Full text of DOI:10.1111/j.0014-3820.2003.tb00347.x

Evolution, 57(6), 2003, pp. 1397–1410
AFLP PHYLOGENY OF MIMULUS SECTION ERYTHRANTHE AND THE EVOLUTION OF
HUMMINGBIRD POLLINATION
1
P
AUL M. BEARDSLEY, ALAN
Y
EN
,
AND
RICHARD G. OLMSTEAD
University of Washington, Department of Botany, Box 355325, Seattle, Washington 98115
1
E-mail: pbeard@u.washington.edu
Abstract. Species in Mimulus section Erythranthe (monkeyflowers) have become model systems for the study of the
genetic basis of ecological adaptations. In this study, we pursued two goals. First, we reconstructed the phylogeny of
species in Erythranthe using both DNA sequences from the ribosomal DNA ITS and ETS and AFLPs. Data from
rDNA sequences support the monophyly of the section, including M. parishii, but provide little support for relationships
within it. Analyses using AFLP data resulted in a well-supported hypothesis of relationships among all Erythranthe
species. Our second goal was to reconstruct ancestral pollination syndromes and ancestral states of individual characters
associated with hummingbird-pollinated flowers. Both parsimony and likelihood approaches indicate that hummingbird
pollination evolved twice in Erythranthe from insect-pollinated ancestors. Our reconstruction of individual characters
indicates that corolla color and some aspects of corolla shape change states at the same point on the phylogenetic
tree as the switch to hummingbird pollination; however, a switch to secretion of high amounts of nectar does not.
Floral trait transformation may have been more punctuational than gradual.
Key words. Amplified fragment-length polymorphisms, Erythranthe, external transcribed spacer, hummingbird-pol-
lination, internal transcribed spacer, Mimulus.
Received February 19, 2002. Accepted January 2, 2003.
Well-resolved species-level phylogenies and estimates of dently at least three times from insect-pollinated ancestors
ancestral states are necessary to interpret and understand the (Beardsley and Olmstead 2002).
evolution of interesting ecological adaptations between spe-
Both Mullerian and Batesian mimicry have been suggested
cies. Ancestral pollination syndromes and character states in as mechanisms by which hummingbird pollination can evolve
hummingbird-pollinated plants have received particular at- in western North America (Grant 1966; Brown and Kodrick-
tention from plant evolutionary biologists. Stebbins (1989) Brown 1979; Bleiweiss 2001). Most hummingbird species in
estimated that 108 of 129 hummingbird-pollinated taxa in western North America are migratory, thus requiring them
the western United States have sister taxa that are insect- to be able to quickly identify food sources in new areas (Grant
pollinated with hummingbird pollination being the derived 1966). According to this hypothesis, hummingbirds learn to
state. In a review of attempts to use phylogenies to determine associate floral color (red) and shape with nectar rewards.
the direction of shifts in pollination syndromes, Weller and Convergence in floral design and rewards thus increases hum-
Sakai (1999) report three studies in which the evolution of mingbird pollination through Mullerian mimicry. Batesian
hummingbird pollination has been investigated (McDade mimicry occurs if rare populations or species possess con-
1
992; Bruneau 1997; Hodges 1997). McDade (1992) found vergent floral displays without offering nectar rewards
that long, decurved corollas associated with pollination by (Brown and Kodrick-Brown 1979).
specialist hermit hummingbirds are ancestral to short corollas
Species in Mimulus section Erythranthe (monkeyflowers)
associated with pollination by generalist hummingbirds. In have become one of the most thoroughly studied model sys-
the Aphelandra pulcherrima complex, short corollas were hy- tems for the study of ecological adaptations (Hiesey et al.
pothesized to have arisen twice. Hummingbird pollination in 1971), including adaptations to different pollinators (Suth-
Erythrina was estimated by Bruneau (1997) to have arisen erland and Vickery 1993; Vickery 1995, Bradshaw et al.
at least four times from ancestral passerine pollination. Aq- 1995, 1998). Five of six species traditionally placed in Er-
uilegia species that are pollinated by hummingbirds were ythranthe are hummingbird-pollinated (M. cardinalis, M. eas-
hypothesized by Hodges (1997) to be derived from taxa with twoodiae, M. verbenaceus, M. nelsonii, and M. rupestris) and
less-specialized, open, radiate corollas. O’Kane and Schaal one is bee-pollinated (M. lewisii; Sutherland and Vickery
(
1998) used a molecular phylogeny to infer that hummingbird 1993). Erythranthe played an important role in Vickery’s
pollination in Lopezia sect. Jehlia was secondarily regained analysis of evolution in species complexes (Vickery 1978)
after a switch to fly pollination. Givnish et al. (2000) also and his studies on color polymorphisms and their effect on
used molecular data to infer the relatively recent evolution pollinator visitation (Sutherland and Vickery 1993) and their
of hummingbird pollination from bee-pollinated ancestors in potential role in reproductive isolation (Vickery 1995). Stud-
Rapateaceae in the western Guyana Shield. Hummingbird ies on the genetic basis for floral differences between M.
pollination was hypothesized to have arisen three times in- lewisii and M. cardinalis (Bradshaw et al. 1995, 1998; Schem-
dependently from bee-pollinated ancestors in tribe Antirrhi- ske and Bradshaw 1999) strongly suggest that quantitative
neae (Veronicaceae, sensu Olmstead et al. 2001; Ghebrehiwet trait loci (QTLs) of large effect have been important for the
et al. 2000). Similarly, in our broad-scale analyses of phy- evolution of traits associated with pollinator preference and,
logenetic relationships within Mimulus and closely related thus, reproductive isolation between these taxa. These studies
genera, we demonstrated that within the newly redefined have also added insight into the genetics of mimicry (Blei-
Phrymaceae, hummingbird pollination has arisen indepen- weiss 2001). Further insight into the evolution of evolution-
1
397
᭧
2003 The Society for the Study of Evolution. All rights reserved.

1
398
PAUL M. BEARDSLEY ET AL.
F
IG. 1. Distribution map of taxa identified as belonging to a monophyletic Erythranthe.
ary processes in Erythranthe has been limited, however, be- (Cunningham et al. 1998). Reconstructing the sequence of
cause no well-resolved phylogeny exists for members of the character evolution within Erythranthe provides an empirical
section, although they have been the subject of many sys- case study of the hypothesis that hummingbird-pollinated
tematic studies (Grant 1924; Hiesey et al. 1971; Vickery and taxa are derived from insect-pollinated ancestors in western
Wullstein 1987).
