The Phylogeny of Linum and Linaceae Subfamily Linoideae, with Implications for Their Systematics, Biogeography, and Evolution of Heterostyly

Abstract

The genus Linum consists of over 180 species, the most famous being L. usitatissimum, the source of linen and linseed oil. The eight genera of Linaceae subf. Linoideae, of which Linum is the largest, exhibit a complex biogeographic distribution, inhabiting all continents except Antarctica. Numerous species in Linoideae are heterostylous, but the ancestral breeding system of the group has not been determined. We present phylogenetic analyses of 44 species representing all eight genera of subf. Linoideae and 37 species of Linum, with data from the chloroplast (ndhF, trnL-F, trnK 3′ intron) and the nuclear ITS, with Hugonia (Linaceae subf. Hugonioideae) as outgroup. Sequences of rbcL from 48 species of Linaceae, including five species from Hugonioideae and seven species from other families of Malpighiales, were analyzed independently. Our results suggest that Linaceae and subf. Linoideae are monophyletic, but Linum is not. Anisadenia, Reinwardtia, and Tirpitzia are found to be the basal members of Linoideae. The rest of the subfamily forms two major lineages: a blue-flowered clade (Linum sections Linum and Dasylinum) and a yellow-flowered clade (Linum sects. Linopsis, Syllinum, and Cathartolinum, and the genera Cliococca, Hesperolinon, Radiola, and Sclerolinon). Diversification of Linoideae may have begun 46–51 mya, probably in Southeast Asia. Linum appears to have arisen in Eurasia, from which it spread to Africa, North America, South America, and Australasia. Our analyses indicate that neither heterostyly nor homostyly can yet be confirmed as the ancestral state in Linoideae or Linaceae, but provide strong evidence that breeding system is evolutionarily labile in this group.

Full text

Systematic Botany (2009), 34(2): pp. 386–405
© Copyright 2009 by the American Society of Plant Taxonomists
386
The genus Linum L. has played an important role in the eco-
nomic and social development of humanity over thousands
of years. Early Mesolithic civilizations realized the value of
a wild flax, perhaps Linum bienne Mill. (= L. angustifolium
Huds.), as a source of useful fibers ( Pengilly 2003 ), although
the oldest archaeological flax seeds, from excavation sites in
Syria, are dated to around 9,000 B.C.E. ( Hillman 1975 ). By
8,000 B.C.E., flax was in use throughout the Fertile Crescent.
Its cultivation for fibers and seeds is believed to have led to
the development of L. usitatissimum L., the modern cultivated
flax, one of the early domesticated plants ( Vaisey-Genser and
Morris 2003 ; Pengilly 2003 ). Flax was the first plant cultivated
for fibers in the Old World ( Zohary and Hopf 2000 ) and was
the basis of one of the earliest plant-based industries: weav-
ing ( Judd 1995 ). Flax seed has been used directly as a food
since its earliest cultivation ( Vaisey-Genser and Morris 2003 ).
The drying properties of the seed oils made them useful addi-
tives in protective coverings and preservative varnishes. The
preservative properties of linseed oil were recognized by the
ancient Egyptians, who used the oil to embalm the dead prior
to wrapping in many layers of fine linen strips ( Vaisey-Genser
and Morris 2003 ). In the last century, the cultivation of flax for
linen dramatically decreased due to the expanded use of cot-
ton and the development of ever-new synthetic alternatives
( Zohary and Hopf 2000 ). Today, the seeds of L. usitatissimum
are the more important product of the species, used nutrition-
ally, medicinally, and as the source of linseed oil, an important
component of many paints, inks, varnishes, and lubricants.
Although the genus owes much of its fame to its single
agricultural member, approximately 180 additional spe-
cies of Linum are distributed throughout the temperate and
subtropical regions of the world. Because of their brightly
colored and attractive flowers, some are garden ornamen-
tals, such as the red-flowered L. grandiflorum Desf., the blue
L. perenne L., and the yellow L. flavum L. In recent years, lig-
nans and α-linolenic acids of Linum have been explored for
their usefulness in treatments for cardiovascular diseases and
cancer, particularly breast cancer ( Rickard-Bon and Thompson
2003 ). These medicinal applications have stimulated renewed
interest in the systematic relationships among Linum species.
Linum has also historically been important for the study of
heterostyly, which was recognized by Darwin in a European
blue flax, Linum perenne ( Darwin 1864 ).
Linum is classified in Linaceae subfamily Linoideae, along
with seven other small genera found primarily in the tem-
perate and subtropical regions of the world, although a few
species occupy habitats within the tropics as well ( Fig. 1 ).
Anisadenia Wall., Linum , Radiola Hill., and Reinwardtia Dumort.
were recognized by Planchon (1847 , 1848 ) in the first detailed
treatments of the family. Tirpitzia Hallier f. was distinguished
from Reinwardtia by Hallier ( 1921 ). The remaining three
genera were segregated from Linum in recognition of their
morphological distinctiveness: Cliococca Bab. ( Rogers and
Mildner 1971 ), Hesperolinon (A. Gray) Small ( Small 1907 ),
and Sclerolinon C. M. Rogers ( Rogers 1966 ). Linum remains
the largest genus of the subfamily, followed by Hesperolinon
with 12 species ( Sharsmith 1961 ). The remaining genera of
Linoideae contain one to three species each.
Numerous infrageneric classifications of Linum have been
published but the genus has never been monographed. The
most recent worldwide treatment is Winkler’s (1931) , which is
largely based on Planchon’s work. Together, Planchon (1847 ,
1848 ) and Winkler ( 1931 ) provided the basis for the five “sec-
tions” of Linum commonly recognized in most regional floris-
tic and taxonomic works, such as revisions of American and
South African linums ( Rogers 1963 , 1981 ; Mildner and Rogers
1978 ), Flora Europaea ( Ockendon and Walters 1968 ), and Flora of
Turkey and the Aegean Islands ( Davis 1967 ). Flora of the U.S.S.R.
( Yuzepchuk 1974 ), which encompasses a large number of spe-
cies occurring from eastern Europe to eastern Asia, utilized a
much more detailed classification system also largely based
on Planchon (1847 , 1848 ). Table 1 provides a comparison of
three historical intrageneric treatments to the current gener-
ally accepted arrangement, along with the distributions and
The Phylogeny of Linum and Linaceae Subfamily Linoideae, with Implications
for Their Systematics, Biogeography, and Evolution of Heterostyly
Joshua McDill, 1, 3, 4 Miriam Repplinger, 2, 3, 4 Beryl B. Simpson, 1 and Joachim W. Kadereit 2
1 Section of Integrative Biology and the Plant Resources Center, The University of Texas at Austin, Texas 78712 U.S.A.
2 Institut für Spezielle Botanik und Botanischer Garten, Johannes Gutenberg-Universität Mainz, Bentzelweg 9a,
55099 Mainz, Germany
3 Authors for correspondence ( josh.m@mail.utexas.edu , reppling@uni-mainz.de )
4 The first two authors contributed equally to this work.
Communicating Editor: Andrea Schwarzbach
Abstract— The genus Linum consists of over 180 species, the most famous being L. usitatissimum , the source of linen and linseed oil. The eight
genera of Linaceae subf. Linoideae, of which Linum is the largest, exhibit a complex biogeographic distribution, inhabiting all continents except
Antarctica. Numerous species in Linoideae are heterostylous, but the ancestral breeding system of the group has not been determined. We pres-
ent phylogenetic analyses of 44 species representing all eight genera of subf. Linoideae and 37 species of Linum , with data from the chloroplast
( ndhF , trnL-F , trnK 3 ′ intron) and the nuclear ITS, with Hugonia (Linaceae subf. Hugonioideae) as outgroup. Sequences of rbcL from 48 species of
Linaceae, including five species from Hugonioideae and seven species from other families of Malpighiales, were analyzed independently. Our
results suggest that Linaceae and subf. Linoideae are monophyletic, but Linum is not. Anisadenia , Reinwardtia , and Tirpitzia are found to be the
basal members of Linoideae. The rest of the subfamily forms two major lineages: a blue-flowered clade ( Linum sections Linum and Dasylinum )
and a yellow-flowered clade ( Linum sects. Linopsis , Syllinum , and Cathartolinum , and the genera Cliococca , Hesperolinon , Radiola , and Sclerolinon ).
Diversification of Linoideae may have begun 46–51 mya, probably in Southeast Asia. Linum appears to have arisen in Eurasia, from which it
spread to Africa, North America, South America, and Australasia. Our analyses indicate that neither heterostyly nor homostyly can yet be con-
firmed as the ancestral state in Linoideae or Linaceae, but provide strong evidence that breeding system is evolutionarily labile in this group.
Keywords— Cliococca , floral polymorphisms , Hesperolinon , Sclerolinon , Radiola .

2009] MCDILL ET AL.: PHYLOGENY OF LINOIDEAE 387
numbers of species in each section. Section Linopsis is often
referred to as sect. Linastrum , but Linopsis is the prior name at
the rank of section ( Rogers 1982 ).
The other subfamily of Linaceae, Hugonioideae ( Dressler
et al. in press ), includes six genera: Durandea Planch. (3–4 spe-
cies, often considered as a section of Hugonia L. ( Van Hooren
and Nooteboom 1984 )), Hugonia (35 species), Hebepetalum
Benth. (three species), Indorouchera Hallier f. (two species),
Philbornea Hallier f. (one species), and Roucheria Planch.
(seven species). Sometimes recognized as a separate fam-
ily, Hugoniaceae ( Cronquist 1988 ), this group differs from
Linoideae in being composed entirely of tropical trees and
lianas with ten fertile stamens and drupaceous fruits ( Jardim
1999 ). Roucheria and Hebepetalum are strictly American in
Fig. 1. A. Geographical distribution of the sections of Linum . B. Geographical distribution of the other genera of Linoideae, and of subfamily
Hugonioideae.

388 SYSTEMATIC BOTANY [Volume 34
distribution, while the other genera are restricted to the Old
World. Altogether, Linaceae thus comprises about 250 spe-
cies in 14 genera. Once considered the core of the epony-
mous order Linales ( Cronquist 1988 ), the Linaceae has been
shown to be part of the large order Malpighiales ( APG II
2003 ). The close relationship of Linoideae to Hugonioideae
has been found consistently in large-scale phylogenetic
studies of Malpighiales and angiosperms in general that
included representatives of both subfamilies, but the clos-
est relative of Linaceae is still unknown. Linaceae have been
shown as potential sister to Violaceae + Passifloraceae +
Turneraceae ( Savolainen et al. 2000 ), Picrodendraceae ( Chase
et al. 2002 ), Balanopaceae + Trigoniaceae + Dichapetalaceae +
Euphroniaceae + Chrysobalanaceae ( Davis and Chase 2004 ),
or Irvingiaceae ( Davis et al. 2005 ). Sampling of Linaceae in
these studies has been limited, with one or two species of
Linum and Reinwardtia representing Linoideae, and one spe-
cies of Hugonia or Durandea from Hugonioideae.
Molecular studies within the Linoideae or Linum itself have
been few, and generally limited in scope. Several investiga-
tors have examined genetic diversity or have mapped genes
in the cultivated flax ( Spielmeyer et al. 1998 ; Ellis et al. 1999 ;
Oh et al. 2000 ; Fu et al. 2002b ; Allaby et al. 2005 ). An analy-
sis of RAPD variation among seven species of blue-flowered
Linum was undertaken to identify the putative wild progen-
itor of L. usitatissimum ( Fu et al. 2002a ). Coates and Cullis
(1987) used RFLPs to map the chloroplast genomes of six spe-
cies of Linum , revealing (among other things) a possible 13 kb
insertion in members of the Linum perenne group. Finally,
Armbruster et al. (2006) analyzed sequences of the nuclear
ribosomal ITS from 16 species of Linum representing all five
currently recognized sections (some represented by single
exemplars), with Radiola linoides Roth. as the outgroup.
We present here what we believe to be the first phyloge-
netic study with extensive sampling focused on Linaceae with
emphasis on subfamily Linoideae and Linum . These lineages
present several interesting questions that can potentially be
answered by molecular phylogenetic analyses.
First, a molecular phylogeny will allow us to address several
fundamental systematic issues. This study will provide a test of
the monophyly of the sections of Linum and improve our knowl-
edge of the relationships among sections. We will also evaluate
the phylogenetic validity of the morphology-based segrega-
tion of Hesperolinon , Sclerolinon , and Cliococca from Linum , and
explore the relationships of these taxa to the other genera of
Linoideae: Anisadenia , Radiola , Reinwardtia , and Tirpitzia .
Second, the almost cosmopolitan distribution of the sub-
family and genus pose interesting biogeographical questions,
particularly with respect to the amphiatlantic and amphitrop-
ical disjunctions found in Linum sections Linum and Linopsis .
The biogeographical origins of the segregate genera Cliococca ,
Hesperolinon, and Sclerolinon with respect to New World
linums are also of interest. Are they derived from geographi-
cally proximate lineages of Linum , as their earlier taxonomic
treatment (as species of Linum ) implies, or do they perhaps
have independent origins from temperate or tropical ances-
try? It is conceivable, for instance, that the distinctive South
American Cliococca originated either directly from a tropical
Table 1. Comparison of three historical intrageneric treatments of Linum and the current treatment (including distribution and number of species).
Our designation of “current treatment” is a combination of taxonomy used in the most recent floristic and taxonomic literature covering the geographic
range of Linum : Davis ( 1967 ), Mildner and Rogers ( 1978 ), Ockendon and Walters ( 1968 ), Rechinger ( 1974 ), Rogers ( 1963 , 1966 , 1969 , 1975 , 1981 , 1982 , 1984 ,
1985 ), Rogers and Mildner ( 1971 ), Yuzepchuk ( 1974 ), and Zohary ( 1972 ).
Current treatment
Planchon 1847 / 1848 Reiche 1897 Winkler 1931 Section or Genus Main Geographic
Distribution No. of Species
(no. sampled)
No.
Heterostylous
(no. sampled)
subg. Syllinum sect. Syllinum sect. Syllinum sect. Syllinum
Griseb.
Europe and
W Asia
ca. 37 (4) ca. 27 (3)
ser. Limoniopsis
ser. Dasylinum sect. Eulinum sect. Eulinum sect. Dasylinum
(Planchon) Juz.
E Europe to
C Asia
ca. 13 (4) ca. 9 (4)
subg. Eulinum sect. Eulinum sect. Eulinum sect. Linum N America,
N Africa,
Eurasia,
Australia, New
Zealand
ca. 50 (9) ca. 20 (3)
ser. Protolinum
ser. Adenolinum
subg. Linastrum sect. Linastrum sects. Cathartolinum &
Linastrum
sect. Linopsis
(Reichenb.)
Engelmann
Europe, SW
Asia, E & S
Africa, N & S
America
ca. 90 (19) ca. 6 (5)
ser. Linopsis
ser. Dichrolinum sect. Eulinum
ser. Halolinum sect. Linastrum
sect. Cathartolinum sect. Cathartolinum sect. Cathartolinum
(Reichenb.) Griseb.
Europe, W Asia 1(1) 0
sect. Cliococca sect. Cliococca sect. Cliococca Cliococca Bab. Chile, Uruguay,
S Brazil
1 (1) 0
— sect. Hesperolinon Hesperolinon Hesperolinon Small Western U.S.A.
(California)
13 (1) 0
— sect. Cathartolinum
Sclerolinon
C.M. Rogers Western U.S.A. 1 (1) 0

