Content uploaded by Félix Forest
Author content
All content in this area was uploaded by Félix Forest on Jan 13, 2014
Content may be subject to copyright.
118
Systematic Botany (2005), 30(1): pp. 118–133
qCopyright 2005 by the American Society of Plant Taxonomists
Teasing Apart Molecular- Versus Fossil-based Error Estimates when Dating
Phylogenetic Trees: A Case Study in the Birch Family (Betulaceae)
F
E
´LIX
F
OREST
,
1,2,6
V
INCENT
S
AVOLAINEN
,
1
M
ARK
W. C
HASE
,
1
R
ICHARD
L
UPIA
,
3
A
NNE
B
RUNEAU
,
4
and P
ETER
R. C
RANE
5
1
Molecular Systematics Section, Jodrell Laboratory, Royal Botanic Gardens, Kew, Richmond, Surrey,
TW9 3DS, U.K.;
2
Centre for Plant Diversity and Systematics, School of Plant Sciences, The University of Reading,
Reading, RG6 6AS, U.K.;
3
Sam Noble Oklahoma Museum of Natural History and School of Geology and Geophysics,
The University of Oklahoma, 2401 Chautauqua Avenue, Norman, Oklahoma 73072–7029;
4
Institut de recherche en biologie ve´ge´tale, Universite´ de Montre´al, 4101 Sherbrooke est, Montre´al, Que´bec,
H1X 2B2, Canada;
5
Royal Botanic Gardens, Kew, Richmond, Surrey, TW9 3AB, U.K.
6
Author for correspondence. Present address: Kirstenbosch Research Centre, South African National
Biodiversity Institute, Private Bag X7, Claremont, 7735, South Africa (forest@sanbi.org)
Communicating Editor: Matt Lavin
A
BSTRACT
.Fossils are widely used as calibration points in molecular-based dating studies, but their placement on a
phylogenetic tree of extant species is always highly problematic. We explore some of the problems linked to calibration with
fossils, in particular their position on the tree, and emphasize the use of multiple calibration points to obtain better estimates.
We use a phylogenetic analysis of Betulaceae based on nuclear ribosomal DNA sequences (5S spacer and ITS) as a case
study and estimated divergence times within the family using the nonparametric rate smoothing method of Sanderson and
five calibration points from the extensive fossil record of this family. To assess the effects of assumptions relating to the
positions of key fossils with respect to stem lineages versus crown groups, we calculated age estimates by placing each
fossil subsequently on the stem lineage node and crown group node, and then determined the median value of the resulting
ten estimates for each node. Using maximum likelihood and DELTRAN and ACCTRAN parsimony branch lengths, we found
that the age of the crown group and stem lineage of Betulaceae vary from 115.2 to 130.6 million years and 211.2 to 302.6
million years (Aptian or before). These results are older than current paleobotanical data. We calculated paleobotanical
confidence intervals using methods based on the occurrence of fossils on a stratigraphic column and the lengths of the gaps
between these occurrences. We apply these methods to the fossil record of Alnus and related extinct genera; however, only
in some cases were molecular- and fossil-based age estimates reconciled.
The sources of error in DNA-based inference of di-
vergence time are numerous. The most commonly
mentioned are those concerning rate variation over
time and among lineages, tree accuracy, inadequate
sampling, ‘noise’, and calibration (e.g., Springer 1995;
Yang 1996; Sanderson 1998; Sanderson and Doyle 2001;
Wikstro¨m and Kenrick 2001; Wikstro¨m et al. 2001; Sol-
tis et al. 2002). In this study, we explore some of the
problems linked to calibration, in particular the posi-
tion of multiple fossil calibration points and the un-
certainty concerning fossil ages. Regardless of the
method and the criteria used to calculate the relative
ages of nodes on a tree, transformation of relative ages
into absolute ages requires at least one calibration
point. This can be obtained from the fossil record (e.g.,
Wikstro¨m et al. 2001), geological events (e.g., Richard-
son et al. 2001), paleoclimatic data (e.g., Baldwin and
Sanderson 1998), or other molecular estimates (e.g.,
Klak et al. 2004).
The fossil record gives the minimum age or upper
bound for a given node because a greater age would
be obtained if older fossils were discovered. Assuming
that the age assigned to a given fossil is accurate, the
sources of error arising from calibration based on this
fossil are two-fold: i) an imposed minimum age that
is too young, and ii) inaccurate positioning of the fossil
on the tree (Wikstro¨m et al. 2001). In the calibration
process, the placement on a tree of a fossil always in-
volves some level of uncertainty, and the importance
of this step in molecular-based dating is often under-
estimated. Placing a fossil on the crown group node
implies that the fossil taxa appeared at the same time
as the living taxa started to diverge, whereas placing
it on the stem (root) lineage node implies that the fossil
appeared at the same time as the clade diverged from
its sister (e.g. see Wikstrom et al. 2001, 2003). In reality,
it is more likely that a given fossil is found somewhere
on the branch between the stem and the crown nodes.
To examine the effects of stem versus crown cali-
bration with multiple fossils, we used Betulaceae, a rel-
atively small family (six genera and about 150 species;
Furlow 1990) with a rich fossil record. Betulaceae have
been extensively studied using different types of mor-
phological and molecular characters (Anderson and
Abbe 1934; Abbe 1935, 1938; Hjelmqvist 1948; Hall
1952; Brunner and Fairbrothers 1979; Kikuzawa 1982;
2005] 119FOREST ET AL.: DATING IN BETULACEAE
T
ABLE
1. The five Betulaceae fossils used as calibration points. Ages based on the geologic time scale of Palmer and Geissman (1999). Lower bounds of the ages or sub-epochs were used to calibrate
the tree (in bold) at the positions indicated on Fig. 2.
Fossils References Structures Ages/Sub-epochs Ages in Ma
Nodes for
calibration
(stem–crown)
AAlnus Budantsev (1982); Crane (1989) Leaves, reproductive structures Paleocene 54.8-65.0 N3–N7
BBetula leopoldae Crane and Stockey (1987) Leaves, infructescences, fruits, pollen,
staminate inflorescences
Middle Eocene 37.0–49.0 N3–N6
Co Corylus Wehr (1995); Pigg et al. (2003) Fruit in husk Middle Eocene 37.0–49.0 N2–N5
Ca Carpinus Pigg et al. (2003) Fruits and involucres Middle Eocene 37.0–49.0 N8–N9
OOstrya Chaney (1927); Manchester and
Crane (1987)
Fruits and involucres Oligocene 23.8–33.7 N8–N10
Hardin and Bell 1986; Crane 1989; Bousquet et al. 1992;
Chen et al. 1999); here we inferred phylogenetic rela-
tionships using ribosomal DNA sequence data and in-
clude representatives from all subgeneric divisions. We
used five fossils (Table 1) from extant taxa to calibrate
trees. Median values were used to summarize the 10
age estimates obtained for each node and were com-
pared to overall paleobotanical evidence including the
extinct genera Palaeocarpinus,Asterocarpinus,andCra-
nea.
The incompleteness of the fossil record leads to in-
evitable underestimation of ages (Springer 1995) and
divergence times estimated from molecular-based
phylogenetic reconstruction often proposed ages dif-
fering from those inferred by the fossil record, some-
time significantly so (e.g., Wikstro¨m et al. 2001; Benton
and Ayala 2003). To address this issue, methods have
been proposed to calculate paleobotanical confidence
intervals for the fossil record (e.g., Strauss and Sadler
1989; Marshall 1990, 1994, 1997; Foote et al. 1999). We
applied two of these methods, based on the occurrence
of given fossils on a stratigraphic column and the
lengths of the gaps between these occurrences, to the
palynological record of Alnus and related extinct gen-
era (Strauss and Sadler 1989; Marshall 1994).
M
ATERIALS AND
M
ETHODS
Sampling. All six extant genera of Betulaceae are represented
in this study. In the case of the genera with infrageneric divisions,
Carpinus,Corylus,Alnus,andBetula, at least one species from each
subgenus and/or section is sampled (see Table 2). The infrageneric
classifications are based on the following taxonomic treatments:
Betula, Winkler (1904), Furlow (1990, 1997); Alnus, Winkler (1904),
Furlow (1979, 1990, 1997); Carpinus, Winkler (1904), Furlow (1990);
Corylus, Bobrov (1936). Vouchers of our material are deposited at
the following herbaria: K, MT, CS, and DU (Table 2).
Ticodendron incognitum, the sole member of Ticodendraceae, was
chosen as outgroup taxon based on a phylogenetic analysis using
matK sequences (Manos and Steele 1997) in which it was sup-
ported as sister to Betulaceae with a bootstrap percentage (BP) of
85. Various morphological and molecular studies also showed this
close relationship of Ticodendron with Betulaceae (Hammel and
Burger 1991, and references therein; Savolainen et al. 2000).
DNA Extraction and Sequencing. For isolation of DNA from
fresh or silica-gel dried leaves, we used the CTAB method of Doyle
and Doyle (1987) with modifications. Silica gel dried leaves (Chase
and Hills 1991) were ground in liquid nitrogen and fresh leaves
in hot CTAB buffer. The Royal Botanic Gardens, Kew, provided
DNA from specimens cultivated there.
The 5S rDNA repeat was amplified with primer P2 (Cox et al.
1992) and primer P5 (59-TGC ACC GGA TCC CAT CAG AA-39;
this study). Both primers are located in the 5S rRNA gene with
their 59end terminating on adjacent positions. P5 was designed
to amplify, in combination with P2, as much as possible of the 5S
rRNA gene. PCR was performed on a 9700 GeneAmp thermocy-
cler from Applied Biosystems (Foster City, California, USA), with
the following program: pre-denaturation of 1 min at 948C followed
by 28 cycles of denaturation of 1 min at 948C, annealing at 538C
for 1 min, extension of 1.5 min at 728C, and a final extension of 2
min at 728C. The 50 mL PCR reaction contained two units of Taq
DNA polymerase (Roche Diagnostics, Laval, Que´bec, Canada), Taq
DNA polymerase buffer (supplied by the manufacturer, including
1.5 mM MgCl
2
), 0.2 mM of each primer and 100 mM of dNTPs.
PCR products were purified with the PCR purification kit from
120 [Volume 30SYSTEMATIC BOTANY
T
ABLE
2. Species included in the phylogenetic analysis of Betulaceae based on rDNA sequences from the 5S spacer and ITS. Infra-
generic classifications are based on various taxonomic treatments (Betula, Winkler 1904, Furlow 1990, 1997; Alnus, Winkler 1904, Furlow
1979, 1990, 1997; Carpinus, Winkler 1904, Furlow 1990; Corylus, Bobrov 1936). For each infrageneric classification, data are presented in
the following sequence: Taxon, Voucher 5S rDNA, Genbank/EMBL accession no. 5S rDNA, Voucher ITS, Genbank/EMBL accession no.
ITS. Note. RBGE, Royal Botanic Gardens, Edinburgh, UK; cult., cultivated; wild, wild collected.