North America (Grant and Grant 1968) and an important
Thus, the first goal of this study was to estimate species- context for understanding the genetic changes between M.
level phylogenies in Mimulus section Erythranthe using two lewisii and M. cardinalis (Bradshaw et al. 1995, 1998).
methods; sequence data from the nuclear ribosomal (nr) DNA
internal and external transcribed spacer (ITS and ETS, re-
spectively; Baldwin et al. 1995; Baldwin and Markos 1998)
regions, and data from amplified fragment length polymor-
phisms (AFLPs; Vos et al. 1995). Recent studies of ITS and
MATERIALS AND METHODS
Sampling
Sampling in the ITS/ETS study was designed to examine
ETS sequences within Mimulus (Beardsley and Olmstead the monophyly of section Erythranthe and to determine ap-
001) indicate that species in Erythranthe are part of a recent propriate outgroups within Mimulus. Included in this analysis
2
radiation, a fact that has made phylogenetic reconstruction are the six species in the currently described section Ery-
solely with ITS and/or ETS more challenging in other groups thranthe, plus 13 other Mimulus species to serve as outgroups
(
Aquilegia [Hodges 1997]; Bidens [Ganders et al. 2000]). and to identify the sister species or clade for Erythranthe
Polymorphic DNA-based markers, such as AFLPs, have been (Table 1). Multiple individuals were sampled for most of the
useful in plant phylogenetic studies (Wolfe and Liston 1998; species in Erythranthe throughout their respective ranges
Caicedo et al. 1999; Xu and Sun 2001; Zhang et al. 2001) (Fig. 1). Most of the material for individuals in Erythranthe
when DNA sequence variation is limited. The second goal came from herbarium specimens. Results from molecular sys-
of this analysis was to reconstruct the evolution of pollination tematic analyses currently underway (P. Beardsley, unpubl.
syndromes in Erythranthe and to infer the ancestral states of data) of nearly every species of Mimulus indicate that our
a suite of traits associated with hummingbird pollination. sample contains all the species that are closely related to
Ancestral character states were optimized using both maxi- Erythranthe.
mum parsimony (Maddison and Maddison 1992) and max-
Due to the variability of AFLP markers and the increasing
imum-likelihood (Pagel 1994). Maximum likelihood methods probability of detected fragments being homoplastic with in-
have the advantage of allowing uncertainty to be quantified creasing genetic distance between taxa, only comparisons

ERYTHRANTHE PHYLOGENY
1399
among closely related taxa are appropriate (Mueller and Wol- purified using Qiaquick spin-columns. DNA samples were
fenbarger 1999). Thus, only taxa in Erythranthe and in its quantified with a spectrophotometer. Two sampling strategies
sister clade were sampled in our AFLP study. Individual ac- were employed. First, AFLP profiles were generated for every
cessions sampled for the AFLP study within Erythranthe dif- individual in the study. Second, after confirming that named
fer from those in the ITS/ETS study because fresh or silica- species were monophyletic, DNA from one to three individ-
dried material is necessary for generating AFLPs. When con- uals, selected to represent the range of diversity present in
ducting analyses with highly polymorphic markers, it is vital the analysis of individuals, was bulked and AFLPs were sep-
to obtain an appropriate frame of reference by sufficiently arately developed and screened for the seven species in Er-
sampling intraspecific variation (Avise 1994). As a result, ythranthe and one outgroup (Wolfe and Liston 1998).
many populations for each taxon were collected, with the
Procedures for AFLPs were as described by Vos et al.
exception of M. nelsonii and the rare M. rupestris, which were (1995) with some modifications. Digestion and ligation: 400–
provided by Bob Vickery (Table 1). An attempt was made 500 ng DNA of each sample was digested and ligated with
to sample across the geographic range of each taxon.
5 U of EcoRI and MseI in a reaction mixture that contained
L 10X RL buffer (100 mM tris-acetate pH 7.6, 100 mM
Mg-acetate, 500 mM K-acetate, 50 mM dithiothreitol), 5
pmol EcoRI adapter, 50 pmol MseI adapter, 1 L 10 mM
ATP, 1 U T4-DNA ligase, and water to a final volume of 50
L. Adapter sequences are the same as those described by
5
␮
ITS/ETS Analysis
␮
The modified cetyltrimethyl ammonium bromide (CTAB)
method of Doyle and Doyle (1987) was used to extract total
genomic DNA, which was further purified using Qiaquick
spin-columns (Qiagen, Valencia, CA). The entire ITS region
was amplified using its4 and its5 primers (Baldwin 1992).
␮
Xu et al. (2000). After digestion and ligation, products were
diluted 1:4 with water.
In the first round of PCR (pre-selective amplification), each
To amplify a portion of the ETS, we used the 3
primer of Baldwin and Markos (1998). The 5 primer, ETS-
B (5 - ATAGAGCGCGTGAGTGGTG-3 ), was designed us-
ing Mimulus sequences as a reference. The polymerase chain
reaction (PCR) conditions were: 35 cycles of 94 C for 1 min,
C for 1 min, and 72 C for 1 min. Polymerase chain re-
Ј 18S-IGS
reaction contained 1.2
action buffer, 0.8 L MgCl2 (25mM), 0.4
mM), 14.32 L water, 0.08 L Promega Taq Polymerase,
.6 L MseI C primer (50 ng/ L), 0.6 L EcoRI
primer (50 ng/ L), and 2.0 L of the diluted digestion/li-
gation product. The PCR conditions were: 94 C for 30 sec,
C for 60 sec, and 72 C for 60 sec, for 28 cycles. The PCR
product was diluted 1:20 with water. Samples used in the
␮
L Promega (Madison, WI) 10X re-
Ј
␮
␮
L dNTPs (10
Ј
Ј
␮
␮
0
␮
ϩ
␮
␮
ϩ A
Њ
␮
␮
5
0
Њ
Њ
Њ
action products were purified using Qiaquick spin-columns
according to the manufacturer’s protocol. Sequences were
generated on an ABI 377 (Applied Biosystems, Foster City,
6
0
Њ
Њ
CA) for both strands of the PCR product. ITS was sequenced bulked DNA study were run individually through the pre-
using the external PCR primers, its4 and its5, and the two selective amplification step. Subsequently, 10 L of the di-
internal primers, its2 and its3. ETS was sequenced using the luted preselective amplification from each of three individ-
8S-E primer of Baldwin and Markos (1998), which is slight- uals was combined. The combined product was used for the
␮
1
ly internal to 18S-IGS primer, and the ETS-B primer. Elec- selective amplification step. Specific control amplifications
tropherograms for each region were compiled and compared were performed to ensure that this dilution of the preselective
using the program Sequencher version 3.0 (Gene Codes Cor- amplification did not influence the number or size of bands
poration, Ann Arbor, MI), from which a consensus sequence recovered in the bulked DNA analysis.
was generated. Consensus sequences for ITS and ETS were
aligned manually.
In the second round of PCR (selective amplification), each
reaction contained 2.0 L Promega 10X Reaction Buffer, 1.6
L MgCl2 (25 mM), 0.4 L dNTPs (10 mM), 11.275
L Promega Taq Polymerase, 0.625 L MseI
Farris et al. 1994, as implemented in PAUP*) was used to XXX primer (50 ng/
L), 1.0 L EcoRI XXX (
assess potential conflicts between the phylogenetic signal primer (1 ng/
L), 3.0
from ITS and ETS. In this test, 500 replicates were analyzed PCR conditions were: 94
with a heuristic search, each with 10 random sequence ad- for 2 min, 1 cycle; 94
C for 1 sec, 64
C for 2 min, 8 cycles; 94 C for 1 sec, 56
C for 2 min, 23 cycles. All samples were processed
␮
Sequence data was analyzed using PAUP* 4.0b3a (Swof-
␮
␮
␮L
ford 1998). The incongruence length difference (ILD) test H2O, 0.1
␮
␮
ϩ
(
␮
␮
ϩ
ϩ
dye)
␮
␮
L diluted preselective product. The
C for 2 min, 65 C for 30 sec, 72
C for 30 sec ( 1.0
C for 30
Њ
Њ
ЊC
ЊC/
Њ
Њ
Ϫ
dition replicates with MULTREES and TBR branch swap-
ping, but limited to 1000 trees. Parsimony searches were
conducted using heuristic searches with 10,000 random se-
quence addition replicates, TBR swapping and MULTREES
to find multiple tree islands, if present (Maddison 1991).
