliminate deleterious mutations much more efficiently than by the mere
steady accumulation of point mutations.  Eukaryotic organisms are the
consummate practicioners of this art, as most of them are obliged to
reproduce sexually at least some of the time.  Sexual reproduction
(speaking generally) is the creation of a diploid organism through
fertilization, the fusion of haploid gametes from one or two organisms.  
The gametes are created by meiosis, a specialized pair of cell divisions
that partition a diploid genome into two equivalent haploid complements.  
The alternating succession of fertilization, which doubles genomes, and
meiosis, which halves them, is the basis for sexual reproduction. Meiosis
can be seen as a variation on the normal cell cycle, with innovations
especially suited for the production of haploid gametes.  (See Figure 1
for an outline of the two types of cell division).  In prophase of the
first meiotic division (meiosis I), the chromosomes undergo homologous
pairing - each chromosome from one parental genome pairs with its
equivalent from the other genome.  They are held rigidly in place against
each other by the synaptonemal complex (SC), a proteinaceous structure.  
The configuration of paired chromosomes allows crossing over, or
recombination, to take place, which results in new chromosomes constructed
of contiguous stretches of the old ones.  The loci of recombination become
chiasmata, physical connections between homologous chromosomes, which
remain after the SC has dissipated.  At metaphase I, centromeres attach to
kinetochore microtubules of the spindle, but with one important
difference:  the kinetochores of sister chromatids behave as a single
kinetochore, always attaching to the same pole.  Kinetochores may attach
to either pole, but tension-based checkpoints (Shah and Cleveland, 2000;
Yu et al., 1999) exist that delay anaphase until the kinetochores of
homologous chromosomes are attached to opposite spindle poles.  When
anaphase commences, the chiasmata also dissipate and the newly recombined
homologous chromosomes are pulled apart from each other into two daughter
cells.  Meiosis I is thus called a reductional division, because the
number of chromosomes is reduced in half.  The daughter cells resulting
from meiosis I carry out the second meiotic division, meiosis II.  
Meiosis II resembles a mitotic division more closely than does meiosis I.  
The major difference is the lack of an S phase, as meiosis II chromosomes
are already replicated.  After meiosis II, there are four daughter cells,
each containing a haploid set of chromosomes which, if recombination has
taken place, are genetically distinct from both parental genomes.  One or
more of these daughter cells eventually become the gametes that undergo
fertilization.

The pairing problem

 Meiosis is thus the essential complement to the genome doubling at
fertilization that occurs in sexually reproducing organisms, as it
provides a mechanism to reduce the genome size in half.  For the daughter
cells and the products of fertilization to be viable, this must be a
precise reduction into two functionally equivalent halves - each cell must
possess a full haploid genome complement.  The solution to this problem of
reduction adopted by the great majority of organisms is, as we have seen,
the pairing of homologous chromosomes during prophase of the first meiotic
division.  When homologous chromosomes are paired, they are able to orient
their kinetochores in opposite directions so that at anaphase, half of the
chromosomes are pulled to each spindle pole.  This simple description of
pairing belies a cell biological process of immense complexity.

The complexity of meiosis can be appreciated by estimating the chances of
its success in the absence of specialized mechanisms.  The result of
meiosis I is two daughter cells, each containing a haploid complement of a
genome.  The chance of obtaining this result through random segregation of
chromosomes is 1 in 2 to the nth power, where n is the haploid number of
chromosomes.  The actual occurrence of correct division is close to 100%,
so it is clear that there is a mechanism operating - the pairing of
homologs - that increases the probability of correct segregation.

The problem does not stop here, however, for now we must ask a deeper
question:  how are the chromosomes able to pair with each other in the
first place?  Again, a perspective from the viewpoint of random chance
gives a rough approximation of the complexity of the problem.  If a
chromosome may pair with any other, then the chance of one chromosome
pairing with its correct partner is 1 in (2n-1) where n is the haploid
number of chromosomes.  The chance for the next chromosome to pair
correctly is 1 in (2n-3), and so on.  With increasing numbers of
chromosomes, the complexity (measured by probability) increases as e to
the power of (log(n!)/2) - faster than an exponential function.  It can be
seen that, by this measure, the "chromosome recognition problem" is more
complicated than the "segregation problem" for haploid numbers of
chromosomes greater than 5.  That is, it would be more likely for a cell
with 5 or more chromosomes to complete meiosis successfully by randomly
segregating its chromosomes, without any pairing, than by first pairing
randomly and then segregating the paired chromosomes to opposite poles.