2009] MCDILL ET AL.: PHYLOGENY OF LINOIDEAE 389
ancestor, or arrived in the southern hemisphere from north
temperate regions. Austral occurrences of representatives of
Linoideae are particularly interesting considering the pan-
tropical distribution of subfamily Hugonioideae. With the
majority of Linum species concentrated in North America and
western Eurasia, the genus also presents another opportunity
to examine the origin of a western Eurasian/North American
disjunction, a distribution type investigated in comparatively
few taxa ( Hohmann et al. 2006 ).
Third, the evolution of heterostyly in the family, subfam-
ily, and genus is of considerable interest. Four types of stylar
polymorphisms are generally recognized in flowering plants:
distyly, tristyly, stigma height dimorphism, and enantiostyly
( Barrett et al. 2000 ), and the first three of these are found in the
Linaceae. Distyly is the most common, reported in approxi-
mately 60 species of Linum (cf., Davis 1967 ; Ockendon and
Walters 1968 ; Zohary 1972 ; Yuzepchuk 1974 ; Rogers 1981 ),
Reinwardtia indica Dumort. ( Lloyd et al. 1990 ; Bahadur et al.
1996 ), Tirpitzia bilocularis Suksathan & Larsen ( Suksathan
and Larsen 2006 ), Indorouchera griffithiana (Planch.) Hallier f.,
and several species of Hugonia ( Lloyd et al. 1990 ; Thompson
et al. 1996 ). Tristyly is reported from Hugonia serrata Lam.
( Thompson et al. 1996 ). Considering species descriptions
and illustrations in several floras, additional di- and tristy-
lous species might exist in Hugonia (c.f., Thompson et al.
1996 ). Stigma height dimorphism is known from L. grandi-
florum ( Darwin 1877 ). In addition to these three manifesta-
tions of heterostyly, a novel type of dimorphism, a reciprocal
“three-dimensional” style dimorphism, has been described in
L. suffruticosum L. ( Armbruster et al. 2006 ). Interestingly, het-
erostyly in Linum appears to be restricted to the Old World: no
New World taxa have been reported as heterostylous ( Rogers
1981 ). The reconstruction of the evolution of heterostyly in
this lineage of Linaceae will add to the number of examples
where the evolution of heterostyly has been studied from a
phylogenetic perspective ( Kohn et al. 1996 ; Conti et al. 2000 ;
Pérez et al. 2003 ; Mast et al. 2004 ; Barrett and Harder 2005 ;
Armbruster et al. 2006 ).
Materials and Methods
Taxon Sampling— A total of 44 taxa (48 for rbcL ) representing all eight
genera of the temperate Linaceae (i.e. subfamily Linoideae) were included
in this study (sampled taxa and voucher specimens are listed in Appendix
1). Of these, 37 samples are from species of Linum , chosen to represent the
five sections of the genus and their geographical centers of diversity. The
other seven genera of Linoideae are represented by one sampled taxon
each. Hugonia busseana Engl., a representative of the tropical subfamily
Hugonioideae, was selected as the outgroup based on previous subfam-
ilial delimitations ( Van Hooren and Nooteboom 1984 ) and phylogenetic
studies that have shown Hugonia to be sister to temperate members of
Linaceae ( Savolainen et al. 2000 ; Davis et al. 2005 ). Additionally, two
species each of Hugonia and Roucheria, and Durandea pentagyna were
sampled for inclusion in an analysis of the relationship of Linoideae to
Hugonioideae in the context of other Malpighiales families using rbcL .
DNA Extraction, PCR, and Sequencing— DNA was extract-
ed from fresh leaves or herbarium collections using a modified Doyle
and Doyle CTAB protocol ( Loockerman and Jansen 1996 ) or the plant
DNA extraction kit NucleoSpin
® Extraktion (Macherey-Nagel, Düren,
Germany). The DNA regions inclu ded in this study are listed in Table 2 ,
along with the primers used to amplify and sequence them.
For the amplification of the nuclear ITS and rbcL (done at
Mainz), PCRs were performed on the basis of the protocol of
Palumbi ( 1996 ). The PCR reaction mix consisted of 2.5 μl 10 ×
BioTherm reaction buffer, 200 μM dNTPs, 1 pM each primer,
2 mM MgCl
2 , 0.025U/μl Biotherm Polymerase (GeneCraft, Köln,
Germany) and 2 μl of DNA extract. PCR reactions were carried out in
a Whatman Biometra
® TGradient TM Thermocycler (Biometra GmbH,
Göttingen, Germany) or a PTC 100
TM (MJ Research, Watertown,
Massachussetts) using the following program: 60 sec at 94°C, followed by
35 cycles of 18 sec at 94°C, 30 sec at 52°C, 60 sec at 72°C and a posttreat-
ment of 78 sec at 55°C and 8 min at 72°C. PCR products were checked
on 0.8% agarose gels and purified directly using the NucleoSpin Extract
purification Kit (Macherey-Nagel).
For the amplification of the three additional chloroplast markers and
some rbcL (done at The University of Texas at Austin), PCR was executed
in 25 μL reaction volumes, including: 2.5 μL 10 × Triton-X reaction buffer,
200 μM dNTPs, 10 pM each primer, 2–3 μL 25 mM MgCl
2 , and 10–100 ng
template DNA. Addition of BSA (3 μL of 3.3% w/v BSA, Savolainen et al.
Table 2. Genomic regions utilized, and primers used to amplify and sequence them.
Region Primer name Primer Sequence (5′-3′) Reference
Chloroplast
rbcL rbcL 1F ATGTCACCACAAACAGAAAC Olmstead et al. 1992
rbcL 579R AAATCAAGTCCACCRCG
rbcL 507F TAYTGGGCTGTACTATTAAAC
rbcL 994R CCTTCAAGTTTACCTAC
rbcL 995F CGTATGTCTGGTGGAGATC
rbcL 3ambigF KCTTATAYTAAAACTTTCCAAGGTCCGCC
rbcL 4R ACCATGATTTTTCTGTCTATCAAATA
rbcL 1460R TCCTTTTAGTAAAAGATTGGGCCGAG Olmstead et al. 1992
ndhF ndhF 5 CAATGGTAGCAGCGGGAATTTTTC Jansen 1992
ndhF 8 ATAGATCCGACACATATAAAATGCGGTT Jansen 1992
trnL - trnF C CGAAATCGGTAGACGCTACG Taberlet et al. 1991
D GGGGATAGAGGGACTTTGAAC Taberlet et al. 1991
E GGTTCAAGTCCCTCTATCCC Taberlet et al. 1991
F ATTTGAACTGGTGACACGAG Taberlet et al. 1991
trnK 3′ intron matK 8 CTTCGACTTTCTTGTGCT Steele and Vilgalys 1994
trnK 2R CTACTCCATCCGACTAGTT Steele and Vilgalys 1994
Nuclear (ITS) ITS A GGAAGGAGAAGTCGTAACAAGG White et al. 1990
ITS C GCAATTCACACCAAGTATCGC White et al. 1990
ITS D CTCTCGGCAACGGATATCTCG White et al. 1990
ITS 4 TCCTCCGCTTATTGATATGC White et al. 1990
18s CCTTMTCATYTAGAGGAAGGAG Muir & Schlötterer 1999
28s CCGCTTATTKATATGCTTAAQ Muir & Schlötterer 1999

390 SYSTEMATIC BOTANY [Volume 34
2000 ) and/or 1.25 μL DMSO to the reaction mix enhanced amplification
from some samples. The thermal cycling program for the amplification of
the chloroplast markers consisted of the following steps: initial denatur-
ation at 95°C for 4 min, followed by 34 cycles with denaturation at 95°C for
30 sec, annealing for 30 sec at 50°C ( trnK -intron), 51°C ( ndhF ), 53° ( trnL-F ),
or 55°C (for rbcL done at The University of Texas at Austin), and extension
at 72°C for 30 sec. Cycles were followed by a final extension step at 72°C
for 6 min. Amplified products were checked on 1.5% agarose gels, and
cleaned using either QiaQuick columns (Qiagen Inc., Valencia, California)
or Centri-Sep columns packed with G-50 Sephadex (Princeton
Separation, Inc., Adelphia, New Jersey), according to the manufacturer’s
instructions.
Amplified fragments were sequenced with forward and reverse dye ter-
minator reactions using the original amplification primers for each marker
and BigDye 3.1 chemistry (Applied Biosystems, Foster City, California).
Purified PCR products for the ITS and a portion of the rbcL dataset were
cycle-sequenced with the Big-Dye Terminator Cycle Sequencing Ready
Reaction Kit (BD 3.1 in 10 μl reactions, Perkin Elmer) using the PCR prim-
ers listed above and following the manufacturer’s protocol. Products were
purified and analyzed on automated sequencers (ABI 373 and ABI 377) by
Genterprise (Mainz, Germany). Sequencing for the remainder of the rbcL
data new in this study, and of the other three chloroplast regions in their
entirety, was done at The University of Texas at Austin on either an MJ
Research BaseStation sequencer (MJ Research, Waltham, Massachussetts,
discontinued) operated by J. McDill, or an ABI 3730 DNA Analyzer at the
Institute for Cell and Molecular Biology Core Facility.
All forward and reverse sequence reads were assembled and edited in
Sequencher 4.1.2 or 4.5 ( GeneCodes Corp. 2005 ), then aligned manually (in
the case of ITS, ndhF, and rbcL ) in MacClade 4.08 ( Maddison and Maddison
2005 ) or using the automated alignment program MUSCLE ( Edgar 2004 )
followed by manual adjustment in MacClade. When sequences for indi-
vidual datasets could not be obtained from specific taxa, those taxa were
excluded from the individual analyses of those datasets. They were
included in the combined analyses, with any missing sequences encoded
as “missing data”.
Phylogenetic Analyses— Analyses using maximum parsimony (MP)
as the optimality criterion were conducted with PAUP* 4.0b10 ( Swofford
2002 ). The heuristic search algorithm was engaged with 1,000 random
addition replicates and tree bisection-reconnection (TBR) branch-swap-
ping, saving and swapping to completion on all minimal trees found.
Bootstrap analyses under the MP criterion for the ITS dataset used 1,000
bootstrap replicates, each with 10 random taxon addition replicates, and
TBR swapping to completion on all minimal trees. For the three non- rbcL
chloroplast regions, MP bootstrap analyses also used 1,000 bootstrap
replicates, the heuristic search algorithm, and TBR swapping, but only
5 random addition replicates per bootstrap replicate, and swapping to
completion on a maximum of 2,000 trees per random addition replicate.
This abbreviated strategy was used because the lack of resolution found
in each of these chloroplast markers when analyzed separately resulted
in prohibitively long bootstrap search times. For the analysis of the ITS,
rbcL , the combined 3-marker chloroplast matrix, and the combined ITS
+ 3-marker chloroplast matrix, the same bootstrap search was conducted
but without the limitation to 2,000 trees per random addition replicate.
Bayesian analyses were performed with MrBayes v.3.1.2 ( Huelsenbeck
and Ronquist 2001 ). Models of nucleotide substitution were selected for
each chloroplast region and for three partitions of the nuclear ITS (ITSI,
5.8s , ITSII) using the Akaike Information Criteria (AIC) implemented
in Modeltest 3.7 ( Posada and Crandall 1998 ). The appropriate form of
substitution model was then applied to each data partition when ana-
lyzed separately and in combination. Prior probability distributions on
all parameters were kept at default values, and all parameters except for
tree topology and branch lengths were unlinked between all partitions.
Two independent analyses were run simultaneously (nruns = 2), with the
built-in convergence diagnostic (average standard deviation of split fre-
quencies) activated (mcmcdiagn = yes), and with a relative burn-in per-
centage of 25 (relburnin = yes, burninfrac = 0.25). Analyses were run for a
minimum of 4 million MCMC generations (sampling rate = 0.01), at which
point they were stopped if the convergence diagnostic had dropped below
0.01. Graphical plots of –lnL sampled throughout the two simultaneous
runs were also examined as an additional confirmation that convergence
had been reached. Each Bayesian analysis was repeated to confirm that
clade posterior probabilities and estimated model parameters were stable
and repeatable.
Analyses using maximum likelihood (ML) were conducted with
PAUP* 4.0 ( Swofford 2002 ) with substitution models covering the chlo-
roplast and nuclear markers independently and in combination as deter-
mined by Modeltest. Maximum likelihood bootstrap values were obtained
from 100 bootstrap replicates, each incorporating five random taxon addi-
tion replicates and TBR swapping to completion on all optimal trees.
All five sequenced regions were initially analyzed separately.
Congruence between datasets was evaluated by visual comparison of
the topologies and levels of clade support arrived at by the three types of
analysis (MP, Bayesian, ML). Congruence of the phylogenetic signals from
the nuclear and chloroplast data was also tested using the Incongruence
Length Difference test (ILD, Farris et al. 1995 ), implemented in PAUP*
as the Partition Homogeneity Test. Five hundred partition homogene-
ity replicates using the heuristic search option with 10 random addition
sequences were implemented for the test. The rbcL matrix was not ana-
lyzed in combination with the other four markers due to differences in
taxon sampling. Taxa whose representation differs in the rbcL matrix, and
the outgroups used for rbcL analyses, are indicated in Appendix 1.
All datasets utilized for this work have been deposited in TreeBASE
(Study number S2090) and DNA sequences are deposited in GenBank
(accession numbers provided in Appendix 1).
Molecular Clock Analysis— Divergence times among major groups in
Linoideae were estimated on the basis of the rbcL dataset. In a first step,
Modeltest ( Posada and Crandall 1998 ) was used to perform a Likelihood
Ratio Test (LRT) in order to test whether the assumption of a molecular
clock was valid. The significant difference found between the scores with
and without a molecular clock at the 0.01 level indicated that a molecular
clock should be rejected. Therefore, a semiparametric method that relaxes
the stringency of the clock assumption by using smoothing methods,
implemented in the program r8s ( Sanderson 2002 ), was used. Starting
point for this was the ML tree topology including ML branch lengths
and the number of positions in the dataset (1,408). Rate smoothing was
done using Penalized Likelihood (PL, Sanderson 1997 , 2002 ). Smoothing
parameters were calculated using the following settings: divtime method:
pl, crossv = yes, cvstart = 0, cvinc = 0.2 and cvnum = 100. The divergence of
Linaceae from Phyllanthus (here sister to Linaceae) was alternately fixed at
the maximum and minimum ages estimated by Davis et al. (2005) for the
divergence of Linaceae from its sister group in their analysis (Irvingiaceae),
based on 4 macrofossil and 11 palynofossil calibration points.
Ancestral State Reconstruction— To investigate the evolution of het-
erostyly in the Linoideae, we utilized parsimony and maximum likelihood
algorithms for reconstructing ancestral states on phylogenies in Mesquite
v1.12 ( Maddison and Maddison 2006 ). The presence/absence of heterostyly
for each taxon was determined from the literature or through examination
of herbarium or living collections when possible (states for each taxon are
provided in Appendix 1). Parsimony reconstruction used equal weighting
of gains and losses of heterostyly and considered all branch lengths as
being equal. Ancestral states were optimized under both “acctran” and
“deltran” assumptions to determine the maximum and minimum num-
bers of gains and losses of heterostyly on the phylogeny. Since likelihood
reconstruction in Mesquite does not allow polymorphic taxa, ancestral
states were optimized with L. tenuifolium (which has both distylous and
homostylous populations), Tirpitzia (with one of three species reported to
be distylous ( Suksathan and Larsen 2006 )), and Hugonia (representative of
Hugonioideae, which has homostylous, distylous, and tristylous species)
alternatively scored as homostylous (Optimization 1) and heterostylous
(Optimization 2). Linum grandiflorum , which exhibits stylar dimorphism,
was scored as heterostylous. The tree used for this reconstruction was the
topology with the best likelihood score sampled during Bayesian analysis
of the combined ITS and the three chloroplast markers. We compared the
likelihood scores under the MK1 (symmetric, rate of gain = rate of loss)
and Asymmetric 2-parameter (gains and losses occurring at different rates)
models to determine the appropriate model and parameters for the ances-
tral state reconstruction. A likelihood ratio test with one degree of freedom
was used to compare the likelihood scores under the two models to deter-
mine whether the more complex, asymmetric model resulted in a signifi-
cantly better likelihood score than the asymmetric model (test statistic =
twice the difference in likelihood scores). Branch lengths of the topology
as estimated by MrBayes were utilized in Mesquite during estimation of
the model parameters and the final estimation of the proportional likeli-
hoods at the internal nodes. A ratio of proportional likelihoods of 7:1 or
greater was considered significant ( Schluter et al. 1997 ; Cunningham 1999 );
this corresponds to a proportional likelihood of 0.875 or greater.
Results
Taxa sampled, with voucher collections and GenBank acces-
sion numbers, are provided in Appendix 1. Basic sequence
information for all markers, including missing data and the