Alnus Mill. Subg. Cremastogyne C.K. Schneid.: A. cremastogyne Burk., Chase 6121 (K) cult., AJ784247, same DNA, AJ783639;
Subg. Clethropsis (Spach) Regel: A. maritima (Marsh.) Muhl. ex Nutt., Forest 97-101 (MT) cult., AJ784245, Savard et al. (1993),
X68135; A. nitida (Spach) Endl., Chase 6122 (K) cult., AJ784243, same DNA, AJ783638; Subg. Alnus Endl., sect. Alnus: A. incana
(L.) Moench subsp. rugosa (Du Roi) Claus., Forest 97-106 (MT) cult., AJ784244, Savard et al. (1993), X68138; Subg. Alnus Endl.,
sect. Phyllothyrsus: A. acuminata Kunth, Lewis 3579 (K) wild, AJ784248, same DNA, AJ783640; Subg. Alnobetula Peterm.: A. viridis
(Vill.) DC subsp. crispa (Ait.) Turril, Forest 97-111 (MT) cult., AJ784246, Savard et al. (1993), X68137. Betula L. Sect. Costatae
Regel: B. nigra L., Forest 97-118 (MT) cult., AJ784255, same DNA, AJ783646; B. insignis Franch., Chase 6148 (K) cult., AJ784254,
same DNA, AJ783645; Sect. Betulaster (Spach) Regel: B. alnoides Buch.-Ham., Forest 97-55 (MT) cult., AJ784249, same DNA,
AJ783641; Sect. Betula: B. populifolia Marsh., Forest 97-116 (MT) cult., AJ784252, same DNA, AJ783644; B. pendula Roth., Forest
97-42 (MT) cult., AJ784253, Savard et al. (1993), X68136; Sect. Humiles W.D.J. Koch; B. pumila L., Charest 3258 (MT) wild, AJ784250,
same DNA, AJ783642; B. humilis Marsh., Forest 97-91 (MT) cult., AJ784251, same DNA, AJ783643. Carpinus L. Sect. Distegocarpus
(Siebold. & Zucc.) Sarg.: C. japonica Bl., Forest 97-19 (MT) cult., AJ784239, same DNA, AJ783635; C. cordata Bl., Chase 6147(K)
cult., AJ784240, Chen et al. (1999), AF081524, AF081525; C. fangiana Hu, Chase 6143 (K) cult., AJ784237, same DNA, AJ783633;
Sect. Carpinus: C. betulus L., Forest 97-35 (MT) cult., AJ784230, same DNA, AJ783636; C. caroliniana Walt. subsp. virginiana
(Marsh.) Fur., Forest 97-36 (MT) cult., AJ784238, same DNA, AJ783634. Ostr ya Scop. O. japonica Sarg, Forest 97-23 (MT) cult.,
AJ784241, same DNA, AJ783637; O. virginiana (Mill.) K. Koch, Forest 97-34 (MT) cult., AJ784242, Savard et al. (1993), X68139.
Ostryopsis Decne. O. davidiana Dcne., Whitcher 97-51 (CS) cult., AJ784236, Chen et al. (1999), AF081526, AF081527. Corylus L.
Subg. Acanthochlamys Spach: C. ferox Wall., RBGE 19930413B cult., AJ784233, Whitcher and Wen (2001), AY006359, AY006360;
Subg. Phyllochlamys Bobr., sect. Diphyllon: C. americana Walt., Forest 96-26 (MT) cult., AJ784234, Whitcher and Wen (2001),
AY006335, AY006336; C. colurna L., Forest 97-49 (MT) cult., AJ784235, Whitcher and Wen (2001), AY006361, AY006362; Subg.
Phyllochlamys Bobr., sect. Monophyllon: C. maxima Mill., Chase 6129 (K) cult., AJ784231, Whitcher and Wen (2001), AY006367,
AY006368; Subg. Siphonochlamys Bobr.: C. cornuta Marsh. subsp. cornuta, Charest sn, (MT) wild, AJ784232, Whitcher and Wen
(2001), AY006341, AY006342. Ticodendraceae. Ticodendron incognitum Go´mez-Laurito & Go´mez, Hammel s.n. (DU) wild,
AJ811536, same DNA, AJ783647.
Qiagen (Mississauga, Ontario, Canada) and eluted in EB buffer
(10 mM Tris, Qiagen). Amplifications of the ITS region used prim-
ers 17SE and 26SE from Sun et al. (1994) following the same pro-
cedure as for the 5S rDNA unit except for the annealing temper-
ature, which was set at 558C instead of 538C, and the number of
cycles, 35 instead of 28.
PCR products of the 5S rDNA repeats were cloned prior to se-
quencing following methods described in Forest and Bruneau
(2000). Once cloned, 5S rDNA repeats were amplified using trans-
formed bacteria colonies as template and primers that anneal to
the vector. Three clones for each species in this study were se-
quenced and included in a phylogenetic analysis (two clones for
B. alnoides and B. humilis). Clones from each species formed mono-
phyletic groups, with two exceptions. In A. incana subsp. rugosa
one clone is part of a trichotomy with the two others and the three
clones of A. acuminata.InB. pumila one clone was the sister to
three clones of B. nigra (data not shown), which might be expected
considering the low variation in 5S rDNA repeat sequences in the
genus. In both instances, one clone of each species was arbitrarily
chosen and included in a combined analysis with ITS sequences.
The ITS region was sequenced directly from the PCR products
usually with primers ITS4 and ITS5 from White et al. (1990). In
some species, sequencing using primer ITS5 was difficult because
of a GC-rich region in the ITS1 spacer, starting 48 bp from the 39
end of the 18S rDNA gene. In those cases, primer ITS2 (White et
al. 1990) was used to obtain missing portions of ITS1 spacer se-
quences. For both 5S rDNA and ITS regions, complementary
strands were sequenced on an ABI 310 automated sequencer (Ap-
plied Biosystems, Foster City, California, USA) following manu-
facturer’s protocols. Seven ITS sequences were obtained from
GenBank (Table 2).
Alignment and Phylogenetic Analyses. The program SE-
QUENCHER 3.1 (Gene Codes Corp., Ann Arbor, Michigan, USA)
was used to assemble complementary strands and verify bases
identified by the software. Sequences were aligned with the CLUS-
TAL X algorithm (Thompson et al. 1997) using the default param-
eters (gap opening set at 10 and gap extension set at 0.05). The
alignment was checked and modified by eye. Although the 5S
rRNA gene was sequenced, its inclusion in preliminary phyloge-
netic analyses resulted in general reduced support (not shown).
Previous studies concluded that data from the 5S rRNA gene alone
do not provide enough resolution to confidently infer phylogenetic
relationships at all taxonomic levels (Halanych 1991; Steele et al.
1991). Thus, the 5S rRNA gene was excluded from subsequent
analyses. Parts of the 5S rDNA intergene spacer were too variable
to permit confident assessments of homology between genera and
thus were also excluded from the analysis. Most of the 5S spacer
sequence of Ticodendron was difficult to align with confidence to
those of Betulaceae, thus it was treated as missing data in the
combined analyses. Alignment of ITS sequences was less problem-
atic, and no regions were excluded due to alignment ambiguities.
The 5.8S gene was not included in the analysis because it was not
available for all taxa (some sequences from GenBank lacked the
5.8S gene). The sequence matrix is available from the TreeBASE
website (study accession number S1121, matrix accession number
M1918). Jukes-Cantor distances were calculated with PAUP* ver-
sion 4.08b (Swofford 2001), and average distances were deter-
mined for each genus to assess the level of sequence divergence
within genera.
Analyses were performed using the maximum parsimony cri-
teria as implemented in PAUP* version 4.0b8 (Swofford 2001). All
maximum parsimony analyses were conducted with the heuristic
search option and tree bisection and reconnection (TBR) branch
swapping with 100 random taxon-addition replicates. All substi-
tutions were equally weighted and unordered (Fitch parsimony;
Fitch 1971). One thousand bootstrap replicates (Felsenstein 1985)
were performed with the same parameters as above. Branches
with BP of 50–74 are considered weakly supported, 75–89 mod-
erately supported, and 90–100 strongly supported.
As an additional measure of support, we reconstructed phylo-
genetic relationships using Bayesian inference as implemented in
the program MrBAYES 2.0 (Huelsenbeck and Ronquist 2001). We
used the DNA evolution model HKY85 1G(Hasegawa et al. 1985;
Yang 1993) with all variables estimated from the data (base fre-
quencies, transition/transversion ratio, and among-site variation).
The more complicated GTR1G model was not significantly dif-
2005] 121FOREST ET AL.: DATING IN BETULACEAE
ferent (R
2
50.999987), thus HKY85 1Gwas preferred because it is
a good compromise between complexity and computing time, es-
pecially during bootstrap resampling (Sanderson and Doyle 2001).
Four simultaneous Markov Chain Monte Carlo (MCMC) chains
were run for 10
6
generations and sampled every 100 generations.
MCMC reached a plateau in less than 50,000 generations, and the
first 500 trees (‘‘burn in’’) were excluded. The rest of these trees
were compiled as a majority-rule consensus in PAUP* version
4.0b8 (Swofford 2001). The frequencies of each clade in this con-
sensus tree are the Bayesian posterior probabilities (PP). Because
Bayesian probabilities have been shown to be too liberal in sim-
ulation studies (Suzuki et al. 2002), we considered strong support
to be posterior probabilities of 0.95 or more (e.g., Wilcox et al.
2002).
Estimating Time of Divergence. One of the most parsimonious
trees found in the combined maximum parsimony analysis was
arbitrarily chosen to infer divergence times. The likelihood ratio
test of Felsenstein (1988) was used to determine if rate heteroge-
neity across lineages was significant for both DNA regions, indi-
vidually and combined. The comparison of the difference between
the likelihood scores of the tree with and without an enforced
molecular clock, multiplied by two, was compared with a x
2
dis-
tribution. If the tree is fully resolved, the number of degrees of
freedom is N-2, where Nis the number of taxa, and corresponds
to the number of internal branches of the tree (Sanderson and
Doyle 2001). However, because a polytomy in the Betula clade oc-
curs in this tree, this reduces the number of internal branches to
23. Therefore, the number of degrees of freedom used in this case
is 23. The likelihood ratio tests performed on both DNA regions,
individually and combined, have significant rate heterogeneity
among lineages (p,0.001). Thus, we applied the non-parametric
rate smoothing (NPRS) method of Sanderson (1997).
Maximum parsimony (MP) branch lengths (both ACCTRAN
and DELTRAN optimizations) and maximum likelihood (ML)
were calculated for one of the most parsimonious trees obtained
in the combined analysis of both DNA regions. ML branch lengths
were optimized using the HKY85 model of DNA evolution (Has-
egawa et al. 1985) with base frequencies and transition/transver-
sion ratio estimated from the matrix. Among-site rate heteroge-
neity was accounted for by applying a gamma distribution with
an alpha shape parameter also estimated from the data (Yang
1993).