Obviously, some regulated process is also required to accomplish correct
discrimination of homologs.  The simplest mechanism would be to use Watson-Crick
base pairing, as homologous chromosomes share homologous sequences in the same
order on the chromosome.  It must be used at some point in meiosis, as intragenic
recombination shows a precise recognition of homology.  DNA-based homology
recognition is quite robust in many systems; for example, homologous
recombination in Saccharomyces cerevisiae is efficient at levels nearing 100% for
small (<1kb) pieces of DNA.  However, this solution can not apply exhaustively to
organisms with large genomes, as not all homologous sequences occur at homologous
loci.  Repetitive sequence families, which constitute a large percentage, even
the majority, of many genomes, can occur on all chromosomes in multiple regions. 
In maize, for instance, 100-200 retroelement families account for over 50% of the
genome (Sanmiguel et al., 1996), making each part of the genome resemble every
other part on the megabase scale (Figure 2).  This would make identification of
DNA homology insufficient to establish the presence of a homologous locus.  This
fact essentially replicates the previous chromosome recognition problem at the
subchromosomal level as many times as there are distinct sequences involved in
homology comparison.

Another constraint on the pairing process is that it must be completed in the
developmental context of the life cycle of the organism.  In angiosperms, meiosis
must be coupled to pollen and megaspore development, which in turn are coupled to
flower development.  In animals, meiosis must be coupled to sperm development and
egg maturation.  This restriction necessitates a timely beginning and end to the
meiotic process.  In a given species, meiosis starts and ends at a stereotypical
time, and requires roughly the same time for each cell to complete it.  This is
true despite the gradual, continuous nature of the morphological changes that
occur in meiotic prophase.  Meiocytes in grass anthers, for instance, proceed
through meiosis in tight synchrony.  This means that the organism cannot rely on
stochastic mechanisms alone for pairing, as such mechanisms would have broad
distributions of finishing times.

There is a final complication to the pairing problem, one that is often
overlooked in discussions of meiosis.  This is the fact that before sequences can
even be compared between potential homologous loci, they must be brought into
close proximity from their prior state of separation (Figure 3a).  In this case,
the null hypothesis is that sites destined for eventual pairing are initially
localized randomly with respect to each other.  The expected mean separation of
two such loci in a sphere is close to the radius of the sphere (Figure 3b).  This
a large distance for a chromosome to move, on the same order as the
spindle-driven movements that occur during metaphase and anaphase.  The volume of
a nucleus may be vast relative to the size of the elements involved in the
homology search:  in maize, the nuclear radius at meiosis is c. 8 µm, whereas a
DNA-protein complex involved in the homology search may have a linear dimension
smaller than 0.2 µm, making the nuclear volume roughly  6.4x104 times larger. 
Given the confined nature of chromatin diffusion, Brownian motion might require
extremely large time scales to bring separated sites together.

Taking all these considerations together, it appears that meiotic chromosome
pairing is an example of an ill-posed problem: a problem which, as stated, has no
solution.  Another biological example of this is the problem of visual
processing: how is usable 3-d information about objects and surfaces extracted
from the excitation of the retina, a 2-dimensional cell layer, when there are an
infinite number of possible interpretations of any given retinal image?  It was
realized (Marr, 1982) that object recognition and depth perception were too
complex to be solved by operations on the 2-dimensional information provided by
the retina - that "bottom-up" mechanisms, which would map a unique 3-d
interpretation to one pattern of retinal activation, could not suffice.  There
had to also be "top-down" constraints acting as well:  in the case of the brain's
visual processing, these are assumptions about the way objects reflect light and
cohere together, that allow visual information to be acted upon correctly.  This
is the general way complex problems are tackled by biological systems (or more
accurately, it is the way such complexity can come to exist in the first place). 
The drawback of this type of solution is a loss of plasticity:  if some
assumptions are violated, the solutions will no longer work.  But it is this fact
that allows experimental manipulation of the system:  the assumptions are
revealed as those aspects whose disruption results in the failure of the process.

The main theme of the work presented in this thesis is the characterization of
"top-down" mechanisms that contribute to the reduction of the immense complexity
of the meiotic pairing problem.