2009] MCDILL ET AL.: PHYLOGENY OF LINOIDEAE 391
likelihood model selected, is summarized in Table 3 . Additional
marker-specific information is provided below as necessary.
rbcL— The rbcL data set comprised 48 species from Linaceae
and seven species of other families of Malpighiales as out-
groups. No sequence fragments were excluded due to ambig-
uous alignment. To calculate the tree used in the molecular
dating analysis, the GTR + I + G model ( Tavaré 1986 ) selected
by AIC was chosen with nucleotide frequencies A = 0.2739,
C = 0.1855, G = 0.2465, T = 0.2911, the rate matrix set to AC =
1.3889, AG = 2.0477, AT = 0.5296, CG = 0.7278, CT = 3.4964 and
GT = 1, a gamma shape parameter of 0.9969 and an assumed
proportion of invariable sites of 0.5336.
ITS— due to uncertainty in alignment, 131 bp of ITS 1 and
2 were excluded from analyses. Several potentially informa-
tive insertions or deletions were observed: Linum tenuifolium ,
L. strictum , L. volkensii , and L. suffruticosum were found to share
a deletion of 19 bp in ITS 1. Members of the Linum perenne
group ( L. austriacum , L. lewisii, L. pallescens , and L. perenne )
share a 4 bp insertion in ITS 1 relative to the all other taxa,
while members of section Dasylinum ( L. pubescens , L. viscosum ,
L. hirsutum , and L. hypericifolium ) share a 10 bp deletion.
ndhF— Only a partial sequence could be obtained for Linum
pubescens ; 168 bp are missing out of 440. No sequence frag-
ments were excluded due to alignment issues.
trnK 3′-intron— This intron could not be amplified from any
of the South American taxa or from Anisadenia pubescens Griff.
(13% missing data). The aligned matrix included sequences
from the remaining 39 taxa. Two characters were excluded
due to alignment issues. A nine bp deletion is shared by
L. album , L. arboreum , L. flavum, and L. nodiflorum (members of
sect. Syllinum ), and a five bp indel distinguishes the members
of sections Linum and Dasylinum from the other taxa sampled.
Ten additional indels ranging from one to 49 bp in length
were unique to the taxa in which they occur.
trnL-F— Sequence of the trnL -intron could not be obtained
from Radiola linoides . The aligned matrix was 1,132 bp in
length, of which 937 bp could be aligned with confidence (195
are excluded from analyses). Two deletions, one of six bp and
the other of 105 bp, are shared by L. hirsutum , L. hypericifolium ,
L. pubescens and L. viscosum (sect. Dasylinum ). A 58 bp deletion
was shared by the members of sections Linum and Dasylinum .
Reinwardtia , Tirpitzia, and Anisadenia share a 12 bp insertion.
A four bp insertion was shared by the North American, South
American, and South African taxa. Several other indel regions
were observed which are autapomorphic or of ambiguous
alignment.
Pairwise ILD tests of the three chloroplast markers did not
indicate significant conflict among them. When they were
combined, MP analysis returned 360 MPT (L = 1,037). The
ILD test also did not indicate significant conflict between the
ITS and the combined 3-marker chloroplast matrix. The com-
bined ndhF , trnK intron, and trnL-F dataset shall henceforth
be referred to as the chloroplast data (CPL). The rbcL gene will
continue to be referred to as rbcL , although it is also located in
the chloroplast genome.
Phylogenetic Results— Phylogenetic analysis of the rbcL
data shows Linaceae to be a well-supported monophyletic
group. The tropical subfamily Hugonioideae is paraphyletic
to subfamily Linoideae which is a monophyletic lineage with
strong support from Bayesian posterior probability (PP) and
bootstrap estimates ( Fig. 2 ). Analysis of rbcL failed to resolve
unambiguously the relationships of the Southeast Asian gen-
era Anisadenia , Reinwardtia, and Tirpitzia , although they are
consistently successive sister to the remaining genera. While
the best topologies recovered showed those three genera
forming a paraphyletic grade subtending the remainder of
Linoideae, the nodes separating them are not well supported.
This paraphyletic grade is also found in the topologies result-
ing from the analysis of ITS ( Fig. 4A ), which provides some
support for a sister relationship between Anisadenia and
Tirpitzia and places Reinwardtia sister to the rest of Linoideae,
although without strong support. The three CPL markers
( Fig. 4B ) do not consistently resolve the relationships of the
Southeast Asian genera. In combination, the ITS + CPL topol-
ogy ( Fig. 5 ) resembles that of ITS alone with respect to the
relationships of these genera.
The remainder of Linoideae are consistently resolved into
two major clades in all analyses: the blue-flowered flaxes,
comprising sections Dasylinum and Linum , and the yellow-
flowered flaxes containing sections Cathartolinum , Linopsis,
and Syllinum as well as Cliococca , Hesperolinon , Radiola, and
Sclerolinon . Thus Linum in its present circumscription is para-
phyletic in relation to these last four genera.
The predominantly blue-flowered sections of Linum clus-
ter into a generally well-resolved and well-supported lineage,
with sect. Dasylinum nested within sect. Linum . The analyses
of rbcL present conflicting support for the placement of the
east Asian L. stelleroides depending on the analytical method
used: parsimony and likelihood bootstrap values indicate low
levels of support (59 ML bootstrap, 51 MP bootstrap) for this
species as the earliest-diverging member of the blue-flowered
lineage, sister to the rest of sections Linum and Dasylinum
Table 3. Sequence data information. CPL: chloroplast loci.
Combined
Parameter rbcL ITS trnK 3′ intron trnL-F ndhF 3 CPL 3 CPL + ITS
Number of sequences 55 45 39 45 45 45 45
Missing data n.a. n.a. n.a. n.a. n.a. 3.70% 2.80%
Aligned length (bp) 1568 704 411 1132 440 1975 2679
bp included in analyses 1568 573 409 937 440 1811 2384
Variable characters 370 325 120 328 123 599 924
Parsimony informative chars. 192 259 45 173 88 324 583
Number of trees (MP) 790 121 378 46916 75090 360 221
MP tree length 735 1215 172 552 237 1037 2231
CI (informative characters only) 0.458 0.439 0.703 0.646 0.655 0.643 0.518
RI (informative characters only) 0.785 0.686 0.91 0.865 0.850 0.857 0.766
L (informative characters only) 539 1135 91 381 200 737 1854
ML model selected (AIC) GTR + I + G GTR + G GTR + G GTR + G GTR + I + G GTR + I + G GTR + I + G

392 SYSTEMATIC BOTANY [Volume 34
Fig. 2. Phylogeny of rbcL from Linaceae and selected outgroup families. Bayesian majority-rule (50%) consensus (posterior probability above
branches, ML bootstrap/MP bootstrap below branches). Alternative placement of Linum stelleroides recovered during ML and MP bootstrap indicated by
dashed line. Open boxes containing taxon names are labeled to indicate the general geographical distributions of the enclosed taxa. Filled boxes indicate
the classification of sampled taxa into the sections of Linum (C = sect. Cathartolinum ), and subfamily Hugonioideae.

2009] MCDILL ET AL.: PHYLOGENY OF LINOIDEAE 393
as sampled ( Fig. 2 ). The Bayesian analysis of rbcL , however,
indicates that L. stelleroides may be sister to sect. Dasylinum
(PP = 74). The CPL and ITS, separately and in combination,
show L. stelleroides as moderately supported sister to a clade
containing the rest of the blue-flowered linums, as found in
the rbcL MP and ML analyses.
The remaining three sections of Linum and the three
New World segregate genera comprise a lineage that is sis-
ter to Radiola linoides ( Figs. 2 , 4 , 5 ). Section Syllinum is evi-
dently monophyletic as sampled, but section Linopsis is
not. Eurasian members of Linopsis consistently group with
Syllinum (also Eurasian); this relationship is supported in
the Bayesian analyses of all datasets but has insignificant
ML and MP bootstrap values in the separate and combined
analyses of ITS and CPL. The relationships of L. catharticum
L. (sect. Cathartolinum ) to other members of this lineage are
unclear. The analysis of rbcL does not resolve the placement
of L. catharticum relative to the Old or New World lineages of
Linopsis and Syllinum ( Fig. 2 ). Bayesian analysis of ITS places
L. catharticum within the Old World Linopsis / Syllinum clade
as sister to sect. Syllinum plus three members of sect. Linopsis,
but this placement receives insignificant bootstrap support
( Fig. 4A ). Analyses of the CPL place L. catharticum as sister to
a clade containing the New World and South African mem-
bers of sect. Linopsis , but with a very low posterior proba-
bility and no significant bootstrap support ( Fig. 4B ). The
combined analyses of ITS + CPL place L. catharticum as sister
to the clade consisting of sect. Syllinum and Eurasian mem-
bers of sect. Linopsis .
The New World and South African members of Linum
sect. Linopsis together with the segregate genera Cliococca ,
Hesperolinon, and Sclerolinon are a consistently recovered
and well-supported group in analyses of rbcL ( Fig. 2 ), CPL
( Fig. 4B ), and ITS + CPL ( Fig. 5 ), but ITS alone does not
support the monophyly of this group ( Fig. 4A ). The South
African and South American species groups each appear
monophyletic as sampled, and are well-supported as such
in all analyses. Cliococca is consistently shown to be sister
to South American members of sect. Linopsis . The relation-
ships of the North American taxa are not well resolved.
Analysis of rbcL results in a topology with Sclerolinon and
the two sampled species of Hesperolinon forming a para-
phyletic grade subtending North American sect. Linopsis
plus the South American and South African groups ( Fig.
2 ). This rbcL topology has only low posterior probabilities
and no bootstrap support for the paraphyly of Hesperolinon .
Sclerolinon and Hesperolinon appear sister to each other in
analyses of ITS and the three chloroplast markers separately
( Fig. 4 ) and in combination ( Fig. 5 ). The CPL analyses ten-
tatively place them as sister group to the North American
L. striatum ( Fig. 4B ). The ITS and ITS + CPL analyses also
support the close relationship of L. vernale and L. rupestre ,
and tentatively place L. kingii as sister to them ( Figs. 4A ,
5 ). The relationships among the South American, South
African, and North American species groups are not well
resolved. Analysis of the three chloroplast markers alone
( Fig. 4B ), and of ITS and those three markers combined
( Fig. 5 ), place the South African group sister to a clade con-
sisting of the North American and South American taxa. ITS
alone fails to resolve the issue ( Fig. 4A ), while rbcL provides
some support for a sister relationship between the South
American and South African lineages ( Fig. 2 ).
The results of the rbcL dating analysis using Penalized
Likelihood are contained in Table 4 , and the dated nodes are
indicated in Fig. 3 .
Heterostyly Ancestral State Reconstruction— Ancestral
state reconstructions using both maximum likelihood and
parsimony methods and two alternative scorings for Hugonia ,
Tirpitzia , and L. tenuifolium are provided in Fig. 6 . The likelihood
ratio test comparing the symmetric (MK1) and Asymmetric
2-parameter models of evolution indicated that the 2-param-
eter model did not result in a significantly better likelihood
score ( Table 5 ), thus ancestral states were subsequently esti-
mated using the MK1 model.
Parsimony Ancestral States— Under Optimization 1
( Hugonia , Tirpitzia , and Linum tenuifolium scored as homosty-
lous, Fig. 6A ), parsimony reconstructs as many as 10 inde-
pendent gains of heterostyly in Linoideae, and a minimum of
five gains with five subsequent losses. In this optimization,
homostyly is unambiguously reconstructed as the ancestral
condition in the most recent common ancestor (MRCA) of the
Linaceae, in the MRCA of subfamily Linoideae, in the MRCA
of all linums sampled, and in the MRCAs of both the yellow-
and blue-flowered lineages. Under Optimization 2 ( Fig. 6B ),
when Hugonia , Tirpitzia , and L. tenuifolium are scored as het-
Fig. 3. Phylogram of the rbcL tree with the best ML score, with dated
nodes indicated (see Table 4 for node age estimates).

