HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (37) Citing Articles via Google Scholar Google Scholar Articles by Alice, L. A. Articles by Campbell, C. S. Search for Related Content PubMed Articles by Alice, L. A. Articles by Campbell, C. S. Agricola Articles by Alice, L. A. Articles by Campbell, C. S.

Phylogeny of Rubus (rosaceae) based on nuclear ribosomal DNA internal transcribed spacer region sequences¹

^a 5722 Deering Hall, Department of Biological Sciences, University of Maine, Orono, Maine 04469-5722

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
We used nuclear ribosomal DNA internal transcribed spacer region (ITS 1 - 5.8S - ITS 2; ITS) sequences to generate the first phylogeny of Rubus based on a large, molecular data set. We sampled 57 taxa including 20 species of subgenus Rubus (blackberries), one to seven species from each of the remaining 11 subgenera, and the monotypic and closely related Dalibarda. In Rubus, ITS sequences are most informative among subgenera, and variability is low between closely related species. Parsimony analysis indicates that Rubus plus Dalibarda form a strongly supported clade, and D. repens may nest within Rubus. Of the subgenera with more than one species sampled, only subgenus Orobatus appears monophyletic. Three large clades are strongly supported: one contains all sampled species of nine of the 12 subgenera; another includes extreme Southern Hemisphere species of subgenera Comaropsis, Dalibarda, and Lampobatus; and a third clade consists of subgenus Rubus plus R. alpinus of subgenus Lampobatus. Rubus ursinus appears to be a hybrid between a close relative of R. macraei (subgenus Idaeobatus, raspberries) and an unidentified subgenus Rubus species. ITS sequences are generally consistent with biogeography and ploidy, but traditionally important morphological characters, such as stem armature and leaf type, appear to have limited phylogenetic value in Rubus.

Key Words: biogeography • classification • internal transcribed spacer (ITS) • phylogeny • Rosaceae • Rubus

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Rubus L. includes ~750 species (Robertson, 1974; Lu, 1983; Gu et al., 1993; Thompson, 1995) and is found on all continents except Antarctica (Focke, 1910, 1911, 1914; Gustafsson, 1942, 1943; Spies and Du Plessis, 1985; Hummer, 1996). The genus is economically and ecologically important as fruit crops, ornamentals, invasive weeds, and in early forest succession (Thompson, 1995; Hummer, 1996; Howarth, Gardner, and Morden, 1997). The most recent global taxonomic treatment of Rubus (Focke, 1910, 1911, 1914) included ~429 species in 12 subgenera (Table 1), the three largest being Idaeobatus (raspberries, 117 species), Malachobatus (115 primarily Asian species), and Rubus (= Eubatus Focke; blackberries, ~132 species). Only three of the remaining nine subgenera have more than six species. Numerous species and infrageneric taxa have been recognized since Focke's monograph, primarily in subgenus Rubus (Bailey, 1941–1945; Gustafsson, 1943; Weber, 1995).

Systematic difficulties also exist at higher infrageneric levels. Two of Focke's (1911, 1914) subgenera contain widely disjunct taxa. Subgenus Dalibarda has one western North American species, one western North American–eastern Asian species, one species each in the Himalayas and Tasmania, and an eastern North American endemic currently placed in the genus Dalibarda (Gleason and Cronquist, 1991; see below). Subgenus Lampobatus was originally divided into sect. Lampobatus, from Mexico, the West Indies, Central–South America, and the Himalayas, and sect. Micranthobatus, from Australia and New Zealand (Focke, 1894). Focke (1911, 1914) united sections Lampobatus and Micranthobatus into subg. Lampobatus, to which he added species from New Guinea and Madagascar. Kalkman (1987) suggested that species of sect. Lampobatus be placed in subg. Rubus and the remaining species in the newly established subg. Micranthobatus (Fritsch) Kalkman. He was, however, "less certain that this is a natural (monophyletic) group than for other Malesian subgenera" (Kalkman, 1987, p. 323).