Identical sequences were excluded from the analysis. Gaps
were scored as missing data. Support for individual branches
was estimated using the bootstrap (Felsenstein 1985), which
was calculated using 1000 replicate heuristic searches with
MULTREES and TBR branch swapping, limited to 1000 trees
per replicate.
cycle), 72
sec, 72
Њ
Њ
Њ
Њ
in random order and subsamples were re-run for internal con-
trol.
Samples were prepared for analysis on an ABI 377 by
combining 0.75
.5 L of loading buffer. Loading buffer is made using 180
L formamide, 50 L GeneScan 500-Tamara size standard,
and 70 L blue loading dye (comes with standard). Samples
␮L of selective amplification product with
1
␮
␮
␮
␮
were run on a 4.0% denaturing polyacrylamide gel. Raw data
were collected using ABI GeneScan software (Applied Bio-
systems). Amplification products were sized in reference to
the size standard using the local Southern method, which
AFLP Analysis
The modified CTAB method of Doyle and Doyle (1987) allowed samples from different gels to be aligned accurately.
was used to extract total genomic DNA, which was further Aligned data were imported into Genographer (ver. 1.4, Mon-

1
400
PAUL M. BEARDSLEY ET AL.

ERYTHRANTHE PHYLOGENY
1401
tana State University, 1998; http://hordeum.oscs.montana.
edu/genographer/) for visualization and scoring. AFLP loci
were analyzed using the ‘‘thumbnail’’ option of Genogra-
pher, which allows comparison of signal strength at each
locus for all samples. Band presence or absence was scored
for each sample and recorded as a binary character. For the
bulked DNA analysis, polymorphisms within species were
not detectable because the samples were combined. Data ma-
trices were constructed and analyzed using PAUP* 4.0b3a
(
Swofford 1998).
Relationships within Erythranthe were explored using par-
simony and neighbor-joining (NJ) analyses. For analysis of
the dataset with all individuals represented, parsimony and
bootstrap searches were conducted as in the ITS/ETS anal-
ysis. For analysis of the bulked dataset, parsimony searches
were conducted using a branch and bound search. Bootstrap
values were calculated using 1000 replicate full branch and
bound parsimony searches with MULTREES and TBR
branch swapping. Distance estimates for the NJ analysis were
calculated using the index of Nei and Li (1979).
Exploration of Character Changes and Ancestral
State Reconstruction
Character data for extant taxa were obtained both from
the literature (Hiesey et al. 1971; Vickery and Wullstein
1
987; Sutherland and Vickery 1993; Thompson 1993; Vick-
ery and Sutherland 1994; Bradshaw et al. 1998) and from
measurements on greenhouse-grown populations of M. par-
ishii. The following characters were studied: pollination
syndrome (scored as a single character for heuristic pur-
poses), nectar production, stamen length, pistil length, de-
gree of reflex of upper and lower petals, and three petal
color characters. The three petal color characters are (1)
presence of carotenoids (yellow in color) in the upper petal,
which is controlled by one gene, the yellow upper petal
(
YUP) locus (Hiesey at al. 1971, Bradshaw et al. 1998) in
M. cardinalis and M. lewisii; (2) presence of carotenoids in
the lower petal, a trait controlled by more than one locus
in these two taxa (Hiesey et al. 1971); and (3) a simple
description of corolla color. Corolla color has been shown
by previous researchers to play a role in pollinator visitation
(
Vickery 1995). Continuous characters were converted to
discrete, binary characters by constructing two bins for each
character (Table 2) for further analyses. Models of character
change were inferred on the optimal tree from the bulked
DNA AFLP study in which each species was represented
by one terminal taxon with the exception that M. lewisii was
represented by both races. Outgroup relationships were de-
termined in the ITS/ETS study. This tree, with added out-
groups, is referred to as the input tree. Ancestral state re-
constructions optimized using maximum parsimony were
calculated using MacClade version 3.0 (Maddison and Mad-
dison 1992). Ancestral pollination syndrome reconstruc-
tions, optimized using maximum likelihood, were per-
formed using Discrete version 4.0 (Pagel 1994; http://
www.ams.reading.ac.uk/zoology/pagel/), which employs a
Markov transition model of evolutionary change. Branch
lengths for the input tree were computed using parsimony
branch lengths from the ITS/ETS study and also from the

1
402
PAUL M. BEARDSLEY ET AL.
T
ABLE 2. Character data used to reconstruct ancestral states. Each character was converted to a binary character using the following
bins: Nectar production 0–1.1
␮
l, low (0); 7–17 ␮l, high (1). Stamen length 0–21 mm, short (0); 32–50 mm, long (1). Style length 0–
2
4 mm, short (0); 32–47 mm, long (1). Petal reflection: 1, reflexed forward; 2, no reflex; 3, reflexed backwards. Carotenoids absent, 0;
present, 1.
Nectar
production
in nature
Upper
petal
reflex
(1–3)
Lower
petal
reflex
(1–3)
Stamen
length
(mm)
Style
length
(mm)
Carotenoids
in upper
petal
Carotenoids
in lower
petal
Pollination
syndrome
Species
(␮l)
M. lewisii (Sierras)
M. lewisii (Rockies)
M. cardinalis
M. parishii
bee (0)
0.6 (0)
0.97 (0)
18.9 (0)
20.2 (0)
34 (1)
16.0 (0)
23.3 (0)
2
2
1.5
2
0
0
0
0
bee (0)
hummingbird (1)
insect/self (0)
hummingbird (1)
hummingbird (1)
hummingbird (1)
hummingbird (1)
insect (0)
insect/self (0)
insect (0)
insect (0)
12.08 (1)
32.7 (1)
3
3
1
1
0
(0)
9
(0)
9.2 (0)
2
2
0
0
M. eastwoodiae
M. verbenaceus
M. nelsonii
1.1 (0)
7.27 (1)
32.8 (1)
37.4 (1)
49.5 (1)
38.4 (1)
inferred (0)
inferred (0)
inferred (0)
inferred (0)
33.6 (1)
2.8
3
2
1
0
38.3 (1)
2
1
1
16.1 (1)
0.99 (0)
inferred (0)
inferred (0)
inferred (0)
inferred (0)
46.8 (1)
3
2
1
1
M. rupestris
36.6 (1)
1
1
1
0
M. filicaulis
inferred (0)
inferred (0)
inferred (0)
inferred (0)
2
2
1
1
M. rubellus
2
2
0/1
0/1
1
0/1
0/1
1
M. bicolor
2
2
M. primuloides
2
2
Nei-Li distances on the neighbor-joining tree constructed the eight taxa within Erythranthe, only 15 parsimony-infor-
using the bulked AFLP data (many outgroups were excluded mative sites exist. Parsimony analyses for all taxa resulted
on this tree because they were not sampled for AFLP data). in 85 most-parsimonious trees of length 432 (consistency
Branches of zero length on the ITS/ETS tree were assigned index [CI]
the arbitrary length of 0.1 because Discrete will not accept consistency index [RC]
zero branch lengths. Because more species were sampled,
Results from the analyses of the ITS and ETS results in-
ϭ
0.792, retention index [RI]
0.691; Fig. 2).
