obtain the scattered intensity I(q), the displacement
u(r) appears in the argument of an exponential and
thus enters I(q) in a nonlinear way. Thus, with sinusoidal
undulation, even though u(r) has only the
single (fundamental) harmonic, multiple diffraction
orders appear versus m for s 1, 2.
45. The dm values quoted here are for fresh samples in
capillaries. dm was found to increase in time with x-ray
exposure and aging at B7 temperatures (up to dm
600  in MHOBOW ). This effect was particularly
noticeable in freely suspended film and freeze fracture
experiments, where there is more exposure to air
during observation and preparation, respectively.
46. J. A. Zasadzinski, J. Phys. 51, 747 (1990).
47. T. Gulik-Krzywicki, M. J. Costello, J. Microsc. 112, 103
(1997).
48. Several recent papers use evidence for polarization
along the layer normal in B7 phases to argue that
they are simply lamellar with triclinic local layer
symmetry (9, 49). However, the phases in question
are shown here to be undulated and polarization
modulated so that the claim of triclinic behavior in a
lamellar smectic may not be justified. In fact, a PM
phase is triclinic essentially everywhere (wherever
there is nonzero layer curvature). It may be that local
triclinic symmetry actually drives the PM, but there is
no evidence for this at present.
49. A. Jákli, D. Kru¨erke, H. Sawade, G. Heppke, Phys. Rev.
Lett. 86, 5715 (2001).
50. It is not understod why dm is larger in films than in
bulk, but this may be due either to surface tension,
which provides an additional elastic resistance to
undulation, or due to the high surface-to-volume
ratio of the films, which enhances hydrolysis and
impurity buildup.
51. S. W. Choi et al., Mol. Cryst. Liq. Cryst. 328, 185
(1999).
52. M. Nakata, N. Chattham, N. A. Clark, H. Takezoe,
unpublished data.
53. C. Y. Young, R. Pindak, N. A. Clark, R. B. Meyer, Phys.
Rev. Lett. 40, 773 (1978).
54. The B7 materials strongly favor formation of freely
suspended filaments (12, 16) rather than films, a
consequence of their 2D lattice structure, but it was
possible to obtain 1- to 10-layer-thick MHOBOW
and 10OAM5AMO10 films at high T in the B7 phase
by either pulling the film very slowly ( 20 m/s) or
by applying an ac field ( 5 V/mm) parallel to the
film plane during the pulling. An initially thin spot will
expand in area over a large fraction of the film after
several hours.
55. R. Pindak, C. Y. Young, R. B. Meyer, N. A. Clark, Phys.
Rev. Lett. 45, 1193 (1980).
56. A. Hauser, A. Schamalfuss, Kresse, Liq. Cryst. 27, 629
(2000).
57. Several B7 phases have been previously identified as
SmCPA on the basis of antiferroelectric properties,
including absence of second-order nonlinear optical
susceptibility. We contend that the underlying structure
is actually SmCSPF and that the antiferroelectricity
is that of the PM stripe pattern.
58. R. Stannarius, C. Langer, W. Weissflog. Phys. Rev. E
66, 031709 (2002).
59. Because the field-induced PM/UL-to-SmCP transition
is an equilibrium energetic effect, the modulation
should in principle return once the field is removed.
However, it did not do so spontaneously in any of the
experiments reported here or in (6). This is likely due
to the fact that in order to reach the threshold field,
the LC was only 1 to 5 m thick, and because the
layer structure shrinkage upon PM expulsion was
locked in by the surfaces, an effect similar to the
irreversible elimination of the chevron structure in
SmCs by field application. The PM/UL-to-SmCP transition
in CITRO could be reversed by waveform selection
(square wave: PM/UL-to-SmCP; triangle wave:
SmCP -to- PM/UL) (10).
60. If p is the pitch of a helical winding of m(x) along a
filament, then at radius from the filament core we
have ( ) (2 /p). Thus, if the pitch is independent
of , then the orientation of m also winds
helically versus at a given x and the PM lattice must
have TGBs.
61. C. R. Safinya, K. S. Liang, W. A. Varady, N. A. Clark, G.
Andersson, Phys. Rev. Lett. 53, 1172 (1984).
62. C. R. Safinya, N. A. Clark, K. S. Liang, W. A. Varady,
L. Y. Chiang, Mol. Cryst. Liq. Cryst. 123, 205 (1985).
63. The elastic constant C of the PM lattice is comparable
to its elastic energy density C K/dm
2 104 J/m3
(see calculation of Eth), about four orders of magnitude
smaller than typical smectic layer compression
moduli. Thus, the typical fluid smectic focal conic
organization of layers is maintained.
64. The winding of the 2D PM lattice on a filament of
fixed cross-sectional layer structure is maintained
topologically by the number of undulation periods
around the filament. However, the PM lattice must
have dislocations because of the curvature of the
smectic layers (Fig. 7, B and C).
65. J. Szydlowska et al., Phys. Rev. E 67, 031702 (2003).
66. The B1 lattice structure shown in Fig. 1H is one of
many possibilities for alternation of polarization and
tilt orientation, which also include PM in the absence
of tilt. Thus, the homochiral, synpolar case (e.g., all
stripe polarizations toward the reader and in magenta)
would necessarity have an oblique 2D lattice,
which has been found in some B1s, e.g., W1044.