Another taxonomic problem is circumscription of Rubus itself. Dalibarda repens was described by Linnaeus (1753), who later placed it in Rubus as R. dalibarda L. Focke (1910) included this species in his Rubus subg. Dalibarda with four other species. In contrast, North American botanists have followed Linnaeus' (1753) original classification of this species and place R. dalibarda in the monotypic genus Dalibarda because of its reduced carpel number, dry fruits, and apetalous flowers (Rydberg, 1913; Bailey, 1941–1945; Fernald, 1950; Gleason and Cronquist, 1991).

Polyploidy and hybridization are prevalent in Rubus. Only subgenera Idaeobatus, Dalibarda, and Anoplobatus are predominantly diploid, whereas Dalibardastrum, Malachobatus, and Orobatus are exclusively polyploid (Thompson, 1995, 1997). Hybridization in Rubus occurs mostly between closely related species (Steele and Hodgdon, 1963, 1970; Naruhashi, 1979, 1990; Kraft, Nybom, and Werlemark, 1995) and in some instances between subgenera (Gustafsson, 1942; Jennings, 1978; Weber, 1995; Alice et al., 1997). For example, Brown (1943) considered R. ursinus to be a cross between an ancestral Pacific blackberry and an eastern North American blackberry. Several intersubgeneric hybrids are horticulturally important (Waugh et al., 1990).

Our objectives are to infer phylogenetic relationships within Rubus plus Dalibarda using molecular data and to compare the implications of our results for Rubus classification with that of Focke (1910, 1911, 1914). Our sample covers much of the taxonomic/morphological diversity within the genus. We used nuclear ribosomal DNA (nrDNA) internal transcribed spacer region (ITS 1 - 5.8S - ITS 2; ITS) sequences because they have been phylogenetically useful in a wide range of taxa at generic and specific levels (Baldwin et al., 1995, and references therein; Downie and Katz-Downie, 1996; Yuan, Küpfer, and Doyle, 1996; Campbell et al., 1997; Potter, Luby, and Harrison, 1997; Eriksson, Donoghue, and Vretblad, in press).

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Plant samples
We sampled 68 accessions from 56 Rubus species, including 20 members of subg. Rubus, 1–7 representatives of each of the remaining 11 subgenera, plus Dalibarda repens (Table 2). Sixteen accessions are from wild plants cultivated in arboreta, botanical gardens, or screenhouses; 18 taxa were sampled from herbarium specimens; and plants of R. nepalensis, R. parvus, and R. tricolor were purchased from a commercial nursery. The remaining 31 accessions were collected from the wild.

Because of low mean ITS sequence divergence between species of Rubus subg. Rubus sect. Rubus (1.21%), only three of the 18 species sampled of this section were included in the final data set, which contains 40 Rubus species, Dalibarda repens, and three outgroups.

Chromosome numbers have been reported for 49 of the 56 Rubus species we sampled plus Dalibarda repens. Ploidy ranges from diploid (x = 7) to dodecaploid (Table 2); 38.8% are diploid, 42.9% are polyploid, and 18.3% have both diploid and polyploid counts. Of the 40 Rubus species plus Dalibarda included in the final ITS data set, 34 have chromosome counts. Ploidy ranges from diploid to octaploid; 50.0% are diploid, 38.2% are polyploid, and 11.8% have both diploid and polyploid counts.

McDade (1992) indicated that inclusion of hybrids in cladistic analysis does not affect topology unless the parents are phylogenetically distant from one another. Using morphological, chemical, and molecular data, Rieseberg and Morefield (1995) concluded that inclusion of Helianthus hybrid species had almost no effect on topology. All Rubus polyploids, except R. ursinus, were included in the final data set because they show levels of nucleotide polymorphism similar to those of diploid species and did not disrupt tree topology in preliminary analyses. We did exclude R. ursinus from most analyses, on the other hand, because ITS sequence polymorphism and morphology indicate that it is an intersubgeneric hybrid (see Dicussion: Possible hybrid taxa).

