Phylogenetic Evidence for Role-Reversals
Mitochondrial DNA in Mytilus (Bivalvia:
Walter R. Hoeh, *cl Donald T. Stewart,*
Eleftherios Zouros* j*Department of Biology, Dalhousie
Biology of Crete, Greece
University;
of Gender-Associated
Mytilidae)
Carlos Saavedra,*2 Brent W. Sutherland,*3
and TDepartment
of Biology, University
and
of Crete and Institute of Marine
Distinct gender-associated
mitochondrial DNA (mtDNA) lineages (i.e., lineages which are transmitted either through
males or through females) have been demonstrated in two families of bivalves, the Mytilidae (marine mussels) and
the Unionidae (freshwater mussels), which have been separated for more than 400 Myr. The mode of transmission
of these M (for male-transmitted)
and F (for female-transmitted)
molecules has been referred to as doubly uniparental
inheritance (DUI), in contrast to standard maternal inheritance (SMI), which is the norm in animals. A previous
study suggested that at least three distinct origins of DUI are required to explain the phylogenetic pattern of M and
F lineages in freshwater and marine mussels. Here we present phylogenetic evidence based on partial sequences of
the cytochrome c oxidase subunit I gene and the 16s RNA gene that indicates that DUI is a dynamic phenomenon.
Specifically, we demonstrate that F lineages in three species of Mytilus mussels, M. edulis, 44. trossulus, and M.
californianus, have spawned separate lineages which are now associated only with males. This process is referred
to as “masculinization”
of F mtDNA. By extension, we propose that DUI may be a primitive bivalve character
and that periodic masculinization
events combined with extinction of previously existing M types effectively reset
the time of divergence between conspecific gender-associated
mtDNA lineages.
Introduction
The phenomenon
of doubly uniparental inheritance
(DUI) of mitochondrial
DNA (mtDNA) has now been
documented
in several phylogenetically
diverse bivalve
taxa. These taxa include marine mussels of the family
Mytilidae
(Myti2u.s edu2is [Skibinski,
Gallagher,
and
Beynon 1994a, 1994b; Zouros et al. 1994a, 1994b], M.
trossulus [Geller 1994; Zouros et al. 1994b; Rawson and
Hilbish 1995; Stewart et al. 19951, M. galloprovincialis
[Rawson and Hilbish 1995; Quesada, Skibinski,
and
Skibinski 1996; Saavedra, Reynero, and Zouros 19971,
Geukensia demissa [Hoeh et al. 19961) and freshwater
mussels of the family Unionidae (Pyganodon fragilis, P.
grandis and Fusconaia Jlava [Liu, Mitton, and Wu 1996;
Hoeh et al. 19961). In these bivalves, males possess two
distinct classes of mtDNA: an M type, so-named because it is transmitted from male parents to their sons,
and an F type that is transmitted
from generation
to
generation through females. Although sons also receive
their mother’s F type mtDNA, they do not transmit it to
their offspring.
Rawson and Hilbish (1995) and Stewart et al.
( 1995) were the first to examine DUI from a phylogenetic perspective.
Using mitochondrial
16s RNA gene
sequences from M. edulis, M. trossulus, and A4. galloprovincialis,
Rawson and Hilbish (1995) showed that
I Present address: Department
of Zoology,
Miami University.
2 Present
of Biology,
University
address:
Department
3 Present address: Department
University of British Columbia.
of Microbiology
of Crete.
and Immunology,
Key words: phylogenetics,
mitochondrial
DNA, doubly uniparental inheritance, cytochrome c oxidase I, 16s RNA, masculinization,
Mytilus, Bivalvia, Mytilidae.
Address for correspondence
and reprints: Donald T. Stewart, Department of Biology, Fairfield University, Fairfield, CT 06430-5 195.
E-mail: [email protected].
Mol. Bid. Evol. 14(9):959-967.
1997
0 1997 by the Society for Molecular Biology and Evolution. ISSN: 0737-4038
the F sequences and the M sequences from all three
species formed two gender-associated
groups. The same
result was observed by Stewart et al. (1995) in a collection of cytochrome c oxidase subunit III (COIII) sequences from M. edulis and M. trossulus. Both studies
concluded that the formation of distinct F and M lineages must have preceded the emergence of the three
species of mussels from their common ancestor.
The discovery of DUI in freshwater mussels (Liu,
Mitton, and Wu 1996) prompted Hoeh et al. (1996) to
examine the phylogeny of M and F mitotypes in both
marine and freshwater mussels. Specifically, these authors compared CO1 sequences from six taxa: M. edulis,
M. trossulus, G. demissa, P. grandis, P. fragilis, and F.
Java. As expected from the previous analysis of 16s
RNA (Rawson and Hilbish 1995) and CO111 sequences
(Stewart et al. 1995), Hoeh et al. (1996) obtained dismitotypes (fig.
tinct M and F groupings for the Myths
1). The six M and F sequences of the three unionid
species (P. grandis, P. fragilis, and F. Java) also formed
two distinct groups defined on the basis of gender, but
as a whole, the freshwater mussel sequences formed a
group that was distinct from the marine mussel sequences. The M and F types present in Geukensia formed a
sister group to the other mytilid sequences. The complete phylogenetic
tree of the 12 sequences indicated at
least three separate divergences
of M and F mitotype
lineages, one in each of the lines leading to Myths
and
to Geukensia and one in the line leading to freshwater
mussels (fig. 1). Hoeh et al. (1996) inferred from these
three F/M mitotype divergence events either that there
have been multiple origins of DUI in bivalves or that
there was a single origin of DUI in the ancestral bivalve
lineage but reversals in the route of mitotype transmission had mimicked de novo divergences
of F and M
lineages. Specifically, Hoeh et al. (1996) referred to the
hypothetical
spawning of an M type from an F type as
a “masculinization”
event. The corresponding
switch
959
960
Hoeh et al.