67. The polarization is sketched as uniform in the B1
phase in (53) and the B1rev phase in (69), but in fact
must be splayed, e.g., as in Fig. 1H.
68. W. L. McMillan, Phys. Rev. A 4, 1238 (1971).
69. W1044 has an oblique 2D reciprocal lattice characterized
by strong [s, m] [1, 1] and [1, 1] reflections,
with complete absence of the s 1, m even
reflections, indicative of an interdigitated real lattice.
The real lattice is as in Fig. 1H but oblique, possibly
due to a uniform rather than alternating molecular
tilt orientation (43).
70. M. Brunet, L. Navailles, N. A. Clark, Eur. Phys. J. E 7, 5
(2002).
71. A. Eremin, S. Diele, G. Pelzl, W. Weissflog, Phys. Rev.
E 67, 020702(R) (2003).
72. J. Ortega, C. L. Folcia, J. Etxebarria, N. Gimeno, M. B.
Ros, Phys. Rev. E 68, 011707 (2003).
73. B. N. Thomas, N. A. Clark, Phys. Rev. E 59, 3040 (1999).
74. S. Pakhomov et al., Proc. Natl. Acad. Sci. U.S.A. 100,
3040 (2003).
75. This work was supported by NSF grant DMR0072989,
NSF Materials Research Science and Engineering
Centers grants 0213918 (University of Colorado)
and 0080034 (University of California, Santa
Barbara), and NASA grant NAG3-2457. Research was
carried out in part at the National Synchrotron Light
Source, supported by U.S. Department of Energy,
Divisions of Materials and Chemical Sciences.
Supporting Online Material
www.sciencemag.org/cgi/content/full/301/5637/1204/
DC1
Figs. S1 to S3
26 March 2003; accepted 4 August 2003
Major Ecological Transitions in
Wild Sunflowers Facilitated
by Hybridization
Loren H. Rieseberg,1
* Olivier Raymond,2
David M. Rosenthal,3
Zhao Lai,1
Kevin Livingstone,1
Takuya Nakazato,1
Jennifer L. Durphy,1
Andrea E. Schwarzbach,4
Lisa A. Donovan,3
Christian Lexer1
Hybridization is frequent in many organismal groups, but its role in adaptation
is poorly understood. In sunflowers, species found in the most extreme habitats
are ancient hybrids, and new gene combinations generated by hybridization are
speculated to have contributed to ecological divergence. This possibility was
tested through phenotypic and genomic comparisons of ancient and synthetic
hybrids. Most trait differences in ancient hybrids could be recreated by complementary
gene action in synthetic hybrids and were favored by selection. The
same combinations of parental chromosomal segments required to generate
extreme phenotypes in synthetic hybrids also occurred in ancient hybrids. Thus,
hybridization facilitated ecological divergence in sunflowers.
The role of hybridization in evolution has
been debated for more than a century. Two
highly polarized viewpoints have emerged.
At one extreme, hybridization is considered
to be a potent evolutionary force that creates
opportunities for adaptive evolution
and speciation (1, 2). In this view, the
increased genetic variation and new gene
combinations resulting from hybridization
promote the development and acquisition
of novel adaptations. The contrasting position
accords little evolutionary importance
to hybridization (aside from allopolyploidy),
viewing it as a primarily
local phenomenon with only transient effects--a
kind of "evolutionary noise" (3­
5). Unfortunately, definitive support for either
viewpoint is lacking. Although footprints
of past hybridization are often detected
by molecular phylogenetic studies (6),
their presence does not indicate that the
hybridization was adaptive. Likewise, the
documentation of fit hybrids in contemporary
hybrid zones (2) is not proof of an
adaptive role for hybridization, because fit
1
Department of Biology, Indiana University, Bloomington,
IN 47405, USA. 2
Laboratoire de Biologie Moleculaire
et Phytochimie, Universite Claude Bernard
Lyon 1, F-69622 Villeurbanne, France. 3
Department of
Plant Biology, University of Georgia, Athens, GA
30602, USA. 4
Department of Biological Sciences, Kent
State University, Kent, OH 44242, USA.
*To whom correspondence should be addressed. Email:
lriesebe@indiana.edu
R E S E A R C H A R T I C L E S
www.sciencemag.org SCIENCE VOL 301 29 AUGUST 2003 1211
hybrid genotypes may be transitory as a
result of recombination (7).
One means by which fit hybrid genotypes
may become established is through
diploid hybrid speciation (8, 9). In this
process, a reproductive barrier arises in a
new hybrid lineage, which protects it from
gene flow with its parental species or other
hybrid genotypes. Models of diploid hybrid
speciation suggest that reproductive isolation
may arise through rapid chromosomal
repatterning, ecological divergence, and/or
spatial separation (8, 10). Ecological divergence
may be of particular importance to
the process, because computer simulations
indicate that hybrid speciation is unlikely in
the absence of niche separation (10, 11),
and all documented examples of diploid
hybrid species are divergent ecologically
from their parental species (12, 13). Indeed,
hybrid species often occur in habitats more
extreme than those of any congener (14).
Evidence that hybridization has played a
key role in adaptation to these novel or
extreme habitats would support the hypothesis
that hybridization can be an important
creative force in evolution.