394 SYSTEMATIC BOTANY [Volume 34
erostylous, parsimony unambiguously reconstructs het-
erostyly as the ancestral state in Linaceae and Linoideae.
In this optimization, parsimony reconstructs homostyly as
the ancestral state in the MRCA of Radiola and the lineage
of Linum et al. to which it is sister. The ancestral states in
the MRCA of blue-flowered linums and the MRCA of all
linums sampled is equivocal in Optimization 2. Parsimony
implies at least three (but as many as five) independent gains
of heterostyly within Linoideae, and at least six (or as many
as eight) reversals to homostyly from heterostylous ances-
tors, depending on the resolution of ambiguous nodes in this
optimization.
Likelihood Ancestral States— In both optimizations,
heterostyly becomes the most likely ancestral state, with
significant support, in three clades: section Syllinum , sect.
Dasylinum , and in the Linum perenne group (excluding L. pall-
escens ). Homostyly is significantly supported in both recon-
structions at all of the internal nodes in the clade of North
American, South American, and South African linums, and in
the MRCA of L. bienne and L. usitatissimum . Optimization 1
adds significant support for homostyly as the ancestral state at
one additional node, the MRCA of L. strictum and L. volkensii .
Neither state is reconstructed with significant support at any
of the other internal nodes of the tree using this likelihood
method.
Discussion
Infrafamilial Relationships— The Linaceae is a well-sup-
ported monophyletic group in our analyses of rbcL . The fam-
ily is also well characterized morphologically by having a
five-merous calyx and corolla, basally fused filaments, nec-
taries associated with the stamens, and a superior ovary with
the number of styles or style branches always corresponding
to the number of carpels. Hugonioideae appear paraphyl-
etic in relation to the monophyletic Linoideae. Hugonioideae
can be distinguished from Linoideae by a number of char-
acters with all members having a diplostemonous androe-
cium of ten fertile stamens, whereas in Linoideae only one
whorl of five antesepalous fertile stamens is found. In most
Linoideae, however, five structures often interpreted as sta-
minodes can be observed between the free parts of the fil-
aments. In Hugonoideae, five nectaries are found at the
base of the antesepalous stamens. Whereas in Indorouchera
Fig. 4. Comparison of A) ITS and B) chloroplast ( ndhF + trnK intron + trnL-F ) phylogenies. Topologies represent 50% majority-rule summaries of
60,000 trees (30,000 from each of two simultaneous runs) sampled during stationarity in Bayesian analyses of each data set. Numbers above branches
represent posterior probabilities expressed as percentages. Branches with posterior probabilities below 50% are collapsed. Numbers below branches are
MP/ML bootstrap percentages. Bootstrap values below 50 are indicated by an asterisk.

2009] MCDILL ET AL.: PHYLOGENY OF LINOIDEAE 395
Fig. 5. Combined ITS + CPL ( ndhF + trnK intron + trnL-F ) topology. 50%-majority-rule consensus of 60,000 trees sampled during stationarity during
6-partition Bayesian analysis of combined ITS and CPL data. Above branches are posterior probabilities as percentages. Below branches are MP and ML
bootstrap values (MP/ML). Sectional affiliation of Linum species are indicated in the filled boxes ( C = sect. Cathartolinum ); open boxes indicate general
geographical distributions of species.

396 SYSTEMATIC BOTANY [Volume 34
and Roucheria these nectaries are diffuse and spread nearly
over the entire base of the filament tube, they are spatially
restricted in the remaining genera ( Link 1989 ). In Linoideae
a reduction in number of nectaries can be observed. In gen-
eral the styles of Hugonioideae are free or connate only at
their base. In contrast, a conspicuous fusion of styles can be
seen in several lineages of Linoideae. The stigmas of all mem-
bers of Hugonioideae are capitate and more or less two-lobed
( Fig. 7A ), whereas in Linoideae a broad range of shapes from
capitate to clavate or linear can be seen ( Fig. 7 ). In Linum ,
stigma shape has been an important character for distinguish-
ing the different sections.
The Southeast Asian genera Anisadenia , Tirpitzia, and
Reinwardtia , successive sisters to the remainder of Linoideae in the
rbcL analysis, share several characteristics with Hugonioideae:
they are all perennials with petiolate leaves and foliate stipules.
In the remaining Linoideae, leaves are sessile and the stipules
either are reduced to a pair of dark brown glands or are absent.
Accordingly, petiolate leaves and foliate stipules can be consid-
ered plesiomorphic in Linoideae.
Relationships of the Sections of Linum and the Segregate
Genera— Our analyses indicate that Linum and two of its five
sections are not monophyletic as currently circumscribed.
Section Dasylinum appears monophyletic, but is nested within
sect. Linum . Radiola is placed squarely between the major lin-
eages of Linum . Sections Cathartolinum and Syllinum are nested
within sect. Linopsis , as are the segregate genera Hesperolinon ,
Sclerolinon, and Cliococca . ( Sharsmith 1961 ; Rogers 1966 , 1975 ),
implying the paraphyly of Linum when the segregates are
recognized.
The Blue-flowered Clade: LINUM sect. LINUM — Section
Linum comprises about 50 species of which 11, covering the
entire geographic range of the section, were included in this
study. Section Linum turned out to be paraphyletic in rela-
tion to sect. Dasylinum because of the position of L. stelleroides
(sect. Linum ), a species found in China, which emerged either
as sister to Dasylinum ( rbcL ) or as sister to Dasylinum plus
the remaining members of sect. Linum (ITS and other chlo-
roplast data). Morphologically, sect. Linum is quite well char-
acterized. All species have lanceolate leaves without stipular
glands, and the majority of species have bright blue petals
and nondecurrent leaves. Within sect. Linum (excl. L. stelleroi-
des ), two well-supported groups can be identified. These are:
1) the L. perenne group, comprising L. austriacum , L. lewisii ,
L. pallescens, and L. perenne and; 2) a clade containing L. bienne ,
L. decumbens , L. grandiflorum , L. marginale , L. narbonense, and
L. usitatissimum . These two groups are morphologically quite
similar but can be distinguished from each other on the basis
of their sepals and stigmas. Members of the L. perenne group
have capitate stigmas and entire sepals with a smooth margin
( Fig. 7B ). In the second clade, stigmas are linear and the sepals
have ciliate margins ( Fig. 7C ).
Morphologically, Linum stelleroides can not be assigned to
either of these groups within section Linum because its stig-
mas are capitate and the sepals have glandular margins as
found in sect. Dasylinum . Linum stelleroides further differs
from the other blue-flowered flaxes in having 2 n = 20 chro-
mosomes - the chromosome base number in the remainder of
sect. Linum is n = 9 ( Linum perenne group, Ockendon 1968 ) or
n = 8 or 15 (clade 2, Martzenitzina 1927 ; Petrova 1972 ). It also
has multiporate pollen grains which otherwise, as far as is
known, occur only in the australasiatic L. monogynum G. Forst.
and L. marginale ( Rogers 1984 ), and in some North American
yellow-flowered flaxes. On the basis of the morphological and
karyological distinctness of L. stelleroides , Yuzepchuk (1974)
recommended its placement in a monotypic sect. Stellerolinon .
Such treatment would be supported by our molecular data
and would make sect. Linum monophyletic (at least in terms
of our current taxon sampling).
The Blue-flowered Clade: LINUM sect. DASYLINUM — Our
phylogenetic reconstructions indicate that sect. Dasylinum is a
well-supported monophyletic group. This finding has morpho-
logical support. Most species of the section are pubescent peren-
nials with more or less terete stems. Leaves are spathulate or
linear and often have stalked marginal glands. Stipular glands
are lacking. Sepals usually have stalked marginal glands, and
stigmas are linear to clavate ( Fig. 7D ). All chromosome numbers
reported are 2 n = 16 ( Ray, 1944 ; Petrova 1972 ). The section is
confined to southern and eastern Europe and Southwest Asia.
The Yellow-flowered Clade: LINUM sect. LINOPSIS — A s
mentioned above, sect. Linopsis is either polyphyletic or para-
phyletic in relations to sects. Syllinum and Cathartolinum a n d
the New World segregate genera. Within currently circum-
scribed sect. Linopsis , several well-supported groups are
identified in our phylogenetic analysis. Linum maritimum,
L. tenue, and L. trigynum (subsect. Halolinum (Planch.) C. M.
Rogers) form a well-supported clade in the ITS, CPL, and
combined analyses ( Figs. 4 , 5 ). This group is characterized
by more or less linear stigmas ( Fig. 7F ) and a chromosome
base number of n = 10 which, as far as is known, is unique
in the yellow-flowered clade. Among the Old World mem-
bers of the yellow-flowered flaxes a second well-supported
group of species can be recognized, comprising L. strictum ,
L. suffruticosum , L. tenuifolium, and L. volkensii . These spe-
cies share capitate stigmas ( Fig. 7E ), hook-shaped epider-
mal cells on the leaves, and a chromosome base number of
n = 9. Although these species share several characters, they
have been placed in different subsections: Linum tenuifolium
and L. suffruticosum in subsect. Dichrolinum (Planch.) C. M.
Rogers, and L. strictum and L. volkensii in subsect. Linopsis
(Rchb.) C. M. Rogers. The four species also share two indels
(four bp and 16 bp) in their ITS.
The Yellow-flowered Clade: LINUM sect. SYLLINUM — The
four species sampled from section Syllinum form a well-sup-
ported monophyletic group in all our analyses, and this is
supported by morphological data. Most species of this sec-
tion are glabrous perennials with much larger leaves than
found in the remaining species of the yellow-flowered flaxes.
Table 4. Penalized Likelihood time estimates for the nodes indicated
with lowercase letters in the rbcL ML phylogram of Fig. 3 . Node ages are
in millions of years. The age of Linaceae was fixed based on the results of
Davis et al. (2005) and these dates are indicated in bold.
Maximum Age Minimum Age
Linaceae (a) 105.00 94.50
Linoideae (b) 51.23 46.05
Linum (c) 46.24 41.63
blue-flowered clade (d) 32.46 29.1
Linum section Dasylinum (e) 14.46 12.92
Linum section Linum (f)
(without L. stelleroides )
25.2 22.53
Linum perenne group (g) 3.78 3.33
yellow-flowered clade (h) 36.02 32.44
Linum section Syllinum (i) 18.75 16.94
Colonization of New World by
yellow-flowered lineage (j)
22.3 20.12

2009] MCDILL ET AL.: PHYLOGENY OF LINOIDEAE 397
Fig. 6. Breeding system ancestral state reconstruction. The topology used is that of the tree with the best likelihood score sampled during the
Bayesian analysis of the combined ITS + CPL data. Circles at branch tips indicate the observed character states in sampled taxa. Branches are shaded
according to parsimony state reconstruction. Pie diagrams at nodes indicate relative degree of support for alternative character states (black = heterostyly,
white = homostyly) as proportional likelihoods calculated under the MK1 (symmetric) model, with a ratio of 7:1 or greater considered significant and
marked with an asterisk. For clarity, trees are drawn here with nodes equally spaced from root to tips and not proportional to estimated branch lengths,
though estimated branch lengths were taken into account in Mesquite during ML model selection and state reconstruction. A) Optimization 1: Hugonia ,
Tirpitzia , and Linum tenuifolium scored as homostylous. B) Optimization 2: Hugonia , Tirpitzia , and L. tenuifolium scored as heterostylous.