Total genomic DNA was isolated using a modified CTAB (hexadecyltrimethylammonium bromide) method (Doyle and Doyle, 1987) from leaf tissue collected fresh and stored at -80°C, leaves dried in silica gel desiccant, or leaves from herbarium specimens. We successfully amplified and sequenced ITS from material collected as early as 1936. Identification of specimens was verified using Focke's monograph or regional keys. Accessions from NCGR (see Table 2) were also verified by M. Thompson, Oregon State University.

Polymerase chain reaction (PCR) and DNA sequencing
PCR amplification of ITS generally followed Baldwin (1992). Double-stranded DNA was directly amplified by symmetric PCR using the ITS5 and ITS4 primers of White et al. (1990). Reaction volumes were 25 µL and contained 1.0 mg/mL bovine serum albumin (New England Biolabs, Beverly, Massachusetts), 50 mmol/L KCl, 10 mmol/L Tris-HCl (pH 8.3), 1.9 mmol/L MgCl₂, 200 µmol/L each deoxynucleotide triphosphate (Stratagene, La Jolla, California), 0.3 µmol/L oligonucleotide primer (Operon Technologies, Inc., Alameda, California), 1.0 unit of AmpliTaq DNA polymerase (Perkin-Elmer, Norwalk, Connecticut), and ~25–60 ng of genomic DNA. PCR was performed in a PTC-100 thermal cycler (MJ-Research, Inc., Watertown, Massachusetts) and consisted of 40 cycles of 1 min at 97°C for template denaturation, 1 min at 48°C for primer annealing, 45 s (increased by 4 s per cycle) at 72°C for primer extension, followed by a final extension of 7 min at 72°C. PCR products were purified by gel electrophoresis in 0.8% SeaPlaque GTG agarose (FMC, Rockland, Maine) followed by band isolation. Each band containing ITS was melted at 65°C and the agarose digested with 1.0–7.5 units of ß-agarase (Sigma, St. Louis, Missouri) at 39°–45°C for 2.5 h. Double-stranded DNA was sequenced using the dideoxy chain termination method using an ABI PRISM(TM) Dye Terminator Cycle Sequencing Ready Reaction Kit with AmpliTaq DNA polymerase FS (Perkin Elmer, Norwalk, Connecticut). Samples were electrophoresed in an ABI 373A automated sequencer in a stretch gel following the manufacturer's instructions (Applied Biosystems, Inc., Foster City, California). Chromatograms were manually edited using Sequence Navigator (Applied Biosystems, Inc., Foster City, California). Primers ITS5 and ITS4 were used to sequence all samples, and in cases of potential nucleotide site polymorphism or ambiguous sequence, primers ITS3 and ITS2 (White et al., 1990) were also used. Approximately 32% of all nucleotides sequenced were derived from a single primer. All other nucleotides were verified with at least two primers. Rubus sequences generated in this study are available from GenBank (Table 2).

Outgroup selection
Two recent molecular phylogenetic studies of Rosaceae included genera of subfamily Rosoideae, to which Rubus belongs. Morgan, Soltis, and Robertson's (1994) rbcL phylogeny included ten Rosoideae s. s. genera and placed Rubus as sister to a clade including Agrimonia, Rosa, Fragaria, Potentilla, and Alchemilla. This clade of six genera is supported by a decay value of 1 and is sister to a well-supported clade containing Fallugia, Geum, and Waldsteinia. Eriksson, Donoghue, and Vretblad's (in press) ITS phylogeny included 18 Rosoideae s. s. genera and placed Rubus as sister to a Fallugia–Geum–Waldsteinia clade. This topology is supported by a decay value of 3 when several Rubus species are included in the analysis. Due to greater support and more extensive sampling of subfamily Rosoideae s. s., we used as outgroups single representative species of Fallugia, Geum, and Waldsteinia (sequences from Eriksson, Donoghue, and Vretblad, in press).