Here, we ask whether hybridization has facilitated
major ecological transitions in annual
sunflowers of the genus Helianthus, a group
Table 1. Comparisons of morphological, life history, and physiological traits
among H. annuus and H. petiolaris and their three hybrid derivative species:
H. anomalus, H. deserticola, and H. paradoxus. Only traits that map to linkage
groups 1 and 10 are shown (see table S1 for comparisons of all 40 traits).
Traits that significantly differ between the two parental species, H. annuus
and H. petiolaris, are indicated with an asterisk. For the ancient hybrid species,
traits that significantly exceed both parental trait values are referred to as
"extreme." Those that are significantly different but intermediate between
the parental species are referred to as "intermediate." Traits that do not differ
significantly from one or both parental species are designated ANN-like,
PET-like, or ANN/PET-like. Trait abbreviations and units of measure are as
follows: DISKDIA, disk diameter (mm); LFAREA, leaf area (cm2
); LFSHAP, leaf
shape (length/width ratio); LFWDTH, leaf width (mm); LIGLGTH, ligule length
(mm); LIGNUM, number of ligules or ray flowers; PHYLGTH, phyllary length
(mm); PHYNUM, phyllary number; RTLG96, root length after 96 hours (mm);
SEEDW, initial seed weight (mg); STEMDIA, stem diameter at 2 cm above the
ground (mm); BUDDAY, days until budding; FLBIO, flower biomass (g);
FLODAY, days until first floret; RGR, relative growth rate (cm/day); B, leaf
boron concentration (ppm); Ca, leaf calcium concentration (ppm); K, leaf
potassium concentration (ppm); Mg, leaf magnesium concentration (ppm);
Mn, leaf manganese concentration (ppm); SLA, specific leaf area (cm2
/g).
Means SD are shown for the two parental species.
Trait
H. annuus
(ANN)
H. petiolaris
(PET)
H. anomalus
(ANO)
H. deserticola
(DES)
H. paradoxus
(PAR)
Morphological traits
DISKDIA* 27.48 5.14 16.24 2.40 13.19 12.04 15.24
Extreme Extreme PET-like
LFAREA* 63.30 36.79 24.97 9.48 19.25 20.21 92.18
Extreme Extreme Extreme
LFSHAP* 1.71 0.27 2.35 0.58 4.47 3.63 2.79
Extreme Extreme PET-like
LFWDTH* 75.9 2.4 42.1 10.9 27.4 28.4 70.8
Extreme Extreme ANN-like
LIGLGTH* 34.7 6.0 28.28 4.30 29.65 25.68 26.25
PET-like PET-like PET-like
LIGNUM* 18.32 3.06 11.59 2.15 10.63 10.25 14.70
Extreme Extreme Intermediate
PHYLGTH* 21.3 3.9 13.0 2.2 23.6 15.1 11.7
Extreme PET-like PET-like
PHYNUM* 27.00 4.61 18.82 2.40 17.19 16.88 18.60
PET-like Extreme PET-like
RTLG96 13.2 6.0 6.4 3.9 13.2 8.6 12.2
ANN-like PET-like ANN-like
SEEDW 5.64 0.37 1.71 0.32 7.39 3.11 4.24
Extreme Intermediate Intermediate
STEMDIA 14.54 1.63 7.55 1.16 6.02 7.74 11.15
Extreme PET-like Intermediate
Life history traits
BUDDAY* 37.85 11.11 34.35 6.91 45.31 29.50 100.06
Extreme Extreme Extreme
FLBIO* 55.7 10.4 10.0 4.8 6.0 4.8 11.8
PET-like PET-like PET-like
FLODAY* 59.35 13.94 48.71 7.24 60.19 42.50 106.00
ANN-like Extreme Extreme
RGR 2.56 0.40 2.37 0.50 2.36 2.34 1.19
PET-like PET-like Extreme
Physiological traits
B* 66.04 19.18 66.54 16.93 61.86 32.10 39.81
ANN/PET-like Extreme Extreme
Ca 16,927.34 7064.30 19,685.13 7358.02 13,436.70 16,843.16 22,338.20
ANN/PET-like ANN/PET-like ANN/PET-like
K* 49,226.90 6831.37 61,708.53 5885.27 44,451.29 49,298.71 38,310.09
Extreme ANN-like Extreme
Mg* 3283.18 582.83 4212.85 542.28 3455.07 4271.48 5698.12
Intermediate PET-like Extreme
Mn 196.69 32.79 216.42 61.76 215.35 189.62 354.12
ANN/PET-like ANN/PET-like Extreme
SLA* 207.88 50.55 320.17 72.87 240.53 268.67 241.85
Intermediate Intermediate Intermediate
R E S E A R C H A R T I C L E S
29 AUGUST 2003 VOL 301 SCIENCE www.sciencemag.org1212
that is ideal for addressing this question. Molecular
phylogenetic evidence indicates that
three of the 11 species in this group (H. anomalus,
H. deserticola, and H. paradoxus) are stabilized
diploid hybrid derivatives of two widespread
species, H. annuus and H. petiolaris
(14). The hybrid species appear to have been
derived independently from the parental species
and are found in the most extreme habitats of
any Helianthus species: H. anomalus occurs on
sand dunes in Utah and northern Arizona, USA;
H. deserticola inhabits dry, sandy soils on the
desert floor in Nevada, Utah, and Arizona; and
H. paradoxus is restricted to a handful of brackish
salt marshes in west Texas and New Mexico.
