The Evolution of Non-Coding Chloroplast DNA and Its Application in Plant Systematics

17 Pages·2000·15.9 MB·

by  Scot A Kelchner

THE EVOLUTION
OF
Scot Kelchner^
A,
NON-CODING CHLOROPLAST
DNA AND
APPLICATION
ITS
PLANT
SYSTEMATICSi
IN
AliS'IKACT
This arlic'le reviews several |)r<)p<)st'(] mechanisms of molecular evolution operating in non-coding regions of ihe
genome
cliloroplast and argues that awareness and identification of tliese mechariisms are essential for improving
alignment and phylogcnctic analysis of non-coding sequence data. The mechanisms are of five categories: slipped-
(1)
strand inispairing; insertions and deletions link<Ml with secondary' structure formations; inversions associated
(2)
(,'?)
with hairpins and stem-loop structures; localized or extra-n»gional intramolecular recombination; and nucleotide
(4) (5)
substitutions. These rruitalions seem to be largely a function of se<|uence structure and palU^rn and may be highly
homoplasious In a parsimony topology; therefore, mutations in non-coding regions of die chloroplasl genome are de-
scribed here as slructure»d, nonrandom, and non-independent events. Established methodologies are based in large part
DNA
on a collective understanding of genie evolution and may need modification when non-coding sequence
appli(*d to
Here
dala. I suggest an approach to die phylogenetic study of non-coding cpDN/V thai incoq)orales identification of
mutational mechanisms in alignment and homology assessment of indels. also discuss rei)ercussions of non-coding
I
sequence evolution for such aspects of phylog<Miy estimation as maximum likelihood, distance, and parsimony analysis,
the inclusion of indels as phyh)genetic characters, and bootstrapping, jackknifing, and "decay" analysis as measures
of clade support.
Key
wonls: alignment, Inlergenic s[)acers, introns, molecular evolution, mutational biases, mutational mechanisms,
phylogenetic secondary
analysis, structure.
There is growing interest in comparative analysis coding regions in the chloroplast: the IrnL-trnF
& &
cpDNA)
of non-coding chloroplast (non-coding se- spacer (e.g., Gielly Taberlet, 1994; Mes t'Hart,
Ham
quences for plant systematic studies low taxo- 1994; van 1994; Sang 1997; Cros
at et al., et al.,
&
nomic levels. Recognition of the limitations of cod- et al., 1998; Bayer Starr, 1998), the trriT-lrnh
DNA
ing (genie) for resolving relationships these spacer (Bohle 1994, 1997; Small
at et al., et al.,
levels inspired the probing of chloroplast introns 1998), the rpoA-petD and rpsll-rpoA spacers (Pe-
&
and intergenic spacers for phylogenetic utility. Un- terson Seberg, 1997), the atpB-rbcL spacer (Go-
&
was
derlying this effort the reasonable premise that lenberg et al., 1993; Hodges Arnold, 1994; Na-
non-coding regions experience limited or no selec- 1995; Samuel 1997; Savolainen
tali et al., et al., et
&
and
tive pressure are likely to evolve at rates far al., 1997; Setoguchi et al., 1997; Hoot Douglas,
& &
surpassing those of genie regions Curtis 1998), the rhcL-psal spacer (Morton Clegg,
(e.g.,
Clegg, 1984; Wolfe et al, 1987; Palmer, 1987, 1993), the psbA-trnH spacer (Aldrich 1988;
et
al.,
&
1991; Olmstead Palmer, 1994; Bohle Sang
et al., et al., 1997), the accD-;>.saI spacer (Small et
1994). There was also an expectation that non-cod- al., 1998), the rp/16-r/?/14 and rp.s8-r/>/14 spacers
ing regions should experience random and inde- (Wolfson et 1991), the intron surrounding ma/K
al.,
&
mode
pendent mutations, both in and distribution. (Johnson Soltis, 1994), the r/>oCl intron (Downie
&
For these s, a remarkable number of plant et al., 1996a, 1996b; Asmussen Liston, 1998;
systematics studies currently in progress include a Downie et al., 1998), the rpll6 intron (Ionian et
al.,
mo &
component
lar of comparative analysis of 1996; Kelchner, 1996; Kelchner Clark, 1997;
1
cpDNA A &
Baum
non-coding sequences. considerable Schnabel Wendel, 1998; 1998;
et
al.,
amount work
of already published has demonstrat- Small et al., 1998), the tniL intron (Sang et al.,
&
ed the potential phylogenetic utility of discrete non- 1997; Bayer Starr, 1998; Kajita et 1998; Bay-
al.,
thank Conny Asmussen, Lyn Cook, Steve Dickie, Jonathan W*Midel, Hich Cronn, and Randy Small many
for
" I
and
interesting insightful discussions on the topic of non-coding secjuence evolution; R. G. Olmstead for his valuable
L
comments
on an earlier version of this paper; R. Rayer, M. D. Crisp, C. West, C. Kelchner, and R. Oxelnum
J. J.
for review of the manuscript; Victoria llollowell for her careful editing and commentar>'; and the anonymous reviewers,
whose comments
insighthil contributed to the clarity and presentation of ideas in this paper.
^Centre CSIRO
for Plant Riodiversily Research. Australian Natioiial Herbarium, Plant Industry, C.PO. Ro\
16()(),
Canberra, A.C.T., 2601, Australia. Second affiliation: Division of Botany and Zoology, The Auslralian National Univer-
sity, (^anlx^rra, Australia, scot. kelchner(£f^pi. csiro.au.
Ann. Missouri Bot. Gaud. 482^98.
87: 2000.