For this discussion, complexity-reduction mechanisms will be separated into 3
levels:  the chromatin (or sub-chromosomal) level, the chromosome level, and the
nuclear level.  These levels are defined by the smallest unit that must be
observed to see the phenomenon in question (Figure 4).  An example of a mechanism
acting at the chromatin level would be the restriction of the homology search to
non-repetitive sequences (sequences which would be sufficient to establish a
homolog's presence) either by their assuming a favorable chromatin conformation,
or by repetitive sequences being "masked" by assuming a non-favorable
conformation.  A chromosome-level mechanism would be a chromosome's assuming of a
favorable shape through condensation or other alteration of chromosome
architecture that can only be seen by looking at an entire chromosome.  The
cloud-like territories occupied by interphase chromosomes would only be able to
compare homology between loci at the surface of the territory.  However, at the
onset of meiotic prophase chromosomes undergo a remarkable morphological
transition:  they extend into long fibers, probably via attachment of chromatin
loops to a linear axial core.  The linear shapes of chromosomes at leptotene have
a maximal surface-to-volume ratio with high persistence length, which ensures
that most contacts between DNA will be interchromosomal, rather than
intrachromosomal.  A nucleus-level mechanism would be the specific arrangement of
chromosome regions within the space of the nucleus.  This dissertation will
discuss two specific examples of nucleus-level mechanisms.  One is the Rabl
configuration (Rabl, 1885), in which the chromosome configuration of anaphase
(centromeres pointing towards poles, telomeres dragging behind) is "frozen"
throughout the next cell cycle.  The other is the bouquet, a meiotic
prophase-specific clustering of telomeres in a small region of the nuclear
envelope.

I hypothesize that these three levels create a context in which the pairing
problem can be approached with great efficiency, as illustrated in the following
sketch:  Nucleus-level organization predisposes homologous sites to
preferentially come into contact with each other, as a stereotyped chromosome
organization will put homologous chromosomes into similar nuclear compartments. 
Chromosome-level organization allows the homology search to proceed efficiently,
as the path of a chromosome is linear, making pairing cooperative, i.e.,
identification of homology in one region will encourage more encounters between
adjacent homologous regions.  Finally, chromatin-level organization ensures that
the homology that is identified is productive; extensive homology between
retroelements, for instance, will not overwhelm the identification of homology
between genes.  This model makes the following predictions:

€Chromosomes will display nonrandom arrangement at some stage prior to pairing.
€Interfering with chromosome arrangement in the nucleus (i.e., disrupting the
Rabl configuration or the bouquet) will lead to problems with homologous pairing.

If (a) holds but (b) does not, the conservation of meiosis-specific chromosome
organization across so many species presents a mystery:  could it be that this
highly organized state is nonfunctional?

Some exceptions to (b) can be found in the literature:  the accommodation of
chromosome inversions by the creation of inversion loops (Burnham, 1962) suggests
a region-autonomous basis to homolog recognition that does not depend on global
localization.  However, these loops do not form in 100% of the cases when they
are able to, and in any case only form when the inversion is of a certain size. 
It appears that inversion loops form when synapsis is initiated within the
inversion, but not when it is initiated outside and spreads through it.  While
global chromosome arrangement is not the sole determinant of correct chromosome
pairing, it is thought to play an important role in increasing the efficiency of
the homology search.

Independent of meiosis-specific considerations, we can identify several
contributions to the organization of chromosomes that may operate to various
degrees in all cells:  Brownian (thermal) motion of chromatin, differential
attachment of certain chromosome regions to the nuclear envelope, and aggregation
of chromatin in intra- or interchromosomal complexes.  A question addressed later
will be whether the chromosome reorganizations observed in meiosis can be
explained as simple modulations of these basic forms of chromosome organization.

The rates of homologous pairing 
 The picture presented so far is that chromosome regions destined for
eventual pairing come together at some rate, are recognized as being
homologous, and stay together from that point on.  Pairing of regions at a
small scale must lead to pairing of entire chromosomes, so there must be a
way for errors to be corrected; i.e., homologous sequences on
non-homologous chromosomes must be allowed to dissociate.  Therefore a
dissociation rate for paired sequences must exist as well.  What must be
true in this model, for the chromosomes to pair correctly, is that the
rate of encounter between correct loci must exceed the general rate of
dissociation.  This is easily done if the dissociation rate is near zero.  
This would allow the homology search to proceed leisurely, assuring that
no productive encounters are wasted and that, provided no mistakes are
made, the correct solution is eventually found.  A near-zero dissociation
rate, however, would also lock in whatever result is first attained; if
this happens to be the wrong situation, meiotic mayhem would result at
anaphase.  One mechanism for bringing correct sequences together at a high
rate is the previous coming together of correct sequences in an adjacent
region.  The efficiency of pairing can be greatly increased by initial
establishment of a few paired loci known to be correct, which then cause
the rest of the chromosomes to "zip up".  Thus the problem reduces to
getting those few initial contacts right.  This seems to be the ideal
place for global chromosome arrangement such as the bouquet to operate, as
the establishment of a few initial contacts near the ends would be
tolerant of massive disorder in centromere-proximal regions.