In contrast, the two parental species are
broadly distributed throughout the central and
western United States and occur in less extreme
environments. Helianthus annuus inhabits mesic,
clay-based soils that are wet in the spring
but may dry out later in the summer. Helianthus
petiolaris occurs in sandier soils with less vegetation
cover. Finally, microsatellite divergence
of the three hybrid species from their parents
suggests that they originated between 60,000
and 200,000 years before the present (15­17).
Thus, they arose recently enough to make partial
experimental replication feasible (18, 19),
yet are ancient enough to exclude anthropogenic
influences in their origin.
We have used several approaches to study
the role of hybridization in ecological adaptation
and speciation in this group, including
detailed phenotypic comparisons of synthetic
and natural sunflower hybrids, selection experiments
in natural sites, quantitative trait locus
(QTL) analyses of synthetic hybrids, and comparisons
of genomic composition between the
synthetic and ancient hybrids. We ask the following
questions: (i) Can the extreme traits
displayed by the ancient hybrid species be
experimentally generated by hybridizing
the parental species? (ii) Are these phenotypes
favored in the natural habitats of the
hybrid species? (iii) What is the genetic
basis of the extreme trait values? (iv) Do
genetic correlations facilitate or impede
ecological divergence? (v) Can the genomic
composition of the ancient hybrid species
be predicted from the QTL analysis of
synthetic hybrids? That is, do the ancient
hybrid species have the predicted combination
of parental chromosomal segments
for producing their extreme phenotypes?
Can trait differences in ancient hybrids
be recreated in synthetic hybrids?
We first asked whether traits that differentiate
the ancient hybrid species from their
parents (20) could be recreated by hybridizing
contemporary populations of the parental
species. This question was addressed
by propagating 20 individuals of each of
the five species plus 400 BC2 hybrids of H.
annuus H. petiolaris in the University of
Georgia greenhouses (21). All plants that
survived to maturity were measured for 22
morphological, 6 life history, and 12 physiological
traits (Table 1) (table S1).
When compared to the parental species, H.
anomalus differed significantly for 20 of the 40
traits ( Table 1) (table S1). Of these 20 traits, 13
significantly exceeded the parental trait values
and seven were intermediate. For H. deserticola,
11 traits were extreme relative to the
parental species and three were intermediate.
Helianthus paradoxus was the most divergent
of the three hybrid species, with 15 extreme and
eight intermediate traits.
Intermediate trait values are most easily
accounted for in hybrid species because hybrids
combine alleles from both parental spe-
ac/ctg292
ORS371
ORS716
ORS509
ORS728
ORS1128
1
MGSEEDW
DISKDIA
LGLGTH
PHYNUM
PHYLGTHBUDDAY
FLODAY
STEMDIA
FLBIO
B
LIGNUM
FLODAY
ANO DES PAR
B
< 10% PVE
> 10% PVE
10 cM
H. annuus QTL allele
toward hybrid species
H. petiolaris QTL allele
toward hybrid species
ORS557
ORS591
ORS256
ORS415
ORS380
ORS3
ORS691
ac/ctg326
10
DISKDIA
LFWDTH
LIGNUM
LFAREA
LFWDTH
LIGNUM
PHYNUM
RTLG96
SLA
STEMDIA
CA
K
MN
RGRSTEMDIA
RGR
MN
K
B
SEEDW
DISKDIA
LFSHAP
DES PARANO
Fig. 1. QTL positions
and correlations among
QTLs on selected linkage
groups for morphological,
life history, and
physiological traits (Table
1) differentiating H.
annuus and H. petiolaris.
Marker names are listed
to the left of each linkage
group, and QTL positions,
1-lod support
intervals, and magnitudes
are indicated by
vertical bars to the immediate
right of each
linkage group. The plus
or minus symbol above
each QTL indicates the
direction of effect of the
H. annuus allele. The
three groups of colored
bars farther to the right
of each linkage group
show the direction of
effect of each QTL with
respect to the three ancient
hybrid species
(ANO, H. anomalus; DES,
H. deserticola; PAR, H.
paradoxus). Blue bars
indicate that the H.
annuus QTL allele is in
the direction of the ancient
hybrid species;
yellow bars indicate
that the H. petiolaris
QTL allele is in the direction
of the ancient
hybrid. Gray bars indicate that the hybrid species is intermediate for
the trait or cannot be differentiated from either parental species.
Overlapping bars that are mostly the same color (excluding the
gray bars) indicate that genetic correlations are favorable for producing
the multitrait phenotype of the corresponding ancient hybrid
species.
R E S E A R C H A R T I C L E S
www.sciencemag.org SCIENCE VOL 301 29 AUGUST 2003 1213
cies. Indeed, the full range of intermediate
values were recovered for all traits in the
synthetic BC2 hybrids (table S1). However,
extreme phenotypic values are frequently reported
in segregating hybrids as well, a phenomenon
referred to as transgressive segregation.
A recent survey of 171 studies of
segregating hybrids revealed that 536
(43.6%) of 1229 traits examined were transgressive
(22). An even higher proportion of
extreme traits was observed in the present
study. Twenty-five (62.5%) or 27 (67.5%) of
the 40 traits assayed in the BC2 hybrids were
transgressive, depending on whether the observed
number of extreme phenotypes was
compared to expectations for two versus
three standard deviations, respectively (table
S1). These are very high percentages for
wild, outcrossing species (22).