Volume Number
87, 4 Kelchner 483
2000
Non-Coding
Chloroplast
DNA
Evolution
er et al., 2000), the rpsl6 intron (Liden et al., 1997; ble for generating sequence diversity in non-coding
Oxelman ndhA
and
et al., 1997), the intron (Small regions of the chloroplast genome. Unfortunately,
et al, 1998). these hanisms are often invoked, but rarely
in-
The
literature above not only reveals profound corporated, into the analysis.
differences between the evolution of non-genic and Recognition of the potential of structured molec-
cpDNA, cpDNA
genic but critically contradicts initial as- ular evolution in non-coding regions im-
to
sumptions of constraint-free evolution in non-cod- prove alignment and assessment of phylogenetic
re-
ing regions. Recurring difficulties associated with lationships is, I believe, critical for the
non-coding sequence data include alternative development of functional molecular systematic
re-
alignment possibilities of insertions and deletions search based on non-coding sequence Toward
data.
(indels), regions of length mutation in which ho- this end, endeavor here to illustrate the following:
I
mology assessment is questionable or impossible, non-coding regions are highly structured and
(1)
and the occurrence of localized "hot spots" of in- their elements evolve non-randomly and non-in-
ferred excessive mutation, frequently to the point dependently; this structure may be used to align
(2)
How
and
of saturation loss of phylogenetic signal. the sequence matrix and better assess homology;
best to proceed with the phylogenetic analysis of the resulting gaps in the aligned matrix may
(3)
such regions should be a topic of considerable con- contain phylogenetically important information and
cem
Golenberg Downie
(see et al., 1993; et al., should be used in a phylogenetic analysis; and
(4)
&
1996a; Kelchner Clark, 1997; Sang et al., 1997; the mode of non-coding sequence evolution de-
Downie may
et al., 1998). scribed here have potentially serious reper-
now
It is evident that sequence evolution in non- cussions for the accuracy of genetic-distance, max-
imum
coding regions of the chloroplast is far more com- likelihood, and parsimony analyses, and for
A
plex than previously supposed. Both introns and bootstrapping and jackknifing techniques. de-
embody
intergenic spacers are thought to a consid- scription of proposed mechanisms of non-coding
erable degree of sequence structure, sometimes in sequence evolution followed by a discussion of
is
DNA
manner
a similar to that of ribosomal (rDNA). the appropriateness of current alignment and anal-
may
This structure generate either regionalized se- ysis procedures, with the expectation that may
it
quence
conservation or mutational hot spots of both provide a more informed approach the applica-
to
nucleotide and
substitutions insertion/deletion tion of non-coding sequence data in plant system-
Sequence-directed
events. initiators of mutational atics research.
may
events persist as "mutational triggers" (Kel- This article is not intended to be a complete re-
&
chner, 1996; Kelchner Clark, 1997), dramatical- view of literature pertaining to the evolution of in-
ly increasing the possibility of reversal or parallel trons and intergenic spacers in genomes of an
all
gain of mutations, particularly length mutations or organism. Instead, serves as a brief review of
it
minute Hence, cpDNA
inversions. there exist tial vi- current literature on non-coding regions,
olations of the assumptions of randomized and in- and summarizes mutational mechanisms suggested
dependent embedded much
character evolution in to occur in these regions. Discussed are some of
of the current phylogenetic methodology for com- the serious implications this manner of molecular
—
parative sequence analysis methodology that is evolution has for the assumptions underlying mod-
based largely on observational comparative study els employed today by plant molecular systematists.
of coding sequence data. Considering that these are
today's commonly employed phylogeny
tools for es-
MECHANISMS or NoN-CoDi\G Sequence
DNA
timalion based on sequences, there has been
Evolltion
remarkably
as yet little controversy in the literature
about their application to non-genic sequence data. The strength of any phylogenetic estimation rests
There are ways to account for mutational patterns on the accuracy of character homology assessment,
DNA.
observed in non-coding Comparative studies Thus, the molecular systematist strives maximize
to
cpDNA
of non-coding sequences during the past character homology by the careful alignment of
DNA
decade in particular (e.g.. Palmer, 1985; Blasko et sequences in a data matrix. Fundamental to
&
vom cpDNA
1988; Stein Hatchel, 1988; Wolfson any alignment procedure of non-coding
al., et se-
&
al., 1991; Golenberg et al., 1993; Gielly Taberlet, quence data should be a familiarity with mutational
1994; Morton, 1995a; Downie et al., 1996a; Kel- mechanisms directing molecular evolution in non-
& &
chner Wendel, 1996; Kelchner Clark, 1997; coding regions. Recognition of these mechanisms
Sang et al., 1997) have allowed inference of spe- as generators of specific mutations can be a pow-
cific underlying mutational mechanisms responsi- erful tool for the placement of gaps and for the
484
Annals
the
of
Garden
Missouri Botanical
assessment of probable homology of insertions and between the probability of inserting subsequent
&
deletions (Kelchner, 1996; Kelchner Clark, length mutations and the probability of removing
1997). sequence from the repeat Whether such an
string.
may
equilibrium is present or not, there be a com-
phenomenon
keeps
petitive that the length of tan-
SUri»l::D-STKAND MISPAIRING (SSM)
dem
repeated sequence units continually in
flux.
A
mechanism
widely reported of length mutation Representation of long repeat strings in non-coding
in non-coding regions of the chloroplast is slipped- sequence alignments would therefore be a "snap-
SSM
strand mispairing (SSM). is thought to be a shot" of sequences experiencing continual inser-
major, even principal, factor in length mutations tions and deletions that
at locality.
within non-coding regions of the chloroplast, mi- follows that a point substitution within a long
It
&
tochondrial, and nuclear genomes (e.g., Levinson string of mononucleotide repeat units could act as
Gutman, 1987; Hancock, 1995; Wolfson
et al., a stabilizing factor, disrupting previous unifor-
its
&
1991; Kelchner Clark, 1997; Sang et al., 1997). mity and lowering the probability of further SSM
Length mutations are important components of non- events. Such a substitution would directly influence
coding sequence evolution and have been suggest- ensuing mutations in the region and one example
is
ed to occur at least as frequently as base substi- of a non-independent character mutation non-
in
some DNA.
tutions in chloroplast non-coding regions coding the situation were reversed, with a
If
&
(Curtis Clegg, 1984; Wolfe 1987; Zurawski non-homogeneous sequence becoming
et al., a string of
& &
SSM
Clegg, 1987; Clegg Zurawski, 1992; Golen- repeat units, the likelihood of an event would
&
berg al, 1993; Gielly Taberlet, 1994; Clegg and could induce non-independent
et further
et al., 1994). mutations by the addition or removal of repeated
Slipped-strand mispairing is thought to proceed sequence by slipped-strand mispairing.