ϭ 0.873, rescaled
ϭ
the ITS/ETS tree contains more outgroups than the AFLP dicate the following: (1) If the desert species M. parishii is
tree. Outgroups included in the ITS/ETS tree were M. pri- included, section Erythranthe forms a monophyletic group
muloides, M. rubellus, and M. filicaulis, whereas the AFLP with a relatively high bootstrap (bs) value (87%). (2) A well-
tree included only M. bicolor. The effect of adding or sub- supported clade (100%) containing M. bicolor, M. filicaulis,
tracting outgroup species on ancestral state reconstruction and M. rubellus is sister to Erythranthe. The monophyly of
was investigated.
the Erythranthe clade and the M. bicolor, M. filicaulis, and
To test hypotheses of whether switches in pollinators are M. rubellus clade is also highly supported (99%). (3) Mimulus
consistent with a gradual or punctuational mode of change, primuloides is strongly supported (99%) as sister to the clade
we explored using different values for a branch-length scaling containing Erythranthe and M. bicolor, M. filicaulis, and M.
parameter (␬; Pagel 1994). A likelihood-ratio test was per-
rubellus. (4) The number of inferred substitutions in ITS and
ETS among species within Erythranthe is small, and branch
lengths on the most-parsimonious trees are very short (0–13
steps). (5) Relationships within Erythranthe are generally un-
resolved. A poorly supported clade (54%) contains M. ru-
pestris, M. nelsonii, M. verbenaceus, and M. eastwoodiae.
Weak support exists (64%) for the sister relationship between
M. verbenaceus and M. eastwoodiae. The relationships among
M. lewisii, M. cardinalis, and M. parishii are unresolved. (6)
At the population level, M. lewisii populations in the Rocky
Mountains seem to be differentiated from those in the Sierra
Nevada. The sequences for the M. cardinalis samples from
Arizona are identical to the M. cardinalis from the Sierra
Nevada and do not group within the M. verbenaceus samples
from Arizona.
formed with ␬ ϭ 0 (all the branch lengths are assumed to be
unit length) and ␬ ϭ the maximum-likelihood (ML) value.
Trait evolution is consistent with a punctuational mode of
change if
␬ (ML) is not significantly different from zero
(
Pagel 1994).
R
ESULTS
ITS/ETS Analysis
The total aligned length of the ITS region was 624 base
pairs (bp) and the aligned ETS sequences were 380 bp in
length. A limited number of short gaps (1–5 bp) were intro-
duced to properly align the sequences. The ETS and ITS
region are closely linked in the rDNA and results of the
partition homogeneity test for ITS versus ETS showed that
the datasets were not significantly different from random pair-
AFLP Analysis
wise partitions of the data (P
ϭ 0.62). Trees were constructed
for each DNA region separately and examined for hard to-
A total of 474 AFLP fragments was scored for 53 acces-
pological incongruencies, and no well-supported incongruen- sions in the portion of the study in which data were gathered
cies were found. Therefore, we combined the datasets in all independently for each individual. The range of fragments
analyses. The combined nrDNA dataset was 1004 bases in collected per accession was 108–194 with a mean of 151.8
aligned length. Including all of the outgroups, 279 sites were fragments. Neighbor-joining analyses of these data are pre-
variable and 180 were parsimony-informative. However, for sented in Figure 3 and led to the following results: (1) All

ERYTHRANTHE PHYLOGENY
1403
F
IG. 2. One of the 85 most-parsimonious trees inferred using combined ITS and ETS data. Numbers subtending particular nodes indicate
bootstrap support. Stars indicate nodes that collapse in the strict consensus tree.
of the accessions of each species, for which more than one tains and M. lewisii Northern race WA 02 from the Cascades).
individual was sampled, clustered together. Three have high The molecular data suggest that these individuals are inter-
levels of support (M. cardinalis 100%, M. parishii 100%, and mediates between the two races, containing different com-
M. eastwoodiae 96%). Mimulus verbenaceus was moderately binations of the race-specific fragments. (3) Erythranthe (plus
supported as monophyletic (71%). (2) Mimulus lewisii was M. parishii) is strongly supported as monophyletic (98%).
moderately supported as monophyletic (70%), but within M. (4) Relationships among species in Erythranthe are moder-
lewisii are two well-supported clades (100% for both) cor- ately to well-supported, with bootstrap values ranging from
responding to the two races recognized by Hiesey et al. 47% (M. parishii, M. lewisii, and M. cardinalis) to 98% (M.
(
1971), which differ geographically and morphologically. eastwoodiae and M. verbenaceus). A clade containing M. ru-
Both of these clades are supported by several diagnostic pestris, M. nelsonii, M. verbenaceus, and M. eastwoodiae was
AFLPs that are unique to each group. However, two M. lewisii recovered, with moderate support (65%). Mimulus rupestris
individuals were not resolved to either of these two clusters and M. nelsonii are resolved as sister species (78%), and M.
(
M. lewisii Northern race N CA 08 from the Siskiyou Moun- cardinalis and M. lewisii are together monophyletic (69%).

1
404
PAUL M. BEARDSLEY ET AL.
F
IG. 3. Inferred neighbor-joining (NJ) tree for all 53 accessions using 474 AFLP fragments scored as present or absent. Numbers above
the branches indicate NJ bootstrap values greater than 50.
The position of M. parishii, M. rupestris, and M. verbenaceus The two races of M. lewisii were each represented by three
are not clearly resolved. Parsimony analyses of these same individuals. In this study, 452 fragments were scored for eight
data recovered the same well-supported clades as the NJ anal- Erythranthe taxa and one outgroup. Neighbor-joining and
yses, but placed M. parishii as sister to the rest of the species parsimony analyses both recovered the same topology, with
in Erythranthe. This relationship, however, was poorly sup- high levels of support for each node (Fig. 4). These analyses
ported (bs
Ͻ
20%).
suggest the following relationships: (1) Species in Erythran-
Because we had evidence for the monophyly of the species the are divided into two major clades, one containing species
in Erythranthe and wished to obtain more data specifically in the Sierras, the Pacific Northwest, and the Rocky Moun-
on relationships among species, additional data were col- tains (M. parishii, both subspecies of M. lewisii, and M. car-
lected in which DNA from three individuals of each species dinalis), and one containing species in the arid Southwest
(
except for the uncommon M. nelsonii and M. rupestris, for and Mexico (M. verbenaceus, M. eastwoodiae, M. nelsonii,
which only one individual was available) was combined and and M. rupestris). (2) The two races of M. lewisii are sister
patterns of AFLP fragments were compared between species. to each other, forming a monophyletic M. lewisii, which is

ERYTHRANTHE PHYLOGENY
1405
(
␣ ϭ ␤, see below). The maximum likelihood occurred with
the branch-length scaling parameter ␬ ϭ 0.001.