There was not a perfect match, however,
between the extreme traits observed in the
three ancient hybrid species and the presence
or direction of transgression in the
BC2 population. In some instances, the lack
of significant transgression appears to be a
statistical artifact of high levels of phenotypic
variation in one of the parental species,
but for other traits the range of phenotypic
variation in the BC2 is truly narrow.
Thus, a more relevant measure of the
possible contribution of hybridization to
phenotypic divergence is the proportion of
extreme traits that are within the range of
variation found in the BC2 population. For
H. anomalus, 11 of 13 (84.6%) extreme
traits were within the range of the BC2
population and thus could be accounted for
by hybridization (table S1). All 11 extreme
traits (100%) in H. deserticola were within
the range of phenotypic variation observed
in the BC2, and 10 of 15 (66.7%) extreme
traits in H. paradoxus could be accounted
for by hybridization (table S1). Traits beyond
the range of the BC2 population might
be transgressive in a reciprocal backcross
population. Alternatively, they may have
arisen through mutational divergence rather
than hybridization. Either way, it is clear
that the majority of extreme traits observed
in the ancient hybrid species could have
been generated through hybridization.
Are the extreme phenotypes adaptive?
Although the three ancient hybrid species are
phenotypically divergent with respect to their
parental species, the trait differences do not
necessarily represent adaptations to the habitats
in which they occur. They could, for
example, have diverged through drift or as a
by-product of adaptive developmental changes
(23). There are two kinds of evidence,
however, that suggest that many of the trait
differences are adaptive.
First, the three hybrid species have
suites of traits that are commonly observed
in unrelated taxa found in the same habitats.
For example, H. anomalus, like other
sand dune endemics, has large seeds, rapid
root growth, and succulent leaves ( Table 1)
(table S1). Large seeds reduce the probability
of burial by moving sand, rapid root
growth helps tap into water reserves, and
succulent leaves reduce water loss in arid
environments (24). Helianthus deserticola
has many classic features of a desert annual,
including rapid flowering, small narrow
leaves, and reduced boron uptake ( Table 1)
(table S1). Rapid flowering ensures reproduction
after heavy seasonal rain, the small
narrow leaves decrease water loss and
avoid fatal overheating (25), and reduced
boron uptake may increase boron toxicity
tolerance in low-rainfall regions. Like
many other salt-loving species (halophytes),
H. paradoxus reduces the toxic
effects of sodium and other mineral ions
through active exclusion (26), internal
sequestration, and increased leaf succulence
(27).
Second, we have conducted a series of
experiments in parallel to those reported here,
in which BC2, parental, and hybrid species
individuals were transplanted into the habitat
of each of the ancient hybrid species. The
results from these experiments have been fully
analyzed only for the H. paradoxus habitat
(26), but they are illustrative of preliminary
findings from the H. anomalus and H. deserticola
sites. As predicted, leaf succulence and
mineral ion uptake are under strong directional
selection in the H. paradoxus habitat, and
the strength of selection for QTLs underlying
these traits (the selection coefficient, s, ranges
from 0.08 to 0.13) is strong enough to
account for the origin of H. paradoxus in
parapatry with its parental species (26). Also,
QTLs from both parental species were significantly
positively selected in the H. paradoxus
habitat, as required for models of diploid
hybrid speciation (10).
What is the mode of gene action responsible
for transgression? To determine
how extreme phenotypes are generated
through hybridization, we genotyped 384
BC2 plants of H. annuus H. petiolaris for
96 molecular markers known to cover most
of the sunflower genome (21). The genomic
location, percentage of variance explained
(PVE), and additive effects of QTLs underlying
the 40 traits scored in the previous
experiment were estimated with composite
interval mapping (21). In addition, all pairs of
marker loci were tested for interaction effects
or epistasis (21).
We detected 185 QTLs for the 40 traits
(Fig. 1) (fig. S1 and table S2). QTL effects
were modest in magnitude, ranging from 4
to 17% PVE. However, it is the directionality
of QTL effects that is most relevant to
the issue of transgressive segregation. More
than 39% of QTL effects were in the opposite
direction of species differences, and 34
of the 40 traits analyzed had at least one
opposing QTL (fig. S1 and table S2). That
is, for these QTLs, the H. annuus allele
produced a more H. petiolaris­like phenotype,
and vice versa.
The presence of QTLs with opposing
effects provides a simple explanation for
transgressive segregation: Segregating hybrids
can combine plus alleles or minus
alleles from both parents, thereby generating
extreme phenotypes ( Table 2). This
"complementary gene action" model accounts
for most of the transgressive phenotypes
observed in the H. annuus H.
petiolaris BC2 population (table S2). All
but five of the transgressive traits had complementary
QTLs. Of the five traits that
lacked opposing QTLs, three had a single
detected QTL, and thus the complementary
gene model could not be tested. For the two
remaining traits (LIGNUM and SLA), putative
antagonistic QTLs were detected that
barely fell beneath the significance threshold.
On the other hand, some traits had
complementary QTLs but did not display
significant phenotypic transgression. This
could be due to the limited sample size of
the BC2 population or to high phenotypic
variance in the parental species.