The tree obtained was made ultrametric using the NPRS meth-
od (Sanderson 1997) as implemented in the program TreeEdit
v1.0a4–61 (Rambaut and Charleston, available at http:/
evolve.zoo.ox.ac.uk/software/TreeEdit). Relative times were
transformed to absolute ages by calibrating the tree with five fos-
sils of Betulaceae used one at the time (Table 1) at both the crown
group node and the stem lineage node (Doyle and Donoghue
1993) using the geological time scale of Palmer and Geissman
(1999). Median values of the resulting 10 estimates were deter-
mined for each node as consensus estimates. For the calculation
of standard deviation on the molecular estimates, the matrix was
bootstrapped one hundred times using the tree used to infer time
divergence as topological constraint. The resulting 100 trees saved
were made ultrametric with the NPRS method as implemented in
the software TreeEdit v1.0a4–61 (Rambaut and Charleston, see
above). Each tree was calibrated separately. Median and standard
deviation were calculated on the absolute ages.
Choice of Fossils as Calibration Points. A
LNUS
. Reproductive
structures unequivocally assigned to Alnus are known from vari-
ous Paleocene localities (e.g., Budantsev 1982; see Crane 1989). The
most complete assemblage of Alnus is from the Middle Eocene
Clarno Formation in central Oregon (37.0–49.0 Ma; Crane 1989).
This group of specimens comprises staminate and pistillate inflo-
rescences and associated leaves, and the staminate inflorescences
bear Alnus-like pollen. These Middle Eocene specimens together
with the several occurrences reported from the Paleocene indicate
that plants showing characters typical of Alnus were present in the
early Paleocene. We calibrated the stem and crown node of Alnus
(nodes 3 and 7: N3 and N7; Fig. 1) using the earliest occurrences
from the Paleocene (54.8–65.0 Ma) unequivocally assigned to the
genus.
B
ETULA
. The oldest and most complete specimens of Betula fos-
sils are from the One Mile Creek locality in southern British Co-
lumbia (Canada) and comprise several leaf remains as well as in-
fructescences, staminate inflorescences, and fruits. These speci-
mens were assigned to a single species, Betula leopoldae Wol fe a n d
Wehr (Crane and Stockey 1987). The well-documented material
from this species even permitted determination of its close rela-
tionship with the extant species of section Costatae (Crane and
Stockey 1987). This section is found to be paraphyletic in our study
and occupies the basal nodes in the genus, thus providing addi-
tional support for the placement of the fossil species at the crown
and stem node of the genus. Although earlier records of Betula
from the Paleocene and Early Eocene exist, none are assigned to
the genus without doubts (see Crane 1989). We used the fossils
assigned to the extinct species B. leopoldae to calibrate the stem and
crown node of genus Betula (N3 and N6; Fig. 1; Middle Eocene,
37.0–49.0).
C
ARPINUS
. The oldest fossil to date assigned to genus Carpinus
is C. perryae from the Middle Eocene (37.0–49.0 Ma) of north-
western North America (Pigg et al. 2003). Carpinus perryae was
described from two nutlets surrounded by a winglike unlobed
bract from the Klondike Mountain Formation of the Republic flora
(Washington, U. S. A.). These fruits have many morphological sim-
ilarities with the extant species, particularly of subgenus Carpinus,
despite the fact that the longitudinal ribs on the nutlet, typical of
Carpinus, were not observed. We used this fossil as calibration
point at the stem and crown nodes of genus Carpinus (N8 and N9;
Fig. 1).
C
ORYLUS
. Thirty-four specimens from the Klondike Mountain
Formation of the Republic flora (Washington, U. S. A.) were used
to describe the species Corylus johnsonii (Pigg et al. 2003), some of
which were first reported elsewhere (Wehr 1995). These specimens
include involucres, solitary or paired nuts, and infructescences of
which some of the most complete are composed of pedunculate
pair of involucres with nuts. The group of specimens presents
some variability having similarities with the extant species C. ferox,
C. wangii,andC. heterophylla (Pigg et al. 2003), the first found to
be sister to the remainder of the genus in recent phylogenetic anal-
yses (e.g., Forest and Bruneau 2000). The specimens of C. johnsonii
are the oldest fossils unequivocally assigned to genus Corylus
(Middle Eocene; 37.0–49.0 Ma). This fossil was used to calibrate
the stem and crown nodes of the genus (N2 and N5; Fig. 1).
O
STRYA
. The earliest occurrence of Ostrya in the fossil record
is from the Oligocene of North America and consists of specimens
of the bladder-like involucres typical of the genus and leaf remains
(Chaney 1927; Manchester and Crane 1987). All observable char-
acteristics of these fossils are comparable to features observed in
extant species of the genus, and thus we used an age of 33.7 Ma
(Oligocene; 23.8–33.7 Ma) to calibrate the stem and crown nodes
of genus Ostrya (N8 and N10; Fig. 1).
Paleobotanical Confidence Inter vals. Confidence intervals on
stratigraphic ranges of Alnus-type pollen grains were calculated
using methods proposed by Strauss and Sadler (1989) and Mar-
shall (1994). A paleobotanical confidence interval is an estimation
of the earliest occurrence of a given taxa in the fossil record based
on the know occurrences of this taxa and the number and size of
the gaps between them in a stratigraphic column. The estimated
lower bound of the paleopalynological record of Alnus will be
compared with the molecular estimates.
Alnus pollen is characterized by three pores and unique arci, a
thickening of the exine between the pores. About 1,000 references
and the Plant Fossil Record database (version 2.2; International
Organisation of Palaeobotany, http://ibs.uel.ac.uk/paleo/pfr2/
pfr.htm) were searched for the occurrence of pollen genera having
a minimum of three pores and arci between them, a feature char-
acteristic of the genus Alnus and related fossil taxa (e.g., Alnipol-
lenites,Paraalnipollenites,andPolyvestibulopollenites). Results were
compiled as presence/absence of this pollen type in one or many
of the divisions (or fossil horizons) of a stratigraphic column, cor-
responding to ages of the Cretaceous and sub-epochs of the Ce-
122 [Volume 30SYSTEMATIC BOTANY
2005] 123FOREST ET AL.: DATING IN BETULACEAE
←
F
IG
. 1. One of the six equally most parsimonious trees resulting from the combined analysis of 5S spacer and ITS DNA
sequences. Parsimony branch lengths (ACCTRAN optimization) are shown above each branch, and bootstrap percentages and
Bayesian posterior probabilities in bold below branches. Nodes are numbered from N0 to N23. Nodes used in calibration with
fossils are indicated (A, Alnus;B,Betula; Ca, Carpinus;Co,Cor ylus;O,Ostrya; see Table 1). Arrowheads mark clades collapsing
in the strict consensus tree.
nozoic. One occurrence only was necessary to indicate the pres-
ence of the pollen type in each horizon although multiple occur-
rences were recorded in the literature in most cases.
The equation of Strauss and Sadler (1989), a5(1 2C)
[21/(H21)]
21, gives the confidence interval, a, for a given confidence level,
C, as a fraction of the age of the total known stratigraphic range
of the taxa under study using the number of fossiliferous horizons,
H, in which the taxa have been recorded. This method assumes
random fossilization and random sampling through the strati-
graphic record.
The method proposed by Marshall (1994) also assumes random
sampling but is less strict concerning random fossilization. The
method uses two equations to determine which gap sizes, deter-
mined by two consecutive occurrences in the stratigraphic column,
correspond to the lower and upper bound of the confidence in-
terval. In addition to the paleobotanical confidence level as in the
Strauss and Sadler (1989) method, the method of Marshall has a
second level of confidence, the confidence probability, correspond-
ing to the probability that the value lies within the range covered
by the confidence interval. Table 1 in Marshall (1994) shows the
result of these two equations for certain numbers of gaps, confi-
dence levels, and confidence probabilities. We used Marshall’s
(1994) table to determine which gap sizes define the lower and
upper bounds of our confidence intervals. The gap sizes are de-
termined in million of years (Ma) using the upper bounds of two
consecutive stratigraphic divisions provided that at least one oc-
currence of the Alnus-type pollen has been recorded during this
interval of time.
Like the Strauss and Sadler (1989) method, the Marshall (1994)
method assumes that there is no significant correlation between
the size of the gaps and their stratigraphic position, i.e. that there
is no increasing or decreasing trend linked to the stratigraphic
position (Marshall 1994). This assumption is evaluated by the
transposition test in which the number of such events is counted.
The number of transpositions, T, corresponds to the number of
times a gap size is larger than subsequent ones in the stratigraphic
sequence. When two or more gap sizes are equal, the following
formula is applied for which mis the number of gaps with equal
sizes: m(m-1)/4. The result of this formula is added to the sum of
transpositions, which is then applied to Table 3 in Marshall (1994)
to determine if there is significant support for an increasing gap
size in the stratigraphic column. Given a hypotheticalstratigraphic
sequence formed of seven consecutive gap sizes of 3, 4, 7, 2, 6, 4,
and 5 Ma, the number of transpositions would be calculated as
follows: for the first gap size in the sequence, 3 Ma, there is one
transposition; for the second, 4 Ma, there is one transposition, etc.
The number of transpositions for each gap size in this example
would be 1, 1, 4, 0, 2, 0, and 0 respectively for a total of eight
transpositions. In addition, in this example, two gap sizes have the
same value, 4 Ma, so the equation mentioned above has to be
applied where m 52: 2 (2–1) / 4 50.5. This gives a total of 8.5
transpositions, which indicates, once applied to Table 3 in Mar-
shall (1994), that there is no significant increase or decrease in gap
sizes (p.0.02) in this hypothetical stratigraphic sequence.
R
ESULTS
Phylogenetic Analyses. The aligned ITS matrix
comprises 475 base positions, but only 462 characters
are included in the analysis, of which 107 are poten-
tially parsimony informative (23.2%). The lengths of
both ITS1 and ITS2 are relatively conserved within Be-
tulaceae, ITS1 ranging from 204 to 221 nucleotides and
ITS2 ranging from 222 to 231 nucleotides. Alnus cre-
mastogyne is unique within Betulaceae in having a 204
nucleotide ITS1 region due to a deletion found 65 nu-
cleotides from the 3’ end of the 18S rDNA gene. In
Ticodendron, the ITS1 and ITS2 regions are 217 and 228
nucleotides, respectively. The final aligned 5S spacer
matrix comprises 729 characters, of which 297 are in-
cluded in the analysis and 197 (66.3%) are potentially
parsimony informative. The 5S spacer varies in length
from 273 to 401 nucleotides. The final combined matrix
comprises 759 characters (8.7% cells scored as missing
data).
The combined analysis produced six most parsi-
monious trees of 785 steps, with a consistency index
(CI) excluding uninformative characters of 0.63 and a
retention index (RI) of 0.84. One of these trees (ran-
domly selected) is shown in Fig. 1. Both subfamilies
of Betulaceae are monophyletic and well supported by
BP and PP (Coryloideae, 96 BP and 1.0 PP; Betuloideae,
79 BP and 1.0 PP). Within Coryloideae, Corylus is sister
to (Ostryopsis (Carpinus,Ostrya)) (Fig. 1). All genera are
monophyletic and have strong bootstrap support ex-
cept Carpinus (65 BP). However, the monophyly of Car-
pinus is supported by a high posterior probability
(0.97; Fig. 1).