DNA
by a localized mispairing of single-stranded As an aid to alignment, SSM-generated inser-
in regions of sequence repeats, as either a string of tions and deletions can be used position and
to
A
mononucleotide repeats or tandemly arranged mul- determine number of gaps. quick study of a re-
tibase repeat units (Palmer, 1991; Wolfson et peat unit or the flanking sequence of a gap may be
al.,
Cummings
Hancock,
1991; et al., 1994; 1995; re- enough to determine if slipped-strand mispairing
i
&
viewed by Levinson Gutman, Diagrams
1987). of the likely progenitor of an observed length muta-
SSM
proposed mechanics can be found Levinson SSM may
in tion. Occasionally, evidence of an event
Gutman
and (1987) and Wolfson et al. (1991). Be- not be apparent, particularly a deleted sequence
if
cause A/T-rich regions of bacterial genomes are not a direct repeat of flanking sequence, or
is its if
particularly susceptible to slipped-strand mispair- a subsequent length mutation due another mech-
to
&
ing (Levinson Gutman, 1987), one could expect anism obscures an earlier SSM event (Kelchner,
a similar effect in the A/T-rich non-coding regions 1996).
genome
of the chloroplast (Wolfson 1991).
et
al.,
SSM A
This is not to imply that acts uniquely on
STEM-LOOP SECONDARY STRUCTURE
T
and nucleotides; aligned non-coding sequenr
('
G
matrices often infer inserted repeats containing Striking both intergenic spacers and introns
to
C G
and nucleotides, sometimes as pure strings of in the chloroplast genome the presence and num-
is
C
mononucleotide
or repeats. ber of probable secondary structures referred as
to
mononucleotide
Strings of repeats, particularly of "stem-loops." Stem-loops are believed occur dur-
to
A
cpDNA,
or T, appear frequently in non-coding ing single-stranding events when inverted repeats
may
and
slipped-strand mispairing potentially gen- meet to form a region of pairing (the stem) sur-
erate length amtations within these strings. The dif- mounted by their interceding sequence (the loop).
ficulty in assessing homology of length variation in Such structures have been widely discussed
for
ri-
DNA, rDNA
long strings of repeats, whether mononucleotide or bosomal with ITS and 18S regions be-
multinut^leotide repeats, derives from the increas- ing of particular interest to the plant systematist
ing potential for further length mutation relative to (see Baldwin et al. (1995), Soltis et al. (1997), and
& &
Owen,
string length (Streisinger 1985; Golenberg Soltis Soltis (1998) for discussion of secondary
&
et al., 1993; Kelchner Clark, 1997; Sang et structures in these regions and their phylogenetic
al.,
SSM
may
Subsequent
1997). activity either gener- implications).
ate additional repeats of the initial sequence or de- Probable stem-loop dary structure com-
is
5
lete sequence susceptible to slipped-strand mis- monly reported in non-coding regions of organellar
pairing. Perhaps an equilibrium might exist genomes Michel 1989; Buroker
(e.g., et al., et al..
Volume Number
4
87, Kelchner 485
2000
Non-Coding
Chloroplast
DNA
Evolution
Ham
1990; Golenberg et al., 1993; van et al., 1994; substitutions and length mutation, the inverted re-
&
Gielly Taberlet, 1994; Natali et al., 1995; Rigaa peated sequence composing the stem is frequently
&
Downie
et al., 1995; et al., 1996b; Kelchner conserved in character (Learn et 1992; Gielly
al.,
& &
Wendel, 1996; Kelchner Clark, 1997; Sang et Taberlet, 1994; Downie 1996a, 1996b;
et
al.,
&
al, 1997; Downie et al, 1998). Gielly and Taberlet Kelchner Clark, 1997), particularly when stems
(1994) reported several probable stem-loops in the are long and possess highly favorable energy of
for-
&
trnL-trnF region of the chloroplast genome, includ- mation values (AG values; see Kelchner Wendel,
A
ing nine highly probable structures within the trnL 1996; Dumolin-Lapfegue sequence
et 1998).
al.,
intron itself. All other introns in the chloroplast ge- involved in stem formation is less available for sub-
nomcs
of land plants are classified as Group II in- stitution and length mutation because paired
is
it
and
trons share a diagnostic secondary structure of with its sister repeat; this can engender non-
six well-defined stem-loop domains (Kohchi domly and non-independently evolving sequence
et
al.,
1988; Michel 1989; Downie 1996b;
et al., et al., units.
RNA
Downie et al., 1998). Diagrams of putative single- Similar to ribosomal and rDNA secondary
&
may
stranded secondary structure of introns be structure Curtiss Voumakis, 1984; Wheeler
(e.g.,
& &
&
found in Michel and Dujon (1983), Michel et al. Honeycutt, 1988; Dixon Hillis, 1993; SoUis
and Downie
(1989), et al. (1998). Soltis, 1998), a nucleotide substitution occurring in
Loop cpDNA
regions of stem-loop secondary structures a stem sequence of a non-coding region
are often associated with hot spots for mutation in could compromise secondary structure formation,
non-coding regions, both of nucleotide substitutions Compensatory mutation may then occur pre
to
&
and indel events (vom Stein Hatchel, 1988; Al- the potential for structure formation (Kelchner,
& &
drich et al, 1988; Golenberg et al., 1993; Gielly 1996; Kelchner Clark, 1997). Although se-
Ham
Taberlet, 1994; van et al., 1994; Clegg et al., quence conservation may be present merely as a
1994; Ferris et al., 1995; Downie et al., 1996b; function of sequence pattern (perhaps the case in
&
Kelchner
Clark, 1997). Indels located in prob- intergenic spacers), the degree of secondary struc-
able loop sequence are frequently inserted or de- ture conservation in a chloroplast Group intron
II
SSM. How-
leted repeat units likely the result of suggests secondary structures are integral to proper
ever, length mutations not attributable to functioning of the intron (Clegg 1986; Learn
et
al.,
slipped-strand mispairing often occur within loop 1992; Downie 1996a). Experimental
et al., et al.,
may
sequences as well and be remnants of recom- evidence has shown some of this structure es-
is
bination events. mechanisms Group
sential for auto-splicing in
I
common
Although indels are most in the termi- and introns (Bonnard 1984; Kohchi
II et al., et al.,
&
may
nal loop, they occur anywhere along a second- 1988; Dujon, 1989; Cech, 1990; Michel Westhof,
ary structure. For example, Kelchner and Clark 1990; Hibbett, 1996).