Alignment of ITS sequences
Boundaries for ITS 1 and ITS 2 in Rubus and Dalibarda repens were determined by comparison with Rosaceae sequences (Campbell et al., 1995). Sequences were aligned visually. Aligned sequences of ITS 1 and ITS 2 for Fallugia, Geum, and the five Rubus species exhibiting the most gaps are shown in Fig. 1. Alignment of ITS 1 sequences within Rubus plus Dalibarda required one two-base gap and six one-base gaps in most sequences. Alignment of ITS 2 sequences necessitated gaps of one to four bases at the 5' end in all species except R. deliciosus, and four one-base gaps in some sequences. Alignment of Rubus and Dalibarda with the outgroups required several one-base gaps and one 12-base gap (positions 118-129) in all ingroup species (Fig. 1). Multiple gaps of varying length were needed to align Geum and Waldsteinia with Fallugia. We determined whether insertions or deletions were responsible for gap regions based on our ITS strict consensus phylogeny.

View larger version (42K): [in this window] [in a new window]   Fig. 1. Aligned sequences of ITS 1 and ITS 2 for Fallugia, Geum , and the five Rubus species exhibiting the most gaps. Rubus species are in lowercase letters and outgroups are in uppercase. Dots represent the same nucleotide present in Fallugia and dashes represent gaps. The four underlined residues in ITS 2 at sites 103–106 are found only in R. macraei and in R. ursinus (not shown) as part of a nucleotide site polymorphism (site 106 appears fixed for the R. macraei nucleotide).

Phylogenies were generated using Fitch parsimony as implemented in PAUP. Because of the number of taxa in this study, we executed HEURISTIC searches including all characters using RANDOM (1000 replicates) stepwise addition of taxa followed by TBR (tree bisection-reconnection) branch swapping. To evaluate the impact of each species on number of trees found, tree length, Consistency Index (CI), Retention Index (RI), and topology, we performed taxon jackknifing (Hillis, Allard, and Miyamoto, 1993) using the HEURISTIC search option with ten replicates of RANDOM stepwise addition of taxa excluding uninformative characters. We also searched for multiple islands of equally parsimonious trees (Maddison, 1991) following methods outlined in Olmstead and Palmer (1994). Gaps were coded as missing data, a unique character state, or binary characters (presence–absence) in separate phylogenetic analyses. Character-state changes were weighted equally; to explore the effect of weighting, transversions were weighted over transitions by 2:1 and 5:1 using the step matrix option in PAUP. The transition–transversion ratio, based on our ITS strict consensus tree and calculated using MacClade (Maddison and Maddison, 1992), is 2.8:1.

Sets of equally parsimonious trees were summarized using strict consensus. Decay indices (Bremer, 1988; Donoghue et al., 1992) and bootstrap values (Felsenstein, 1985) with 500 replicates, saving up to 200 trees per replicate, were calculated as measures of support for individual clades. Decay analyses were performed with AutoDecay (Eriksson and Wikstrom, 1996) and the reverse constraint option in PAUP. To test further our results, we used topological constraint trees in PAUP to determine the monophyly cost (i.e., the number of steps beyond the most parsimonious necessary for monophyly) of Rubus (excluding Dalibarda) and each subgenus for which we sampled more than one species (except subg. Orobatus). Search methods were the same as those employed in taxon jackknifing. Pairwise divergence, adjusted for missing data, was calculated for all taxa in PAUP. We also mapped changes in leaf type (simple or compound), stem armature (absent, bristles, prickles, or bristles and prickles), ploidy level, and biogeographic region onto our ITS strict consensus tree using MacClade. For simplicity, we recognized only two leaf-type character states.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
ITS length, GC content, sequence divergence, and nucleotide site variation
ITS 1 ranges from 254 to 257 base pairs (bp) in Rubus and Dalibarda and from 229 (Geum and Waldsteinia) to 270 bp (Fallugia) in the outgroups (Table 3). ITS 2 varies from 207 to 212 bp in all taxa sampled. Mean guanine + cytosine (GC) content is slightly higher in ITS 2 than in ITS 1 for Rubus, Dalibarda, and the outgroups. Mean pairwise divergence of sequences is higher in ITS 2 than in ITS 1 in Rubus and Dalibarda and almost identical between the ingroup and the outgroups.