Dating. Molecular estimates based on both ML
and MP branch length and calibrated using five fossils
(Table 1) are shown in Table 3, and Fig. 2 summarizes
these estimates for five nodes. In general, MP estimates
are older than the ML estimates, both with ACCTRAN
and DELTRAN optimizations, but ACCTRAN opti-
mization tends to give older ages (Fig. 2; Table 3).
However, there are exceptions, depending on which
node is examined and which fossils are used for cali-
bration. When the fossil of Corylus is used to calibrate
the tree on the crown node of Corylus (N5), the ML
estimates are older than the MP estimates for all the
nodes (Fig. 2; Table 3). Most of the nodes also have
older ML estimates if the same fossil is used to cali-
brate the tree at the stem node of Corylus (N2). More-
over, for some terminal nodes (e.g., N22), the ML es-
timates are older than the MP estimates for all calibra-
tion points (Fig. 2).
Discrepancies between molecular estimates also
vary greatly depending on the fossil used as calibra-
tion point and its position on the tree (crown versus
stem). For example, the ML estimates attributed to the
crown of Betulaceae range from 229.6 Ma (Corylus fos-
124 [Volume 30SYSTEMATIC BOTANY
T
ABLE
3. Age estimates and median values based on maximum likelihood (ML) and maximum parsimony (ACCTRAN and DELTRAN) branch lengths for the 23 nodes of the Betulaceae tree and
the stem node of Betulaceae (N0) with calibration performed with five fossils used both at the crown node (crown) and stem node (stem). Letters in parentheses refer to Table 1. The confidence
interval of the 10 estimates was calculated with a confidence level of 95%. All ages are in millions years.
Node
no. Analysis
Alnus (A)
Crown Stem
Betula (B)
Crown Stem
Corylus (Co)
Crown Stem
Carpinus (Ca)
Crown Stem
Ostrya (O)
Crown Stem Median Confidence
interval
N0 ML 331.8 179.9 419.4 135.6 583.8 211.4 363.0 303.7 301.5 208.9 302.6 81.9
ACCTRAN 304.1 162.2 315.2 122.3 324.9 137.2 275.6 238.2 252.0 163.8 245.1 47.8
DELTRAN 253.3 148.8 303.9 112.1 312.9 135.5 270.6 209.3 213.0 144.0 211.2 45.4
N1 ML 130.5 70.8 164.9 53.3 229.6 83.1 142.8 119.4 118.6 82.1 119.0 32.2
ACCTRAN 162.1 86.5 168.0 65.2 173.2 73.2 146.9 126.9 134.3 87.3 130.6 25.5
DELTRAN 138.2 81.2 165.8 61.2 170.7 73.9 147.6 114.2 116.2 78.5 115.2 24.7
N2 ML 76.9 41.7 97.2 31.4 135.3 49.0 84.1 70.4 69.9 48.4 70.2 19.0
ACCTRAN 108.6 57.9 112.6 43.7 116.0 49.0 98.4 85.0 90.0 58.5 87.5 17.1
DELTRAN 91.6 53.8 109.9 40.6 118.1 49.0 97.8 75.7 77.0 52.1 76.4 16.9
N3 ML 119.9 65.0 151.5 49.0 210.9 76.4 131.1 109.7 108.9 75.5 109.3 29.6
ACCTRAN 121.8 65.0 126.3 49.0 130.1 55.0 110.4 95.4 101.0 65.6 98.2 19.1
DELTRAN 110.7 65.0 132.8 49.0 136.7 59.2 118.2 91.5 93.1 62.9 92.3 19.8
N4 ML 71.2 38.6 90.1 29.1 125.4 45.4 78.0 65.2 64.7 44.9 65.0 17.6
ACCTRAN 87.6 46.8 90.9 35.2 93.6 39.6 79.4 68.6 72.6 47.2 70.6 13.8
DELTRAN 76.3 44.8 91.5 33.8 94.2 40.8 81.5 63.0 64.1 43.3 63.6 13.7
N5 ML 27.9 15.1 35.2 11.4 49.0 17.8 30.5 25.5 25.3 17.5 25.4 6.9
ACCTRAN 45.9 24.5 47.5 18.4 49.0 20.7 41.6 35.9 38.0 24.7 37.0 7.2
DELTRAN 39.7 23.3 47.6 17.6 49.0 21.2 42.4 32.8 33.4 22.5 33.1 7.1
N6 ML 38.7 21.0 49.0 15.8 68.1 24.7 42.4 35.5 35.2 24.4 35.4 9.6
ACCTRAN 47.3 25.2 49.0 19.0 50.5 21.3 42.8 37.0 39.2 25.5 38.1 7.4
DELTRAN 40.8 24.0 49.0 18.1 50.5 21.9 43.6 33.8 34.3 23.2 34.1 7.3
N7 ML 65.0 35.3 82.2 26.6 114.4 41.4 71.1 59.5 59.1 40.9 59.3 16.1
ACCTRAN 65.0 34.7 67.4 26.1 69.4 29.3 58.9 50.9 53.9 35.0 52.4 10.2
DELTRAN 65.0 38.2 78.0 28.8 80.3 34.8 69.4 53.7 54.7 36.9 54.2 11.6
N8 ML 58.5 29.0 67.7 21.9 94.2 34.1 58.6 49.0 48.6 38.7 48.8 13.1
ACCTRAN 62.6 33.4 64.8 25.2 66.8 28.2 56.7 49.0 51.8 33.7 50.4 9.8
DELTRAN 59.3 34.8 71.1 26.3 73.2 31.8 63.3 49.0 49.9 33.7 49.5 10.6
N9 ML 44.8 24.3 56.6 18.3 78.8 28.6 49.0 41.0 40.7 28.2 40.9 11.1
ACCTRAN 54.1 28.8 56.1 21.7 57.8 24.4 49.0 42.3 44.8 29.1 43.6 8.5
DELTRAN 45.9 26.9 55.0 20.3 56.7 24.5 49.0 37.9 38.6 26.1 38.3 8.2
N10 ML 37.1 20.1 46.7 15.2 65.3 23.6 40.6 34.0 33.7 23.4 33.9 9.1
ACCTRAN 40.7 21.7 42.2 16.4 43.4 18.4 36.9 31.8 33.7 21.9 32.8 6.4
DELTRAN 40.1 23.5 48.1 17.7 49.5 21.4 42.8 33.1 33.7 22.8 33.4 7.2
N11 ML 42.4 23.0 53.5 17.3 74.5 27.0 46.3 38.8 38.5 26.7 38.7 10.4
ACCTRAN 46.5 24.8 48.2 18.7 49.7 21.0 42.2 36.4 38.6 25.1 37.5 7.3
DELTRAN 41.9 24.6 50.3 18.6 51.8 22.4 44.8 34.6 35.2 23.8 34.9 7.5
N12 ML 19.4 10.5 24.5 7.9 34.1 12.3 21.2 17.7 17.6 12.2 17.7 4.8
ACCTRAN 19.6 10.5 20.4 7.9 21.0 8.9 17.8 15.4 16.3 10.6 15.9 3.1
DELTRAN 20.8 12.2 24.9 9.2 25.7 11.1 22.2 17.2 17.5 11.8 17.4 3.7
2005] 125FOREST ET AL.: DATING IN BETULACEAE
T
ABLE
3. Continued.
Node
no. Analysis
Alnus (A)
Crown Stem
Betula (B)
Crown Stem
Corylus (Co)
Crown Stem
Carpinus (Ca)
Crown Stem
Ostrya (O)
Crown Stem Median Confidence
interval
N13 ML 16.4 8.9 20.7 6.7 28.9 10.5 18.0 15.0 14.9 10.3 15.0 4.1
ACCTRAN 20.8 11.1 21.6 8.4 22.2 9.4 18.9 16.3 17.3 11.2 16.8 3.3
DELTRAN 18.7 11.0 22.4 8.3 23.1 10.0 20.0 15.4 15.7 10.6 15.6 3.3
N14 ML 18.9 10.2 23.9 7.7 33.2 12.0 20.6 17.3 17.1 11.9 17.2 4.7
ACCTRAN 29.7 15.8 30.8 11.9 31.7 13.4 26.9 23.2 24.6 16.0 23.9 4.7
DELTRAN 26.6 15.6 31.9 11.8 32.8 14.2 28.4 21.9 22.3 15.1 22.1 4.8
N15 ML 15.3 8.3 19.4 6.2 26.9 9.8 16.8 14.0 13.9 9.6 14.0 3.8
ACCTRAN 23.7 12.7 24.6 9.5 25.4 10.7 21.5 18.6 19.7 12.8 19.2 3.7
DELTRAN 17.6 10.3 21.1 7.8 21.8 9.4 18.8 14.6 14.8 10.0 14.7 3.2
N16 ML 12.3 6.7 15.6 5.0 21.7 7.9 13.5 11.3 11.2 7.8 11.3 3.0
ACCTRAN 18.9 10.1 19.5 7.6 20.1 8.5 17.1 14.8 15.6 10.2 15.2 3.0
DELTRAN 14.1 8.3 16.9 6.2 17.4 7.6 15.1 11.7 11.9 8.0 11.8 2.5
N17 ML 30.1 18.3 38.0 19.3 59.9 19.2 32.9 27.5 27.3 18.9 27.4 7.9
ACCTRAN 36.4 19.4 37.8 14.7 38.9 16.4 33.0 28.5 30.2 19.6 29.4 5.7
DELTRAN 30.2 17.7 36.2 13.4 37.3 16.1 32.2 24.9 25.4 17.1 25.2 5.4
N18 ML 20.1 10.9 25.4 8.2 35.4 12.8 22.0 18.4 14.6 12.7 16.5 5.0
ACCTRAN 23.1 12.3 23.9 9.3 24.6 10.4 20.9 18.1 19.1 12.4 18.6 3.6
DELTRAN 19.1 11.2 22.9 8.5 23.6 10.2 20.4 15.8 16.1 10.8 16.0 3.4
N19 ML 16.1 8.7 20.4 6.5 28.4 10.3 17.6 14.7 14.6 10.1 14.7 4.0
ACCTRAN 18.3 9.8 19.0 7.4 19.6 8.3 16.6 14.4 15.2 9.9 14.8 2.9
DELTRAN 15.2 8.9 18.2 6.7 18.8 8.1 16.2 12.6 12.8 8.6 12.7 2.7
N20 ML 51.2 27.8 64.7 20.9 90.1 32.6 56.0 46.9 46.5 32.2 46.7 12.6
ACCTRAN 42.3 22.6 43.8 17.0 45.1 19.1 38.3 33.1 35.0 22.8 34.1 6.6
DELTRAN 46.9 27.6 56.3 20.8 58.0 25.1 50.1 38.8 39.5 26.7 39.2 8.4
N21 ML 38.5 20.9 48.7 15.8 67.8 24.6 42.2 35.3 35.0 24.3 35.2 9.5
ACCTRAN 28.8 15.4 29.9 11.6 30.8 13.0 26.1 22.6 23.9 15.5 23.3 4.5
DELTRAN 25.8 15.2 31.0 11.4 31.9 13.8 27.6 21.3 21.7 14.7 21.5 4.6
N22 ML 32.5 17.6 41.0 13.3 57.1 20.7 35.5 29.7 29.5 20.4 29.6 8.0
ACCTRAN 21.5 11.4 22.3 8.6 23.0 9.7 19.5 16.8 17.8 11.6 17.3 3.4
DELTRAN 21.5 12.7 25.8 9.5 26.6 11.5 23.0 17.8 18.1 12.2 18.0 3.9
N23 ML 26.9 14.6 34.0 11.0 47.3 17.1 29.4 24.6 24.4 16.9 24.5 6.6
ACCTRAN 18.2 9.7 18.9 7.3 19.5 8.2 16.5 14.3 15.1 9.8 14.7 2.9
DELTRAN 18.4 10.8 22.1 8.1 22.8 9.9 19.7 15.2 15.5 10.5 15.4 3.3
126 [Volume 30SYSTEMATIC BOTANY
F
IG
. 2. Variation of the molecular estimates based on ML,
ACCTRAN (ACC), and DELTRAN (DEL) branch lengths for
five nodes of the tree as numbered in Fig. 1 (crown nodes of
Betulaceae (N1), Coryloideae (N2), Betuloideae (N3), Corylus
(N5) and the A. nitida /A. cremastogyne node, N22). Estimates
resulting from the calibration on the crown node (C) and stem
node (S) are indicated for each fossils and branch length op-
timization (black circles, ML; gray circles, ACC; black squares,
DEL). Horizontal lines are median values of the 10 estimates
for each node and branch length optimization (ML, ACC and
DEL). Ages are in millions of years on y-axis.