Although most transgressive phenotypes
can be accounted for by complementary
genes, epistatic interactions appear to contribute
as well. Significant epistasis was
observed for 18 traits, 14 of which showed
significant transgression (tables S1 and
S3). These observations are broadly consisTable
2. Hypothetical example of transgressive segregation due to the complementary action of genes
with additive effects.
QTLs
Phenotypic values
Species A
(AA genotype)
Species B
(BB genotype)
Transgressive
F2
Transgressive
F2
1 1 ­1 1 (AA) ­1 (BB)
2 1 ­1 1 (AA) ­1 (BB)
3 1 ­1 1 (AA) ­1 (BB)
4 ­1 1 1 (BB) ­1 (AA)
5 ­1 1 1 (BB) ­1 (AA)
Total 1 ­1 5 ­5
R E S E A R C H A R T I C L E S
29 AUGUST 2003 VOL 301 SCIENCE www.sciencemag.org1214
tent with evidence from studies of cultivated
plants, in which complementary gene
action is the primary cause of transgressive
segregation, with less frequent contributions
from epistasis and overdominance
(22, 28).
Do genetic correlations facilitate or
impede ecological divergence? The phenotypic
comparisons and QTL experiments
described above indicate that most trait
differences in the ancient hybrid species
can be recovered by hybridizing their parental
species. To generate the complex,
multitrait phenotype of a hybrid species,
however, it is necessary to combine all trait
differences into a single individual. This
will only be possible if closely linked or
pleiotropic QTLs have effects that are in
the same direction with respect to the hybrid
species phenotype. Thus, for each of
the hybrid species, we asked whether correlations
among closely linked or pleiotropic
QTLs were suitable for producing their
phenotype (21). This was accomplished by
testing whether the positions [i.e., 1-lod
(logarithm of the odds ratio for linkage
1) support intervals] of QTLs with effects
in the same direction with respect to a
given hybrid species phenotype overlapped
more frequently than would be expected by
chance (Fig. 1) (fig. S1). Congruence in the
direction of QTL effects was measured
with the coefficient of association, a
standard measure of association that can
vary from ­1 to 1 (29).
Overlapping QTLs most often had effects
in the same direction with respect to
the phenotypes of the ancient hybrid species
(Fig. 1) (fig. S1). The coefficient of
association was 0.32 (N 256, P 0.001)
for H. paradoxus, 0.47 (N 256, P
0.001) for H. anomalus, and 0.53 (N 256,
P 0.001) for H. deserticola. Note that
these are conservative estimates of congruence
levels, because the 1-lod support intervals
for some QTLs were large (Fig. 1)
(fig. S1 and table S2). Incongruence of
QTL directions for QTLs within 2 cM of
each other are infrequent; only 4, 5, and 10
such instances of incongruence were observed
for H. anomalus, H. deserticola, and
H. paradoxus, respectively.
These results indicate that genetic correlations
likely facilitated the origin of the three
ancient hybrid species and the acquisition of
their multitrait phenotypes, particularly for H.
anomalus and H. deserticola. It is noteworthy
that both species carry multiple chloroplast
DNA haplotypes from their parental species
(15, 16), which suggests that they may be
multiply derived in nature. In contrast, less
frequent positive genetic correlations make it
difficult to recreate the multitrait phenotype
of H. paradoxus, and this species appears to
have a single origin (17).
Can the genomic composition of the
ancient hybrid species be predicted from
QTL analyses of synthetic hybrids? The
experiments described above show that hybridization
could have facilitated ecological
divergence, but do not prove that it
actually did so. An alternative hypothesis,
for example, is that the ecological differences
that characterize the three ancient
hybrid species arose through mutational divergence
and were incidental to hybrid origin.
To determine whether hybridization
actually did facilitate ecological divergence,
we compared the genomic composition
of the three ancient hybrid species with
predictions from the QTL analyses of
the BC2 population of H. annuus H.
petiolaris.
To estimate genomic composition, we
generated high-resolution genetic linkage
maps, based primarily on microsatellite and
amplified fragment length polymorphism
(AFLP) markers (21), for each of the three
ancient hybrid species. Seventeen linkage
ORS1128
ORS509
ORS728
ORS716
ORS371
ac/ctg292
1 (B)
anomalus
ORS557
ORS691
ac/ctg326
ORS591
ORS3
ORS380
10 (C)
ORS256
ORS415
anomalus
ORS240
ORS840
ORS547
5 (T)
anomalus
ORS671
ORS702/2
ORS702/1
ORS814
ORS163
7 (E)
ORS782
anomalus
ORS887
ORS882
ta/gc192
ac/ctg214
ORS176
ORS617
9 (A)
anomalus
ac/ctg292
ORS509
ORS1128
ORS371
ORS716
ORS728
1
deserticola
ORS887
ORS882
ORS617
9
ORS176
ta/gc192
ac/ctg214
deserticola
ORS547
ORS840
ORS240
5
deserticola
ORS814
ORS163
ORS782
ORS671
ORS702/1
ORS702/2
7
deserticola
ORS591
ORS557
ORS691
ac/ctg326
ORS256
ORS380
ORS3
10
deserticola
ORS509
ac/ctg292
ORS371
ORS716
ORS1128
ORS728
1
paradoxus
ORS547
ORS840
ORS240
5
paradoxus
ac/ctg214
ORS887
ORS882
ta/gc192
ORS617
ORS176
9
paradoxus
ORS671
ORS782
ORS702/1
ORS814
ORS702/2
ORS163
7
paradoxus
ORS380
ORS591
ORS256
ORS415
ORS3
ORS691
ac/ctg326
10
paradoxus
H. petiolaris marker or predicted QTL
H. annuus marker or predicted QTL
25 cM
Fig. 2. Comparison of the genomic composition of the ancient hybrid species with predictions
from the QTL analyses (Fig. 1) (fig. S1) for five collinear chromosomes (see fig. S2 for the
remaining 12 chromosomes). Linkage groups are designated by number according to the
standard nomenclature for sunflower (21). Letter(s) in parentheses after linkage group
designations for H. anomalus enable comparisons with previous genetic mapping studies of this
species (18, 30). Positions of molecular markers used in the QTL analyses are shown to the left
of each linkage group, and the distribution of species-specific parental markers (table S4) is
shown in the center of each linkage group, with markers derived from H. annnus in blue and
those from H. petiolaris in yellow. Bars to the right of each linkage group indicate the
parentage of chromosomal segments, as predicted from the QTL analyses.