To test the hypothesis of whether switches in pollinators
are consistent with a gradual or punctuational mode of
change, a likelihood-ratio test was performed of ␬ ϭ 0 and
␬
ϭ the maximum-likelihood (ML) value of 0.001 (likelihood
values calculated assuming ␣ ϭ ␤). The two likelihoods (L
ϭ Ϫ12.32259 and L ϭ Ϫ12.32261 were not significantly
different. For comparison, when ␬ ϭ 1, L ϭϪ15.082.
The influence of outgroups was assessed by removing all
but one of them, and recalculating ML values. The most likely
ancestral states were identical to those calculated using four
outgroups, although the probability of each reconstruction
differed. The probability at six of the eight nodes decreased
(
a maximum of 14.6% decrease), whereas two nodes in-
creased. The ML assignments of rates was ␣ ϭ 0.23, ␤ ϭ
.00052, similar to the results using four outgroups. Assum-
ing that ␣ ϭ ␤, the parameter was again estimated to be
very low and not significantly different from zero. Ancestral
pollination states, transformation rates, and the parameter
0
␬
␬
were also estimated using the branch lengths derived from
the NJ analysis of the bulked AFLP data but were nearly
identical to the results using ITS/ETS branch lengths.
F
IG. 4. Neighbor-joining tree estimated using 452 AFLP fragments
obtained using bulked DNA from each taxon. The numbers sub-
tending the nodes indicate bootstrap support estimates. The first
number used neighbor-joining and the second number used parsi-
mony.
D
ISCUSSION
Usefulness of AFLP Data
Plant systematists possess few tools for effectively re-
solving species-level phylogenies in clades that have radiated
relatively recently. Results from DNA sequence and AFLP
sister to M. cardinalis. The M. lewisii/M. cardinalis clade is analyses in Mimulus sect. Erythranthe indicate that AFLPs
sister to M. parishii. (3) Mimulus eastwoodiae is sister to M. are useful for resolving phylogenies that are not well-resolved
verbenaceus, and M. rupestris is sister to M. nelsonii. These using DNA sequencing when sequence variation is limiting.
two clades are sister to each other.
Vickery and Wullstein (1987) used six different approaches
to assess the relationships among the six traditional species
in Erythranthe, including alpha taxonomy, numerical tax-
onomy, experimental hybridization, chemotaxonomy, allo-
zyme comparisons, and DNA/DNA hybridizations. None of
these approaches yielded a well-resolved estimate of rela-
tionships, and most emphasized the many similarities within
Erythranthe. We analyzed relationships within Erythranthe
by collecting data on two DNA regions commonly used in
species-level systematics, the nrDNA ITS and ETS. Analyses
of these data provided some insight into relationships and
confirmed that this group of species has radiated relatively
recently, but overall demonstrated little resolution. The rig-
Character Analyses
Character states obtained from the literature and from our
own measurements are presented in Table 2. A summary of
ancestral state reconstructions optimized using maximum
parsimony is shown in Figure 5. For pollination syndrome,
the ancestor of M. lewisii and M. cardinalis is unambiguously
reconstructed as being insect-pollinated, as is the ancestor of
the Erythranthe clade (results not shown).
The following characters were found to be invariable
among the hummingbird-pollinated taxa: stamen and pistil
length (long), flower color (red), carotenoids in the upper orous application of AFLP data, however, provided a strong-
petal (petal), and orientation of the upper petals (reflexed). ly-supported and fully-resolved phylogenetic hypothesis at
Characters that varied among the hummingbird-pollinated the species level and added additional insights at the popu-
taxa were nectar production, carotenoids in the lower petal, lation level. The pattern of relationships suggested by se-
and orientation of lower petals.
Ancestral state reconstructions for pollination syndrome
quence data is consistent with those inferred from AFLP data.
The use of AFLPs holds many advantages over other mark-
using maximum likelihood and ITS/ETS branch lengths are ers (e.g., RAPDs, microsatellites). They require no previous
shown in Figure 6. Differences in likelihood of the two pos- sequence knowledge, show very high levels of repeatability
sible ancestral states were considered significant when they (Vos et al. 1995), and many loci can be assayed in a relatively
exceeded the conventional cutoff point of two log units (Ed- short time due to the large number of fragments scored per
wards 1972; Pagel 1999). For these analyses, we assumed gel (Hill et al. 1996; Mueller and Wolfenbarger 1999). AFLPs
that the rate of transformation from hummingbird pollination were also shown to generate tenfold more informative bands
to insect pollination was the same as the rate of transfor- per primer in comparison to RAPDs (Sharma et al. 1996).
mation from insect pollination to hummingbird pollination The method of data collection we used includes an internal

1
406
PAUL M. BEARDSLEY ET AL.
F
IG. 5. Ancestral state of hummingbird pollination and individual characters associated with hummingbird pollination estimated using
maximum parsimony. Tree used for ancestral state estimation is that inferred from AFLP data for the bulked DNA.
size standard in each lane, making band size calling much
Analysis of AFLP data for phylogenetic reconstruction de-
easier. Each lane is stored as an individual data file, which serves more attention. Although further model development
facilitates comparisons across different gels. The method is is encouraged, our own analyses suggest that with increasing
highly repeatable, relatively inexpensive, and three different amounts of AFLP data, all commonly used methods of re-
primer combinations can be scored using one gel.
constructing relationships with fragment data converge on

ERYTHRANTHE PHYLOGENY
1407
nelsonii group is supported by the presence of anther lobes
that are reflexed and horseshoe-shaped at the tip of the fil-
ament. This character is particularly useful to distinguish M.
cardinalis from M. verbenaceus in Arizona, where there has
been confusion in identifying these two taxa (Kearney and
Peebles 1960). Support for this group can also be found in
analyses of pollen morphology (Argue 1980).
The results of this study also shed light on the two races
of M. lewisii. Mimulus lewisii is found from the Kings Canyon
region in the southern Sierra Nevada mountains of California,
north through the Cascade Ranges, to southern coastal Alas-
ka, and east to Alberta, Idaho, Montana, Wyoming, Utah,
and Colorado (Hiesey et al. 1971). Differences exist in corolla
color, size, shape of leaves, and in stem and branching char-
acteristics between M. lewisii populations in the Sierra Ne-
vada (Sierra Nevada race) and those in the Cascade and
Rocky Mountains (Northern race). Two studies have indi-
cated that partial postzygotic barriers exist between these two
systems of populations, putatively due to two pairs of recip-
rocal translocations. (Hiesey et al. 1971; Vickery and Wull-
stein 1987). The results of our analysis of 474 AFLP frag-
ments indicate two very well-supported clusters (100%) that
correspond to the two races. All of the M. lewisii from the
Sierra Nevada fall into one cluster and all the M. lewisii from
F
IG. 6. Estimates of ancestral pollination states using maximum
likelihood. Dark shading indicates hummingbird pollination and
white indicates insect pollination. The shaded portion in each pie the Cascades and the Rockies fall into another, with two
diagram corresponds to the calculated probability of that recon-
exceptions. Mimulus lewisii Northern race NCA 08 from the
Siskiyou Mountains in northern California and M. lewisii
Northern race WA 02 from Poe Mountain in the Cascades
in Washington appear to be intermediate between the two
races. The presence of intermediates between the races and
the lack of substantive evidence for reproductive isolation in
nature lead us to retain M. lewisii as one species at the present
time.
struction. Stars indicate nodes in which one mode of pollination
was significantly preferred. Branch lengths are not scaled in this
diagram, because outgroups were not all sampled as part of the
AFLP analysis.
the same tree and these estimates are robust. For example,
estimates of relationships using maximum parsimony in
which fragment gain and loss are weighted equally, and using
Dollo parsimony, which makes the assumption that a char-
acter can be lost many times but gained only once, differ in
the resolution of only one node (results not shown).