sil placed on the crown node, N5) to 53.3 Ma (B. leo-
poldae placed on the stem lineage, N3). When MP is
used to calculate branch lengths, B. leopoldae placed on
the crown node of Betula (N6) gives older ages, 168.0
Ma and 165.8 Ma for ACCTRAN and DELTRAN, re-
spectively. Calibration of the tree with B. leopoldae on
the stem node of Betula (N3) is also the younger age
for the crown node of Betulaceae (N1; 65.2 Ma with
ACCTRAN and 61.2 Ma with DELTRAN). The results
of bootstrapped matrices used to calculate standard
deviations on the ML estimates for the crown nodes of
Betulaceae (N1) and Alnus (N7) using the fossil of B.
leopoldae placed on the crown node of Betula are shown
in Fig. 3. Under these parameters, the average of the
bootstrap estimate distribution for the crown node of
Betulaceae (N1) is 152.7 Ma (standard deviation of 24.4
Ma; Fig. 3A) and 78.4 Ma (standard deviation of 15.8
Ma; Fig. 3B) for the crown node of Alnus (N7). These
average estimates are similar to the estimates obtained
by calibration under the same parameters (164.9 Ma
for the crown node of Betulaceae, N1, and 82.2 Ma for
the crown node of Alnus, N7; see Table 3).
Median values were used to compile the ten esti-
mates resulting from the calibration of the tree with
the five fossils at crown and stem nodes. Each graph
in Fig. 2 shows the difference between the estimates at
the crown and stem nodes using the five fossils (see
Table 1). These show that the average value between
crown/stem calibration for any particular node cor-
responds broadly to the median value of the estimates
given when the tree is calibrated with either the fossils
of B. leopoldae (fossil B) or Alnus (fossil A). On the other
hand, calibrations performed with Corylus (fossil Co)
and Carpinus (fossil Ca) give older estimates than the
median values, and calibrations performed with Ostrya
(fossil O) give younger ages (see Fig. 2 and Table 3 for
further details).
Paleobotanical Confidence Intervals. The earliest
record of an Alnus pollen type is of the extinct pollen
genus Paraalnipollenites from the Turonian of North Bo-
hemia (Konzalova 1971). The next earliest record is
from the Coniacian of Japan and was attributed to the
extinct genus Alnipollenites (Miki 1977; Takahashi et al.
1999). Muller (1981) considered this record as the ear-
liest pollen record for Betulaceae, although he recorded
its presence in the Santonian, now interpreted to be
Coniacian of Japan (Takahashi et al. 1999). In situ pol-
len grains found in a fossilized flower from the Late
Santonian and attributed to the extinct genus Bedellia
(Sims et al. 1999) was the only occurrence of this pollen
type recorded from this period. Pollen from Bedellia is
of the Normapolles-type and has the typical arci of
Alnus grains, although they are weakly developed
(Sims et al. 1999). Pollen having a minimum of three
pores and typical arci between them were recorded in
all ages/sub-epochs since the Turonian. The strati-
graphic range of the Alnus-type pollen record is 93.5
Ma and corresponds to the lower boundary of the Tu-
ronian; the number of fossiliferous horizons is 18.
Gap sizes from the end of the Holocene to the end
of the stratigraphic range of the pollen of Alnus and
related extinct genera in the Turonian are shown in
Table 4. Prior to the use of these gaps to determine a
2005] 127FOREST ET AL.: DATING IN BETULACEAE
F
IG
. 3. Histograms of the bootstrap frequencies for the crown node of Betulaceae (N1) and Alnus (N7) when calibrated with
the macrofossil of B. leopoldae (B) at the crown node of Betula (N6) with ML branch-length estimates. A. Crown node of
Betulaceae, N1 (median 150.7 Ma; standard deviation 24.4 Ma). B. Crown node of Alnus, N7 (median 75.9 Ma; standard
deviation 15.8 Ma).
confidence interval on the fossil record, one had to en-
sure that gap sizes were not correlated with their
stratigraphic position using the transposition test. The
number of transpositions for the Alnus-type occur-
rences is T552.5. Using Table 3 in Marshall (1994), we
determined that there is no significant support for in-
creasing gap sizes with age in this sequence (0.05 ,p
,0.10) and the method of Marshall (1994) can be used
(but see Marshall 1994, and Discussion here).
Using the equation of Strauss and Sadler (1989), we
calculated that the time of origin of the Alnus clade
was between 93.5 Ma and 97.4 Ma (Cenomanian) with
a confidence level of 50% and, between 93.5 Ma and
102.8 Ma (Albian to Cenomanian) with a confidence
level of 80% and between 93.5 Ma and 111.5 Ma (Al-
bian to Cenomanian) with a confidence level of 95%.
Using Marshall’s method (1994) with a confidence
probability of 95%, there is 50% chance that the Alnus
clade originated between 96.7 Ma and 99.7 Ma (Albian
to Cenomanian) and 80% chance that it originated be-
tween 98.7 Ma and 105.7 Ma (Albian to Cenomanian).
Paleobotanical confidence intervals are shown on Fig.
4. The upper bound was impossible to determine with
a confidence level of 95% because available data were
insufficient.
D
ISCUSSION
Phylogenetic Relationships. Results of phylogenet-
ic analyses based on combined ITS and 5S spacer se-
quences are consistent with findings of previous stud-
ies of intergeneric relationships (Crane 1989; Bousquet
et al. 1992; Chen et al. 1999). However, the relation-
ships of Carpinus and Ostrya remain uncertain. A
study based on ITS sequences showed that Carpinus
was not monophyletic: Carpinus section Distegocarpus
was sister to a clade formed by section Carpinus and
the genus Ostrya (Chen et al. 1999). Analyses including
additional ITS sequences for Ostrya and Carpinus also
showed that Carpinus is paraphyletic, as for Ostrya
(Yoo and Wen 2002). Based on morphology, however,
Carpinus was monophyletic (Yoo and Wen 2002).
Carpinus section Distegocarpus is found to be para-
phyletic both in the separate (not shown) and com-
bined analyses presented here (Fig. 1) and in the mo-
lecular analysis of Yoo and Wen (2002). The 5S spacer
tree contradicts the ITS tree with regard to the rela-
tionships of Carpinus and Ostrya, with Carpinus being
monophyletic (78 BP; tree not shown). In the combined
analysis, Carpinus is also monophyletic with a high
posterior probability (0.97), but with weak bootstrap
support (65 BP). Additional data are needed to clarify
relationships of Ostrya and Carpinus.
The relationships found here between subgenera
within Corylus are consistent with previous findings,
which showed that subgenus Acanthochlamys is sister
to the remainder of genus Corylus and that subgenera
Siphonochlamys and Phyllochlamys are sister taxa (Forest
and Bruneau 2000). Other studies based on ITS se-
quences proposed similar relationships but with less
resolution at the base of the Siphonochlamys-Phyllochla-
mys clade (Erdogan and Mehlenbacher 2000; Whitcher
and Wen 2001).
In Betula, paraphyly of section Costatae has been
found by previous studies based on ITS sequences
(Bousquet et al. 1992; Chen et al. 1999), although this
contrasts with results based on wood anatomical char-
acters. Hall (1952) suggested that section Costatae was
closely related to section Betula, the former possessing
more advanced wood characters, and that both were
intermediate between section Betulaster and section
Humiles. Hall (1952) also suggested that section Betu-
laster is the most advanced. Lack of resolution cur-
rently limits the inference of relationships among spe-
cies above section Costatae in Betula (Fig. 1).
Alnus subgenus Clethropsis was once thought to com-
prise the most primitive species in Alnus (Murae 1964),
128 [Volume 30SYSTEMATIC BOTANY
T
ABLE
4. Gap sizes for the palynological fossil record of Alnus and related genera. Gap sizes are calculated using consecutive lower
bounds and represent the time interval between a given stratigraphic division and the following one. Note that there are two gaps of
5.2 Ma, which are arbitrarily assigned rank 10 and 11; this does not have any effect on subsequent calculations because they have the
same length. The number of transpositions, T, is calculated for each gap, starting from the Recent. Age ranges are based on the geological
time scale of Palmer and Geismann (1999).
Stratigraphic divisions
with Alnus-type
pollen recorded Age range
(Ma) Gap size
(Ma) Rank
(ascending) T
Holocene 0.01 1.79 2 1
Pleistocene 0.01–1.8 1.8 3 1
Late Pliocene 1.8–3.6 1.7 1 0
Early Pliocene 3.6–5.3 5.9 13 9
Late Miocene 5.3–11.2 5.2 10 6
Middle Miocene 11.2–16.4 7.4 16 10
Early Miocene 16.4–23.8 4.7 9 5
Late Oligocene 23.8–28.5 5.2 11 5
Early Oligocene 28.5–33.7 3.3 6 2
Late Eocene 33.7–37.0 12.0 17 7
Middle Eocene 37.0–49.0 5.8 12 4
Early Eocene 49.0–54.8 6.2 14 4
Late Paleocene 54.8–61.0 4.0 7 2
Early Paleocene 61.0–65.0 6.3 15 3
Maastrichtian 65.0–71.3 12.2 18 3
Campanian 71.3–83.5 2.3 4 0
Santonian 83.5–85.8 3.2 5 0
Coniacian 85.8–89.0 4.5 8 0
Turonian 89.0–93.5
but Furlow (1979) suggested that it has a more derived
position in the genus. In two previous ITS studies, sub-
genus Clethropsis appears either sister to the remainder
of the genus (Chen et al. 1999) or more derived (Bous-
quet et al. 1992) within Alnus, depending on the inclu-
sion of A. glutinosa, which was the only sampling dif-
ference between the two studies (A. glutinosa was in-
cluded in Chen et al. 1999 but not in Bousquet et al.