(1997) detected what appeared to be an entire de- Identification of probable secondary structure
letion of a small sub-loop positioned partway up the can be valuable when aligning and analyzing non-
stem of an rpll6 intron stem-loop in Oryza sativa. coding sequences by improving gap positioning and
Such side loops, when present, may be removed in the appraisal of character homology. Gaps flanked
some taxa without compromising by and G
the favorability of inverted repeats regions
relatively rich in
a stem formation. Occasionally, small segments of and C content are suspect as possible stems of see-
the stem itself will be deleted, decreasing the stem ondary structures. As noted, regions of chaotic
though perhaps an would
length, not to extent that length mutations are correlated with loops, so the
annihilate possible secondary structure formation. boundaries of a chaotic region frequently
will cor-
Very large loops are often associated with regions respond with inverted repeats that can form a stem,
of chaotic or "labile" length variation characteristic even they do not directly neighbor the chaotic
if
cpDNA
many OLIGO
of non-coding sequence matrices Computer programs such
region. as (Ry-
&
MULFOLD
(e.g., Golenberg et al., 1993; Downie et al„ 1996a; chlik Rhoads, 1989), (Jaeger et al.,
&
Baum
Soltis et al., 1996; Kelchner Clark, 1997; 1989; Zuker, 1989), and GCG's Stemloop (Genetics
Homology
et 1998). assessment here can be Computer Group, Madison, Wisconsin) can
al., assist
and
difficult or impossible, the conservative ap- in the detection of secondary structure in non-cod-
A
proach of removing these regions from the data ma- ing sequences. search can be conducted by hand,
trix before phylogenetic analysis is frequently particularly if a published data set exists for the
adopted. region. Free energy of formation values (AG) can
In contrast to the loop of stem-loop secondary be calculated with some of the prior software as an
structures being highly susceptible to nucleotide appraisal of the likelihood of formation of a partic-
486
Annals
the
of
Garden
Missouri Botanical
& & &
ular secondary structure (see Kelchner Wendel 1987; Zurawski Clegg, 1987; Clegg Zu-
al.,
AG &
(1996) lor an example where values were ap- rawski, 1992; Golenberg 1993; Gielly Ta-
et
al.,
plied to parallel inversion events in their data). berlet, 1994; however, see Small et 1998).
al.,
AT
Percent content quite variable in non-cod-
is
MIM
cpDNA
INVERSIONS
IK ing regions, though generally higher
is
it
than the average value for the chloroplast genome
Minute inversions of four to six base pairs have
&
(Shimada Downie
Sugiura, 1991; 1996a;
at al.,
been linked small stem-loop secondary
to struc-
AT
Small et al., 1998). Because of their high con-
tures conmionly referred to as hairpins (Kelchner
make
non-genic regions must
tent, a significant
&
Wendel, 1996). Hairpins consist of a stem com-
A
contribution to the high overall frequency of and
posed of nearly adjacent inverted repeats producing
T
in the chloroplast genome. Kajita (1998)
et al.
stem-loop
a structure with a particularly small loop.
AT 67%
reported an content of the IrnL-tniF
in
may become
This loop inverted by recombination,
spacer and IrnL intron, Kelchner and Clark (1997)
and the inversion may be so small that either
it
AT
70.5%
reported composition
in the intron of
&
escapes notice during alignment (Kelchner Wen-
chloroplast gene rpll6 bamboos, and Small
in et al.
&
del, 1996; Kelchner Clark, 1997), or the inverted
AT
77.1%
found an
(1998) incredible content
in
sequence matches particular bases of the uninvert-
the intergenic spacer trnT-trnL in G<\ssypium. Un-
ed sequence, resulting in a confusing array of mi-
AT
doubtedly, unequal tendency toward
this rich-
nute gaps Golenberg
(see et 1993).
al.,
DNA
ness in non-genic chloroplast has several as
Identifying minute inversions can require careful
undetermined
yet implications for phylogenetic
when
sequence
attention aligning data, particularly
analysis of non-coding sequence data. At a mini-
alternative gap weighting schemes of an align-
if
mum,
introduces a strong base composition bias
it
ment program have been
not rigorously explored.
into the analysis.
Candidates hidden
for a inversion severa ad-
1
may
Substitutions demonstrate rather high levels
jacent nucleotide
substitutions, a series of tiny
cpDNA
of homoplasy in non-coding regions due
to
gaps, or a gap that demonstrates no repeat aspect
the frequency of inferred multiple-hit sites (nucle-
to its sequence structure. Alternatively, one could
otide sites experiencing multiple substitution
investigate these probable secondary structures by
events). Multiple-hit occur even very low
sites at
hand
or with a secondary structure computer pro-
estimates of percent sequence divergenc-e (Kel-
gram. Failure recognize minute inversions a
to in
&
chner, 1996; Kelchner Clark, 1997), suggesting
sequence data set has several repercussions for
accepted coding
that the region estimates of
phylogenetic analysis, discussed fully in Kelchner
10-15%"
"around sequence
divergence optimal
for
and Wendel and summarized
(1996) here Anal-
in
may
phylogenetic signal be inadequate measures
Non-Coding Sequence
ysis of Data.
for phylogenetic utility of a non-coding region.