Biogeography
Major biogeographic patterns within Erythranthe can be
interpreted in light of the inferred phylogeny. All of the most
closely related outgroups to Erythranthe are found in the
Sierra Nevada of California. Assuming that present distri-
butions reflect ancestral distributions, the most parsimonious
geographic location for the common ancestor of the Ery-
thranthe clade is in the Sierra Nevada. Early in the diversi-
fication of Erythranthe, a migration of the common ancestor
of M. verbenaceus, M. eastwoodiae, M. rupestris, and M. nel-
Relationships within Erythranthe
Species in Erythranthe have been a part of landmark studies
on plant adaptation (Hiesey et al. 1971; Vickery 1978) and
the subject of many biosystematic studies (Grant 1924; Hie-
sey et al. 1971; Vickery and Wullstein 1987). However, con-
siderable confusion exists as to which species belong within
Erythranthe. Both Grant (1924) and Greene (1885) suggested
that M. parishii was closely related to M. lewisii, yet the sonii to deserts of the American Southwest and Mexico is
absence of rigorous phylogenetic inference methods led sub- inferred to have taken place. The distinct geographic ranges
sequent workers to disregard their insight. Our results indi- of these four species imply that allopatric speciation was the
cate that Erythranthe includes the following taxa: M. cardi- mechanism for further diversification in this region. The com-
nalis, M. lewisii, M. parishii, M. verbenaceus, M. eastwoodiae, mon ancestor of M. lewisii, M. cardinalis, and M. parishii is
M. nelsonii, and M. rupestris.
inferred to have existed in the Sierra Nevada where it sub-
The AFLP data resolve two major groups of species in sequently diversified both morphologically and ecologically.
Erythranthe; one contains M. cardinalis, M. lewisii, and M. The three extant species differ in habitat preferences, ele-
parishii, and the other contains M. verbenaceus, M. eastwood- vation, and pollinators (Hiesey et al. 1971; Thompson 1993;
iae, M. rupestris, and M. nelsonii. The fact that M. cardinalis Bradshaw et al. 1995). Mimulus lewisii is inferred to have
is more closely related to M. lewisii than it is to the other expanded its range from the Sierra Nevada to the north and
red-flowered, hummingbird-pollinated taxa has support from east into the Cascade and Rocky Mountains. Mimulus car-
studies of crossing relationships (Vickery and Anderson dinalis expanded its range east and south of the Sierra Nevada
1
967; Vickery 1969; Hiesey et al. 1971). The monophyly of to include the coastal ranges, the transverse ranges, and the
the M. verbenaceus, M. eastwoodiae, M. rupestris, and M. Baja peninsula. Populations of M. cardinalis have also es-

1
408
PAUL M. BEARDSLEY ET AL.
tablished in high elevations in southeastern Arizona, marking species, yet red corollas always exist in the hummingbird-
a second establishment of Erythranthe taxa in the south- pollinated taxa, consistent with the results of Schemske and
western deserts. The close relationship of the Arizona pop- Bradshaw (1999), who found that flower color contributes to
ulations of M. cardinalis to those in the Sierra Nevada is pollinator-mediated reproductive isolation in M. lewisii and
confirmed in the ITS/ETS and AFLP studies. Mimulus par- M. cardinalis, and the hypotheses for the evolution of hum-
ishii is distributed along streams in the desert mountain re- mingbird pollination through Batesian mimicry (Bleiweiss
gions of southern California.
2001).
Schemske and Bradshaw (1999) found an allele that in-
Reconstructing Ancestral Pollination Syndromes
creases nectar production increases the rate of hummingbird
pollination twofold when homozygous. However, nectar pro-
duction, in both the greenhouse and in nature, varied in Er-
ythranthe more than tenfold (Vickery and Sutherland 1994).
Two of the lowest producers of nectar were hummingbird-
pollinated M. eastwoodiae and M. rupestris, both of which
had values similar to the bee-pollinated M. lewisii. The com-
mon ancestor of the M. verbenaceus, M. eastwoodiae, M.
rupestris, and M. nelsonii clade is estimated using parsimony
to have produced low amounts of nectar. One extra step is
required if the high nectar production evolved in the ancestor
of this clade and was subsequently lost twice.
Reconstructions of ancestral pollination syndromes using
both parsimony and maximum likelihood suggest that the
ancestor of the Erythranthe clade and the ancestor of M.
lewisii and M. cardinalis were insect-pollinated. Given our
estimates of ancestral states, hummingbird pollination
evolved independently twice from insect-pollinated ancestors
in Erythranthe. Statistical tests using maximum-likelihood
models also indicate that pollinator transformations in Ery-
thranthe more closely fit a model of punctuational evolution
than a model of gradual evolution. This result is consistent
with findings that QTLs of large effect are responsible for
floral differences between M. lewisii and M. cardinalis (Brad-
shaw et al. 1995, 1998; Schemske and Bradshaw 1999).
Studies by Bradshaw and Schemske (Bradshaw et al. 1995,
One interpretation of the low nectar values in M. eas-
twoodiae and M. rupestris is that they are functioning as
Batesian mimics; attracting hummingbirds with red, tubular
corollas but offering no nectar reward. Relative to other hum-
1
998; Schemske and Bradshaw 1999) have demonstrated that
changes in floral morphology that affect pollinator visitation mingbird-pollinated plants, Batesian mimics are expected to
have a dramatic effect on reproductive isolation. Assuming be rare. Consistent with the mimicry hypothesis, M. rupestris
that the ancestral state was insect pollination, both transfor- is exceedingly rare, known from only one population. Mi-
mations to hummingbird pollination in this clade could be mulus eastwoodiae is relatively infrequent in hanging gardens
interpreted as adaptations to increase reproductive isolation throughout the canyon country of southeastern Utah, north-
in the diverging ancestral populations. However, selection eastern Arizona, and southwestern Colorado. Found in sim-
for reproductive isolation can only occur when populations ilar ecological conditions are the putatively hummingbird-
are sympatric. The common ancestor of M. verbenaceus, M. pollinated Castilleja chromosa, Ipomopsis aggregata, Pen-
eastwoodiae, M. rupestris, and M. nelsonii is inferred to have stemon barbatus, and P. eatonii. The frequency of low nectar
existed in the deserts of the American Southwest and Mexico, populations of M. eastwoodiae should be further investigated.
whereas the ancestor of M. lewisii, M. parishii, and M. car- A second interpretation is that M. eastwoodiae and/or M.
dinalis clade is inferred to have existed in the Sierra Nevada. rupestris maintain the plesiomorphic condition for nectar pro-
Thus, the transformation to hummingbird pollination in the duction.
ancestor of the M. verbenaceus, M. eastwoodiae, M. rupestris,
The phylogenetic estimation of the transformation to re-
and M. nelsonii clade was more likely due to selection for flexed upper petals is correlated with the switch in pollination
the most effective pollinator (Stebbins 1970) than for repro-
ductive isolation per se.
syndromes, whereas reflexed lower petals are not so corre-
lated. This result was somewhat surprising given that the
lower petals are those that would be involved in forming a
landing platform for bees. Flowers with a landing platform
are reported to be preferred by bumblebees (Percival 1979).