398 SYSTEMATIC BOTANY [Volume 34
Stipular glands are often, but not always, present. The leaves
are usually decurrent, with the leaf margins and midveins
forming prominent wings along the stems. The margin of the
keeled sepals is ciliate and sometimes covered with glands
( Fig. 7G ). The predominantly yellow petals fuse after initia-
tion and the stigmas are linear and taper into the style ( Fig.
7G ). Section Syllinum occurs mainly in Southwest Asia and
the Mediterranean area.
The Yellow-flowered Clade: LINUM sect. CATHARTOLINUM —
Section Cathartolinum is the only monotypic section of Linum .
Linum catharticum is generally distinguished from the other sec-
tions by its combination of decussate phyllotaxis, small white
petals, and annual habit. A greatly expanded circumscription
was recognized by Small ( 1907 ) as the genus Cathartolinum
Rchb., which included L. catharticum (as Cathartolinum cathar-
ticum Small), L. digynum A. Gray (as Cath. digynum , later to
become Sclerolinon ), and all of the North American yellow-
flowered linums. This expanded Cathartolinum was accepted
at the sectional rank sporadically by later workers, such as
Winkler ( 1931 ) and Diederichsen and Richards (2003) , while
others (notably Rogers, in his extensive work on the genus)
consider most of the species of Small’s Cathartolinum as mem-
bers of Linum sect. Linopsis (= sect. Linastrum ). The relation-
ship of L. catharticum is not well supported in our analyses,
but results suggest that additional North American, South
American, Eurasian, and African species would have to be
included in an expanded concept of Cathartolinum in order
for it to be monophyletic. However, there are no obvious mor-
phological synapomorphies shared by L. catharticum and the
members of sections Linopsis or Syllinum to which it appears
related in our analyses. Its chromosome number of 2 n = 16
( Petrova 1972 ) is unique in the yellow-flowered clade; only
members of sections Dasylinum and Linum are otherwise
known to have diploid chromosome numbers of 2 n = 16
( Ray 1944 ). A chromosome number of 2 n = 57, reported from
L. catharticum by Martzenitzina ( 1927 ), seems highly ques-
tionable. Linum catharticum is a common and widespread
species in Europe and also is found in eastern Canada and
the north-eastern U.S. In North America, L. catharticum is
restricted to formerly glaciated regions and is regarded as a
relatively recent introduction to that continent ( Harris 1968 ).
Some of the features used to distinguish L. catharticum are
shared by some North American taxa, notably the opposite
leaves and annual habit, which can be found in Sclerolinon ,
some Hesperolinon , and some Linum .
The Yellow-flowered Clade: HESPEROLINON and SCLE-
ROLINON — With the caveat that more extensive sampling of
North American Linum may alter this result, we find a close
relationship between Hesperolinon and Sclerolinon , support-
ing the morphological affinities of the two genera. In both,
the number of carpels is reduced to three (most Hesperolinon )
or two (in two species of Hesperolinon , and in Sclerolinon ).
Sclerolinon digynum has some unique features, including a
bicarpellate fruit which splits into four 1-seeded nutlets at
maturity ( Rogers 1966 ), which further serve as the basis for
its segregation. Hesperolinon is distinguished by peculiar cup-
or flap-like appendages on the petal claws ( Sharsmith 1961 ).
The two genera are members of the clade that includes the
American and South African species of sect. Linopsis . They
appear to be associated with North American species, but
relationships within this clade are not well resolved or sup-
ported in our analyses. While Hesperolinon and Sclerolinon
are distinct morphologically from most linums, several
features may be indicative of their origin from North American
Linum . Notably, opposite or whorled leaves also occur in sev-
eral Linum species found in the southwestern United States
and in Mexico ( Rogers 1984 ). A study of floral pigments
also revealed some similarities between Hesperolinon and
Southwestern linums ( Rogers 1975 ), including the presence
of at least one compound (apigenin-c-glycoside) that is not
found elsewhere in Linum . Petal claw appendages similar
to those of Hesperolinon are also found in L. kingii and other
North American species of Linum ( Sharsmith 1961 ).
The Yellow-flowered Clade: CLIOCOCCA — The third New
World segregate, Cliococca , like Sclerolinon , is a monotypic genus
erected in recognition of its unique morphology ( Babington
1842 ; Rogers and Mildner 1971 ). Described by Lamarck in 1791,
Linum selaginoides Lam. was indeed a peculiar Linum , as illus-
trated by the fact that many herbaria contain vegetative speci-
mens of Cliococca selaginoides which, due to their small size and
small, densely-packed, needle-like leaves, were initially deter-
mined to be species of Selaginella P. Beauv. While Babington
(1842) recognized it as a genus, Planchon (1848) retained the
species in Linum , where it remained for 123 yr until segregated
by Rogers and Mildner ( 1971 ) during their revision of South
American Linum . The segregation was justified with characters
such as indehiscent capsules, small style branches, and sepals
that extend beyond the petals. Solitary flowers, as found in
Cliococca , occur rarely in the rest of Linoideae and, excepting
depauperate specimens of several species, have only been con-
sistently observed in Linum hudsonioides Planch. and L. imbri-
catum (Raf.) Shinners from North America. Rogers (1985) also
noted that while the pollen of Cliococca is tricolpate, a com-
mon character in Linum and found also in South American
linums, its exine sculpturing is unique. We find that Cliococca
is moderately supported as sister to the clade containing South
American species of Linum sect. Linopsis ( L. littorale , L. macraei,
L. prostratum, and L. oligophyllum ). This relationship rules out a
direct origin of Cliococca from any hugonioid tropical ancestor.
The Yellow-flowered Clade: RADIOLA — Radiola linoides was
also originally described by Linnaeus as a species of Linum (as
L. radiola L.). It has been generally maintained since as a separate
genus based on its obvious distinctiveness from most Old World
linums: it is a tiny annual with tetramerous flowers. However,
many species in this lineage are annual as well, and at least one
other species, L. keniense T. C. E. Fr. from East Africa (sect. Linopsis ,
not sampled here), is tetramerous, so that Radiola does not fall too
far outside the range of variation seen in its relatives. A conser-
vative taxonomic revision could return the segregated species to
Linum and would require no new nomenclatural combinations.
Alternatively, it could also be argued that one or more genera
could be recognized in the lineage that is sister to Radiola , but
additional phylogenetic work with much more extensive sam-
pling, as well as rigorous morphological comparison of all of the
species involved, are recommended before any such changes
are proposed.
Table 5. Maximum likelihood model calculation for ancestral state
optimization. For each optimization, the character states assigned to
Hugonia, Tirpitzia, and L. tenuifolium, respectively, are given in paren-
theses. 0 = homostylous, 1 = heterostylous.
Model -Ln L scores
MK1 Asymm. Diff. p
Optimization 1 (0, 0, 0) 29.127 27.623 1.504 0.08
Optimization 2 (1, 1, 1) 27.519 26.876 0.643 0.27

2009] MCDILL ET AL.: PHYLOGENY OF LINOIDEAE 399
Fig. 7. A) Stigmas of Hugonia sp. (Hugonioideae; drawn from Troll spirit collection 3726 (MJG)). Stigma (abaxial and adaxial view) and sepal of B)
Linum perenne (sect. Linum , L. perenne group, drawn from MJG 041531), C) L. usitatissimum (sect. Linum , MJG 040924), D) L. hirsutum (sect. Dasylinum , MJG
040935). E) Stigma and sepal of L. tenuifolium (sect. Linopsis subsect. Dichrolinum , MJG 041528), F) stigma (abaxial and adaxial view) of L. maritimum (sect.
Linopsis subsect. Halolinum , MJG 040937). G) Stigma (abaxial and adaxial view) and sepal of L. flavum (sect. Syllinum , MJG 041532), Scale = 0.5 mm.

400 SYSTEMATIC BOTANY [Volume 34
Biogeography— Subfamilies Hugonioideae and Linoideae
differ not only in morphology but also in their modern geo-
graphical distribution. Whereas Hugonioideae occur in
subtropical and tropical regions, Linoideae are restricted
mainly to the temperate regions of the world. The earliest-
diverging members of Linoideae are Anisadenia , Tirpitzia,
and Reinwardtia, occurring from south-central to Southeast
Asia, and the basal members of the blue- and yellow-flow-
ered lineages of Linum as sampled are Asian and European,
respectively. The distribution of Reinwardtia , Anisadenia, and
Tirpitzia (from the tropics of Southeast Asia into northern
India along the Himalayan Mountains), their similarities to
members of Hugonioideae in plesiomorphic characters such
as presence of petioles and foliate stipules, their leaf anatomy
( Van Welzen and Baas 1984 ) and habit (woody in Reinwardtia
and Tirpitzia , the latter arborescent), and the possible para-
phyly of Hugonioideae, all suggest an origin of Linoideae
from a tropical Hugonioid ancestor, perhaps in Southeast
Asia. Linoideae is estimated to have originated from such
an ancestor between 51 and 46 mya. Further sampling of the
remaining genera of Hugonioideae and collection of data
that can confidently resolve the relationships at the base of
Linoideae and within the Hugonioideae are required to test
more adequately this hypothesis of the time and area of origin
of Linoideae. More generally, Davis et al. (2005) postulated
an origin of all families of Malpighiales in tropical rain for-
ests during the Mid-Cretaceous (112–94 mya). Interestingly,
only few families of this large order have colonized temper-
ate regions. Besides Linaceae these include Euphorbiaceae,
Hypericaceae, Salicaceae, and Violaceae.
Eurasia has usually been considered the ancestral area for
Linum , due to the fact that all five of its sections, and the major-
ity of species in most of them, occur in the Mediterranean
region and western Asia ( Rogers 1982 ). Sections Syllinum and
Dasylinum are entirely confined to the Mediterranean area and
Eurasia, respectively. Linum catharticum (sect. Cathartolinum )
is known to occur widely in Europe. Its presence in North
America (eastern Canada) only in areas that were heavily
glaciated during the Pleistocene, and occupation of “weedy”
habitats there, led Rogers to conclude that its transatlantic dis-
junction is anthropogenic ( Rogers 1969 ). Linum s. l. (includ-
ing Cliococca , Hesperolinon , Radiola , and Sclerolinon ) may have
originated during the Eocene (46–42 mya) and split into the
yellow-flowered and blue-flowered clades shortly there-
after. In view of only low bootstrap support for the mono-
phyly of Linum s. l. in the separate analyses of the chloroplast
and nuclear data sets ( Fig. 4 ) it is also possible that these two
lineages originated independently from ancestors shared
with one or more of the Southeast Asian genera ( Anisadenia ,
Reinwardtia , and Tirpitzia ).
Biogeography of the Blue-flowered Clade— The blue-
flowered flaxes (sections Dasylinum and Linum ) seem to have
originated during the Oligocene (32–29 mya) in temperate Asia.
Today, sect. Dasylinum is restricted to Southwest Asia and the
Mediterranean region, whereas sect. Linum is distributed across
the entire temperate northern hemisphere and Australasia.
In section Linum , two species ( L. marginale , sampled here,
and L. monogynum ) are known to be native to Australia and
New Zealand ( Rogers and Xavier 1971 ), and three ( L. lewisii
sampled here, L. pratense Small, and L. rzedowskii Arreguín)
to North America ( Mosquin 1971 ; Rogers 1984 ; Arreguín-
Sánchez 1985 ). With regard to the presence of L. marginale in
Australia and of
L. monogynum in New Zealand, it could be
hypothesized that these two species migrated there from the
area of origin of Linoideae (Southeast Asia). In light of the
phylogeny, however, such scenario seems unlikely because
of the close relationship of L. marginale to European species.
Linum marginale and L. monogynum were considered closely
related to each other by Rogers ( 1984 ), and most similar
to Linum hologynum Rchb. from the Balkan Peninsula (not
included in our study). These three species all have fused
styles and pantoporate pollen grains, which are unusual
features in the blue-flowered flaxes and are found (sepa-
rately) in only two other species: fused styles in L. tmoleum
Boiss. of Turkey (also not included here), and pantoporate
pollen grains in L. stelleroides of China (which is shown
not to be a close relative of L. marginale ). Linum hologynum
and L. monogynum share a haploid chromosome number of
n = 42 ( Hair and Beuzenberg 1959 ; Petrova 1972 ; Mugnier
1983 ). Unfortunately L. hologynum could not be included in
our analyses to test the hypothesis that this species might
represent a link between European and Australasian taxa.
There is little evidence for how the disjunction between the
Australasian linums and their apparent European relatives
could have been established, although one or several long-
distance dispersal events are a possibility.
The North American blue-flowered taxa have been con-
sidered members of the Linum perenne group ( Ockendon
1968 ) based on morphological data. Our phylogenetic results
indeed show Linum lewisii nested within the L . perenne group.
Hypotheses for the movement of members of sect. Linum into
North America include migration from Asia via a Bering land
bridge or from Europe via a transatlantic route. The Bering
route is thought to have been suitable for the transfer of tem-
perate plants from the Mid to Late Tertiary until around 3.5
mya ( Wen 1999 ). North Atlantic land bridges are thought to
have been available somewhat earlier, during the Paleocene
and Eocene ( Tiffney 1985 ), and cut off by the Oligocene, 33
mya ( Tiffney and Manchester 2001 ). The diversification time
estimated for the L . perenne group as sampled, 3.8–3.3 mya,
would seem to rule out the transatlantic migration of an
ancestor of L. lewisii and the other North American blue-flow-
ered species, as the North Atlantic land bridge was not lon-
ger available by the time this lineage originated in Eurasia.
Linum lewisii is a widespread species in North America and
its distribution ranges from northern Mexico along the Rocky
Mountains into Alaska. More easterly situated populations in
Canada were considered the result of recent anthropogenic
introductions ( Mosquin 1971 ). Considering the essentially
western North American distribution of Linum lewisii , the
occurrence of representatives of the L. perenne group in east-
ern Siberia, and the age of this clade, migration of the ancestor
of L. lewisii into North America via Beringia would indeed be
possible, as earlier suggested by Harris (1968) . Colonization
of North America from Asia via a Bering land bridge has
been postulated for several genera, e.g. Aralia , Boykinia ,
Calycanthus , Caulophyllum , Gleditsia, and Magnolia (summa-
rized by Donoghue and Smith 2004 ). In summary, migration
into North America via the Bering Land Bridge appears the
most consistent explanation of our findings although long-
distance dispersal from Europe cannot be ruled out.
Biogeography of The Yellow-flowered Clade— The
most geographically dispersed lineage of Linum consists of
Linum sections Linopsis, Syllinum, and Cathartolinum , plus the
three New World segregate genera, i.e. the clade that is sister
to Radiola . This clade is estimated to have originated during