In Rubus and Dalibarda there are two phylogenetically informative gaps in ITS 1: a two-base insertion (positions 47–48) and a one-base deletion (position 111) in R. rosifolius and R. minusculus (Fig. 1). One-base insertions in ITS 1 (position 218) in R. deliciosus, R. trilobus, and R. arcticus are homoplastic. In ITS 2, four of the five gaps are potentially informative phylogenetically. The first gap occurs after a series of cytosine residues near the 5' end (Fig. 1, positions 15–18) and was not used in our analysis because unambiguous positional homology could not be determined and several accessions exhibit length variability. The second gap (position 26) cannot be classified as an insertion or deletion, or as plesiomorphic or apomorphic, because it varies in the outgroups. The presence of a cytosine residue is, however, diagnostic of clades B and F (see below for composition of clades), subg. Orobatus, and R. pectinellus + R. nepalensis. The third gap is a deletion (position 90) that is potentially homoplastic and synapomorphic for clades B and C, and subg. Orobatus. The fourth gap (position 135) is a one-base insertion and synapomorphic for clade B.

Phylogeny of Rubus
Phylogenetic signal in the final ITS data set is significant (P < 0.01) based on the value of the g₁ statistic (-0.983). Heuristic searches including only ITS characters and gaps coded as missing data generated 208 equally parsimonious trees requiring 445 evolutionary steps (strict consensus in Fig. 2). Excluding uninformative sites, the CI is 0.579, and the RI is 0.745. Regression analysis of log-transformed CIs against number of taxa predicts an expected CI for 44 taxa of 0.344 (Sanderson and Donoghue, 1989). Thus, levels of homoplasy in our ITS data set are lower than expected.

View larger version (37K): [in this window] [in a new window]   Fig. 2. Strict consensus of 208 equally parsimonious trees of length 445 containing 40 Rubus taxa, Dalibarda repens , and three outgroups based on ITS sequences with gaps coded as missing data. Rubus species are in lowercase. Bootstrap values, based on 500 replicates saving a maximum of 200 trees per replicate, are above branches, and decay values are below branches. Major clades are labeled A–F above branches. Rubus subgenera codes: An-Anoplobatus, Cb-Chamaebatus, Cm-Chamaemorus, Co-Comaropsis, Cy-Cylactis, Da-Dalibarda, Ds-Dalibardastrum, Id-Idaeobatus, La-Lampobatus, Ma-Malachobatus, Or-Orobatus , and Ru-Rubus .

Three clades with more than three species within the ingroup (A, B, and C; Fig. 2) are supported by bootstrap values >93% and decay values greater than 3. Clade A contains all sampled species of nine of the 12 subgenera and excludes R. chamaemorus of the monotypic subg. Chamaemorus, and all but one sampled member each of subg. Anoplobatus (R. trifidus) and subg. Dalibarda (R. gunnianus). Clade B includes R. geoides of subg. Comaropsis from southern South America, R. gunnianus of subg. Dalibarda from Tasmania, and three species of subg. Lampobatus from Australia and New Zealand. Clade B is one branch of a weakly supported trichotomy with R. nivalis of subg. Chamaebatus plus two subg. Orobatus species. Clade C consists of four subg. Rubus species, representing two of the six sections (but not R. ursinus of sect. Ursini, not shown), plus R. alpinus of subg. Lampobatus.