Finally, small inversions associated with hairpins
Precise understanding of mechanisms underlying
may
be highly susceptible to reversal and parallel-
DNA
multiple-hit substitutions in non-coding
is
within a study group, even the interspecific
at
lacking. However, attributes of the molecular evo-
& &
level (Kelchner Wendel, 1996; Kelchner
non-coding manner
lution of regions influence the
Clark, 1997; Sang 1997; Dumolin-Lapegue
et al.,
of nucleotide mutation or the distribution of nucle-
et al., 1998). This susceptibility to reversal or par-
otide substitution events in an intron or intergenic
due
allelism to the persistence of the mutational
is
—
spacer. Stem sequence and loop regions may
dif-
&
trigger (Kelchner Clark, 1997) the nearly ad-
—
ferentially permit mutations, resulting in non-ran-
jacent inverted repeats after the initial inversion
domly and non-independent
distributed nucleotide
event.
substitutions. Statistical significance of differential
may
mutation rates in loops relative to stems be
NUCLKOTIDE SUBSTITUTIONS
tested for an adequate distribution model (see 01m-
Nucleotide substitutions rally reported stead et al. s (1998) test for stochastic mutation in
common
as being more in non-coding than in cod- the chloroplast genes ndhF and rhcL), yet has rare-
&
cpDNA
ing regions (Wolfe et al., 1987; Zurawski Clegg, ly, if ever, been performed on non-coding
& &
1987; Olmstead Palmer, 1994; Hoot Douglas, data
sets.
1998; however, see Sang et al., 1997, for an excep- In addition to secondary structure affecting the
number
random
tion). Surprisingly, a of studies report nu- distribution of nucleotide substitutions,
cleotide substitutions as being just equal to or less there may be constraints on the type of mutation
frequent than length mutations in closely related an individual experiences. For example,
site there
&
taxonomic groups (Curtis Clegg, 1984; Wolfe at is a correlation between transilion/transversion ra-
Volume Number 4 Kelchner 487
87,
2000 Non-Coding
Chloroplast
DNA
Evolution
and neighboring base composition in non-cod- certain inserted or deleted tandem-repeat length
tios
ing regions (Morton, 1995a, b; Morton et aL, 1997; mutations (Palmer, 1985; Blasko et al., 1988).
Savolainen et al., 1997). The correlation suggests However, Wolfson et al. (1991), Sang et al. (1997),
A T SSM
that nucleotides flanked by and/or will dem- and Kelchner and Clark (1997) suggested
is
more mechanism
onstrate a significant tendency toward transversion a likely for length mutation in
mutations. Such a tendency limits possible nucle- their studies of chloroplast introns and intergenic
otide replacements at these sites, increasing the spacers,
chance of parallelism and reversals, particularly
if
One
would
the site experiences multiple hits. also ALIGNMENT
be more com-
expect transversion substitutions to
mon AT many
in data sets of high content. There are philosophies for sequence align-
much
ment, and of the literature centers on the
proper application of computer software for this
INTRAMOLECULAR RECOMBINATION
The
purpose. structure present in a non-coding
cpDNA
Intramolecular recombination on an extra-re- sequence makes an excellent example for
it
gional or genomic scale has been suggested be- discussing what believe to be the fundamental
I
tween adjacent or nearby repeats in the chloroplast problem of most computer alignment programs: de-
genome (Howe, 1985; Palmer et aL, 1985; Palmer fining the nucleotide as a discrete and independent
The
et al., 1987; Blasko et al., 1988; Ogihara et al., character. identification of secondary structure
&
Kanno and mechanisms may
1988; Milligan et 1989; Hirai, 1992; mutational in the data greatly
al.,
& &
Kanno
1993; Morton Clegg, 1993; Hoot improve on current algorithmic alignments of gaps.
et
al.,
Palmer, 1994). In the context of non-coding se- and thus on assessment of character homology.
Many
quence comparison, such a large-scale recombi- have found software, particularly versions
CLUSTAL
Thompson
nation involving the particular region of study could of (Higgins et 1992; et
al.,
result in indels of surprising size that contain se- 1994), to be of help at least initially with the
al.,
quence content not readily identifiable in origin. alignment of non-coding sequences. The alignment
may
Recombination events operate on a then subjected an "improvement by hand"
finer is to to
scale within a discrete non-coding region. Occa- position gaps (e.g., Samuel et al., 1997; Downie et
&
sionally one infers extensive deleted sequence in al., 1998; Bayer Starr, 1998; Kajita et al., 1998).
an alignment with no apparent mechanistic expla- This procedure saves time the sequences are sim-
if
when become numerous
nation, presence of a small or moderately sized in- ilar in length, but indels
version, or a large insertion showing congru- in the data matrix the difficulties of alignment dra-
little
ence with surrounding sequence pattern. Such matically increase. This because most alignment
is
mutations suggest intramolecular recombination, software initially regards each character in the ma-
and they frequently occur in the loop regions of trix as an independent unit, unless otherwise spec-
probable secondary structures. Sequences involved ified by particular position or gap weighting
in stem-loops may be particularly susceptible to re- schemes defined by the user. The software is in-
when
combinatlon events due the conserved inverted capable of determining mutations other than
to
repeats and mutationally flexible loop. Therefore, substitutions have arisen, such as non-independent
such structures could experience interactive recom- insertions, deletions, or inversions correlated with
SSM
bination with other stem-loops, particularly with and secondary structure. Appropriate weight-
those existing in complementary sequence position. ing for these mutations that could be incorporated
Recombination involving the entire loop of a sec- into an alignment algorithm at present, unde-
is,
may
ondary structure occur, particularly in struc- veloped.