3
r
L
+
F. flava M
P. grandis M
P. fragilis M
F. flava F
P. grandis F
-k
P. fragilis F
r
1
/L.
G. demissa M
damissa F
-2
M. edulis M
M. trossulus
M
f
M. edulis F
M. trossulus
F
FIG. l.-Topology
from Hoeh et al. (1996) indicating three divergence events between M and F mitotypes (shown by arrows) in a
collection of freshwater and marine mussels. The tree suggests that one
divergence event occurred in an ancestor to the freshwater mussel species P. grandis, P. fragilis, and F. Jlava, another in the lineage leading
to the marine mussel G. demissa, and one in a common ancestor to
M. edulis and M. trossulus.
from an M to an F type would be referred to as a “feminization”
event. This hypothesis was based on observations that the fidelity of DUI is not always perfect.
That is, females containing an M type and males lacking
an M type have occasionally
been observed (Fisher and
Skibinski
1990; Zouros et al. 1992, 1994b; Saavedra,
Reynero, and Zouros 1997). Such anomalous individuals
could provide information
about how an mtDNA type
associated with otie gender lineage could switch its route
of transmission
to the opposite gender lineage.
To further evaluate the alternate hypotheses indicated above, we have characterized
additional mitotype
lineages within Mytilus for CO1 and 16s RNA and phylogenetically
analyzed these sequences in conjunction
with the CO1 sequences presented in Hoeh et al. (1996)
and the 16s RNA sequences presented in Rawson and
Hilbish (1995). From these analyses, we present evidence that M mitotypes in the genus Mytilus are derived
from multiple ancestral lineages. We then demonstrate
how the reversal of the transmission
route of F genomes
(i.e., masculinization
events) best accounts for the observed phylogenetic
pattern, and we suggest a sequence
of events to explain the groupings of M and F types at
different levels of the bivalve mitochondrial
phylogeny.
Materials and Methods
Sequencing Protocol
Total DNA was isolated from either gonad or
spawned gametes of both sexes of Mytilus californianus.
This material was used to amplify a 710-bp fragment of
the mitochondrial
cytochrome
oxidase subunit I gene
(COI) using the primers LCO1490 and HC02198
(for
primer sequences and details of the amplification
protocol see Folmer et al. [1994] and Hoeh et al. [ 19961).
These primers were also used to generate 622 bp of CO1
sequence either by cycle sequencing
(AmpliCycle
Se-
quencing Kit, Perkin Elmer) or by direct sequencing
(Sequenase version 2.0 DNA Sequencing
Kit and Reagent Pack, USB) of the PCR product. Approximately
590 bp of sequence was obtained using either primer.
Consequently,
85%-90% of the total sequence obtained
was confirmed by sequencing in both directions.
We have also obtained additional CO1 sequences of
two M. trossulus mitotypes and one M. edulis mitotype
(designated
M-O, F-O, and M-A, respectively).
M-O
and M-A were classified as M types because they were
isolated from the gonads of males (Saavedra, Reynero,
and Zouros 1996). Restriction mapping (data not shown)
indicated that these putative mitochondrial
DNA types
were of a size consistent with mitochondrial
genomes
and therefore were not the result of transfer of mitochondrial genes to the nucleus. Although the restriction
fragment profiles also indicated that M-A and M-O were
considerably
divergent from the most common M types
of M. edulis and M. trossulus, respectively, these mitotypes were not previously sequenced because of technical difficulties (see Stewart et al. 1995). Briefly, because males are heteroplasmic
for an M and an F type
and because of the sequence similarity of M-O and M-A
with their conspecific F types (see below), it was difficult to isolate the CO111 amplification
products of M-O
and M-A from the amplification
products of their accompanying
F types. The newly obtained CO1 sequences were manually aligned, using MacClade (Maddison
and Maddison
1992), with CO1 sequences previously
analyzed by Hoeh et al. (1996).
We also amplified and obtained sequence from a
527-bp portion of the mitochondrial
16s RNA gene
from two male and two female M. californianus specimens using the primer pair 16s AR and 16s BR (Palumbi et al. 199 1; Rawson and Hilbish 1995). For an
outgroup, we sequenced the F type from two female
Geukensia demissa. The amplification protocol followed
Rawson and Hilbish (1995) and the product was sequenced directly as described above. These sequences
were aligned against 458 bp of 16s RNA sequence of
A4. edulis and M. trossulus M and F types extracted from
the GenBank database (U22864, U22865, U22879, and
U22882; Rawson and Hilbish 1995).
The CO1 and 16s RNA sequences generated herein
have been submitted to the GenBank database (accession numbers
U68770-U68772
[16S RNA]
and
U68773-U68777
[COI]).
Phylogenetic
Reconstruction
In this paper, we present three sets of phylogenetic
reconstructions.
The first set was conducted to test the
validity of the deeper nodes of our previously reported
topology (see Hoeh et al. 1996, fig. 1). To this end, we
conducted
maximum-parsimony
(MP) (PAUP version
3.1.1; Swofford
1993) and neighbor-joining
(NJ)
(MEGA version 1.02; Kumar, Tamura, and Nei 1993)
analyses on the inferred amino acid sequences of our
CO1 data set (i.e., 11 mytilid and 6 unionid sequences).