R E S E A R C H A R T I C L E S
www.sciencemag.org SCIENCE VOL 301 29 AUGUST 2003 1215
groups were recovered for each of the three
hybrid species, which corresponds to their
haploid chromosome number (Fig. 2) (fig.
S2). For H. anomalus, the present study
added 318 markers to the previously published
701-marker map (30), bringing the
total number of markers mapped to 1019.
The 1019 markers spanned 1908.3 cM, with
an average spacing between markers of
1.90 cM. The linkage maps for H. deserticola
and H. paradoxus reported here comprise
672 and 771 markers, respectively.
Map lengths and average marker spacing
are 1229 cM and 1.88 cM, respectively, for
H. deserticola, and 1420.5 cM and 1.88
cM, respectively, for H. paradoxus.
The most likely parental species origin
of each mapped marker was then determined
by surveying five natural populations
each of H. annuus and H. petiolaris
(21). In all, 427, 290, and 325 of the markers
mapped in H. anomalus, H. deserticola,
and H. paradoxus, respectively, could be
assigned to one or the other parental species
and thus were informative with respect
to genomic composition (table S4).
We then used the coefficient of association
(21) to compare the genomic distribution
of parental species markers in each of the
ancient hybrids with predictions from the
BC2 QTL data. Species-specific parental
markers mapped in each of the hybrid species
either matched or did not match the identity
of the predicted parental species donor for
that genomic region (Fig. 2) (fig. S2).
Congruence between predicted and actual
genomic composition was high (Fig. 2)
(fig. S2), with coefficient values of 0.56
(N 150, P 0.001) for H. paradoxus,
0.58 (N 193, P 0.001) for H. anomalus,
and 0.65 (N 94, P 0.001) for H.
deserticola. Because of limited map resolution
for species-specific markers, we cannot
determine the actual lengths of parental
chromosomal segments. Possibly they are
very large, as previously suggested for H.
anomalus (30). If this were the case, it
would suggest that hybrid speciation occurred
rapidly, resulting in the fixation of
large parental chromosomal segments that
have remained largely intact since their
initial establishment. Alternatively, the hybrid
species' genomes may be considerably
more fine-grained than shown here, and the
clustering of markers derived from the
same parental species may result from the
preponderance of genetic material from that
parent in a given genomic region (and not
from the presence of a fully intact chromosomal
segment). Either way, the ancient
hybrid species appear to have the predicted
combination of parental genetic material
for producing their extreme phenotypes.
Previously (18), it was shown that three
synthetic hybrid lineages of H. annuus
H. petiolaris converged onto a combination
of chromosomal blocks that was similar to
that found in H. anomalus, although some
differences remained. The present data set
accounts for aspects of the H. anomalus
genotype that could not be explained by
fertility selection alone. For many genomic
regions, however, there was concordance
between the predicted genomic composition
from fertility selection and the QTL
analysis of phenotypic differences, implying
that genes causing hybrid inviability or
sterility may overlap with those responsible
for phenotypic differences.
Conclusions. Our results corroborate the
view that hybridization can play an important
creative role in adaptive evolution, and suggest
a simple genetic mechanism (complementary
gene action) by which this may occur. Whether
these conclusions can be extrapolated to taxa
other than Helianthus is less clear. Hybridization
is frequent in many organismal groups (2),
particularly plants (31), fish (32), and birds
(33), and transgressive segregation appears to
be the rule rather than the exception in interspecific
crosses (22). Thus, an important and
widespread role for hybridization in adaptive
evolution is plausible. For hybridization to contribute
to adaptation, however, fit hybrid genotypes
must escape from "the mass of unfit
recombinants" in a hybrid population [(7), although
see (34)]. Mechanisms by which this
may occur include asexual reproduction, selfing,
diploid hybrid speciation (discussed herein),
and the introgression of advantageous alleles
(7). Unfortunately, little is known about
the frequency of the latter two mechanisms,
which are most applicable to sexual, outcrossing
species.