The argument from developmental time 
 Another argument for the requirement of order is the observed tight
regulation of pairing within the developmental time scale of the organism,
as mentioned previously.  In grasses especially, meiosis is synchronous
within each anther, i.e., the time from beginning of pairing to its
completion is nearly the same for each of the hundreds of meiocytes in an
anther.  This would not be expected from a random process, however.  If
the initial configuration of homologous loci is random, and the search
process is also random, there would be a broad distribution of finishing
times, with some cells predisposed to finding the solution quickly, others
taking longer and longer.  As this is not the case, either the initial
configuration of homologous loci, or the search mechanism, or both, must
not be random.

(1)	For the search mechanism to be nonrandom, there must be a way for
information about which loci are homologous to influence the mechanism that
causes movement of actively searching loci.  The bouquet is a coarse example of
this, as homologous loci are in effect moved together.  However, they are not
brought together specifically in preference to nonhomologous loci.

(2) 	For the initial configuration to be nonrandom, there must be a bias in
the location of pairable sites on the chromosome such that a bulk,
noninformational process can result in their favorable positioning.  This would
be the case if, for instance, the density of pairable sites were greater towards
the ends - as has been reported for genes(Gill et al., 1996) - so that the Rabl
configuration suffices to set up an organization capable of timely progression
through pairing..

A model for pairing based on simple diffusion of initially random loci cannot
account for the efficiency and timeliness of meiosis.  This is borne out in
simulations, as seen in chapters 4 and 5.  Instead, combinations of (1) and (2)
above have been cited for almost every organism in which meiosis has been
observed, though different organisms combine the two strategies to different
extents.  To take two extremes, Drosophila enters meiosis with its chromosomes
already paired from somatic processes; while maize has been reported to enter
meiosis with no discernible organization whatsoever.  Though (1) seems to
predominate in maize, I will argue that (2) also makes a significant
contribution, and that in fact pairing without (2) ought to be extremely rare, if
observed at all.

The Stages of Meiotic Prophase

 As meiotic prophase progresses, chromosomes undergo a number of
rearrangements in shape and position.  The major divisions between
morphological states (identified by light microscopy), correspond to the
synapsis state of the chromosomes (identified by electron microscopy),
indicating a strong correlation between morphology and pairing.  In
grasses, these stages are particularly well-defined and easy to
discriminate via light microscopy (Dawe et al., 1994), (Aragón-Alcaide et
al., 1998).  The stages in maize (top) and wheat (bottom) are illustrated
in Figure 5.  The stages of most interest to this study are premeiotic
interphase, leptotene, and zygotene, as events occurring in those stages
determine the nature of the homology search.  The ability to determine a
cell's precise meiotic substage is crucial to drawing conclusions from
examining chromosomes in a fixed state.

The Chromatin Level of Organization

 As the bulk of the thesis will deal with the chromosome and nuclear
levels of organization, I will pause here to briefly summarize what is
known about chromatin organization in meiosis.  The earliest signs of
altered chromatin conformations are in premeiotic S phase, as shown by
studies in yeast (Ohta et al., 1998).  Recombination hotspots display
increased micrococcal nuclease sensitivity during meiosis.  In yeast
carrying null mutations of clb5 and clb6, which encode B-type cyclins, DNA
replication is skipped in S phase (Smith et al., 2001).  In these mutants,
hotspots are no longer nuclease-hypersensitive.  These mutants also fail
to initiate any recombination.  Therefore a chromatin context is
established in S phase which is correlated with the ability to undergo
later events of meiosis.  Another meiosis-specific modification is found
in sister chromatid cohesion.  In mitotic chromosomes, sister chromatids
are held together via cohesin complexes containing Scc1p (Klein et al.,
1999); in meiosis, this protein is replaced by Rec8p (Buonomo et al.,
2000).  As in mitosis, the cohesin complex is degraded at the
metaphase-anaphase transition; in meiosis this allows homologs to separate
from each other, as they are held together by sister chromatid cohesion
distal to sites of crossing over.  However, the sister centromeres of
meiosis I chromosomes must not disjoin at anaphase (Toth et al., 2000);
therefore, they must somehow be protected from degradation.