minute mod- The method Wheeler
tures with long stems, resulting in or Elision of et al. (1995) at-
erate-sized inversions in both intron and intergenic tempts to improve gap placement and indel homol-
&
spacer regions (Natali et 1995; Kelchner ogy by alignment software. The Elision method uses
al.,
&
Wendel, 1996; Kelchner Clark, 1997; Sang et standard alignment algorithms to produce a series
Such homoplasious competing alignments based on varying gap
1997). incidents are often of
al.,
& &
(Kelchner Wendel, 1996; Kelchner Clark, weighting schemes. These competing alignments
1997; Sang et al., 1997; Dumolin-Lapegue et al., are then combined in a single matrix and an anal-
1998) due to the persistence of the mutational trig- ysis is performed, with the effect that support is
ger; in this case, the hairpin stem. increased for aligned regions that most frequently
among
Intramolecular recombination a notable appear the various gap-weighting schemes,
is alter-
native to slipped-strand mispairing as a source for This method aims at objectivity, but makes no im-
488
Annals
of the
Garden
Missouri Botanical
provement on the alignment algorithm's inabihty alignment issues: examples from non-coding
to
assess mutation types other than independent point cpDNA DATA
substitutions. Mutations in non-coding regions are
influenced by surrounding sequence structure and
&
Here present examples (Kelchner Wendel,
I
frequently occur not as independent base mutations
&
1996; Kelchner Clark, 1997; Kelchner, unpub-
but as linked multinucleotide mutation events, like
lished data) to illustrate the inference of mutational
the insertion of a repeat unit (Kelchner, 1996; Kel-
cpDNA
mechanisms non-coding sequences and
in
&
chner The many
Clark, 1997). likelihood that
demonstrate the practice of applying mechanistic
non-coding mutations are derived from sequence
_*jji*j*
»*u*
_.j explanations alignment and homolocv assess-
to
r
^
iragments that are m• serted, deleted, mverted, or
...
ment. Nuclieoti• dies i• n li ower-case bi oli di pri• nt are in-
ol,1herwi.se rearranged1, negat.es xtihe assumptXi* on r
ol
. . ,
ferred insertions; underlined leotides indicate
discrete, independent nucleotide characters under-
probable progenitor sequence of an insertion
*^^ or,
lying alignment algorithms, as well as any ex-
all
Examples 4 and
in 5, call attention to a particular
tension of those algorithms like the Elision method.
sequence
mterest.
of
minimum,
At a those using sequence alignment
A
common
type of insertion in non-coding
programs homology
establish putative of char-
to
^P^^A
a direct repeat of a neighboring sequence
is
acters in their data matrix should experiment with
&
1^" Golenberg Hoot
al, 1993;
et
a wide variety of gap-weighting options. These op- C'^yP^ g^P'
may mu- Douglas, 1998). These often take the form of vari-
tions, however, not reveal the underlying
able-length strings of a mononucleotide repeat unit
mechani occasionmg sequence
tational rear-
^
(Example
rangements in chloroplast non-coding regions. They 1).
may, however, facilitate the rapid alignment of seg-
ments of the matrix that share consistent sequence EXAMPLE
1.
and
integrity thus pinpoint regions of variable
TTAAAAAAAAA TTGA
1.
length that require special consideration.
Alternatively, some have avoided alignment pro- TTAAAAAAAAAA--TTGA
2.
grams entirely and describe aligning sequences by
TTAAAAAAAA TTGA
& 3.
hand Golenberg 1993; Hodges Ar-
(e.g., et al.,
&
nold, 1994; Kelchner Clark, 1997). This ap- TTAAAAAAAAAAAATTGA
4.
proach facilitates a careful study of the matrix as
forms and increases the researcher's familiarity
it
Homology can be highly uncertain for these re-
with mutations in the sequences. However, align-
nucleotides. Therefore, such regions are
ei-
ment by hand, especially when dealing with con- P*^^*^^
removed from consideration as potential phy-
^^""^
siderably divergent taxa or with the presence of a
logenetic characters conservative approach) or
(a
number
great of length mutations, can be tedious
included as coded gap characters corresponding to
and time consuming.
becoming
Kelchner and Clark (1997) suggested that aware- ^'^'^^^ ^^ ^he repeat string (often highly
homoplasious
in the context of a resulting topology).
ness of the proposed mutational mechanisms active
Uncertainty of homology exacerbated by potential
in non-coding regions can be useful for inferring is
PCR
inaccuracies of enzymatic processes during
and and
positioning gaps ultimately assessing
in
amplification and sequencing, which can gen-
homology. Golenberg were also
(1993) the
et al. first to
erate variable-length repeat strings independent of
detail a criterion for aligning gaps in non-coding
When
cpDNA sequence
matrices. Based on their example, Kel- the template's constitution. strings
chner (1996) and Kelchner and Clark (1997) mod- of adjacent mononucleotide repeats are highly var-
dth
ified the alignment criterion for chloroplast rpllG iable in length in a matrix and h or excee
intron sequences. Hoot and Douglas (1998) also re- range demonstrated above, they become more likely
SSM
vised Golenberg et al.s (1993) method of gap align- to experience further mutation. For this rea-
ment, framing the beginnings of a nomenclatural son, is perhaps most reasonable to remove such
it
procedure for defining gap categories. Although a areas from consideration in a phylogenetic analysis.
nomenclatural system is not requisite for gap treat- Insertions can also be muhinucleotide repeat
ment in a phylogenetic analysis, may be useful units of a neighboring sequence, as demonstrated
it
Example
in collating information of inferred mutational in 2 by the inserted repeat unit ataaa
&
;hanisms if universally applied in non-coding ("Type lb" gap; Golenberg et al., 1993; Hoot
DNA
studies. Douglas, 1998).
Volume Number
4
87, Kelchner 489
2000
Non-Coding
Chloroplast
DNA
Evolution
—
EXAMPLE
common
2. repeat unit a mutation type non-cod-
in
B C
ing regions. alignment options or were used
If
ATAAAACAAA GAGCG
1.
for phylogenetic analysis, the content of the inser-
ATAAAATAAAataaaGAGCG
2. tion would be of unexplainable origin (though
still
and
possible) the potential of incorrectly assessing
ATAAAATAAA
GAGCG
3.
nucleotide homology in the region may be consid-
ATAAAATAT^ GAGCG erable.