The MP analysis was conducted with and without invoking a step matrix to weight amino acid substitutions
relative to the minimum number of changes required to
Masculinization
switch from one amino acid to another. NJ trees were
constructed
using an uncorrected
amino acid distance
matrix and matrices for which distances were estimated
using gamma parameters
(a) of 0.5, 1.0, and 2.0, respectively.
In the second set of analyses, we focus on the relationship among the nine M and F Mytilus sequences
using the CO1 nucleotide data. The methods used included
MP, NJ and maximum
likelihood
(ML)
(DNAML, PHYLIP version 3.5~; Felsenstein 1993). The
ML analysis was performed using 100 random terminal
mitotype addition sequence runs with the global rearrangement option in force. The third set of analyses was
conducted
on a subset of the mytilid sequences for
which a combined CO1 and 16s RNA nucleotide data
set was available. Although the CO1 and 16s RNA sequences for a given mitotype (e.g., M. trossulus M) were
not obtained from the same individual, it is assumed that
the two sequences that were pooled to form a composite
mitotype share a more recent common ancestor with
each other than they do with any other sequence included in the analysis. Again, all three methods of tree reconstruction
were used. Percent bootstrap support for
the resulting nodes was evaluated using 1,000 replications for all of the MP and NJ analyses and 100 replications for the ML analysis (Hillis and Bull 1993).
In the second set of analyses, we used methods that
take into consideration
different rates of sequence divergence among sites and different transition (TS) and
transversion
(TV) ratios across codon positions within
the CO1 gene. The ML and NJ analyses also corrected
for differences
in frequencies
of the four nucleotides.
Specifically,
the NJ tree was reconstructed
from a distance matrix estimated using the Tamura-Nei model (Tamura and Nei 1993) and a gamma coefficient of 1 .O. In
the third set of analyses, we were also able to compensate for the difference in the overall level of sequence
divergence between the CO1 and 16s RNA genes in the
MP and ML analyses. These weighting schemes were
implemented
to give more weight to presumably
more
conservative
changes (Hillis, Huelsenbeck,
and Cunningham 1994) and/or to estimate, as accurately as possible, the evolutionary
distance between mitotypes (Kumar, Tamura, and Nei 1993).
The approximate
number of changes at each CO1
codon position was obtained by examining
a preliminary tree of mytilid relationships
constructed in PAUP
(without invoking differential
weighting).
The number
of changes
at each position
was calculated
using
MacClade version 3.05 (Maddison and Maddison 1992).
The relative difference in levels of sequence divergence
between the CO1 and 16s RNA genes was calculated
by taking the mean of the ratio between all pairwise
p-distances
for the CO1 and 16s RNA gene sequences
as calculated in MEGA. Similarly, MEGA was used to
calculate the TS/TV ratio for all pairwise comparisons
mitotypes. The mean TS/TV
among the M and F Myths
ratio calculated from this matrix was then used to calculate the relative weight given to transversion-versustransition substitutions
as suggested by Hillis, Huelsenbeck, and Cunningham
(1994).
of mtDNA
in Mytilus
961
We used the character optimization
algorithm in
MacClade
on our best estimate of the relationships
among all mytilid mitotypes to reconstruct the gender
association of ancestral bivalve mitotypes. Essentially,
we used the principle of parsimony to infer whether an
ancestral mitotype was transmitted through a male or a
female lineage. To facilitate the reconstruction
of ancestral character states, we assumed that DUI evolved from
standard maternal inheritance, which is presumably the
ancestral condition in animals.
Results
Phylogenetic
Relationships
The weighted and unweighted
MP and NJ trees
generated from the inferred CO1 amino acid sequences
(trees not shown) were similar to the topology presented
in Hoeh et al. (1996) (fig. 1). That is, distinct genderassociated groupings of M and F mitotypes were indicated within the freshwater mussels and the marine mussel Geukensia demissa. These groupings were also supported in 100% of the bootstrapped replicates in the MP
and NJ analyses. However, while a group consisting of
sequences was supported in 100% of the
the Myths
bootstrapped trees, the branching order within this clade
varied between the NJ and MP analyses. In all cases,
however, the A4. californianus M type was basal to the
Mytilus sequences. To resolve relationships
remaining
among the various M and F Mytilus sequences, we conducted two sets of additional analyses using (1) Geukensia as a functional outgroup (Maddison, Donoghue,
and Maddison 1984), (2) nucleotide rather than amino
acid sequences, and (3) the methods described above
that compensate for among-site rate variation.
Using an unweighted
parsimony analysis, we determined that there are approximately
3.7 and 20.4 times
as many changes at first and third codon positions, respectively, as at second codon positions. This rank order
of rates of change (i.e., r3 > rl > Y*) is commonly seen
in protein-coding
genes (Yang 1996). Three categories
of rates, with a frequency of 0.333 each, were incorporated into the DNAML analysis. To assign differential
weights in PAUP, which requires whole numbers, we
divided each of the above rates by 20.4 and took the
inverse to obtain an approximate weighting scheme of
5, 20, and 1 for changes at first, second, and third positions, respectively.
Similarly, because of a fourfold
transition bias at first positions, transversions
were assigned four times the weight of transitions for that codon
position.
Transitions
and transversions
occurred
in
roughly equal frequencies at second positions; therefore,
no differential
weighting was required. At third positions, however, transition substitutions were approaching
saturation in some comparisons.
To accommodate
for
this, we arbitrarily assigned third-position
transversions
10 times the weight of third-position
transitions to minimize the potential impact of multiple transition substitutions at a site.