The Rabl Organization

 Chromosome segregation at anaphase results in the polarization of
chromosomes because sister centromeres are pulled in opposite directions
and the rest of the chromosome trails behind. In some instances, the
anaphase arrangement of chromosomes persists into the following interphase
(see Figure 1, col. 4); this is known as the Rabl organization (Dernburg
et al., 1995). Observations of whole chromosomes in the Rabl configuration
show that they occupy elongated territories, stretching from one end of
the nucleus to the other. This is a departure from the cloud-shaped
territory expected for a free-floating polymer in solution (Marko and
Siggia, 1997) and suggests the operation of one or more active constraints
on their morphology.  The presence of the Rabl organization is known to
vary greatly between species and among tissues or developmental stages of
an organism.  Some cells lose the Rabl organization soon after entering
interphase, whereas others retain the organization through to the next
mitosis. In a study of a variety of plants, genome size and chromosome
length were postulated to be two possible reasons for this variation (Dong
and Jiang, 1998). The Rabl organization was observed in wheat, rye,
barley, and oats (Avena sativa), all of which have C values above 4,800
Mbp. Chromosomes of sorghum (Sorghum bicolor) and rice (Oryza sativa),
both with genomes under 1,000 Mbp, lacked the Rabl configuration. Maize,
which at 3,000 Mbp is intermediate in genome size, displayed neither
entirely Rabl nor entirely random chromosome organization. Previous
studies of other genera (Allium, Vicia, Arabidopsis, Brassica, Solanum,
and Pisum) supported their model (Dong and Jiang, 1998), although it does
not appear to extend to the animals, as Drosophila melanogaster, a
small-genome organism, displays a striking Rabl organization.

One hundred twenty years after its initial observation, there are still many
unanswered questions about the Rabl configuration. The Rabl organization is a
direct consequence of the anaphase configuration of the chromosomes, and how it
is established is not in question. More intriguing is the mechanism by which the
chromosome polarization is retained through the following interphase. Is it
actively maintained, e.g. through centromere and telomere interactions with
opposite halves of the NE, or is it passively maintained, being the chromosome
arrangement with the lowest free energy in the interphase nucleus?

How is the Rabl Configuration Maintained?

 The Rabl organization of chromosomes in the interphase nucleus is
relatively fixed over time and in the nuclear space, and the telomeres
appear to provide important anchorage points. Organization of the
chromosomes in the interphase nucleus may rely solely on
chromosome-chromosome and chromosome-NE interactions. The nuclear lamins
appear to be excellent candidates for providing chromosome attachment in
animals, though the presence of nuclear lamins in plants and fungi is
controversial. The nuclear lamins are a class of intermediate filaments
that form a polymer network on the inner face of the NE as well as foci
throughout the nucleoplasm. Lamins can bind DNA, chromosomes, and histones
in vitro (Wilson, 2000).  Evidence from mice and D. melanogaster suggests
that loss of a nuclear lamin causes chromatin to detach from the NE
(Wilson, 2000). It is unclear how the specificity for telomere-lamin
interactions might arise. One possibility is suggested by evidence that
histones are spaced differently in the telomeres relative to the rest of
the chromatin. Telomeric nucleosomes (histones and associated DNA) are
spaced 15 to 30 bp closer in telomeres than in the rest of the genome
(Fajkus et al., 1995). Because histones can bind directly to the nuclear
lamins, tighter nucleosome spacing in the telomeres could increase the
number of histone-lamin interactions in a given length of DNA, potentially
stabilizing the attachment of telomeres, though this has not been
experimentally investigated.  A microtubule-based mechanism for the
maintenance of Rabl-like organization in budding yeast interphase has been
proposed (Jin et al., 2000), as the organization observed is disrupted by
nocodazole.