4,
Any
gap
of the positions in this particular ex-
An
inserted repeat of this nature could be exten-
ample would
not affect a topology generated from
may
sive in length and be difficult to recognize as
may
these four taxa, but gap positioning have a
a repeat unit during alignment example, have
(for
I
more
significant effect in a larger matrix of distantly
73 bp
identified a inserted repeat [unpublished
The
related taxa. position of the gap in alignment
data] in the trnT-trnL intergenic spacer Myopor-
in 3A
and may
detection of the repeat unit also be
A
by
aceae). repeat unit very nature shares nu-
its
relevant in determining a weighting scheme
for
and
cleotide content order with flanking sequence;
these non-independent
characters.
may
therefore, multiple gaps be inferred by pairing
may
Length mutations overlap with one another
segments an
of inserted repeat with progenitor
its
to create a progressive-step indel. In the more ex-
sequence. This particularly problematic the
is in-
if
treme homology
cases, appraisal of in these regions
sertion or progenitor has experienced subse-
its
can be very or impossible (Palmer
difficult, et
al.,
quent nucleotide
substitutions. &
Downie
1985; 1996b; Kelchner
et al., Clark,
Even when
a single gap inferred, positioning
is
Example 4
1997). demonstrates a probable pro-
may
of the gap hide evidence
that the insertion
is
which
gressive-step indel in two possible place-
Example
a repeat unit. 3 reproduced from Kel-
is TTGA.
ments
exist for the repeat Note that the un-
chner and Clark and demonstrates how
(1997) a
derlined sequence a direct repeat of the
is
may
repeat unit be obscured in a sequence matrix. TCGTAATTGA
preceding sequence
in the matrix.
EXAMPLE
EXAMPLE
3.
4.
A.
AATCGTAATTGA AACAGA
-
1. -
GGTTATGA ATTAACA
1. AATCGTAATTGA AACAGA
2.
GGTTATAA ATTAACA AATCGTAATTGA TCGTAATTGA AACAGA
2.
3.
GGTTATAA ATTAACA AATCGTAATTGA TCGTAATTGA AACAGA
3. 4.
GGTTATAA ATTAACA AATCGTAATTGA TCGTAATTGA ttgaAACAGA
4.
5.
TTGA
B If part of the underlined in sequences 3
and 4 moved from
is its current position to align
GGTTAT GA ATTAACA
1.
with ttga sequence
in the possibility that the
5,
sequence
ttga a direct repeat of the preceding
is
GGTTAT AA ATTAACA
2.
may
sequence be obscured; however, alignment
this
GGTTATAA ATTAACA choice would not be impossible. As the preceding
3.
sequence to the underlined 10 bp repeat does not
GGTTATAA ATTAACA
we
4. contain this additional ttga repeat, can infer
that two separate events have given an
rise to initial
C
10 bp insertion in sequences 3 and followed by
4,
GGTTA TGA ATTAACA an additional 4 bp insertion In sequence Wheth-
5.
1.
TTGA
er ttga itself or the preceding the sub-
is
GGTTA TAA ATTAACA
2. sequent inserted mutation is impossible to deter-
mine. In this case, either alternative alignment of
GGTTATAA ATTAACA
3. TTGA
the unit would cause no effect in a phylo-
genetic analysis; most important here
is to dis-
it
GGTTATAA ATTAACA
4.
cern the two length mutation events. any poten-
If
Alignment possibilities A, B, and C were equally tially informative nucleotide substitutions were
W
CLUSTAL
probable using (Thompson et al., present in either of the repeat units in Example 4,
A
Only
1994). alignment reveals the insertion a these substitutions should be excluded from a phy-
is
490 Annals
the
of
Garden
Missouri Botanical
logenetic analysis on the basis that nucleotide ho- sequence Sequences 2 and 3 probably share a
1.
mology of the repeats not discernable. similar origin as a repeat of the preceding sequence
is
The example above homology may TTAAT. The Example
suggests that events, aligned as they are in
A
be indicated by the length of insertions or deletions 6A, are probably non-homologous. re-alignment
in a gap, although such an assumption is not with- could be performed to accommodate the two sepa-
out risk. Example 5 below demonstrates multiple rate indel events (Example 6B), even though the
possible alignments of the gait repeat unit (repre- insertions are of the same length and the alignment
&
sented individually by sequences and with infers an additional gap (see Hoot Douglas,
2, 3, 4)
sequence
the insertion in 1998).
1.
EXAMPLE
EXAMPLE
6B.
5.
L
AGATTGATTGATTATTATACTGATTATGC GGTTAAT tctat TCTATCT
C
1
.
GGTTAAT ttaat TCTATCT
2.
CAGATT ATGC
2.
GGTTAAT ttaat TCTATCT
CAGATTgatt ATGC 3.
3.
GGTTAAT TCTATCT
4.
CAGATT gatt ATGC
4,
GGTTAAT TCTATCT
CAGATT ATGC 5.
5.
There a hazard that minute inversions (Kel-
is
Again, actual homology impossible assess
to
is
&
chner Wendel, 1996) can be completely ob-
GATT
with confidence, for there exist three repeat
scured in a matrix they introduce no gaps during
if
units in the insertion in sequence 1. In cases like
alignment, particularly alternative gap-weighting
if
homology on
often inferred the basis
this, is oi
schemes have not been rigorously pursued. pre-
If
minimum number
and gaps
length of indel of re-
sent and unrecognized in a data matrix, minute in-
quired to position the repeat. Hence, the gatt re-
may
versions overweigh a particular mutation by
peats in sequences and 4 would be aligned
2, 3,
interpreting the single mutation event (an inversion)
one above the other and on one side of the gap to
apomorphies
as multiple of adjacent nucleotide
When
number
reduc^e the of inferred indel events.
Example below
substitutions. 7 illustrates a situa-
coding indels as characters, this would be a rea-
tion in which sequences 2 and 3 share the inversion
sonable solution in lieu of other evidence for indel
TTGG CCAA &
(from Kelchner Wendel, 1996).
to
origin, and the repeat gatt would be treated as ho-
EXAMPLE
mologous ior those sequences that contain
it. 7.
may
Equal be
length of insertions not strong ev-
TAATATT TTGG AATATTA
1.
homology
idence of their (Kelchner, 1996; Kel-
& &
chner Clark, 1997; Hool Douglas, 1998). Con- T7VATATT CCAA AATATTA
2.