The matrix of pairwise evolutionary
distances (estimated using the Tamura-Nei model with a gamma coefficient of 1.0) among all the mytilid CO1 sequences is
962
Hoeh et al.
Table 1
Evolutionary
Sequences
Distances
F-ed
F-ed. . . .
M-A . . .
F-tr . . . .
F-O. . . .
M-O . . .
M-ed.. .
M-tr. . . .
F-ca. . . .
M-ca . . .
F-de . . .
M-de...
0.0277
0.2050
0.2102
0.2276
0.3135
0.2730
0.2259
0.3256
0.6353
0.6509
(below diagonal,
standard
errors above diagonal)
for All Pairwise
Comparisons
of Mytilid CO1
M-A
F-tr
F-O
M-O
M-ed
M-tr
F-ca
M-ca
F-de
M-de
0.0073
0.0277
0.0299
0.0283
0.0310
0.005 1
0.0310
0.033 1
0.006 1
0.0058
0.0399
0.0440
0.0391
0.0405
0.0404
0.033 1
0.0342
0.03 11
0.0315
0.0327
0.0388
0.0289
0.0284
0.0249
0.0258
0.0273
0.0385
0.0315
0.0366
0.0367
0.0376
0.0391
0.0423
0.0501
0.0411
0.0354
0.0696
0.0689
0.0720
0.0722
0.0749
0.0798
0.0772
0.0600
0.0733
0.0742
0.0765
0.0686
0.070 1
0.0735
0.0896
0.070 1
0.0675
0.0688
0.0304
0.2250
0.233 1
0.2472
0.3373
0.283 1
0.2197
0.3256
0.6278
0.6547
0.0150
0.0203
0.3 165
0.2562
0.1930
0.3353
0.6438
0.6276
0.0186
0.3276
0.2608
0.1987
0.347 1
0.6505
0.6412
0.3292
0.2712
0.2108
0.3719
0.665 1
0.6665
0.2930
0.2964
0.4368
0.7199
0.7730
0.263 1
0.3519
0.6913
0.6430
0.3115
0.5627
0.6174
NoTI%--Estimates
are based on the Tamura-Nei model (Tamura and Nei 1993) using a gamma coefficient
(M-A and M-O) with the standard conspecific M and F sequences are underlined.
presented in table 1. These distances were used in the
NJ analysis.
For the 16s RNA nucleotide sequences used in the
combined COI/16S RNA data set, transitions were three
times more common than transversions.
Accordingly,
transversions
were given three times the weight of transitions in the MP analysis. Similarly, since the 16s RNA
gene was approximately
two thirds as divergent as the
CO1 sequences
on average, we superimposed
a 3:2
weighting scheme on the 16s RNA and CO1 portions
of the combined data set. In the same manner, to compensate for differences in the rates of change of the two
gene regions and differences
among codon positions
within COI, four categories of rates (2.0, 3.7, 1.0, and
20.4) were specified in the DNAML analysis of the
combined data set. The value 2.0 was chosen because
the level of divergence of a 16s RNA site is, on average,
intermediate
between first and second CO1 codon positions. The values 0.427, 0.191, 0.191, and 0.191 were
also supplied to indicate the relative frequencies of 16s
RNA sites and first, second, and third CO1 codon positions, respectively.
The phylogenetic
trees resulting from the aforementioned analyses are shown in figures 2 (COI) and 3
(CO1 and 16s RNA). From the CO1 data alone, four
features of the Mytilus phylogeny were invariable:
(1)
A4. californianus M was always a sister group to the
remaining Mytilus sequences. This relationship was supported by bootstrap values of 91%-100%;
(2) The male
type M-A was always affiliated with the M. edulis F
(bootstraps
= 100%). (3) The M. trossulus M-O and
F-O sequences formed a clade of their own which clustered with the standard M. trossulus F lineage (bootstraps L 99%). (4) The standard M types of M. edulis
and M. trossulus were always grouped together (bootstrap values 1 83%). An important difference among
the trees was the position of the M. californianus F type.
Using MacClade, we calculated that the numbers of substitutions associated with each of these topologies (without invoking differential
weighting of alternative character transformations)
were 691 (MP), 690 (ML), and
689 (NJ). (The two most-parsimonious
CO1 trees constructed without differential weighting [trees not shown]
were 688 steps in length, and the position of the M.
0.665 1
0.6066
of 1.0. Comparisons
0.2556
of newly masculinized
types
californianus F type varied in each case.) Based on such
small differences,
it is difficult to prefer one of these
topologies over the others.
Analysis of the combined COI/16S RNA data set
helped resolve some of the aforementioned
ambiguities.
First, the ML, MP and NJ analyses all indicated that the
M. edulis and A4. trossulus mitotypes (regardless of gender association) are descendants of a common ancestor
shared with the M. californianus F mitotype (fig. 3).
This set of relationships was supported by bootstrap values of 73%, 88%, and 91% in the MP, ML and NJ analyses, respectively.
Relationships
among the four remaining A4. edulis and A4. trossulus mitotypes varied
slightly between the NJ and the MP and ML trees (the
latter two trees being identical). The NJ trees both for
the CO1 data alone (fig. 2C) and for the combined data
set (fig. 3C) gave the division between M and F lineages
of M. edulis and M. trossulus that has previously been
proposed both by Stewart et al. (1995) based on CO111
gene sequences
and by Rawson and Hilbish (1995)
based on 16s RNA sequences. In contrast, the MP and
ML trees suggested a clade consisting of the M. edulis
F type as a sister group to a clade consisting of the A4.
edulis and M. trossulus M types (fig. 3A and B). This
portion of the topology was weakly supported, however,
with bootstraps values of only 57%-62%.