Our results also suggest a largely unexplored
mechanism for large and rapid adaptive
transitions, such as the colonization of
discrete and divergent ecological niches. Entry
into discrete niches is theoretically difficult,
because it may require simultaneous
changes at multiple traits and/or genes. Hybridization
offers a means by which this
difficulty may be overcome because, unlike
mutation, it provides genetic variation at hundreds
or thousands of genes in a single generation.
Moreover, new allelic variation introduced
by hybridization has already been tested
by selection in the genetic background of
one of the parental species and thus is less
likely to be deleterious.
Evolutionary biologists have struggled
to link results from experimental microevolutionary
studies of contemporary populations
directly to evolutionary changes occurring
in the distant past (34). Our results
bridge this gap by showing that large morphological
and ecological differences in
three ancient hybrid sunflower species can
be recreated through contemporary hybridization,
and that these ancient hybrid species
have the predicted combination of parental
chromosomal blocks for producing
their phenotypes. Hence, hybridization did
indeed facilitate major ecological transitions
in wild sunflowers.
References and Notes
1. E. Anderson, Introgressive Hybridization (Wiley, New
York, 1949).
2. M. L. Arnold, Natural Hybridization and Evolution
(Oxford Univ. Press, New York, 1997).
3. W. H. Wagner, Taxon 19, 146 (1970).
4. E. Mayr, Am. J. Bot. 79, 222 (1992).
5. D. W. Schemske, Evolution 54, 1069 (2000).
6. J. F. Wendel, J. J. Doyle, in Molecular Systematics of
Plants II: DNA Sequencing, D. E. Soltis, P. S. Soltis, J. J.
Doyle, Eds. (Kluwer, Boston, 1998), pp. 265­296.
7. N. H. Barton, Mol. Ecol. 10, 551 (2001).
8. V. Grant, Plant Speciation (Columbia Univ. Press, New
York, 1981).
9. D. Greig, E. J. Louis, R. H. Borts, M. Travisano, Science
298, 1773 (2002).
10. C. A. Buerkle, R. J. Morris, M. A. Asmussen, L. H.
Rieseberg, Heredity 84, 441 (2000).
11. E. M. McCarthy, M. A. Asmussen, W. W. Anderson,
Heredity 74, 502 (1995).
12. R. J. Abbott, Trends Ecol. Evol. 7, 401 (1992).
13. L. H. Rieseberg, Annu. Rev. Ecol. Syst. 28, 359 (1997).
14. L. H. Rieseberg, Am. J. Bot. 78, 1218 (1991).
15. B. L. Gross, A. E. Schwarzbach, L. H. Rieseberg, Am. J.
Bot., in press.
16. A. E. Schwarzbach, L. H. Rieseberg, Mol. Ecol. 11,
1703 (2002).
17. M. E. Welch, L. H. Rieseberg, Evolution 56, 2126
(2002).
18. L. H. Rieseberg, B. Sinervo, C. R. Linder, M. Ungerer,
D. M. Arias, Science 272, 741 (1996).
19. L. H. Rieseberg, Evolution 54, 859 (2000).
20. D. M. Rosenthal, A. E. Schwarzbach, L. A. Donovan, O.
Raymond, L. H. Rieseberg, Int. J. Plant Sci. 162, 387
(2002).
21. See supporting data on Science Online.
22. L. H. Rieseberg, M. A. Archer, R. K. Wayne, Heredity
83, 363 (1999).
23. S. J. Gould, R. C. Lewontin, Proc. R. Soc. London Ser.
B 205, 581 (1979).
24. A. Danin, J. Arid Environ. 21, 193 (1991).
25. A. C. Gibson, Bioscience 48, 911 (1998).
26. C. Lexer, M. Welch, J. L. Durphy, L. H. Rieseberg, Mol.
Ecol. 12, 1225 (2003).
27. M. E. Welch, L. H. Rieseberg, Am. J. Bot. 89, 472
(2002).
28. M. C. deVicente, S. D. Tanksley, Genetics 134, 585
(1993).
29. R. R. Sokal, F. J. Rohlf, Biometry (Freeman, New York,
ed. 3, 1995).
30. M. C. Ungerer, S. Baird, J. Pan, L. H. Rieseberg, Proc.
Natl. Acad. Sci. U.S.A. 95, 11757 (1998).
31. N. C. Ellstrand, R. Whitkus, L. H. Rieseberg, Proc. Natl.
Acad. Sci. U.S.A. 93, 5090 (1996).
32. C. L. Hubbs, Syst. Zool. 4, 1 (1995).
33. P. R. Grant, B. R. Grant, Science 256, 193 (1992).
34. P. R. Grant, B. R. Grant, Science 296, 707 (2002).
35. Supported by NIH grant R01 G59065 and NSF grant
DEB 9806290 (L.H.R.). We thank M.-J. Kim for assistance
with genotyping, J. Lance for assistance with
phenotyping, and S. Knapp for microsatellite primer
sequences and mapping information.
Supporting Online Material
www.sciencemag.org/cgi/content/full/1086949/DC1
Materials and Methods
Figs. S1 and S2
Tables S1 to S4
References
19 May 2003; accepted 10 July 2003
Published online 7 August 2003;
10.1126/science.1086949
Include this information when citing this paper.
R E S E A R C H A R T I C L E S
29 AUGUST 2003 VOL 301 SCIENCE www.sciencemag.org1216