Possible Functions of Rabl Organization

 The Rabl configuration imposes a striking degree of order on interphase
chromosomes, isolating specific chromosome regions as small, well-defined
domains within the nucleus. This isolation would be important if factors
necessary for maintaining certain chromatin configurations need to be
sequestered. It has been strongly suggested that in wheat and many other
organisms that the density of genes on the chromosome increases near the
telomeres. To the extent that this is true, positioning telomeres in the
nucleus is equivalent to positioning genes. Gene position in the nucleus
has been found to affect expression in many systems. However, a study in
wheat (Abranches et al., 1998) has shown that active transcription sites
do not show obvious localization patterns in nuclei, but are randomly
distributed. The Rabl orientation may still contribute to this pattern in
a non-obvious way. Another hint as to a possible function of the Rabl
orientation comes from live imaging studies of heterochromatic foci in
cell lines of muntjac (Muntiacus muntjak). Manders et al. (Manders et al.,
1999) demonstrated that at the G2-M transition (the entry into mitosis)
when chromosomes are condensing, chromosomes do not undergo much internal
reorganization (movement) to reach the structure they will have at
metaphase. This would not be the case in a non-Rabl cell when individual
chromosomes must rearrange from a cloud-shaped to a rod-shaped territory.
Because non-Rabl cells tend to have smaller chromosomes, this indicates
that the Rabl orientation may be a way of dealing with the difficult task
of forming large metaphase chromosomes.

Introduction to the Bouquet

 The bouquet is the clustering of chromosome ends on the NE during meiotic
prophase, coincident with the initiation of homologous chromosome
synapsis. The bouquet has been extensively described in many species in
all eukaryotic groups (Bass et al., 1997; Loidl, 1990; Zickler and
Kleckner, 1998). In particular, there are no documented cases of plant
species that lack the bouquet stage. However, both the mechanism of
bouquet formation and its function in meiosis remain unknown. It has been
proposed as an aid to presynaptic alignment of homologous chromosomes
because it brings all chromosome ends into a common nuclear subregion and
makes them all roughly codirectional. The bouquet in mice, humans
(Scherthan et al., 1996), and maize (Bass et al., 2000) appears to occur
after large rearrangements of the chromosomes. In hexaploid wheat, the
bouquet appears to be a tightening of the already present Rabl
(Aragón-Alcaide et al., 1997; Schwarzacher, 1997). The surface area of the
NE occupied by telomeres also varies between organisms, ranging from
extremely tight where little intertelomere space is visible (rye and
wheat) to a more loose clustering (maize and lily). It would be
interesting to correlate the tightness of the bouquet with other nuclear
features such as chromosome length, genome size, and presence of Rabl
configuration in interphase. As of this moment, there are no rules for
predicting bouquet morphology.

How is the Bouquet Formed?

 The similarity of the bouquet to the Rabl conformation has long been
noted. However, it is clear that the two are not the same. Some organisms
(Arabidopsis thaliana) display a bouquet but not Rabl, whereas others (D.
melanogaster) display Rabl but no apparent bouquet. In the case of wheat,
the Rabl organization of interphase is tightened to form the telomere
bouquet. The Rabl organization and the bouquet show different degrees of
telomere clustering. The sequence of maize bouquet formation indicates
that telomere clustering does not require a preexisting Rabl organization.  
Maize exhibits a Rabl organization prior to the last premeiotic cell
division.  The Rabl organization is lost during the following interphase,
and during meiotic prophase the telomeres cluster in the bouquet (Bass et
al., 1997). The Rabl organization appears to utilize the centromeres in
addition to the telomeres in some circumstances, whereas the bouquet
appears to rely solely on the telomeres (Martínez-Pérez et al., 1999). The
differences observed between the Rabl organization and the bouquet suggest
that the mechanisms of telomere positioning between the two are also
different. A defining factor of the bouquet in animal cells is the spatial
relationship between the centrosome (the animal microtubule organizing
center (MTOC) and the telomeres during the bouquet (Zickler and Kleckner,
1998). The telomere cluster occurs at a site on the inner nuclear membrane
adjacent to the centrosome position near the outer nuclear membrane.
Plants, not having a defined MTOC, are nonetheless capable of having very
tight bouquets (Loidl, 1990), suggesting that the MTOC position does not
define the telomere cluster position in plants. The functional
relationship between centrosomes and the telomeres during the bouquet in
animals has not been investigated; thus, its importance is unclear.
However, a similar spatial relationship exists during the fission yeast
(Schizosaccharomyces pombe) bouquet, which has proved amenable to
investigation. S. pombe exhibits a prominent telomere cluster during
meiotic prophase. All 12 telomeres cluster in the limited region adjacent
to the spindle pole body (SPB, the fungal MTOC; (Chikashige et al., 1994).