Sutherland and Vickery (1993) report that hummingbirds
show a preference for reflexed petals.
In conclusion, the common ancestor of M. cardinalis and
M. lewisii is estimated to have been insect-pollinated, with
a relatively short style and stamen, no carotenoids in the
lower petal, erect upper and lower petals, and relatively low
nectar volumes. The reconstruction of flower color and ca-
Characters Associated with Hummingbird
Pollination in Erythranthe
In addition to estimating the ancestral pollination syn-
drome, we separately estimated the ancestral state of indi-
vidual characters putatively associated with hummingbird
pollination to gain insight into the coevolution of these traits.
Overall, our results show that the characters that are consis-
tently correlated with hummingbird pollination in Erythran-
the are exerted styles and stamens, red corollas, and reflexed
upper petals, whereas high nectar production and reflexed rotenoid accumulation is ambiguous. In many ways, then, the
lower petals were not consistent. The fact that long stamens common ancestor would be expected to look similar to extant
and pistils were consistently and exclusively found in all M. lewisii. The common ancestor of the clade comprising M.
hummingbird-pollinated species was expected. Both of these verbenaceus, M. eastwoodiae, M. rupestris, and M. nelsonii
morphological traits are important for the effective removal is estimated to have been hummingbird-pollinated, possess-
of pollen from the anther and the deposition of pollen on the ing a relatively long style and stamen, a red corolla due to
stigma, and thus directly influence fitness. Flower color in the presence of carotenoids in the upper petal (but lacking
Erythranthe and its sister clade varies considerably between carotenoids in the lower petal), upper petals reflexed back-

ERYTHRANTHE PHYLOGENY
1409
Doyle, J. J., and J. L. Doyle. 1987. A rapid isolation procedure for
small quantities of leaf tissue. Phytochem. Bull. 19:11–15.
Edwards, A. W. F. 1972. Likelihood. Cambridge Univ. Press, Cam-
bridge, U.K.
wards, lower petals erect, and producing low amounts of
nectar.
The origin of hummingbird pollination in the lineage lead-
ing to M. cardinalis is consistent with an origin via Mullerian Farris, J. S., M. Kallersj o¨ , A. G. Kluge, and C. Bult. 1994. Testing
mimicry. Hummingbird pollination in the M. verbenaceus,
M. eastwoodiae, M. rupestris, and M. nelsonii clade is con-
sistent with an origin via Batesian mimicry followed by the
evolution of increased nectar production and, as population
numbers grew, a switch to Mullerian mimicry. In addition to
experimental manipulations, results from other phylogenetic
studies with explicit character state reconstruction are nec-
essary before general conclusions about the relative impor-
significance of incongruence. Cladistics 10:315–319.
Felsenstein, J. 1985. Confidence limits on phylogenies: An approach
using the bootstrap. Evolution 39:783–791.
Ganders, F. R., M. Berbee, and M. Pirseyedi. 2000. ITS base se-
quence phylogeny in Bidens (Asteraceae): evidence for the con-
tinental relatives of Hawaiian and Marquesan Bidens. Syst. Bot.
2
5:122–133.
Ghebrehiwet, M., B. Bremer, and M. Thulin. 2000. Phylogeny of
tribe Antirrhineae (Scrophulariaceae) based on morphological
data and ndhF sequence data. Plant Syst. Evol. 220:223–239.
tance of individual characters within the suite of traits as- Givnish, T. J., T. M. Evans, M. L. Zjhra, T. B. Patterson, P. E.
sociated with hummingbird pollination can be made.
Berry, and K. J. Sytsma. 2000. Molecular evolution, adaptive
radiation, and geographic diversification in the amphiatlantic
family Rapateaceae: Evidence from ndhF sequences and mor-
phology. Evolution 54:1915–1937.
A
CKNOWLEDGMENTS
Grant, A. L. 1924. A monograph of the genus Mimulus. Ann. MO.
Bot. Gard. 11:99–389.
Grant, K. A. 1966. A hypothesis concerning the prevalence of red
coloration in California hummingbird flowers. Am. Nat. 100:
85–97.
Grant, K. A., and V. Grant. 1968. Hummingbirds and their flowers.
Columbia Univ. Press, New York.
Greene, E. L. 1885. Studies in the botany of California and parts
adjacent. I. Bull. Calif. Acad. Sci. 1:66–127.
We thank A. Angert, A. Colwell, A. Denton, D. Schemske,
S. Schoenig, M. Tulig, and R. Vickery, Jr. for plant material
and H. D. Bradshaw, Jr., M. Pagel, P. A. Reeves and D.
Schemske for helpful discussions. Funding was provided by
grants to PMB from the American Society of Plant Taxon-
omists and National Science Foundation grant DEB 9727025
to RGO.
Hiesey, W. M., M. A. Nobs, and O. Bjorkman. 1971. Carnegie Inst.
Washington Publ. 628:1–213.
Hill, M., H. Witsenboer, M. Zabeau, P. Vos, R. Kesseli, and R.
Michelmore. 1996. PCR-based fingerprinting using AFLPs as a
tool for studying genetic relationships in Lactuca spp. Theor.
Appl. Genet. 93:1202–1210.
L
ITERATURE CITED
Argue, C. L. 1980. Pollen morphology in the genus Mimulus (Scro-
phulariaceae) and its taxonomic significance. Am. J. Bot. 67:
6
8–87.
Avise, J. C. 1994. Molecular markers, natural history and evolution. Hodges, S. A. 1997. A rapid radiation via a key innovation in
Chapman and Hall, New York.
Aquilegia. Pp. 391–405 in T. Givnish and K. Sytsma, eds. Mo-
lecular evolution and adaptive radiations, Cambridge Press,
Cambridge, U.K.
Baldwin, B. G. 1992. Phylogenetic utility of the internal transcribed
spacers of nuclear ribosomal DNA in plants: an example from
the Compositae. Mol. Phylogenet. Evol. 1:3–16.
Kearney, T. H. and R. H. Peebles. 1960. Arizona flora. Univ. of
California Press, Berkeley, CA.
external transcribed spacer (ETS) of 18S–26S rDNA: Congru- Maddison, D. R. 1991. The discovery and importance of multiple
Baldwin, B. G., and S. Markos. 1998. Phylogenetic utility of the
ence of ETS and ITS trees of Calycadenia (Compositae). Mol.
Phylogenet. Evol. 10:449–463.
Baldwin, B. G., M. J. Sanderson, J. M. Porter, M. F. Wojciechowski,
C. S. Campbell, and M. J. Donoghue. 1995. The ITS region of
islands of most-parsimonious trees. Syst. Zool. 40:315–328.
Maddison, W. P., and D. R. Maddison. 1992. MacClade: analysis
of phylogeny and character evolution. Ver. 3.0. Sinauer, Sun-
derland, MA.
nuclear ribosomal DNA: a valuable source of evidence on an- McDade, L. A. 1992. Pollinator relationships, biogeography, and
giosperm phylogeny. Ann. MO. Bot. Gard. 82:247–277. phylogenetics. Bioscience 42:21–26.