Given the
above, the fully resolved NJ tree (fig. 2C) is regarded
as the best estimate of mitotype relationships
and, as
such, will be used herein to interpret the evolutionary
dynamics of DUI.
Reconstruction
of Ancestral
Bivalve Mitotypes
Gender
Associations
of
Figure 4 is a synthesis of phylogenetic information
from this study and Hoeh et al. (1996). Superimposed
on this tree is the most parsimonious
reconstruction
of
ancestral gender states as calculated in MacClade under
the assumption that DUI (specifically, the generation and
inheritance
of M mitotypes) is a derived condition in
bivalves, and therefore the ancestral gender state can be
coded as E Masculinization
events are therefore indicated at nodes 4, 5, and 6 and inferred to have taken
place at nodes 1, 2, and 3.
Masculinization
A
of mtDNA in Mytifus
963
A
M. trossulus M
100
Ir
r
L M. trossulus F
M. trossulus F
100
M. trossulus F-O
L
100
-M.
-
M. californianus F
M. trossulus M-O
90
califomianus
M. californianus M
F
I
M. californianus M
7
G. demissa M
-
G. demissa F
G. demissa F
rr
M. edulis F
62
6
M. edulis M
r-
M. trossulus F-O
M. trossulus M-O
M. trossulus M
ri
96
LM. trossulus F
-M.
californianus
F
L M. trossulus F
L
FM.
edulis M
M. trossulus M
F--
M. californianus
M. californianus
A
F
M
G. demissa F
-
M. edulis F
65_
C
91
100
M. trossulus F
_
M. edulis M
96
M. trossulus M
M. trossulus M-O
C
-M.
californianus
F
M. trossulus F
M. californianus
M
G. demissa F
i,
100
-
M. trossulus M
-M.
californianus
F
M. californianus
1-G.
.
FG.
M
FIG. 3.-The
best estimates of the phylogenetic
relationships
among the mytilid M and F mitotypes obtained from analyzing 622
bp of CO1 and 458 bp of 16s RNA nucleotide sequences by three
methods: (A) maximum parsimony, (B) maximum likelihood, and (C)
neighbor-joining.
Numbers indicate bootstrap support for each node.
See text for details of the phylogenetic
methodology.
Geukensia demissa was used as the outgroup.
demissa M
demissa F
0'
011
0.1
FIG. 2.-The
best estimates of the phylogenetic
relationships
among the mytilid M and F mitotypes obtained from analyzing 622
bp of CO1 sequence data by three methods: (A) maximum parsimony,
(B) maximum likelihood, and (C) neighbor-joining.
Numbers indicate
bootstrap support for each node. See text for details of the phylogenetic
methodology.
Geukensia demissa was used as the outgroup.
Discussion
Hoeh et al. (1996) recently demonstrated three distinct divergence events of M and F mitotypes in several
distantly related bivalve species and proposed two explanations which might account for this pattern. If the
formation of distinct M and F types is taken as an indication of a de novo initiation of DUI, then this complex phenomenon
would have evolved at least three
964
Hoeh et al.
:““--- f. flava M
: ...........f
i___--P. grandis M
li
i-
P. fragilis M
F. flava F
P. grandis F
P. fragilis F
2 r G. demissa M
L G. demissa F
,_........................
Mmcalifornianus
M
M. californianus
F
31
I-.. M. edulis M
*...........f
I---M. frossulus M
4:'
5 !-
IF-
7
r
M. edulis M-A
M. edulis F
M. rrossulus F
LL
6 y- M. trossulus M-O
M. trossulus F-O
FIG. 4.-Character
optimization of ancestral mitotype gender associations inferred from the best estimates of phylogenetic
relationships among mytilid mitotypes (NJ tree, fig. 2C) combined with the
phylogenetic
information
of Hoeh et al. (1996) (fig. 1) and the assumption that DUI (i.e., the generation and inheritance of M mitotypes)
is a derived condition in the Mollusca. Male-transmitted
(M) mitotypes
are shown in a dotted line. In addition, ancestral mitotypes inferred to
be inherited via males are also indicated by a dotted line. Numbers
indicate nodes on the tree where masculinization
events (see text) are
indicated (4, 5, and 6) or may have taken place (1, 2, and 3). (Note:
branch lengths are not drawn to scale.)
times in the history of bivalves (fig. 1). This assumes
that once established, the M and F types do not switch
roles. In a review of DUI in mussels, Hurst and Hoekstra
(1994) proposed that for such a system to remain stable,
the M and/or F types must remain faithful to their own
lineages. Alternatively,
DUI may be a unique plesiomorphic trait in bivalves but reversals in the route of
transmission
of gender-associated
types may occur periodically. These hypothetical
transitions were referred
to as “masculinization”
or “feminization”
events by
Hoeh et al. (1996). The strongest evidence in support of
masculinization
events comes from the phylogenetic
affinity of the M. trossulus
type M-O and the M. e&&s
type M-A with their respective conspecific F types (fig.