Strong evidence for the conservation of bouquet function has come recently with
the discovery of bouquets in S. cerevisiae (Trelles-Sticken et al. 1999).  A
remarkable degree of conservation of morphology was revealed through chromosome
painting in this organism, including a cloud-to-fiber transition that preceded a
tight clustering of telomeres near the spindle pole body.  More importantly, the
bouquet could be disrupted through mutations in the telomere-binding protein
ndj1, allowing its function to be tested.  Although meiosis completed normally in
ndj1- strains, synapsis was delayed by up to 2 hours.  As pairing is usually
completed within 3 hours, this represents almost a doubling in the time required
for chromosomes to find each other.  Thus even in a small-genome organism like
budding yeast, the bouquet appears to be important for the efficiency of pairing.
 Further studies to elucidate the mechanism of bouquet formation in yeast has
revealed that kar3, a kinesin-like motor with a mutant allele that is defective
in meiotic prophase (Bascom-Slack et al. 1997), is not essential for bouquet
formation (Scherthan et al., 2001).

Another example of a mutation disrupting the bouquet is kms1 in S. pombe (Niwa et
al. 2000).  Normally, the telomeres cluster at the SPB at the onset of mating. 
In the kms1- mutant, the SPB becomes fragmented over the surface of the nuclear
envelope at the beginning of meiosis.  Subsequently, the telomeres become
localized to the individual subclusters of the SPB.  This leads to an increase in
non-allelic recombination.  Additionally, a minichromosome which is normally
unable to pair with the region of its derivation near the centromere due to its
localization at the bouquet site is freed from this constraint in the kms1-
mutant, and is seen correctly paired.  Thus the arrangement of telomeres at the
bouquet in S. pombe imposes strong constraints on the ability of chromosomes to
pair and recombine.

Hints to the status of the bouquet as a specific meiotic process come from
recombination-deficient mutants.  The Spo11- mutant of Sordaria completely lacks
presynaptic alignment of chromosomes found in pre-bouquet wildtype cells (Denise
Zickler, personal communication), yet forms morphologically normal bouquets.  As
Spo11p is required for recombination initiation, this demonstrates that
recombination is not necessary to form the bouquet.  Additionally, maize
possesses several meiotic mutants (dsy498, mms25:dsyCS) which do not homologously
synapse, lack cytologically detectable Rad51 complexes, markers of recombination
(Rockmill et al., 1995), but have morphologically normal bouquets (Inna
Golubovskaya, personal communication).  Thus any model in which the bouquet
emerges from interactions between chromosomes must postulate a
recombination-independent means of chromosome interaction.

The synaptonemal complex and the bouquet

 The SC (see introduction) is an important protein component of meiotic
chromosomes. It forms a core extending the length of each chromosome:  
when two homologous chromosomes synapse, their SCs become joined except at
the most distal, telomeric regions. The ends of the SC (the telomeres)
appear to contact the inner nuclear membrane. Specialized, thickened SC
ends at telomere attachment sites have been documented in a large number
of animals (Esponda and Giménez-Martín, 1972) and in a few higher plants
(Holm, 1977). The thickened SC structures consistently appear to be
conical in shape, the wide end of the cone being the attachment to the
inner nuclear membrane. In a number of species, the inner and outer
membranes appeared to have a layer of increased electron density at the
telomere attachment site (Esponda and Giménez-Martín, 1972;; Rasmussen,
1976; Holm, 1977). In silkworms (Bombyx mori), it has been suggested that
the deposition of the electron-dense material onto both faces of the NE
precedes the attachment of the SC ends to the NE (Rasmussen, 1976).

Importantly, these attachment sites are seen before the bouquet is formed,
raising the possibility of their involvement.  If both the inner and outer layers
of the attachment plaques are modified, this could act as a bridge between the
contents of the nucleus and the cytoskeleton.  Such a link clearly exists in the
case of S. pombe, which undergoes dynein-dependent nuclear oscillations during
meiotic prophase (Chikashige et al., 1994).  The site of nucleus attachment to
the cytoskeleton is the SPB.  If this type of arrangement is conserved among
other organisms, it would suggest the involvement of the cytoskeleton in
establishing the bouquet.