2009] MCDILL ET AL.: PHYLOGENY OF LINOIDEAE 401
the early Oligocene (36–32 mya) in Europe or Southwest Asia,
and subsequently spread into parts of the northern hemi-
sphere, Africa and South America. A European/Southwest
Asian origin for this clade has previously been proposed
based on morphological evidence, with species distributed
around the Mediterranean in particular exhibiting charac-
ters that have been regarded as plesiomorphic, such as free
styles and incomplete false septa ( Rogers 1982 ). In contrast
to the blue-flowered clade, the members of this yellow-flow-
ered clade are less represented in eastern Eurasia, and show
greater diversity in the Americas and in Africa.
The yellow-flowered clade appears to have been present in
North America by at least 20 mya, and as early as 36 mya.
This estimate is based on the rbcL topology, which places
Hesperolinon and Sclerolinon as the basal members of the lin-
eage that diversified in North America and from which the
South American and South African groups are derived. Given
uncertainties inherent in molecular dating estimates, and the
fact that sampling from additional North American Linum
may push this date earlier, we can not rule out migration of
this lineage via a North Atlantic land bridge, which is thought
to have been a viable route until the end of the Eocene, approx-
imately 33 mya ( Tiffney and Manchester 2001 ). Migration via
a Bering land bridge, the most likely colonization route in
sect. Linum , seems unlikely to have been a migration route for
yellow-flowered Linum, in view of the absence of the yellow-
flowered species in East Asia. Long-distance dispersal from
Europe or Asia, however, can not be ruled out as a possible
alternative to overland migration for the movement of yel-
low-flowered Linum into North America.
The placement of Sclerolinon and Hesperolinon as the basal
members of the American-South African lineage in the anal-
ysis of rbcL seems anomalous in light of our other analyses
and previous works. These genera have generally been con-
sidered as among the most “derived” and “specialized” in
the Linaceae, based on their unique morphologies and lim-
ited distributions in Western North America. Hesperolinon
essentially occupies open habitats in the chaparral of the
coastal ranges and Sierra Nevada foothills of California, par-
ticularly on outcrops of serpentine rock. Sclerolinon occurs in
vernally wet habitats in the mountains and basins from cen-
tral eastern California north to Washington and Idaho. To be
sure, rbcL does not provide strong support for relationships
within this lineage of Linum as sampled, and it is possible
that future sampling of North American Linum , particularly
those that inhabit the eastern part of the continent, will pro-
vide better understanding of the biogeographic history of
this lineage.
Regarding the origin of African Linum , Rogers ( 1981 , 1982 )
hypothesized that the species found in South Africa (includ-
ing the more widely distributed L. thunbergii Eckl. & Zeyh.)
are linked to European ancestors through East African spe-
cies such as L. volkensii . This hypothesis is contradicted by
our analyses, which indicate a dual origin of African Linum .
Linum volkensii , which occurs along the East African rift from
Ethiopia to Tanzania and is also reported from Cameroon,
does appear to be derived from European ancestry, and is
shown as sister to L. strictum , a species of Southern Europe, in
our analyses. The four South African linums sampled, how-
ever, have an independent origin, and are shown here either as
sister to all American species (CPL and ITS + CPL, Figs. 4B , 5 )
or as sister to the temperate South American clade ( rbcL , Fig. 2 ).
Accordingly, Linum in South Africa may represent part of an
amphiatlantic disjunction between the New and Old Worlds.
Rogers ( 1981 , 1982 ) identified several morphological simi-
larities between South African and American species, nota-
bly whorled leaves, which occur in L. quadrifolium of South
Africa and in several species of Central and North America.
Morphological similarities and the close phylogenetic relation-
ships between American and South African taxa does bring to
mind the question of whether vicariance may have played a
role in their diversification, but our estimate of the time of ini-
tial diversification of the yellow-flowered clade (36–32 mya)
indicates that the split between African and American species
may have occurred more than 60 million years after Africa
and South America were well separated ( McLoughlin 2001 ).
Long-distance dispersal from either Europe or America is
another process that should be considered as having a poten-
tial role in the origins of South African Linum . Similarly, rbcL
does not provide strong evidence for the origins of South
American Linum , although the phylogeny does suggest that
they, like the South African group, may share recent com-
mon ancestry with North American taxa. The ITS, CPL, and
combined ITS + CPL analyses likewise do not satisfactorily
resolve relationships within this lineage.
Evolution of Heterostyly— When considered in a macroevo-
lutionary context, heterostyly has generally been thought to be
the ancestral condition in Linum ( Rogers 1982 ), based on two
sources of evidence. First, heterostyly is found in four of the five
sections of the genus but only in Old World species, primarily
those found in the circum-Mediterranean region of Europe and
western Asia, which were considered to be the ancestral area
in which Linum diversified. Second, heterostyly was known
to occur in other genera of Linaceae, namely Reinwardtia and
several members of the Hugonioideae. While this is somewhat
circumstantial evidence for determining the “polarity” of the
heterostyly/homostyly transitions in Linum , it is consistent
with opinions on the evolutionary development and inheri-
tance of heterostyly, and heterostyly has thus generally been
considered to have been “lost” in homostylous Linum .
Distyly in Linum (as in most other genera) involves the
combination of diallelic sporophytic self-incompatibility with
dimorphisms in several morphological characters, including
style length, anther height, and pollen size ( Barrett et al. 2000 ),
controlled by at least three different genes ( Dowrick 1956 ;
Richards 1986 ) that comprise the heterostyly “supergene.” It
may therefore be reasonable to conclude that the evolution of
heterostyly in a lineage is “difficult” due to the requirement
that appropriate mutations in all of the necessary characters
arise with appropriate dominance relationships and become
linked in a heritable unit. The loss of heterostyly, on the other
hand, might easily result from random recombination within
that unit, which has been concluded to have occurred in sev-
eral taxa, including Primula ( Mast et al. 2006 ) and Amsinckia
( Schoen et al. 1997 ). Such “breakdown” of heterostyly, possibly
due to recombination, was also described in Linum ( Nicholls
1985 ), but without a phylogenetic context. The breakdown
of heterostyly due to recombination is also probable in the
L. perenne complex ( L. leonii Schultz; Repplinger and Kadereit,
unpubl. results).
The only previous study of the evolution of heterostyly in
Linum within a phylogenetic framework was by Armbruster
et al. (2006) , who concluded that heterostyly evolved several
times independently within the lineages of Linum (parsimony
reconstruction based on a phylogeny of ITS sampled from
16 Old World species (nine heterostylous) primarily from

402 SYSTEMATIC BOTANY [Volume 34
sect. Linopsis , rooted with R. linoides ). Our results show that
the determination of heterostyly or homostyly as the ances-
tral state in Linum , using MP and ML techniques for estimat-
ing ancestral states on phylogenies, cannot be made with a
high level of confidence even with the greater taxon sampling
used in this study. The ancestral states reconstructed at the
basal nodes of Linoideae using parsimony depend greatly on
how the outgroup and polymorphic taxa within the ingroup
are scored. When the three polymorphic taxa sampled here
are scored as homostylous, then homostyly is reconstructed
at the base of Linoideae. When they are scored as heterosty-
lous, then heterostyly becomes the most parsimonious state
assigned at the base of Linoideae. The description of a new
distylous species of Tirpitzia by Suksathan and Larsen (2006)
requires that heterostyly be considered the potential ancestral
state for that genus. If Tirpitzia were simply scored as homo-
stylous, as it would be if only the two homostylous species
were known, then the parsimony ancestral state for Linoideae
is reconstructed as homostylous (when Hugonia is scored as
homostylous) or equivocal (when Hugonia is scored as hetero-
stylous, results not shown). It therefore appears that the esti-
mation of the breeding system ancestral state in Linum and
Linoideae is tied to the questions of phylogenetic relationships
and ancestral state of species of Tirpitzia and Hugonioideae.
In Hugonioideae only few detailed studies of flower mor-
phology ( Lloyd et al. 1990 ; Thompson et al. 1996 ) have been
published, and it is not well known how many species of this
subfamily are heterostylous.
While we cannot answer the question of the ancestral state
in Linum , our results do provide strong support for transi-
tions in breeding system at three places on our phylogeny:
the gain of heterostyly in the South African L. comptonii from
unequivocal homostylous ancestry, and independent losses
of heterostyly in L. nodiflorum and L. lewisii , which are shown
to be derived from well-supported heterostylous ancestry.
Coupled with experimental evidence for the loss of hetero-
styly in L. tenuifolium ( Nicholls 1985 ) and L. leonii (Repplinger
and Kadereit, unpubl. results), our results support the notion
that the breeding system is somewhat labile in the Linaceae,
as concluded before by Armbruster et al. (2006) .
Reconstruction of ancestral character states is prone to sev-
eral types of uncertainty, stemming from the limitations of
taxon sampling, ancestral state reconstruction method used,
and phylogenetic tree uncertainty, and the advantages and
disadvantages of different reconstruction methods have been
summarized by Cunningham (1999) . Both likelihood and par-
simony methods of ancestral state reconstruction are sensitive
to the relative frequencies of the character states in question.
The proportion of heterostylous taxa sampled in our anal-
yses (17 out of 44 total taxa of Linoideae, or about 38%) is
slightly higher than the proportion of Linoideae known to
be heterostylous (approximately 63 out of a total of 209 spe-
cies, or 30%). The taxon sampling and phylogenetic results
we present provide a framework to which additional data
can be added; heterostylous and homostylous taxa added in
future analyses will not join the phylogeny randomly, but will
likely fall into lineages for which we find strong support. As
indicated in Table 1 , sections Syllinum and Dasylinum are pre-
dominantly heterostylous and collectively contain the major-
ity of unsampled heterostylous species. Additional sampling
from these sections, which we find to be monophyletic in the
molecular phylogenies and which have good morphological
support as well, will probably add more heterostylous ter-
minals within those two clades. Additional losses or gains of
heterostyly within those sections may be reconstructed using
parsimony, depending on the species relationships discov-
ered. In section Linopsis , the vast majority of species (ca. 84)
are homostylous, and of the six known heterostylous species,
five are sampled in our analyses. Sampling of L. heterostylum
from South Africa, the only heterostylous species of Linopsis
not sampled here, has the potential to reveal an additional
gain of heterostyly, depending on the relationship of that spe-
cies to L. comptonii . Of particular importance in future work
on this topic will be section Linum , which includes 17 unsam-
pled heterostylous species and a similar number of unsam-
pled homostylous taxa.
Additional sampling from sections Linopsis , Dasylinum , and
Syllinum will affect the ancestral state reconstruction using
maximum likelihood as well. Adding large numbers of taxa
from lineages in which breeding system does not frequently
change, such as the large, homostylous American and South
African species groups of sect. Linopsis , and the primarily
heterostylous sections Dasylinum and Syllinum , would tend
to decrease the rate parameters estimated by the likelihood
reconstruction method used here by increasing the branch
length over which the states do not change. This may have
several consequences for the ancestral state reconstruction,
perhaps altering the model selected or increasing the num-
ber of internal nodes at which ancestral states can be esti-
mated with significant support. A final understanding of the
evolution of heterostyly in this group will necessarily require
complete taxon sampling; addition of taxa may alter the tree
topology and distribution of heterostyly/homostyly thereon,
alter likelihood model parameters, and change the character
states reconstructed at specific nodes. The previous discus-
sion of uncertainties in the phylogeny of Linoideae presented
here may also cast some doubt on our reconstructions of the
evolutionary history of heterostyly in this lineage. However,
most of the nodes that are poorly resolved or weakly sup-
ported phylogenetically occur in lineages that appear to be
essentially entirely homostylous (e.g. the New World and
African clades). The areas of the topology most concerned
with the evolution of heterostyly are fairly well resolved
and well supported. Although we have not used them here
because we do not believe tree uncertainty to be significantly
affecting our limited conclusions, Bayesian methods of inte-
grating ancestral state reconstructions over a large sample of
tree topologies can be used to assess the effect of phylogenetic
uncertainty ( Pagel et al. 2004 ; Vanderpoorten and Goffinet
2006 ). In addition to their sensitivity to taxon sampling, it has
been shown that both parsimony and likelihood-based ances-
tral state reconstructions can strongly support incorrect con-
clusions about the evolution of breeding systems when the
underlying molecular genetics of those systems are not taken
into account ( Igic et al. 2006 ).
We have here provided a basic overview of the phylog-
eny of Linaceae subf. Linoideae and Linum , showing support
for several well-defined lineages including the monophyl-
etic lineages of blue-flowered and yellow-flowered linums.
Sections Syllinum and Dasylinum are natural lineages well
supported by morphological and molecular data, but they are
nested within sections Linopsis and Linum , respectively. Also
nested within Linum are the segregate genera Hesperolinon ,
Sclerolinon , Cliococca , and Radiola , which should be returned
to Linum in a conservative taxonomic revision. We find sup-
port for the traditional view that Linum initially diversi-