2). The derived position of these paternally transmitted
mitochondrial
genomes within M. edulis and M. trossulus relative
to the maternally
transmitted
mitotypes
coupled with the close genetic similarity between these
conspecific M and F types suggests that the paternally
inherited mitotypes were recently spawned from female
mitotypes (fig. 2 and table 1). Similarly, the phylogenetic position of the M. trossulus and M. edulis M types
relative to the M. californianus F type (fig. 3) also indicates another instance of masculinization
of female
mtDNA, assuming that this is the true phylogeny
of
these genomes (see below). By extension, we propose
that the same process could have occurred repeatedly in
the history of the bivalves. This hypothesis is summarized in figure 4, which is a synthesis of phylogenetic
information from this study and Hoeh et al. (1996) combined with the most parsimonious
reconstruction
of ancestral gender states. Transitions
4, 5, and 6 are only
compatible with masculinization
events. Under the assumption that DUI arose once in an ancestral bivalve
lineage (prior to node A, fig. 4), additional masculinization events must have occurred at position 1 within
the Unionidae and at positions 2 and 3 within the Mytilidae.
As previously
described
(Rawson
and Hilbish
1995; Stewart et al. 1995, 1996), the “standard”
M mitotypes in A4. edulis and M. trossulus arose from an M
mitotype which existed in the common ancestor of these
two species. In contrast, it may be inferred from the
comparatively
small genetic divergences
between the
M-O and M-A mitotypes and their respective conspecific F types (table 1) that the former two mitotypes
arose relatively recently. Furthermore, these newly masculinized M mitotypes are coexisting with the “old” M
mitotypes in a state of polymorphism
within these species. M-O occurs at a frequency of 0.24 in M. trossulus
males and M-A occurs at a frequency of 0.26 in A4.
edulis males sampled from Lunenburg Bay, Nova Scotia
(Stewart et al. 1995). By comparison, only one class of
male types has been found in the other species so far.
This is not surprising given that only species of the M.
edulis complex (i.e., M. edulis, M. trossulus, and M.
galloprovincialis) have been examined in numbers large
enough to reveal the existence of polymorphism.
However, the fact that the mitotypes from two unionid genera
affiliate according to gender, whereas those from two
genera of the Mytilidae (i.e., Mytilus and Geukensia) do
not, poses the question of whether there are taxonomic
differences in the stability of DUI. As stated by Hoeh
et al. (1996), the time of the split between M and F
mitotypes in the freshwater mussels (and therefore the
length of time for which DUI has operated in a stable
fashion) may be much longer than the time indicated for
the Mytilidae. Given that the taxonomic split between
the Pyganodon and Fusconaia lineages occurred at least
100 MYA, Hoeh et al. (1996) deduced that the M and
F mitotypes in these species are at least 100 Myr old.
In contrast, Rawson and Hilbish (1995) estimated the
time of the split between the most common M and F
mitotypes in the M. edulis species complex at 5.3 MYA.
Hoeh et al. (1996) estimated that the formation of the
M and F mitotypes in Geukensia was of a similar age.
The newly masculinized
M mitotypes in Mytilus are
clearly of even more recent origin. In contrast to the
large divergences
between the standard conspecific M
and F mitotypes in M. edulis and A4. trossulus (ca. 25%30%), the newly masculinized
mitotypes have diverged
from their conspecific female types by less than 2.8%
(table 1).
The sequence of events that led to masculinization
remains unknown. However, one possible mechanism is
supported by recent studies of mtDNA transmission
in
Mytilus (Zouros et al. 1994b; Saavedra, Reynero, and
Zouros 1997). In these crosses, some males failed to
pass either the M or F mitotype to their sons. A small
percentage of sons from other males also failed to inherit
mtDNA from their fathers. The overall rate of M-negative males in these crosses was about 20%. Although
the information existing so far is limited to a few cases,
Masculinization
M-negative
males produce
sperm that contains
the
mtDNA of the mother (Saavedra, Reynero, and Zouros
1997). Thus, there is direct evidence that in the absence
of an M type in a male, the F molecule will assume the
role of the M molecule in sperm. It is not yet known,
however, if M-negative
males are actually fertile or, if
so, whether they produce sons.
Although Hoeh et al. (1996) proposed that role reversals could theoretically occur in either direction, it is
important
to note that the demonstrated
reversals in
routes of transmission
in the genus Myth
as well as
the possible reversals in the remaining bivalve taxa are
all compatible
with masculinization
events. However,
our hypothesis that only changes from F to M occur
must be qualified given the difficulties of phylogenetic
reconstruction
among sequences with such variable rates
of character transformations
(Huelsenbeck
and Hillis
1993; Hillis, Huelsenbeck,
and Cunningham
1994).
Nonetheless,
other interpretations
of Mytilus mitotype
relationships
do not appear to be plausible. For instance,
if the standard M types of M. edulis and M. trossulus
are placed in a clade with the A4. californianus
M mitotype (thereby eliminating
one masculinization
event),
the length of the tree increases considerably
(from 689
to 710 steps). Furthermore,
to accept an alternate topology such as either of those presented in figure 2A or B,
we must invoke a relatively more complicated biological
explanation
that incorporates
some form of hybridization and introgression
of mitochondrial
DNA across species boundaries
to account for the pattern of mitotype
relationships.
Based on the results of a survey of 201
“pure” and “hybrid” M. edulis and M. trossulus mussels from the wild, Saavedra et al. (1996) concluded that
there are intrinsic barriers to the exchange of mtDNA
between these two species. The bias in favor of masculinization
as opposed to feminization
events is also
consistent with the data currently available concerning
within-individual
mitotype distributions
in Mytilus. The
number of M-positive females is much smaller than that
of M-negative males, and the amount of M mitotype in
the former is very small, whereas in the latter the F
molecule is 100%. Furthermore,
there is, as yet, no evidence of eggs containing
M molecules but, as noted,
there is evidence that M-negative males produce sperm
that contains the F molecule (Saavedra, Reynero, and
Zouros 1997). While alternate interpretations
invoking
feminization
events do not appear to be in accord with
our limited data from the wild or from laboratory crosses, they cannot, for the moment, be completely discounted.