Possible Functions of the Bouquet

 Although the timing and universality of the bouquet suggest a direct and
important role in chromosome pairing, finding evidence for this hypothesis
has proven difficult. It has been recognized that if homology searching is
done primarily near the ends of chromosomes, then limiting the telomeres
to the NE would constrain the homology search to a smaller space, making
it more efficient (Loidl, 1990; Dernburg et al., 1995).  Additionally, the
colinearity of chromosomes in the immediate vicinity of the bouquet would
promote the cooperativity of pairing, as less work is needed to bring
chromosome regions together.  The physical anchoring of a chromosome end
to the nuclear envelope also creates a closed topological domain, a
prerequisite to a model of recombination interference based on imposition
and local relief of tension (Zickler and Kleckner, 1998).  However, this
does not explain the need for nonhomologous telomeres to be in such close
proximity.  Additional theories involve functions quite similar to those
posited for the Rabl orientation, i.e. creating a chromosomal compartment
necessary for some chromosome region-specific function.  Such functions
might include recombination initiation or SC formation, both of which have
chromosome region-specific biases.  Additionally, the extremely tight
bouquets observed in wheat has the effect of concentrating chromosomes in
a small space: 84 chromosome ends are corralled into an area of roughly
30µm2.

The bouquet would promote the alignment of terminal regions of chromosomes, but
not interstitial regions.  Depending on the precise morphology (i.e., persistence
length) of chromosomes, the regions positively affected by the bouquet may be as
small as a few percent of the chromosome length.  The consensus views on synapsis
initiation (Zickler and Kleckner, 1999) indicates that although synapsis
predominantly initiates near chromosome ends, interstitial initiation is also
frequent.  Therefore, the bouquet cannot be necessary for all synapsis
initiation.  However, it may still be necessary for the efficiency of chromosome
synapsis:  the few interstitial sites of initiation that do occur might not
suffice to pair all the chromosomes in every meiotic cell.

Having argued for a central role in the promotion of pairing in meiosis,
it must be pointed out that it is also possible for the bouquet to be a
epiphenomenal outcome, rather than a cause, of the homology search.  In
chapter 4 I will discuss a model in which chromosome-autonomous
interactions suffice to organize a clustering of telomeres, provided that:

1)telomeres are attached to the nuclear envelope. 2)interactions between
chromosomes result in their being temporarily pulled together by a small amount.
3)there is a gradient along the length of the chromosome for likelihood of
interaction, high near telomeres and low near centromeres.

In this model, no special bouquet-promoting force is needed; a morphologically
normal bouquet emerges as the result of interactions along the entire length of
the chromosome.  However, it does not account for all cases of bouquet formation
observed.  I will later argue that the failure of this model to explain all cases
implicates the existence of at least some specialized machinery specifically
necessary for correct bouquet formation, and that chromosome-chromosome
interactions do not suffice to form a bouquet (though they may contribute to it).

The use of grasses as a model system

 All levels of organization mentioned above are orchestrated ultimately by
molecular mechanisms, but particular organisms are better-suited to the
exploration of certain levels than others.  While organisms amenable to
molecular genetics, such as yeasts and Drosophila, have shed much light on
chromatin structure, the structure of chromosomes themselves and their
organization within the nucleus requires an organism more amenable to
cytological analysis.  For this purpose I have used grasses, especially
maize (Zea mays) and wheat (Triticum aestivum) as models for the study of
meiosis.  The advantages of these organisms are: 1) Meiotic cells are easy
to collect and prepare for analysis 2)Meiosis proceeds synchronously
within a single anther, allowing large sample sizes of a given stage to be
simultaneously examined 3)Chromosomes and nuclei are large (100 µm long
and 16µm in diameter, respectively) compared to the maximum theoretical
resolution of light microscopy (0.2 µm), allowing spatial analysis to be
done with accuracy and confidence. 4)The long history of grasses in
cytological analysis has produced a large number of freely available
chromosome derivatives

Maize and wheat are two distantly related species in the grass family (Poaceae),
a lineage of angiosperms first attested in the fossil record from c. 65 million
years ago.  The phylogenetic relationship of maize and wheat is thus comparable
to that of humans and mice.  Although the basic structure of meiosis is the same
between the two organisms, there are some important differences, especially in
nuclear organization.  In the following sections, I describe experiments using
maize and wheat, and computer simulations based on rye, a close relative of
wheat. 

That's the introduction...hopefully soon I will have the whole thesis as a pdf online.
Pete Carlton
07/03/01