2009] MCDILL ET AL.: PHYLOGENY OF LINOIDEAE 403
fied in Europe and western Asia. Finally, with regard to the
evolution of heterostyly, we use parsimony and likelihood-
based estimates of ancestral states in concert to show that,
while neither heterostyly nor homostyly can yet be ruled out
as the ancestral state in Linoideae and Linum , breeding sys-
tem appears to have transitioned on at least three occasions
in Linum , with both losses and an independent gain of het-
erostyly being confidently reconstructed. A full understand-
ing of the phylogeny and evolution of breeding systems in
Linum will require not only additional (hopefully complete)
taxonomic sampling, but better knowledge of the underlying
genetics of self incompatibility and other characters associ-
ated with heterostyly in the genus.
Acknowledgments We thank the curators and staff of the herbaria
at The University of Texas at Austin (TEX), the Missouri Botanical Garden
(MO), the Johannes Gutenberg-Universität Mainz (MJG), and Harvard
University (GH) for allowing us to sample from specimens of many of
the taxa included in this study. We also thank Iraj Mehregan (Mainz),
Natalie Schmalz (Mainz), Marcus Quint (Mainz), and Matthias Kropf
(Wien) for providing leaf material, and Li-Bing Zhang (St. Louis) and
Jim F. Smith (Boise) for sending material from the Chinese Academy of
Science, Beijing herbarium (PE) and the Snake River Plains Herbarium at
Boise State University (SRP), respectively. Dr. Ken Wurdack generously
provided rbcL sequences for Anisadenia pubescens and Roucheria calophylla .
Leaf material from Reinwardtia indica in cultivation at the Conservatory of
Flowers, San Francisco (California, U.S.A.), was obtained with the help
of Jim Henrich. Ghazaleh Moayedi and Sandra Sipe assisted with labora-
tory work at The University of Texas at Austin. Doris Franke (Mainz) is
gratefully acknowledged for the preparation of Fig. 7 . Funding for work
done at The University of Texas came in part from a National Science
Foundation Dissertation Improvement Grant (DEB-0508802) to JRM
and BBS, and from the C. L. Lundell Chair of Systematic Botany (BBS).
We also thank the two anonymous reviewers for their comments and
suggestions.
Literature Cited
Allaby, R. G. , G. W. Peterson , D. A. Merriwether, and Y. B. Fu . 2005 .
Evidence of the domestication history of flax ( Linum usitatissimum )
from genetic diversity of the sad2 locus . Theoretical and Applied
Genetics 112 : 58 – 65 .
APG II . 2003 . An update of the Angiosperm Phylogeny Group classifica-
tion for the orders and families of flowering plants: APG II . Botanical
Journal of the Linnean Society 141 : 399 – 436 .
Armbruster, W. S. , R. Pérez-Barrales , J. Arroyo , M. E. Edwards and
P. Vargas . 2006 . Three-dimensional reciprocity of floral morphs in
wild flax ( Linum suffruticosum ): a new twist on heterostyly . The New
Phytologist 171 : 581 – 590 .
Arreguín-Sánchez, M. 1985 . Una nueva especie de Linum (Linaceae) del
valle de México . Phytologia 57 : 262 – 263 .
Babington, C. C. 1842 . Description of a new genus of Lineae . Transactions
of the Linnaean Society of London 19 : 33 – 34 .
Bahadur, B. , N. P. Reddy, and N. Ramaswamy . 1996 . Heterostyly and pol-
len dimorphism in Reinwardtia indica Dum. (Linaceae) . Journal of
Palynology 32 : 67 – 77 .
Barrett, S. C. H. , L. K. Jesson, and A. M. Baker . 2000 . The evolution and
function of stylar polymorphisms in flowering plants . Annals of
Botany 85 suppl. A : 253 – 265 .
Barrett, S. C. H. and L. D. Harder . 2005 . The evolution of polymorphic
sexual systems in daffodils ( Narcissus ) . The New Phytologist 165 :
45 – 53 .
Chase, M. W. , S. Zmarzty , M. D. Lledó , K. J. Wurdack , S. M. Swensen,
and M. F. Fay . 2002 . When in doubt, put in Flacourtiaceae: a molecu-
lar phylogenetic analysis based on plastid rbcL DNA sequences . Kew
Bulletin 57 : 141 – 181 .
Conti, E. , E. Suring , D. Boyd , J. Jorgensen , J. Grant, and S. Kelso . 2000 .
Phylogenetic relationships and character evolution in Primula L.: the
usefulness of ITS sequence data . Plant Biosystems 134 : 385 – 392 .
Coates, D. and C. A. Cullis . 1987 . Chloroplast DNA variability among
Linum species . American Journal of Botany 74 : 260 – 268 .
Cronquist, A. 1988 . The evolution and classification of flowering plants . Ed. 2 .
Bronx : The New York Botanical Garden .
Cunningham, C. W. 1999 . Some limitations of ancestral character-state
reconstruction when testing evolutionary hypotheses . Systematic
Biology 48 : 665 – 674 .
Darwin, C. 1864 . On the existence of two forms, and on their reciprocal
sexual relations, in several species of the genus Linum. Journal of the
Linnean Society, London . Botany 7 : 69 – 83 .
Darwin, C. 1877 . The different forms of flowers on plants of the same species .
London : Murray .
Davis, C. C. and M. W. Chase . 2004 . Elatinaceae are sister to Malpighiaceae;
Peridiscaceae belong to Saxifragales . American Journal of Botany 91 :
262 – 273 .
Davis, C. C. , C. O. Webb , K. J. Wurdack , C. A. Jaramillo, and M. J.
Donoghue . 2005 . Explosive radiation of Malpighiales supports a
mid-Cretaceous origin of modern tropical rain forests . American
Naturalist 165 : E36 – E65 .
Davis, P. H. 1967 . Linaceae . Pp. 425 – 450 in Flora of Turkey and the Aegean
Islands . ed. P. H. Davis Edinburgh : Edinburgh University Press .
Diederichsen, A. and K. Richards . 2003 . Cultivated flax and the genus
Linum L . Pp. 22 – 54 in Flax: the genus Linum , eds. A. D. Muir and N. D.
Westcott . London : Taylor and Francis .
Donoghue, M. J. and S. A. Smith . 2004 . Patterns in the assembly of temper-
ate forests around the Northern Hemisphere . Philosophical Transactions
of the Royal Society of London. Series B, Biological Sciences 359 :
1633 – 1644 .
Dowrick, V. P. J. 1956 . Heterostyly and homostyly in Primula obconica .
Heredity 10 : 219 – 236 .
Dressler, S. , M. Repplinger, and C. Bayer . I n press . Linaceae . in The families
and genera of vascular plants vol. 10 , ed. K. Kubitzki Berlin : Springer
Verlag .
Edgar, R. C. 2004 . MUSCLE: multiple sequence alignment with high accu-
racy and high throughput . Nucleic Acids Research 32 : 1792 – 1797 .
Ellis, J. G. , G. J. Lawrence , J. E. Luck, and P. N. Dodds . 1999 . Identification
of regions in alleles of the flax rust resistance gene L that deter-
mine differences in gene-for-gene specificity . The Plant Cell 1 1 :
495 – 506 .
Farris, J. S. , M. Kallersjö , A. G. Kluge , and C. Bult . 1995 . Constructing a sig-
nificance test for incongruence . Systematic Biology 44 : 570 – 572 .
Fu, Y.-B. , G. Peterson , A. Diederichsen, and K. W. Richards . 2002a . Genetic
diversity within a range of cultivars and landraces of flax ( Linum
usitatissimum L.) as revealed by RAPDs . Genetic Resources and Crop
Evolution 49 : 167 – 174 .
Fu, Y.-B. , G. Peterson , A. Diederichsen, and K. W. Richards . 2002b . RAPD
analysis of genetic relationships of seven flax species in the genus
Linum L . Genetic Resources and Crop Evolution 49 : 253 – 259 .
GeneCodes Corp . 2005 . Sequencher 4.5: software for DNA sequence assembly .
Ann Arbor : Gene Codes Inc .
Hair, J. B. and E. J. Beuzenberg . 1959 . Contributions to a chromosome
atlas of the New Zealand flora . New Zealand Journal of Science
2 : 148 – 156 .
Hallier, H. 1921 . Beiträge zur Kenntnis der Linaceae (DC. 1819) Dumort .
Beihefte zum Botanischen Centralblatt Abt. II. 39 : 1 – 178 .
Harris, B. D. 1968 . Chromosome numbers and evolution in North American
species of Linum . American Journal of Botany 55 : 1197 – 1204 .
Hillman, G. 1975 . The plant remains from Tell Abu Hureyra: A prelimi-
nary report . Proceedings of the Prehistoric Society 41 : 70 – 73 .
Hohmann, S. , J. W. Kadereit, and G. Kadereit . 2006 . Understanding
Mediterranean-Californian disjunctions: molecular evidence from
Chenopodiaceae-Betioideae . Taxon 55 : 67 – 78 .
Huelsenbeck, J. P. and F. Ronquist . 2001 . MrBayes: Bayesian inference of
phylogenetic trees . Bioinformatics 17 : 754 – 755 .
Igic, B. , L. Bohs, and J. R. Kohn . 2006 . Ancient polymorphism reveals uni-
directional breeding system shifts . Proceedings of the National Academy
of Sciences USA 103 : 1359 – 1363 .
Jansen, R. K. 1992 . Current Research: ndhF PCR and Sequencing Primers .
Plant Molecular Evolution Newsletter 2 : 13 .
Jardim, A. 1999 . A revision of Roucheria Planch. & Benth. (Hugoniaceae) .
M. S. thesis. Saint Louis, Missouri : University of Missouri-St. Louis .
Judd, A. 1995 . Flax-Some historical considerations . Pp. 1 – 10 in Flaxseed in
human nutrition , eds. S. C. Cunnane and L. U. Thompson . Champaign,
Illinois : AOCS Press .
Kohn, J. R. , S. W. Graham , B. Morton , J. J. Doyle, and S. C. H. Barrett .
1996 . Reconstruction of the evolution of reproductive characters in
Pontederiaceae using phylogenetic evidence from chloroplast DNA
restriction-site variation . Evolution 50 : 1454 – 1469 .
Link, D. A. 1989 . Die Nektarien der Geraniales. Morphologie, Anatomie, Histologie,
Blütenökologische Bedeutung und Konsequenzen für die Systematik . Ph. D.
thesis. Mainz, Germany : Johannes Gutenberg-University .

404 SYSTEMATIC BOTANY [Volume 34
Lloyd, D. G. , C. J. Webb, and R. Dulberger . 1990 . Heterostyly in species
of Narcissus (Amaryllidaceae) and Hugonia (Linaceae) and other dis-
puted cases . Plant Systematics and Evolution 172 : 215 – 227 .
Loockerman, D. J. and R. K. Jansen . 1996 . The use of herbarium material
for DNA studies . Pp. 205 – 220 in Sampling the green world , ed. S. H .
Sohmer New York : Columbia University Press .
Maddison, D. R. and W. P. Maddison . 2005 . MacClade 4.08. Sunderland :
Sinauer Associates .
Maddison, W. P. and D. R. Maddison . 2006 . Mesquite: a modular system for
evolutionary analysis . Version 1.12 at http://mesquiteproject.org .
Martzenitzina, K. K. 1927 . The chromosomes of some species of the genus
Linum L. Bulletin of Applied Botany, of Genetics and Plant Breeding 17 :
253 – 264 .
Mast, A. R. , D. M. S. Feller , S. Kelso, and E. Conti . 2004 . Buzz-pollinated
Dodecatheon originated from within the heterostylous Primula subge-
nus Auriculastrum (Primulaceae): a seven-region cpDNA phylogeny
and its implications for floral evolution . American Journal of Botany
91 : 926 – 942 .
Mast, A. R. , S. Kelso, and E. Conti . 2006 . Are any primroses ( Primula ) prim-
itively monomorphic? The New Phytologist 171 : 605 – 616 .
McLoughlin, S. 2001 . The breakup history of Gondwana and its impact
on pre-Cenozoic floristic provincialism . Australian Journal of Botany
49 : 271 – 300 .
Mildner, R. A. and C. M. Rogers . 1978 . Revision of the native South
American species of Linum (Linaceae) . Phytologia 39 : 343 – 390 .
Mosquin, T. 1971 . Biosystematic studies in North American species of
Linum , section Adenolinum (Linaceae) . Canadian Journal of Botany 49 :
1379 – 1388 .
Mugnier, C. 1983 . Linaceae. IOPB Chromosome number reports LXXIX .
Taxon 32 : 320 – 324 .
Muir, G. and C. Schlötterer . 1999 . Chapter 11. Limitations to the phylo-
genetic use of ITS sequences in closely related species and pop-
ulations: a case study in Quercus petraea (Matt.) Liebl . in Which
DNA marker for which purpose? ed. E. M. Gillet Hamburg : Wiede-
busch . http://webdoc.sub.gwdg.de/ebook/y/1999/whichmarker/
index.htm .
Nicholls, M. S. 1985 . The evolutionary breakdown of distyly in Linum
tenuifolium (Linaceae) . Plant Systematics and Evolution 150 : 291 – 302 .
Ockendon, D. J. 1968 . Biosystematic studies in the Linum perenne group .
The New Phytologist 67 : 787–813 .
Ockendon, D. J. and S. M. Walters . 1968 . Linaceae . Pp. 589 – 603 in Flora
Europaea vol. 2: Rosaceae to Umbelliferae , eds. T. G. , Tutin, V. H.
Heywood , N. A. Burges , D. M. Moore , D.H. Valentine , S. M. Walters,
and D. A. Webb . Cambridge : University of Cambridge Press .
Oh, T. J. , M. Gorman, and C. A. Cullis . 2000 . RFLP and RAPD mapping
in flax ( Linum usitatissimum ) . Theoretical and Applied Genetics 101 :
590 – 593 .
Olmstead, R. G. , H. J. Michaels , K. M. Scott, and J. D. Palmer . 1992 . Monophyly of
the Asteridae and identification of their major lineages inferred from DNA
sequences of rbcL . Annals of the Missouri Botanical Garden 79 : 249 – 265 .
Pagel, M. A. , A. Meade and D. Barker . 2004 . Bayesian estimation of ances-
tral character states on phylogenies . Systematic Biology 53 : 673 – 684 .
Palumbi, S. 1996 . Nucleic Acids II: The polymerase chain reaction . Pp.
407 – 514 in Molecular systematics , Ed. 2 . eds. D. M. Hillis, C. Moritz
and B. K. Mable . Sunderland : Sinauer Associates .
Pengilly, N. L. 2003 . Traditional food and medicinal uses of flaxseed . Pp.
252 – 267 in Flax: the genus Linum , eds. A. D. Muir and N. D. Westcott .
London : Taylor and Francis .
Pérez, R. , P. Vargas, and J. Arroyo . 2003 . Convergent evolution of flower
polymorphism in Narcissus (Amaryllidaceae) . The New Phytologist
161 : 235 – 252 .
Petrova, A. V. 1972 . IOPB Chromosome number reports XXXV . Taxon 21 :
161 – 166 .
Planchon, J. E. 1847 . Sur la famille des Linées . The London Journal of Botany
6 : 588 – 603 .
Planchon, J. E. 1848 . Sur la famille des Linées . The London Journal of Botany
7 : 165 – 186 , 473–501, 507–528.
Posada, D. and K. A. Crandall . 1998 . Modeltest: testing the model of DNA
substitution . Bioinformatics 14 : 817 – 818 .
Ray, C. 1944 . Cytological studies in the flax genus, Linum . American Journal
of Botany 31 : 241 – 248 .
Rechinger, K. H. 1974 . Linaceae . Pp. 1 – 19 in Flora Iranica: Flora des iranischen
Hochlandes und der umrahmenden Gebirge vol. 106 , ed. K. H. Rechinger
Graz : Akademische Druck- und Verlagsanstalt .
Reiche, K. 1897 . Linaceae . Pp. 27 – 35 in Die natürlichen Pflanzenfamilien 3 .
Teil, Abt. 3, eds. H. G. A. Engler and K. A. E. Prantl . Leipzig :
Engelmann .
Richards, A. J. 1986 . Plant breeding systems . London : Chapman and Hall .
Rickard-Bon, S. E. and L. U. Thompson . 2003 . The role of flaxseed lignans
in hormone-dependent and independent cancer . Pp. 181 – 203 in Flax:
The genus Linum , eds. A. D. Muir and N. D. Westcott . London : Taylor
and Francis .
Rogers, C. M. 1963 . Yellow flowered species of Linum in eastern North
America . Brittonia 15 : 97 – 122 .
Rogers, C. M. 1966 . Sclerolinon , a new genus in the Linaceae . Madroño 23 :
153 – 159 .
Rogers, C. M. 1969 . Relationships of the North American species of Linum
(flax) . Bulletin of the Torrey Botanical Club 96 : 176 – 190 .
Rogers, C. M. 1975 . Relationships of Hesperolinon and Linum (Linaceae) .
Madroño 23 : 153 – 159 .
Rogers, C. M. 1981 . A revision of the genus Linum in southern Africa .
Nordic Journal of Botany 1 : 711 – 722 .
Rogers, C. M. 1982 . The systematics of Linum sect. Linopsis (Linaceae) .
Plant Systematics and Evolution 140 : 225 – 234 .
Rogers, C. M. 1984 . A further note on the relationships of the European
Linum hologynum and the Australian species of Linum (Linaceae) .
Plant Systematics and Evolution 147 : 327 – 328 .
Rogers, C. M. 1985 . Pollen morphology of the monotypic genus Cliococca
(Linaceae) . Grana 24 : 121 – 123 .
Rogers, C. M. and R. Mildner . 1971 . The reevaluation of the genus Cliococca
(Linaceae) of South America . Rhodora 73 : 560 – 565 .
Rogers, C. M. and K. S. Xavier . 1971 . Pollen morphology as an aid in deter-
mining relationships among some widely separated Old World spe-
cies of Linum . Grana 11 : 55 – 57 .
Sanderson, M. J. 1997 . A nonparametric approach to estimating diver-
gence times in the absence of rate constancy . Molecular Biology and
Evolution 14 : 1218 – 1231 .
Sanderson, M. J. 2002 . Estimating absolute rates of molecular evolution
and divergence times: a penalized likelihood approach . Molecular
Biology and Evolution 19 : 101 – 109 .
Savolainen, V. , M. W. Chase , S. B. Hoot , C. M. Morton , D. E. Soltis , C. Bayer ,
M. F. Fay , A. Y. de Bruijn , S. Sullivan , and Y. Qiu . 2000 . Phylogenetics
of flowering plants based on a combined analysis of plastid atpβ and
rbcL gene sequences . Systematic Botany 49 : 306 – 362 .
Schluter, D. , T. Price , A. O. Mooers , and D. Ludwig . 1997 . Likelihood of
Ancestor States in Adaptive Radiation . Evolution 51 : 1699 – 1711 .
Schoen, D. J. , M. O. Johnston , A. M. L’Heureux, and J. V. Marsolais . 1997 .
Evolutionary history of the mating system in Amsinckia (Boraginaceae) .
Evolution 51 : 1090 – 1099 .
Sharsmith, H. K. 1961 . The genus Hesperolinon . University of California
Publications in Botany 32 : 235 – 314 .
Small, J. K. 1907 . Linaceae . Pp. 67 – 87 in North American Flora 25 (1) . New
York : New York Botanical Garden .
Spielmeyer, W. , A. G. Green , D. Bittisnich , N. Mendham, and E. S. Lagudah .
1998 . Idendification of quantitative trait loci contributing to Fusarium
wild resistance on an AFLP linkage map of flax ( Linum usitatissimum ) .
Theoretical and Applied Genetics 97 : 633 – 641 .
Steele, K. P. and R. Vilgalys . 1994 . Phylogenetic analyses of Polemoniaceae
using nucleotide sequences of the plastid gene matK . Systematic
Botany 19 : 126 – 142 .
Suksathan, P. and K. Larsen . 2006 . A new species of Tirpitzia (Linaceae)
from Thailand . Thai Forest Bulletin 34 : 201 – 205 .
Swofford, D. L. 2002 . PAUP*, Phylogenetic analysis using parsimony (*and
other methods) . Version 4.0. Sunderland : Sinauer Associates .
Taberlet, P. , L. Gielly , G. Pautou , and J. Bouvet . 1991 . Universal primers for
amplification of three non-coding regions of chloroplast DNA . Plant
Molecular Biology 17 : 1105 – 1109 .
Tavaré, S. 1986 . Some probabilistic and statistical problems in the analysis of
DNA sequences . Lectures on Mathematics in the Life Sciences 17 : 57 – 86 .
Thompson, J. D. , T. Pailler , D. Strasberg, and D. Manicacci . 1996 . Tristyly in
the endangered Mascarene island endemic Hugonia serrata (Linaceae) .
American Journal of Botany 83 : 1160 – 1167 .
Tiffney, B. H. 1985 . The Eocene North Atlantic land bridge and its impor-
tance in Tertiary and modern phytogeography of the Northern
Hemisphere . Journal of the Arnold Arboretum 66 : 243 – 273 .
Tiffney, B. H. and S. R. Manchester . 2001 . The use of geological and pale-
ontological evidence in evaluating plant phylogeographic hypothe-
ses in the Northern Hemisphere Tertiary . International Journal of Plant
Sciences 162 : S3 – S17 .
Vaisey-Genser, M. and D. H. Morris . 2003 . History of the cultivation and
uses of flaxseed . Pp. 1 – 21 in Flax: the genus Linum , eds. A.D. Muir and
N.D. Westcott . London : Taylor and Francis .
Vanderpoorten, A. and B. Goffinet . 2006 . Mapping uncertainty and phy-
logenetic uncertainty in ancestral character state reconstruction: an