Example 6A.
sider the insertions in
TAATATT CCAA AATATTA
3.
EXAMPLE
6A.
TAATATT TTGG AATATTA
4.
GGTTAAT tctat TCTATCT
1.
TAATATT TTGG AATATTA
5.
GGTTAAT TCTATCT
2.
If the inversion is of sufficient length to introduce
multiple gaps in the matrix (see Golenberg
et al.,
GGTTAAT ttaat TCTATCT
3.
1993; Sang el 1997), two possibilities can oc-
al.,
GGTTAAT TCTATCT
4. cur: the gaps will be misaligned to parts of the in-
verted sequence sharing spurious sequence simi-
GGTTAAT TCTATCT
5.
larity with the uninverted sequences; there will
or,
6A
Alignment of the insertions in Example re- be inference of an inserted sequence of unknown
sults in the probably mistaken homology of indels origin the inverted nucleotides), which
(in reality,
in sequences 2 and 3 with that of sequence The corresponds with a deletion in the homologous un-
1.
insertion in sequence 1 likely arose from an in- inverted sequences. Each possibility will lead to
may
serted repeat of the sequence to the right of the inaccurate assessment of homology and poten-
gap, TCTAT. This would be a more parsimonious tially have a considerable effect on phylogeny es-
number
explanation, in terms of total of mutation timation.
many
events, than to infer a single inserted repeat fol- Regions in the matrix demonstrating in-
lowed by two adjacent nucleotide substitutions in dependent variable-length insertion and deletion
Volume Number
4
87. Kelchner 491
2000
Non-Coding
Chloroplast
DNA
Evolution
events will likely be associated with secondary mutations, and multiple-hit point substitutions, any
may
structures, specifically with loop regions of stem- of which obscure evolutionary
history.
loops (Kelchner, 1996; Downie et al, 1996b; Kel- Inversions may show high levels of parallel-
(3)
&
chner
Clark, 1997). Identification of flanking se- ism and reversal, and their phylogenetic may
utility
form not be particularly robust. Undetected minute
in-
locate the boundaries for the gion and aid in versions may be buried within a data matrix and
aligning the Indels. Discerning probable SSM-sus- consequently treated as multiple base substitution
informat synapomorphies
instead of a single mutational
ence and
of parallel reversed insertions or dele- event.
may
tions. Nucleotide substitutions be under pe-
(4)
Perhaps methods gap
of or character weighting culiar constraints not fully understood. There ev-
is
and alignment based on mechanisms
of mutation id of a bias in non-coding regions involving
1
can be incorporated into software designed for non- transition/transversion substitution ratios due to the
A
coding sequence alignment, particularly by includ- influence of neighboring bases. particular base
AG
an may
ing evaluation of values for proba})le sec- experience substitution events multiple times
ondary However,
structures. the diversity of rates in closely related lineages, reaching saturation long
and types of molecular evolution in non-coding re- before the expected saturation level for the remain-
gions may be profound. As with coding DNA, we ing sequence. A base-composition bias toward A/T
are far from understanding all forces directing non- content clearly present in non-coding cpDNA.
is
coding molecular evolution to a degree that we can. Selective pressures exerted on non-coding
re-
with any may
certainty, assign probabilities to individual gions be largely a function of the physical
mutations. sequence and
structure of the possible functionality
Considering that alignment of sequence data is of introns and intergenic spacers. Reliance on
fundamental to the entire phylogeny estimation pro- methodology developed coding sequence, which
for
cess, authors should more fully describe the steps includes estimates of constraints on coding se-
taken to align their sequence data in order to pro- quence evolution, transition/transversion and
ratios,
vide necessary information for the assessment of mutation probabilities, inappropriate the
is for
proposed
their reconstructions of phylogenies. analysis of non-coding regions.
Phylogenetic estimations based on genetic
dis-
cpDNA
Analysis of Non-Coding Sequence Data tance measures of non-coding sequences
must be approached with
care. Superficial appli-
The mechanisms of evolution described above cation of models for maximum likelihood (ML; Fel-
number &
have
a of significant implications for the senstein, 1981) or neighbor-joining (NJ; Saitou
phylogenetic analysis of non-coding sequence data. Nei, 1987) could easily produce erroneous phylo-
Among
these are the following: genetic estimations several key assumptions un-
if
Shpped-strand mispairing can be
(1) the result derlying the methodology are violated.
of persistent mutational triggers (especially when For example, most models consider a nucleotide
&
the trigger sequence is located in the stem of a site as the unit of evolution (Ritland Eckenwald-
stem-Ioop secondary structure). This can introduce 1992), a consideration contradicted by
er, that
is
homoplasy
from parallelisms and reversals into any the mode of non-coding sequence evolution. Sim-
phylogenetic estimations that include gap-coded models based on commonly
plistic the calculated
characters in the matrix. Multiple indel events in a Kimura estimates (Kimura, 1980) and Jukes-Cantor
&
may
localized region obscure homology of length estimates (Jukes Cantor, 1969) assume an equal
25%
Non-independence
mutations. of these mutations frequency each
for nucleotide type throughout
introduces the issue of relative weighting of nucle- the sequence and generate base mutation proba-
otide characters linked in a repeat unit, each from assumption. Because non-coding
if bilities this
cpDNA
base is treated as a character in an analysis. Weight regions can demonstrate much higher A/T
of the unit taken as a single character is also an content, this assumption clearly contradicted.
is
issue if the unit is included in the analysis as a Furthermore, transition/transversion ratios in non-
coded gap
character. coding regions can considerably from coding
differ
&
Secondary structure shows nonrandom mu- ones Hoot Douglas, and may
(2) (see 1998),
tation in the form of compensatory mutation and vary between discrete non-coding regions of the
possible homogenization of sequence necessary for chloroplast genome. Among-site mutation
rate het-
stem formation. Loop sequence is available for mul- erogeneity is highly probable, especially if regions
tiple mutations in the form of inversions, length of conservation and hot spots for mutation exist in