1992). Here, subgenus Clethropsis is paraphyletic (Fig.
1).
Fossils Versus Molecules. In general, the median
values calculated in our study are older than what has
been estimated from the fossil record. This is espe-
cially the case for the crown of Betulaceae. The oldest
fossil occurrence of the family, the extinct pollen genus
Paraalnipollenites, is from the Turonian (89.0–93.5 Ma;
Konzalova 1971), and the median values of the molec-
ular estimates are 119.0 Ma, 130.6 Ma, and 115.2 Ma
for ML, ACCTRAN, and DELTRAN, respectively,
which are older than we would expect based on the
palynological succession of the mid-Cretaceous. Esti-
mates for subfamily Betuloideae are older than those
inferred from the fossil record, and the same is ob-
served for subfamily Coryloideae (Fig. 4; Table 3).
Median ages given to the crown nodes of Betula
(N6), Corylus (N5), and Carpinus (N9), and to a certain
extent Alnus (N7), are younger than those inferred
from the fossil record (ML estimates of 35.4 Ma, 25.4
Ma, 40.9 and 59.3 Ma, respectively, whereas the fossil
record indicates 49.0 Ma for Betula,Corylus,andCar-
pinus, and 65.0 Ma for Alnus). In one respect, these
younger ages are in accordance with the estimates
based on the fossil record since these fossils can be
placed between the stem and crown nodes. As men-
tioned earlier, if a fossil possesses some characters
placing it in a given clade, it may not be possible with
incomplete fossils to know if it possesses all charac-
teristic synapomorphies of this clade. Thus, if the
crown node of a clade is younger than the earliest fos-
sil that can be attributed to this clade, it could mean
that the fossil species is a stem representative of the
clade, maybe even the sister of the extant species. The
oldest fossil unequivocally assigned to genus Corylus
is from the Middle Eocene (41.3–49 Ma; Wehr 1995;
Pigg et al. 2003), which is too old to be included in the
Corylus crown according to the three molecular esti-
mates but is more appropriate for its stem lineage. A
similar situation pertains for the fossil infructescences
of Alnus. The extinct species, B. leopoldae,couldbe
placed on the stem lineage of the clade formed by ex-
tant species of Betula. Based on the results of our phy-
logenetic analysis and as previously suggested by
Crane and Stockey (1987), B. section Costatae occupies
the basal nodes within the genus. Betula leopoldae has
characters that place it near species of B. section Cos-
tatae (Crane and Stockey 1987), which supports its
placement either along the stem lineage of Betula or in
the grade of extant species that diverged early in the
diversification of the genus.
Extinct Genera. Other well-documented fossils
have been discovered and attributed to Betulaceae but
were not used as calibration points for this study be-
2005] 129FOREST ET AL.: DATING IN BETULACEAE
F
IG
. 4. Median values of the molecular estimates and paleobotanical confidence intervals. All ages in million years (Ma).
A. Stratigraphic column of geological ages and sub-epochs used for the calculation of the paleobotanical confidence intervals
(L, late; M, middle; E, early). B. Stratigraphic range of Alnus pollen-type and paleobotanical confidence interval calculated with
method of Strauss and Sadler (1989) with confidence levels of 50%, 80% and 95%. C. Stratigraphic range of Alnus pollen-type
and paleobotanical confidence interval calculated with method of Marshall (1994) and confidence levels of 50% and 80%. D-F.
Chronograms of Betulaceae based on ML and MP (ACCTRAN and DELTRAN) branch lengths (D, E, and F, respectively).
cause they belong to extinct genera, and their relation-
ships with extant genera are not clear. Nevertheless,
they are valuable for comparisons of molecular and
fossil-based estimates. The most complete records of
these extinct genera are from the widely distributed
genus Palaeocarpinus, found in Paleogene formations
(mostly Paleocene) from Europe (Crane 1981), North
America (Crane et al. 1990; Sun and Stockey 1992;
Wehr 1995; Manchester and Chen 1996; Wehr and
Manchester 1996; Pigg et al. 2003), and Asia (Man-
chester and Guo 1996; Akhmetiev and Manchester
2000; Golovneva 2002). With nutlets similar to those of
Carpinus and Ostrya and a pair of involucral bracts,
Palaeocarpinus clearly belongs to Coryloideae, from
which two other extinct genera are known. Asterocar-
pinus, characterized by a star-shaped involucre, is
thought to have appeared at about the same time as
Carpinus and Ostrya (Manchester and Crane 1987). Os-
tryopsis is the only genus of Betulaceae with no fossil
record so far, but various vegetative and reproductive
structures having similar characters were found in the
Paleocene of Wyoming (USA) and attributed to the ex-
tinct genus Cranea (Manchester and Chen 1998). Pre-
liminary cladistic analyses based on a restricted set of
morphological characters showed that species of Paleo-
carpinus form a polytomy at the crown node of tribe
Carpineae (N4; Forest 1999). These analyses also
placed Cranea as sister to Carpineae and Asterocarpinus
as sister to the Carpinus-Ostrya clade (Forest 1999).
The median values of the estimates for the crown
node of Coryloideae (N2) are 70.2 Ma, 87.5 Ma, and
76.4 Ma for ML, ACCTRAN, and DELTRAN branch
lengths, respectively. These ages are older than the es-
timates based on the oldest fossils of extant genera of
the subfamily (Corylus fruits from the middle Eocene;
Wehr 1995; Pigg et al. 2003). The ML estimates agree
with the fossil record when fossils of the extinct genera
Paleocarpinus and Cranea from the Paleocene are taken
into account (stem and crown of Carpineae are 70.2
and 65.0 Ma, respectively); MP estimates are older. Es-
timates given to the crown node of Carpineae (N4) are
also in concordance with the putative relationships of
this tribe and the extinct genera, Paleocarpinus and Cra-
nea (65.0 Ma for ML, 70.6 Ma for ACCTRAN, and 63.6
130 [Volume 30SYSTEMATIC BOTANY
Ma for DELTRAN). The median value of the crown
node of the Carpinus-Ostrya clade (N8) is older (ML
estimates of 48.8 Ma) than the age of the fossil Aster-
ocarpinus from the Oligocene (23.8–33.7 Ma), its puta-
tive sister taxon. Manchester and Crane (1987) stated
that based on fruit and leaf characters, Asterocarpinus
is more closely related to Carpinus, but the ML estimate
of the crown node of Carpinus (N9) is also older (40.9
Ma) than the fossil of Asterocarpinus. However, Aster-
ocarpinus could have originated earlier than its first oc-
currence in the fossil record.
Rate Heterogeneity. The younger ages attributed to
the crown nodes of Betula (N7), Alnus (N6), and Cor-
ylus (N5) could in part be due to substitution rate in-
equality among lineages (see Sanderson and Doyle
2001). Here, differences in molecular rates among lin-
eages could either be a deceleration in Corylus,Betula,
and Alnus or acceleration in tribe Carpineae. There are
several lines of evidence that favor the deceleration hy-
pothesis. First, the molecular estimates for Carpineae
are broadly concordant with the fossil estimates, and
that is not the case for Corylus,Betula,andAlnus,in
which they are younger. Second, the branch lengths of
the stem lineages of Corylus,Betula,andAlnus relative
to the length of the branches within their respective
crowns are much longer than what is observed in tribe
Carpineae (Fig. 1). Third, species of Corylus and Betula
show less sequence divergence (average Jukes-Cantor
distances of 0.040 and 0.027, respectively) than species
of other genera in Betulaceae (Carpinus, 0.096; Ostrya,
0.062). Alnus maritima is on a long branch, which re-
sults in average Jukes-Cantor distances between spe-
cies of Alnus of 0.068 if A. maritima is included and
0.042 when it is excluded. The use of multiple calibra-
tion points can reduce the impact of rate heterogeneity
over time, and methods that take into account simul-
taneously many calibration points are more likely to
be more robust than those with just one calibration
point; however rate heterogeneity certainly remains an
important problem for the use of molecular data to
estimate divergence times.
Paleobotanical Confidence Intervals and Palynolog-
ical Records. Like estimations of divergence based on
molecular data, determination of confidence intervals
for the fossil record relies on a number of assumptions.
The main difference between the two methods used
here to calculate the confidence intervals with the fossil
record of Alnus-type pollen concern the random fos-
silization assumption. The earlier method (Strauss and
Sadler 1989) assumed that fossiliferous horizons are
randomly distributed, whereas Marshall’s (1994) is less
restrictive regarding this aspect. However, this comes
with the requirement of more fossiliferous horizons to
achieve an increase in confidence levels.
Both Strauss and Sadler (1989) and Marshall (1994)
pointed out that their respective methods are appro-
priate only for continuous rather than discrete sam-
pling. Discrete sampling produces more conservative
paleontological confidence intervals because it under-
estimates the richness of the fossil record (Marshall
1994). In reality, continuous sampling, of drill cores for
example, could produce many more fossiliferous ho-
rizons and thus more restricted paleontological confi-
dence intervals. In compiling data from the literature
and the Plant Fossil Record of the Alnus pollen-type,
we took into account the occurrences reported from
cores and similar continuous sampling strategies, but
overall this should not affect the results as we present
them here. The main concern is that there is a degree
of discreteness in the data because they were com-
partmentalized into subdivisions of the geological
time scale. Geological time-scale subdivisions were
used because we considered that level to be a reason-
able trade-off between reliable global stratigraphic cor-
relation (i.e., integration of fossil occurrences from
multiple continents) and accurate paleontological con-
fidence intervals.
Both methods (Strauss and Sadler 1989; Marshall
1994) assumed that there is no correlation between size
of gaps and stratigraphic position (Marshall 1994). The
use of stratigraphic subdivisions as fossiliferous hori-
zons could become problematic. For example, subdi-
visions nearer the present are finer in the geological
time scale and in some cases this could produce a
gradual decrease in gap sizes toward the top of the
stratigraphic column (Marshall 1994). If there is a sig-
nificant correlation between gap sizes and stratigraph-
ic position, this does not necessarily mean that use of
paleontolgical confidence intervals should be avoided.
Rather, it should remind us to interpret the results
more carefully and serve as a rough estimate of incom-
pleteness of the fossil record (Marshall 1994). In the
case of the alnoid pollen record, there is no significant
correlation between gap sizes and stratigraphic posi-
tion.