Beardsley, P. M., and R. G. Olmstead. 2002. Redefining Phryma- Mueller, U. G., and L. L. Wolfenbarger. 1999. AFLP genotyping
ceae: the placement of Mimulus, tribe Mimuleae, and Phryma.
Am. J. Bot. 89:1093–1102.
Bleiweiss, R. 2001. Mimicry on the QT(L): genetics of speciation
in Mimulus. Evolution 55:1706–1709.
and fingerprinting. Trends Ecol. Evol. 14:389–394.
Nei, M., and W. H. Li. 1979. Mathematical model for studying
genetic variation in terms of restriction endonucleases. Proc.
Natl. Acad. Sci. USA. 76:5269–5273.
Bradshaw, H. D., Jr., S. M. Wilbert, K. G. Otto, and D. W. Schem- O’Kane, S. L., Jr.,and B. A. Schaal. 1998. Phylogenetics of Lopezia
ske. 1995. Genetic mapping of floral traits associated with re-
productive isolation in monkeyflowers (Mimulus). Nature 376:
(Onagraceae): Evidence from chloroplast DNA restriction sites.
Syst. Bot. 23:5–20.
7
62–765.
Olmstead, R. G., C. W. dePamphilis, A. D. Wolfe, N. D. Young,
W. J. Elisens, and P. A. Reeves. 2001. Disintegration of the
Scrophulariaceae. Am. J. Bot. 88:348–361.
Bradshaw, H. D., Jr., K. G. Otto, B. E. Frewen, J. K. McKay, and
D. W. Schemske. 1998. Quantitative trait loci affecting differ-
ences in floral morphology between two plant species of mon- Pagel, M. 1994. Detecting correlated evolution on phylogenies: a
keyflower (Mimulus). Genetics 149:367–382.
Brown, J. H., and A. Kodrick-Brown. 1979. Convergence, com-
general method for the comparative analysis of discrete char-
acters. Proc. R. Soc. Lond. B. Biol. Sci. 255:37–45.
petition, and mimicry in a temperate community of humming- ———. 1999. The maximum likelihood approach to reconstructing
bird-pollinated flowers. Ecology 60:1022–1035.
Bruneau, A. 1997. Evolution and homology of bird pollination syn-
dromes in Erythrina (Leguminosae). Am. J. Bot. 84:54–71.
ancestral character states of discrete characters on phylogenies.
Syst. Biol. 48:612–622.
Percival, M. S. 1979. Floral biology. Pergamon, Oxford, U.K.
Caicedo, A. L., O. Gaitan, M. C. Duque, O. Toro Chica, D. G. Schemske, D. W., and H. D. Bradshaw, Jr. 1999. Pollinator pref-
Debouck, and J. Thome. 1999. AFLP fingerprinting of Phaseolus
lunatus L. and related wild species from South America. Crop
Sci. 39:1497–1507.
erence and the evolution of floral traits in monkeyflowers (Mi-
mulus). Proc. Natl. Acad. Sci. USA. 96:11910–11915.
Sharma, S. K., M. R. Knox, T. H. N. Ellis. 1996. AFLP analysis
of the diversity and phylogeny of Lens and its comparison with
RAPD analysis. Theor. Appl. Genet. 93:751–758.
Cunningham, C. W., K. E. Omland, and T. H. Oakley. 1998. Re-
constructing ancestral character states: a critical reappraisal.
Trends Ecol. Evol. 13:361–366.
Stebbins, G. L. 1970. Adaptive radiation of reproductive charac-

1
410
PAUL M. BEARDSLEY ET AL.
teristics in angiosperms. I. Pollination mechanisms. Annu. Rev. Vos, P., R. Hogers, M. Bleeker, M. Reijans, T. van de Lee, M.
Ecol. Syst. 1:307–326.
Hornes, A. Frijters, J. Pot, J. Peleman, M. Kuiper, and M. Za-
beau. 1995. AFLP: A new technique for DNA fingerprinting.
Nucleic Acids Res. 23:4407–4414.
—
——. 1989. Adaptive shifts toward hummingbird pollination. Pp.
3
9–60 in J. H. Bock and Y. B. Linhart, eds. The evolutionary
ecology of plants. Westview Press, Boulder, CO.
Weller, S. G., and A. K. Sakai. 1999. Using phylogenetic ap-
proaches for the analysis of plant breeding system evolution.
Annu. Rev. Ecol. Syst. 30:167–199.
Wolfe, A. D., and A. Liston. 1998. Contributions of PCR-based
methods to plant systematics and evolutionary biology. Pp. 43–
86 in D. E. Soltis, P. S. Soltis, and J. J. Doyle, eds. Molecular
systematics of plants II DNA sequencing. Kluwer, Boston, MA.
Sutherland, S. D., and R. K. Vickery, Jr. 1993. On the relative
importance of floral color, shape, and nectar rewards in attracting
pollinators to Mimulus. Great Basin Nat. 53:107–117.
Swofford, D. L. 1998. PAUP*: phylogenetic analysis using parsi-
mony. * Sinauer, Sunderland, MA.
Thompson, D. M. 1993. Mimulus. Pp.1037–1051 in J. C. Hickman,
ed. The Jepson manual. Univ. of California Press, Berkeley, CA. Xu, F., and M. Sun. 2001. Comparative analysis of phylogenetic
Vickery, R. K. Jr. 1969. Crossing barriers in Mimulus. J. Jpn. Genet.
4(suppl 1):325–336.
relationships of grain amaranths and their wild relatives (Am-
aranthus; Amaranthaceae) using internal transcribed spacer, am-
plified fragment length polymorphism, and double-primer fluo-
rescent intersimple sequence repeat markers. Mol. Phylogenet.
Evol. 21:372–387.
4
—
——. 1978. Case studies in the evolution of species complexes
in Mimulus. Evol. Biol. 11:404–506.
—
——. 1995. Speciation in Mimulus, or, can a simple flower color
mutant lead to species divergence? Great Basin Nat. 55:177–180.
Xu, R. Q., N. Tomooka, and D. A. Vaughan. 2000. AFLP markers
for characterizing the Azuki bean complex. Crop. Sci. 40:
Vickery, R. K., Jr.,and D. G. Anderson. 1967. Experimental hy-
bridizations in the genus Mimulus (Scrophulariaceae). VI. Sec-
tion Erythranthe. Proc. Utah Acad. Sci. Arts Lett. 44:321–333.
Vickery, R. K., Jr.,and S. D. Sutherland. 1994. Variance and re-
plenishment of nectar in wild and greenhouse populations of
Mimulus. Great Basin Nat. 54:212–227.
8
08–815.
Zhang, L-B., H. P. Comes, and J. W. Kadereit. 2001. Phylogeny
and quaternary history of the European montane/alpine endemic
Soldanella (Primulaceae) based on ITS and AFLP variation. Am.
J. Bot. 88:2331–2345.
Vickery, R. K., Jr.,and B. M. Wullstein. 1987. Comparison of six
approaches to the classification of Mimulus sect. Erythranthe
(
Scropulariaceae). Syst. Bot. 12:339–364.
Corresponding Editor: J. Willis