Role reversal of an F mitotype to yield a newly
masculinized
M mitotype is presumably the first step in
a process which resets, initially to zero, the amount of
sequence divergence
(and apparent time since divergence) between a species’ M and F mitotypes. The second step is a transient state of polymorphism
in which
both new and old M mitotypes coexist in the population.
The third step is replacement
of the old M mitotype by
a newly masculinized
type (see Hoeh et al. 1996, fig.
3). We can illustrate how these stages would have proceeded for the species of the genus Mytilus (fig. 5). First,
M. calijomianus
M
F
M. edulis
M
F M-A
of mtDNA
in Mytilus
M. trossulus
965
Taxon
M M-O F
FIG. 5.-Hypothetical
reconstruction
of speciation, mitotype masculinization,
and mitotype extinction events which could explain the
observed phylogenetic relationships of the M and F mitotypes sampled
in extant members of the genus A4yriZus.The heavy solid lines outline
organismal phylogeny. The light solid lines represent F mitotype phylogeny while the dashed lines represent M mitotype phylogeny. Chronology: A, An ancestral Mytilus lineage displays DUI. B, Allopatric
speciation occurs such that an ancestral species bifurcates to produce
two descendant lineages. One of these leads to the M. edulis/M. crossulus complex and the other to M. californianus. C, In the line leading
to the M. edulislhf trossulus complex and before the separation of h4.
edulis and M. trossulus, the F mitotype lineage spawned an M mitotype
via a masculinization
event. D, The common ancestor of 44. edulis and
M. trossulus would then have passed through a transitional state of
polymorphism
for the newly masculinized
M mitotype and the preexisting M mitotype. E, The original M mitotype in the M. edulis/M.
trossulus complex went on to extinction. F, Allopatric speciation leading to M. edulis and M. trossulus. G, Masculinization
event giving rise
to the M. edufis M-A mitotype. H, Masculinization
event giving rise
to the M. trossulus M-O mitotype. The M. trossulus F-O mitotype has
been omitted to simplify the presentation.
assume a standard mode of allopatric
speciation
in
which an ancestral DUI-containing
species (fig. 5A) bifurcates to produce two descendant lineages (fig. 5B).
One of these leads to the M. eduZis/M. trossulus comPresumplex, and the other leads to M. californianus.
ably, each of these descendant
lineages contained homologous (orthologous) M and F mitotypes. In the line
leading to the M. eduZislA4. trossulus complex and before the separation of M. edulis and M. trossulus, the F
lineage spawned an M mitotype via a masculinization
event (fig. 5c). The common ancestor of M. edulis and
M. trossulus would then have passed through a transi-
966
Hoeh et al.
tional state of polymorphism
for the newly masculinized
M mitotype and the preexisting
M mitotype (fig. 5D).
Finally, the original M mitotype presumably went on to
extinction (fig. 5E), since we have not, as yet, found an
M mitotype in the M. eduZislA4. trossulus complex that
is a sister mitotype to the A4. californianus M mitotype.
As mentioned
above, a consequence
of this process is
that the apparent origin of DUI (i.e., divergence of F
and M mitotypes) in the A4. eduZis/M. trossulus species
complex is moved forward in time to the point of the
most recent masculinization
event.
From a population genetics point of view, we want
to know if the replacement
of the old M lineage by a
newly masculinized
lineage is a stochastic event, or if
the latter has intrinsic advantages over the former. It is
well established that the M lineage evolves faster than
the F lineage, and we have presented evidence that the
reason for this is relaxed selection against the M type
(Stewart et al. 1996). This mechanism might favor the
accumulation
of slightly deleterious
mutations
which
would predispose the old M type to extinction under
competition with a new M type. On the other hand, this
could lead to very frequent masculinization
events. This
is clearly not the case in freshwater mussels. We also
note that if masculinizations
occur very often and if
newly spawned molecules quickly replace the old molecules, this would not allow for the mutational
divergence of M and F molecules. In such systems, the existence of DUI could remain undetected.
Acknowledgments
We thank the following for assistance in the procurement and/or maintenance
of specimens: W. Borgeson, D. Cook, D. Hedgecock,
C. Herbinger, E. Kenchington, J. Maunder, R. Noseworthy, and R. Trdan. A.
Ball, S. Baldauf, D. Cook, and M. Dillon kindly provided assistance
with laboratory
procedures
and (or)
provided access to the facilities of the Marine Gene
Probe Laboratory, Dalhousie University. We also thank
Andrew Martin and an anonymous reviewer for numerous constructive
comments on this manuscript. This research was supported by grants from the Natural Sciences and Engineering
Research Council of Canada
(NSERC) to E.Z. and by the Research Development
Fund (Dalhousie
University)
to W.R.H. and D.T.S.
W.R.H. and D.T.S. were supported by postdoctoral fellowships from NSERC. C.S. was supported by postdoctoral fellowships
from the Conselleria
de Education,
Xunta de Galicia (Spain) and the Ministerio de Educacion y Ciencia (Spain).
LITERATURE CITED
FELSENSTEIN, J. 1993. PHYLIP: phylogeny inference package.
Version 3.5~. University of Washington, Seattle.
FISHER, C., and D. 0. E SKIBINSKI. 1990. Sex-biased mitochondrial DNA heteroplasmy in the marine mussel Mytilus.
Proc. R. Sot. Lond. 242:149-156.