Only the 95% paleobotanical confidence intervals
calculated using the method of Strauss and Sadler
(1989) overlap with the median value of the ML esti-
mates for the stem node of Alnus (N3), which corre-
sponds to the crown node of Betuloideae (Fig. 4). The
distribution-free paleobotanical confidence intervals
calculated according to Marshall’s method (1994) with
50% and 80% confidence level are younger than the
ML estimate (Fig. 4).
Paleobotanical confidence intervals calculated here
are likely to be too conservative because these methods
do not take into account quality and density of the
fossil record. Marshall’s (1997) generalized method
was based on that of Strauss and Sadler (1989) with
the addition of a ‘‘recovery potential function’’ that
took into account bias in collecting and preservation.
This approach allows for the determination of pale-
2005] 131FOREST ET AL.: DATING IN BETULACEAE
ontological confidence intervals when the fosiliferous
horizons are non-randomly distributed and should
only be used when the stratigraphic ranges are based
on continuous or nearly continuous sampling (Mar-
shall 1997). More restricted paleontological confidence
intervals could potentially be determined using this
method by also considering the richness of the fossil
record in the calculations.
Molecules, Fossils and the Origin of Betulaceae.
Molecular estimates based on ML and MP with DEL-
TRAN branch length optimization predict that Betu-
laceae started to diversify during the Aptian (121–112
Ma), whereas MP with ACCTRAN branch length op-
timization place the origin of the family even earlier,
in the Hauterivian (132–127 Ma). The median value of
the ML molecular estimate is some 25 million years
older than the first known occurrence of Betulaceae in
the fossil record and also before the first appearance
of triporate pollen grains of the Normapolles type in
the late Cenomanian (99.0–93.5 Ma; Muller 1981). The
paleobotanical confidence intervals of the pollen rec-
ord of Alnus and related genera vary depending on the
method used and the level of confidence applied, but
the results basically indicate that the stem node of ge-
nus Alnus originated in the Albian (112–99 Ma). The
ML estimate for the stem node of Alnus is also found
in the Albian and the crown node estimate is more in
line with fossil estimates. The fact that Betulaceae spe-
cies are wind-pollinated, which usually involves abun-
dant production of pollen, coupled with the richness
of the family in the fossil record both in pollen and
macrofossils, are good indicators that the family (as
well as genus Alnus) could not have been much older
than its first appearance in the palynological record
(i.e., Turonian). The discrepancy between molecular-
based age estimates and the angiosperm fossil record
as a whole still needs to be resolved, but the use of
paleobotanical confidence intervals and multiple cali-
bration points in molecular dating studies might even-
tually bring these estimates to a consensus on the age
of angiosperms groups.
A
CKNOWLEDGEMENTS
. We thank R. Charest, G. Lewis, P. Ma-
nos, T. Pennington, and I. Whitcher for providing material, the
Jardin Botanique de Montre´al and the Arnold Arboretum for as-
sistance, and P. Herendeen, M. Lavin, S. Manchester and an anon-
ymous reviewer for comments on the manuscript. This work was
supported by the Natural Sciences and Engineering Research
Council of Canada (NSERC), the Fonds de recherche sur la nature
et les technologies (Que´ bec) to A. Bruneau, the Institut de Re-
cherche en Biologie Ve´ge´ tale and Deep Time (US NSF DEB-
0090283 to D. Soltis, P. Soltis, D. Dilcher, and P. Herendeen) to F.
Forest.
LITERATURE CITED
A
BBE
, E. C. 1935. Studies in the phylogeny of the Betulaceae I.
Floral and inflorescence anatomy and morphology. Botanical
Gazette 97: 1–67.
———. 1938. Studies in the phylogeny of the Betulaceae II. Ex-
tremes in the range of variation of floral and inflorescence
morphology. Botanical Gazette 99: 431–469.
A
KMETIEV
,M.A.andS.R.M
ANCHESTER
. 2000. A new species of
Paleocarpinus (Betulaceae) from the Paleogene of the eastern
Sikhote-Alin. Paleontological Journal 34: 469–474.
A
NDERSON
, E. and E. C. A
BBE
. 1934. A quantitative comparison of
specific and generic differences in the Betulaceae. Journal of
the Arnold Arboretum 15: 43–49.
B
ALDWIN
,B.G.andM.J.S
ANDERSON
. 1998. Age and rate of di-
versification of the Hawaiian silversword alliance (Compos-
itae). Proceedings of the National Academy of Sciences, USA 95:
9402–9406.
B
ENTON
,M.J.andF.J.A
YALA
. 2003. Dating the tree of life. Science
300: 1698–1700.
B
OBROV
, E. G. 1936. History and systematics of genus Corylus.Sov-
etskaia Botanica 4: 11–39 [in Russian].
B
OUSQUET
, J., S. H. S
TRAUSS
,andP.L
I
. 1992. Complete congruence
between morphological and rbcL-based molecular phyloge-
nies in birches and related species (Betulaceae). Molecular Bi-
ology and Evolution 9: 1076–1088.
B
RUNNER
,F.andD.E.F
AIRBROTHERS
. 1979. Serological investiga-
tion of the Corylaceae. Bulletin of the Torrey Botanical Club 106:
97–103.
B
UDANTSEV
, L. Y. 1982. Alnus Mill, species identified from leaves.
Pp. 120–128 in Magnoliophyta Fossilia USRR 2, Ulmaceae-Betu-
laceae, ed. A. Takhtajan. Leningrad: Nauka [in Russian].
C
HANEY
, R. W. 1927. Geology and paleontology of the Crooked
River basin, with special reference to the Bridge Creek flora.
Carnegie Institution of Washington Publications 346: 45–138.
C
HASE
,M.W.andH.G.H
ILLS
. 1991. Silica gel: an ideal dessicant
for preserving field-collected leaves for use in molecular
studies. Tax on 40: 215–220.
C
HEN
, Z.-D., S. R. M
ANCHESTER
, and H.-Y. S
UN
. 1999. An inte-
grated analysis of DNA sequences, morphology and paleo-
botany for the phylogenetic relationships of the Betulaceae.
American Journal of Botany 86: 1168–1181.
C
OX
,A.V.,M.D.B
ENNETT
, and T. A. D
YER
. 1992. Use of the poly-
merase chain reaction to detect spacer size heterogeneity in
plant 5S-rRNA gene clusters and to locate such clusters in
wheat (Triticum aestivum L.). Theoretical and Applied Genetics
83: 684–690.
C
RANE
, P. R. 1981. Betulaceous leaves and fruits from the British
Upper Paleocene. Botanical Journal of the Linnean Society 83:
103–136.
———. 1989. Early fossil history and evolution of the Betulaceae.
Pp. 87–116 in Evolution, systematics and fossil history of the Ha-
mamelidae, vol. 2: ‘Higher’ Hamamelidae, eds. P. R. Crane and S.
Blackmore. Oxford: Systematics Association, Clarendon
Press.
——— and R. A. S
TOCKEY
. 1987. Betula leaves and reproductive
structures from the Middle Eocene of British Columbia, Can-
ada. Canadian Journal of Botany 65: 2490–2500.
———, S. R. M
ANCHESTER
, and D. L. D
ILCHER
. 1990. A prelimi-
nary survey of fossil leaves and well-preserved reproductive
structures from the Sentinel Butte formation (Paleocene) near
Almond, North Dakota. Fieldiana, Geology, New Series, 20.
D
OYLE
,J.A.andM.J.D
ONOGHUE
. 1993. Phylogenies and angio-
sperm diversification. Paleobiology 19: 141–167.
D
OYLE
J. J. and J. L. D
OYLE
. 1987. A rapid DNA isolation procedure
for small quantities of fresh leaf tissue. Phytochemical Bulletin.
19: 11–15.
E
RDOGAN
, V. and S. A. M
EHLENBACHER
. 2001. Phylogenetic rela-
tionships of Corylus species (Betulaceae) based on nuclear ri-
bosomal DNA ITS region and chloroplast matK gene sequenc-
es. Systematic Botany 25: 727–737.
F
ELSENSTEIN
, J. 1985. Confidence limits on phylogenies: an ap-
proach using the bootstrap. Evolution 39: 783–791.
132 [Volume 30SYSTEMATIC BOTANY
———. 1988. Phylogenies from molecular sequences: inference
and reliability. Annual Review of Genetics 22: 521–565.
F
ITCH
, W. M. 1971. Towards defining the course of evolution: min-
imum change for a specified tree topology. Systematic Zoology
20: 406–416.
F
OOTE
,M.,J.P.H
UNTER
,C.M.J
ANIS
, and J. J. S
EPKOSKI
. 1999.
Evolutionary and preservational constraints of origins of bi-
ologic groups: divergence times in eutherian mammals. Sci-
ence 283: 1310–1314.
F
OREST
, F. 1999. Implications des analyses phyloge´ne´tiques et bioge´o-
graphique de l’espaceur des ge`nes ribosomaux 5S chez le genre Cor-
ylus L. (Betulaceae). M.Sc. thesis, Departement de Sciences
Biologiques, Universite´ de Montre´ al, Canada.
——— and A. B
RUNEAU
. 2000. Phylogenetic analysis, organiza-
tion, and molecular evolution of the nontranscribed spacer of
the 5S ribosomal RNA genes in Corylus (Betulaceae). Inter-
national Journal of Plant Sciences 161: 793–806.
F
URLOW
, J. J. 1979. The systematics of the American species of
Alnus (Betulaceae). Rhodora 81: 1–121, 151–248.
———. 1990. The genera of Betulaceae in the southeastern United
States. Journal of the Arnold Arboretum 71: 1–67.
———. 1997. Betulaceae, birch family. Pp. 507–538 in Flora of North
America north of Mexico vol. 3, ed. Flora of North America
Editorial Committee. New York: Oxford University Press.
G
OLOVNEVA
, L. B. 2002. Paleocarpinus (Betulaceae) from the Paleo-
gene of Spitsbergen and transatlantic floristic migrations. Pa-
leontological Journal 36: 422–428.
H
ALANYCH
, K. M. 1991. 5S ribosomal RNA sequences inappro-
priate for phylogenetic reconstruction. Molecular Biology and
Evolution 8: 249–253.
H
ALL
, J. W. 1952. The comparative anatomy and phylogeny of the
Betulaceae. Botanical Gazette 113: 235–270.
H
AMMEL
,B.andW.G.B
URGER
. 1991. Neither oak nor alder, but
nearly: the history of Ticodendraceae. Annals of the Missouri
Botanical Garden 78: 89–95.
H
ARDIN
,J.W.andJ.M.B
ELL
. 1986. Atlas of foliar surface features
in woody plants, IX—Betulaceae of eastern United States.
Brittonia 38: 133–144.
H
ASEGAWA
, M., H. K
ISHINO
,andT.Y
ANO
. 1985. Dating the hu-
man-ape splitting by a molecular clock of mitochondrial
DNA. Journal of Molecular Evolution 21: 160–174.
H
JELMQVIST
, H. 1948. Studies on the floral morphology and phy-
logeny of the Amentiferae. Botaniska Notiser S2: 5–171.