2009] MCDILL ET AL.: PHYLOGENY OF LINOIDEAE 405
example in the moss genus Brachytheciastrum . Systematic Biology 55 :
957 – 971 .
Van Hooren, A. M. N. and H. P. Nooteboom . 1984 . Linaceae and Ctenol-
ophonaceae especially of Malesia, with notes on their demarcation
and relationships with Ixonanthaceae . Blumea 29 : 547 – 563 .
Van Welzen, P. C. and P. Baas . 1984 . A leaf-anatomical contribution to the
classification of the Linaceae complex . Blumea 29 : 453 – 479 .
Wen, J. 1999 . Evolution of eastern Asian and eastern North American dis-
junct distributions in flowering plants . Annual Review of Ecology and
Systematics 30 : 421 – 455 .
White, T. J. , T. Burns , S. Lee, and, J. Taylor . 1990 . Amplification and direct
sequencing of fungal ribosomal RNA genes for phylogenetics .
Pp. 315 – 322 in PCR protocols: a guide to methods and applications , eds.
M. Innis, D. Gelfand , J. Sninsky and T. White . San Diego : Academic
Press .
Winkler, H. 1931 . Linaceae . Pp. 82 – 130 in Die natürlichen Pflanzenfamilien
19a , eds. H. G. A. Engler and K. A. E. Prantl . Leipzig : Engelmann .
Yuzepchuk, S. V. 1974 . Linaceae . Pp. 67 – 112 in Flora of the U.S.S.R. 14 , eds.
B. K. Shishkin and E. G. Bobrov . Leningrad : Botanical Institute of the
Academy of Sciences of the USSR . Translated by the Israel Program
for Scientific Translations, Jerusalem.
Zohary, D. and M. Hopf . 2000 . Domestication of plants in the Old World .
Oxford : Oxford University Press .
Zohary, M. 1972 . Flora Palaestina Part Two: Platanaceae to Umbelliferae .
Jerusalem : The Israel Academy of Sciences and Humanities .
Appendix 1. Sampled taxa, presence of heterostyly/homostyly (given
only for Linaceae taxa used in ancestral state reconstruction), voucher
specimen information (collection locality, collection number, herbarium
(accession number provided for specimens at MJG), and GenBank acces-
sion numbers ( rbcL , ITS, ndhF , trnK 3 ′ intron, trnL-F ); — = sequence not
obtained. Specimens sampled for rbcL only are indicated by †. Previously
published rbcL sequences downloaded for inclusion in our analyses are
indicated by ‡. Family is noted for non-Linaceae taxa included in the rbcL
analyses.
Anisadenia pubescens Griff., Homostylous, China, Bartho lomew et al. 1984
Sino-American Bot. Exped. No. 1011 (GH), FJ169557, FJ169513, FJ160772, —,
FJ160856. Austrobuxus megacarpus P.I. Forst. (Euphorbiaceae), AY380343‡.
Balanops vieillardi Baill. (Balanopaceae), AF089760‡. Cliococca selaginoides
(Lam.) C.M.Rogers & Mildner, Homostylous, Brazil, A. Krapovickas & C.L.
Cristóbal 34169 (MO), FJ169558, FJ169540, FJ160774, —, FJ160858. Couepia
robusta Huber (Chrysobalanaceae), AF089757‡. Durandea pentagyna
K.Schum., A, AY788173‡. B, Papua New Guinea, W. Takeuchi 4523 (MO)†,
FJ169559. Hesperolinon drymarioides Small, California, Dibble and Griggs 14
(MO)†, U.S.A., FJ169560. Hesperolinon micranthum Small, Homostylous,
California,
van der Werff and Clark 4574 (MO), U.S.A., FJ169561, FJ169542,
FJ160775, FJ160818, FJ160859. Hugonia busseana Engl., Homostylous (some
other species of Hugonia are Heterostylous), Malawi, D.A. Link (MJG
MW0029), FJ169512, FJ169512, FJ160773, FJ160817, FJ160857. Hugonia cas-
taneifolia Engl., Kenya, Gilbert and Kuchar 5859 (MO)†, FJ169562. Hugonia
gabunensis Engl., Gabon, Lee White 0908 (MO)†, FJ169563. Linum africa-
num L., Homostylous, South Africa, W.J. Hanekom 2630 (MO), FJ169564,
FJ169551, FJ160777, FJ160820, FJ160861. Linum album Kotschy ex Boiss.,
Heterostylous, Iran, I. Meheregan (MJG 040948), —, FJ169547, FJ160792,
FJ160831, FJ160876. Linum arboreum L., Heterostylous, Turkey, U. Hecker
(MJG TR0191), FJ169565, FJ169537, FJ160793, FJ160832, FJ160877. Linum
austriacum L., Heterostylous, (MJG 040933), FJ169566, FJ169522, FJ160799,
FJ160838, FJ160883. Linum berlandieri Hook., Texas, U.S.A., B.L. Turner
21-427 (TEX)†, FJ169567. Linum bienne Mill., Homostylous, Italy, (MJG
040947), FJ169568, FJ169527, FJ160797, FJ160836, FJ160881. Linum campan-
ulatum L., (MJG 040928)†, FJ169569. Linum catharticum L., Homostylous,
Germany, (MJG 040944), FJ169570, FJ169533, FJ160796, FJ160835, FJ160880.
Linum comptonii C.M.Rogers, Heterostylous, South Africa, Elsie Esterhuysen
35882 (MO), FJ169572, FJ169550, FJ160778, FJ160821, FJ160862. Linum
decumbens Desf. (MJG 040932)†, FJ169573. Linum esterhuysenae C.M.Rogers,
Homostylous, South Africa, C.M. Rogers 13695 (MO),—, FJ169549,
FJ160779, FJ160822, FJ160863. Linum flavum L., Heterostylous, Romania,
H. Weber (MJG R00024), FJ169574, FJ169538, FJ160794, FJ160833, FJ160878.
Linum gracile Planch., Homostylous, South Africa, M. Quint (MJG 040942),
FJ169587, FJ169556, FJ160776, FJ160819, FJ160860. Linum grandiflorum Desf.,
Heterostylous, (MJG 040930), FJ169575, FJ169525, FJ160798, FJ160837,
FJ160882. Linum hirsutum L., Heterostylous, (MJG 040935), FJ169583,
FJ169520, FJ160788, FJ160827, FJ160872. Linum hypericifolium Salisb.,
Heterostylous, Georgia, M. Merello et al. 2116 (MO), —, FJ169519, FJ160789,
FJ160828, FJ160873. Linum kingii S.Watson, Homostylous, Idaho, U.S.A.,
A. Davis & S. Long s.n. (SRP), FJ169576, FJ169555, FJ160780, FJ160823,
FJ160864. Linum lewisii Pursh, Homostylous, Texas, U.S.A., B.L. Turner
21-263 (TEX), FJ169577, FJ169523, FJ160800, FJ160839, FJ160884. Linum
littorale St.Hil., Homostylous, Argentina, A. Schinini y C. Quarín 14510
(MO), —, FJ169543, FJ160781, —, FJ160865. Linum macraei ssp. macraei
Benth., Homostylous, Chile, P. Aravena 33388 (MO), FJ169578, FJ169544,
FJ160782,—, FJ160866. Linum marginale A.Cunn. ex Planch., Homostylous,
Australia, R. Taylor 127 (MO), —, FJ169528, FJ160804, FJ160843, FJ160888.
Linum maritimum L., Heterostylous, Italy, (MJG 040937), FJ169579,
FJ169535, FJ160811, FJ160850, FJ160895. Linum narbonense L., (MJG
040923)†, FJ169580. Linum nodiflorum L., Homostylous, Israel, (MJG
040940), FJ169581, FJ169539, FJ160795, FJ160834, FJ160879. Linum oli-
gophyllum Willd., Homostylous, Peru, D.N. Smith et al. 9426 (MO), —,
FJ169546, FJ160783, —, FJ160867. Linum pallescens Bunge, Homostylous,
Soviet Union, (MJG 040941), —, FJ169521, FJ160801, FJ160840, FJ160885.
Linum perenne L., Heterostylous, (MJG 040929), FJ169582, FJ169524,
FJ160802, FJ160841, FJ160886. Linum c.f. prostratum Domb. ex Lam., Peru,
Weigend et al. 7324 (MSB)†, FJ169571. Linum prostratum var. parvum
(Johnst.) Mildner, Homostylous, Peru, R. Mildner 52 (WUD), —, FJ169545,
FJ160784, —, FJ160868. Linum pubescens Banks & Sol., Heterostylous,
Syria, W. Licht (MJG SYR351), FJ169585, FJ169518, FJ160790, FJ160829,
FJ160874. Linum rupestre (A.Gray) Engelm. ex A.Gray, Homostylous,
Texas, U.S.A., B.L. Turner 21-589 (TEX), FJ169586, FJ169553, FJ160785,
FJ160824, FJ160869. Linum stelleroides Planch., Homostylous, China, K.J.
Guan et al. PE 5 , FJ169588, FJ169516, FJ160805, FJ160844, FJ160889. Linum
striatum Walter, Homostylous, Tennessee, U.S.A., V.E. McNeilus 98-404
(TEX), FJ169589, FJ169554, FJ160786, FJ160825, FJ160870. Linum stric-
tum L., Homostylous, Italy, W. Licht (MJG 040945), FJ169590, FJ169530,
FJ160806, FJ160845, FJ160890. Linum suffruticosum L., Heterostylous,
Spain, C. Navarro et al. CN 2339 (TEX), FJ169591, FJ169532, FJ160807,
FJ160846, FJ160891. Linum sulcatum Riddell, Texas, U.S.A., Saunders 1343
(TEX), FJ169592. Linum tenue Desf., Heterostylous, Spain, N. Schmalz (MJG
040938), FJ169593, FJ169548, FJ160808, FJ160847, FJ160892. Linum tenuifo-
lium L., Homostylous and Heterostylous populations, Germany, M. Kropf
(MJG 040934), FJ169594, FJ169529, FJ160809, FJ160848, FJ160893. Linum
trigynum L., Homostylous, (MJG 040946), FJ169595, FJ169536, FJ160810,
FJ160849, FJ160894. Linum usitatissimum L., Homostylous, (MJG 040924),
FJ169596, FJ169526, FJ160803, FJ160842, FJ160887. Linum vernale Wooton,
Homostylous, Texas, U.S.A., D.S. Correll & Helen B. Correll 38582 (TEX), —,
FJ169552, FJ160812, FJ160851, FJ160896. Linum viscosum L., Heterostylous,
(MJG 040939), FJ169584, FJ169517, FJ160791, FJ160830, FJ160875. Linum
volkensii Engl., Homostylous, Tanzania, R.E. Gereau & C.J. Kayombo 4662
(MO), FJ169597, FJ169531, FJ160813, FJ160852, FJ160897. Medusagyne
oppositifolia Baker (Medusagynaceae), Z75670‡. Ochna serrulata (Hochst.)
Walp. (Ochnaceae), AB233908‡. Phyllanthus liebmannianus Müll.Arg.
(Phyllanthaceae), Z75676‡. Radiola linoides Roth., Homostylous, Germany,
D. Hartl s.n. (MJG), FJ169598, FJ169534, FJ160815, FJ160854, FJ160899.
Reinwardtia indica Dum., Heterostylous, cultivated, McDill s.n. 2002
(TEX), FJ169599, FJ169514, FJ160814, FJ160853, FJ160898. Roucheria calo-
phylla Planch., Brazil, Alencar 593 (NY)†, FJ169600. Roucheria schomburgkii
Planch., Peru, Manuel Rimachi Y. 8416 (MO)†, FJ169603. Sclerolinon digynum
(A. Gray) C.M. Rogers, Homostylous, California, U.S.A., M.S. Taylor 3935
(TEX), FJ169601, FJ169541, FJ160787, FJ160826, FJ160871. Tapura amazonica
Poepp. & Endl. (Dichapetalaceae), AF089763‡. Tirpitzia sinensis (Hemsl.)
Hall., Homostylous (other species Heterostylous), China, D.A. Link (MJG
RG0002), FJ169602, FJ169515, FJ160816, FJ160855, FJ160900.