FOLMER, O., M. BLACK, W. HOEH, R. LUTZ, and R. VRIJENHOEK. 1994. DNA primers for amplification of mitochondrial cytochrome c oxidase subunit I from diverse metazoan
invertebrates. Mol. Mar. Biol. Biotechnol. 3:294-299.
GELLER, J. B. 1994. Sex-specific mitochondrial
DNA haplotypes and heteroplasmy
in Mytilus trossulus and Mytilus
galloprovincialis
populations. Mol. Mar. Biol. Biotechnol.
3:334-337.
HILLIS, D. M., and J. J. BULL. 1993. An empirical test of bootstrapping as a method for assessing confidence in phylogenetic analysis. Syst. Biol. 42: 182-192.
HILLIS, D. M., J. I? HUELSENBECK, and C. W. CUNNINGHAM.
1994. Application and accuracy of molecular phylogenies.
Science 264:67 l-677.
HOEH, W. R., D. T STEWART, B. W. SUTHERLAND,and E. ZouROS. 1996. Multiple origins of gender-associated
mitochondrial DNA lineages in bivalves (Mollusca: Bivalvia). Evolution 50:2276-2286.
HUELSENBECK, J. I?, and D. M. HILLIS. 1993. Success of phylogenetic methods in the four-taxon case. Syst. Biol. 42:
247-264.
HURST, L. D., and R. E HOEKSTRA. 1994. Shellfish genes kept
in line. Nature 368:811-812.
KUMAR, K., K. TAMURA, and M. NEI. 1993. MEGA: molecular
evolutionary genetics analysis. Version 1.02. Pennsylvania
State University, University Park.
LIU, H.-P, J. B. MITTON, and S.-K. Wu. 1996. Paternal mitochondrial DNA differentiation
far exceeds maternal DNA
and allozyme differentiation
in the freshwater mussel, Anodonta grandis grandis. Evolution 50:952-957.
MADDISON, W. I?, M. J. DONOGHUE, and D. R. MADDISON.
1984. Outgroup analysis and parsimony. Syst. Zool. 33:
83-103.
MADDISON, W. I?, and D. R. MADDISON. 1992. MacClade: analysis of phylogeny and character evolution. Version 3.01.
Sinauer, Sunderland, Mass.
PALUMBI, S. R., A. l? MARTIN, S. ROMANO, W. 0. MCMILLAN,
L. STICE, and G. GRABOWSKI. 1991. The simple fool’s guide
to PCR. Department of Zoology, University of Hawaii, Honolulu.
QUESADA, H., D. A. G. SKIBINSKI, and D. 0. E SKIBINSKI.
1996. Sex-biased heteroplasmy and mitochondrial DNA in
galloprovincialis
Lmk. Curt-. Genet.
the mussel Myths
29~423-426.
RAWSON, I? D., and T. J. HILBISH. 1995. Evolutionary relationships among the male and female mitochondrial DNA linedulis species complex.
Mol. Biol.
eages in the Myths
Evol. 12:893-901.
SAAVEDRA, C., M. REYNERO, and E. ZOUROS. 1997. Male-dependent doubly uniparental inheritance of mitochondrial
DNA and female-dependent
sex-ratio in the mussel Myths
Genetics 145: 1073-1082.
galloprovincialis.
SAAVEDRA, C., D. T. STEWART, R. R. STANWOOD,and E. ZouROS. 1996. Species-specific
segregation of gender-associated mitochondrial DNA types in an area where two mussel
species (Myths
edulis and Myths
trossulus) hybridize.
Genetics 143: 1359-l 367.
SKIBINSKI, D. 0. E, C. GALLAGHER, and C. M. BEYNON.
1994~. Mitochondrial
DNA inheritance. Nature 368: 8 17818.
-.
1994b. Sex-limited mitochondrial DNA transmission
in the marine mussel Myths edulis. Genetics 138:801-809.
STEWART, D. T., E. KENCHINGTON,R. K. SINGH, and E. ZouROS. 1996. Degree of selective constraint as an explanation
of the different rates of evolution of the gender-specific mitochondrial DNA lineages in the mussel Myths. Genetics
143:1349-1357.
STEWART, D. T., C. SAAVEDRA, R. RIGBY, A. 0. BALL, and E.
ZOUROS. 1995. Male and female mitochondrial
DNA lineages in the blue mussel (Myths
edulis) species group.
Mol. Biol. Evol. 12:735-747.
Masculinizationof mtDNA in Myths
SWOFFORD,D. L. 1993. PAUP: phylogenetic analysis using
parsimony. Version 3.1.1. Smithsonian Institution, Washington, D.C.
TAMURA, K. and M. NEI. 1993. Estimation of the nucleotide
substitutions in the control region of mitochondrial DNA in
humans and chimpanzees. Mol. Biol. Evol. 10512-526.
YANG, Z. 1996. Among-site rate variation and its impact on
phylogenetic analyses. TREE 11:367-372.
ZOUROS,E., A. 0. BALL, C. SAAVEDRA,and K. R. FREEMAN.
1994~. Mitochondrial DNA inheritance. Nature 368:8 18.
-.
967
1994b. An unusual type of mitochondrial DNA inheritance in the blue mussel Mytilus. Proc. Natl. Acad. Sci.
USA 91:7463-7467.
ZOUROS,E., K. R. FREEMAN,A. 0. BALL, and G. H. POGSON.
1992. Direct evidence for extensive paternal mitochondrial
DNA inheritance in the marine mussel Mytilus. Nature 359:
412-414.
CARO-BETH STEWART, reviewing
Accepted
May 30, 1997
editor