H
UELSENBECK
,J.P.andF.R.R
ONQUIST
. 2001. MrBAYES: Bayesian
inference of phylogenetic trees. Bioinformatics 17: 754–755.
K
IKUZAWA
, K. 1982. Leaf survival and evolution in Betulaceae. An-
nals of Botany 50: 345–353.
K
LAK
,C.,G.R
EEVES
,andT.H
EDDERSON
. 2004. Unmatched tempo
of evolution in southern African semi-desert ice plants.Nature
427: 63–65.
K
ONZALOVA
, M. 1971. Paraalnipollenites Hills and Wallace 1969, in
the Turonian of the upper Cretaceous of North Bohemia.
Vestnik Ustredniho Geologickeho 46: 39–40.
M
ANCHESTER
, S. R. and Z.-D. C
HEN
. 1996. Paleocarpinus aspinosa sp.
nov. (Betulaceae) from the Paleocene of Wyoming, U.S.A. In-
ternational Journal of Plant Sciences 157: 644–655.
——— and ———. 1998. A new genus of Coryloideae (Betulaceae)
from the Paleocene of North America. International Journal of
Plant Sciences 159: 522–532.
——— and P. R. C
RANE
. 1987. A new genus of Betulaceae from
the Oligocene of western North America. Botanical Gazette
148: 263–273.
——— and S.-X. G
UO
. 1996. Paleocarpinus (extinct Betulaceae) from
northwestern China: new evidence for Paleocene floristic con-
tinuity between Asia, North America, and Europe. Interna-
tional Journal of Plant Sciences 157: 240–246.
M
ANOS
,P.S.andK.P.S
TEELE
. 1997. Phylogenetic analyses of
‘‘higher’’ Hamamelididae based on plastid sequence data.
American Journal of Botany 84: 1407–1419.
M
ARSHALL
, C. R. 1990. Confidence intervals on stratigraphic rang-
es. Paleobiology 16: 1–10.
———. 1994. Confidence intervals on stratigraphic ranges: partial
relaxation of the assumption of randomly distributed fossil
horizons. Paleobiology 20: 459–469.
———. 1997. Confidence intervals on stratigraphic ranges with
nonrandom distributions of fossil horizons. Paleobiology 23:
165–173.
M
IKI
, A. 1977. Late Cretaceous pollen and spore floras of northern
Japan: composition and interpretation. Journal of the Faculty of
Science, Hokkaido University, Ser. IV, Geology and Mineralogy 17:
399–436.
M
ULLER
, J. 1981. Fossil pollen records of extant angiosperms. Bo-
tanical Review 47: 1–142.
M
URAE
, S. 1964. Phytotaxonomical and geobotanical studies on the
genus Alnus in Japan. III. Taxonomy of whole world species
and distribution of each section. Bulletin of the Government For-
estry Experimental Station, Japan 171: 1–107.
P
ALMER
,A.R.andJ.G
EISMANN
. 1999. 1999 Geologic time scale.
The Geological Society of America (available at www.
geosociety.org/science/timescale/timescl.pdf).
P
IGG
, K. B., S. R. M
ANCHESTER
, and W. C. W
EHR
. 2003. Corylus,
Carpinus,andPalaeocarpinus (Betulaceae) from the middle Eo-
cene Klondike Mountain and Allenby formations of north-
western North America. International Journal of Plant Sciences
164: 807–822.
R
ICHARDSON
,J.E.,F.M.W
EITZ
,M.F.F
AY
,Q.C.B.C
RONK
,H.P.
L
INDER
,G.R
EEVES
, and M. W. C
HASE
. 2001. Rapid and recent
origin of species richness in the Cape flora of South Africa.
Nature 412: 181–183.
S
ANDERSON
, M. J. 1997. A nonparametric approach to estimating
devergence times in the absence of rate constancy. Molecular
Biology and Evolution 14: 1218–1231.
———. 1998. Estimating rate and time in molecular phylogenies:
beyond the molecular clock? Pp. 242–264 in Molecular system-
atics of plants II, DNA sequencing, eds D. E. Soltis, P. S. Soltis,
and J. J. Doyle. Boston: Kluwer Academic Publishers.
——— and J. A. D
OYLE
. 2001. Sources of error and confidence
intervals in estimating the age of angiosperms from rbcL and
18S rDNA data. American Journal of Botany 88: 1499–1516.
S
AVARD
, L., M. M
ICHAUD
,andJ.B
OUSQUET
. 1993. Genetic diver-
sity and phylogenetic relationships between birches and al-
ders using ITS, 18S rRNA, and rbcL gene sequences. Molecular
Phylogenetics and Evolution 2: 112–118.
S
AVOLAINEN
, V., M. F. F
AY
,D.C.A
LBACH
,A.B
ACKLUND
,M.
VAN
DER
B
ANK
,K.M.C
AMEROUN
,S.A.J
OHNSON
,M.D.L
LEDO
,
J.-C. P
INTAUD
,M.P
OWELL
,M.C.S
HEANAN
,D.E.S
OLTIS
,P.
S. S
OLTIS
,P.W
ESTON
,W.M.W
HITTEN
,K.J.W
URDACK
,and
M. W. C
HASE
. 2000. Phylogeny of the eudicots: a nearly com-
plete familial analysis based on rbcL gene sequences. Kew Bul-
letin 55: 257–309.
S
IMS
,H.J.,P.S.H
ERENDEEN
,R.L
UPIA
,R.A.C
HRISTOPHER
,andP.
R. C
RANE
. 1999. Fossil flowers with Normapolles pollen from
the Upper Cretaceous of southeastern North America. Review
of Palaeobotany and Palynology 106: 131–151.
S
OLTIS
,P.S.,D.E.S
OLTIS
,V.S
AVOLAINEN
,P.R.C
RANE
,andT.G.
B
ARRACLOUGH
. 2002. Rate heterogeneity among lineages of
tracheophytes: integration of molecular and fossil data and
evidence for molecular living fossils. Proceedings of the Nation-
al Academy of Sciences, USA 99: 4430–4435.
S
PRINGER
, M. S. 1995. Molecular clocks and the incompleteness of
the fossil record. Journal of Molecular Evolution 41: 531–538.
S
TEELE
, K. P., K. E. H
OLSINGER
,R.K.J
ANSEN
,andD.W.T
AYLOR
.
1991. Assessing the reliability of 5S rRNA sequence data for
phylogenetic analysis in green plants. Molecular Biology and
Evolution 8: 240–248.
2005] 133FOREST ET AL.: DATING IN BETULACEAE
S
TRAUSS
,D.andP.M.S
ADLER
. 1989. Classical confidence intervals
and Bayesian probability estimates for ends of local taxon
ranges. Mathematical Geology 21: 411–427.
S
UN
, F. and R. A. S
TOCKEY
. 1992. A new species of Paleocarpinus
(Betulaceae) based on infructescences, fruits, and associated
staminate inflorescences and leaves from the Paleocene of Al-
berta, Canada. International Journal of Plant Sciences 153: 136–
146.
S
UN
, Y., D. Z. S
KINNER
,G.H.L
IANG
, and S. H. H
ULBERT
. 1994.
Phylogenetic analysis of Sorghum and related taxa using in-
ternal transcribed spacers of nuclear ribosomal DNA. Theo-
retical and Applied Genetics 89: 26–32.
S
UZUKI
,Y.,G.V.G
LAZKO
, and M. N
EI
. 2002. Overcredibility of
molecular phylogenies obtained by Bayesian phylogenetics.
Proceedings of the National Academy of Sciences, USA 99: 16138–
16143.
S
WOFFORD
, D. L. 2001. PAUP* 4.0: Phylogenetic analysis using par-
simony (and other methods). Sunderland: Sinauer Associates.
T
AKAHASHI
, M., P. R. C
RANE
, and H. A
NDO
. 1999. Fossils flowers
and associated plant fossils from the Kamikitaba Locality
(Ashizawa Formation, Futaba Group, Lower Coniacian, Up-
per Cretaceous) of northeast Japan. Journal of Plant Research
112: 187–206.
T
HOMPSON
,J.D.,T.J.G
IBSON
,F.P
LEWNIAK
,F.J
EANMOUGIN
,and
D. G. H
IGGINS
. 1997. The ClustalX windows interface: flexible
strategies for multiple sequence alignment aided by quality
analysis tools. Nucleic Acids Research 25: 4876–4882.
W
EHR
, W. C. 1995. Early Tertiary flowers, fruits, and seeds of
Washington State and adjacent areas. Washington Geology 23:
3–16.
——— and S. R. M
ANCHESTER
. 1996. Paleobotanical significance
of flowers, fruits, and seeds from the Eocene of Republic,
Washington. Washington Geology 24: 25–27.
W
HITCHER
,I.N.andJ.W
EN
. 2001. Phylogeny and biogeography
of Corylus (Betulaceae): inference from ITS sequences. System-
atic Botany 26: 283–298.
W
HITE
T. J. , T. B
RUNS
,S.L
EE
,andJ.T
AYLOR
. 1990. Amplification
and direct sequencing of fungal ribosomal RNA genes for
phylogenetics. Pp. 315–322 in PCR protocols: a guide to methods
and applications, eds. M. Innis, D. Gelfand, J. Sninsky, and T.
White. San Diego: Academic Press.
W
IKSTRO
¨M
,N.andP.K
ENRICK
. 2001. Evolution of Lycopodiaceae
(Lycopsida): estimating divergence times from rbcL gene se-
quences by use of nonparametric rate smoothing. Molecular
Phylogenetics and Evolution 19: 177–186.
———, V. S
AVOLAINEN
, and M. W. C
HASE
. 2001. Evolution of the
angiosperms: calibrating the family tree. Proceedings of the
Royal Society of London, Series B 268: 2211–2220.
———, ———, and ———. 2003. Angiosperm divergence times:
congruence and incongruence between fossils and sequence
divergence estimates. Pp. 142–165 in Telling the evolutionary
time: molecular clocks and the fossil record, eds. P. C. J. Donoghue
and M. P. Smith. London: Taylor and Francis.
W
ILCOX
,T.P.,D.J.Z
WICKL
,T.A.H
EATH
, and D. M. H
ILLIS
. 2002.
Phylogenetic relationships of the dwarf boas and a compar-
ison of Bayesian and bootstrap measures of phylogenetic
support. Molecular Phylogenetics and Evolution 25: 361–371.
W
INKLER
, H. 1904. Betulaceae. Pp. 1–149 in Das Pflanzenreich, Regni
vegetabilis conspectus 19 (IV, 61), ed. A. Engler. Leipzig: En-
gelmann.
Y
ANG
, Z. 1993. Maximum likelihood estimation of phylogeny from
DNA sequences when substitution rates differ over sites. Mo-
lecular Biology and Evolution 10: 1396–1401.
———. 1996. Among-site rate variation and its impact on phylo-
genetic analyses. Trends in Ecology and Evolution 11: 367–372.
Y
OO
, K.-O. and J. W
EN
. 2002. Phylogeny and biogeography of Car-
pinus and subfamily Coryloideae (Betulaceae). International
Journal of Plant Sciences 163: 641–650.