For purpose of discussion, three other clades with more than three species are named (D, E, and F; Fig. 2). Clade D contains representatives of three subgenera: R. minusculus of subg. Cylactis; R. rosifolius and R. crataegifolius of subg. Idaeobatus; and R. trifidus of subg. Anoplobatus. Clade E includes four species of subg. Idaeobatus, R. idaeus, R. macraei, R. occidentalis, and R. phoenicolasius, plus R. saxatilis of subg. Cylactis. Clade F has three of the four species of subg. Malachobatus sampled plus R. tricolor of subg. Dalibardastrum.

For each subgenus in which we sampled more than one species (except the monophyletic subg. Orobatus), we used topological constraint trees in PAUP to force monophyly. Of the nine subgenera we tested, four produced minimum-length trees two to six steps longer than Fig. 2, and the remaining five subgenera each required at least 15 additional steps (Table 1). Therefore, it is unlikely that Focke's (1910, 1911, 1914) subgenera Anoplobatus, Cylactis, Dalibarda, Idaeobatus, and Lampobatus are monophyletic. If highly divergent species are not constrained to belong to their respective subgenus, monophyly cost is reduced.

Taxon jackknifing identified removal of R. nepalensis as the exclusion that most markedly reduces the number of trees recovered and increases resolution in clade A. Removal of this species yields four equally parsimonious trees of length 441. All nodes are resolved in the strict consensus tree (Fig. 3), except for one trichotomy each in clades C and F. However, the newly resolved nodes are poorly supported with bootstrap values <50% and decay values of 1 (not shown in Fig. 3). Clade F is weakly united with R. pectinellus of subg. Chamaebatus. Clade F + R. pectinellus is sister to a clade containing two groups. The first group has R. lineatus of subg. Malachobatus + clade E. The second group contains three large clades. One clade includes ((R. arcticus + R. pubescens) + (R. humulifolius + clade D)). This first clade is sister to (clade C + (clade B + (R. nivalis + subg. Orobatus))).

View larger version (33K): [in this window] [in a new window]   Fig. 3. Strict consensus phylogram of four equally parsimonious trees of length 441 containing 39 Rubus taxa (excluding R. nepalensis ), Dalibarda repens , and three outgroups based on ITS sequences with gaps coded as missing data. The branch leading to Rubus plus Dalibarda has 31 nucleotide substitutions. Rubus species are in lowercase. Major clades are labeled A–F on branches. See Fig. 2 caption for Rubus subgenera codes.

If transversions are weighted over transitions by 2:1, the strict consensus tree is basically identical to Fig. 2. With a 5:1 weighting scheme there is some increase in resolution; clades A through F are maintained and clade C is allied with (clade F + R. lineatus). In an unweighted analysis with gaps coded as binary characters, clade C is associated with (clade B + (R. nivalis + subg. Orobatus)). Other differences using a 5:1 weighting scheme include: (1) R. odoratus + R. parviflorus are separated from R. trilobus + R. deliciosus; (2) R. chamaemorus is united with R. pedatus; and (3) R. crataegifolius + (R. rosifolius + R. minusculus) are separate from R. trifidus and R. humulifolius. These differences are weakly supported.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
ITS divergence and phylogenetic information in Rubus
In Rubus and Dalibarda, length and GC content of ITS 1 and ITS 2 (Table 3) are similar to those of other reported angiosperm sequences (Baldwin et al., 1995). Although several small gaps (1–4 bp) are present in each spacer, nucleotide substitution appears to be the main source of variability. Pairwise divergence of Rubus and Dalibarda sequences ranges from 0.4 to 9.4% in ITS 1 and from 0.0 to 16.0% in ITS 2. These ITS 1 divergence values are similar, or slightly lower, than those reported by Baldwin et al. (1995) in Astragalus, Fouquieria, Gilia sect. Giliandra, and Viburnum. ITS 2 divergence is slightly higher in Rubus and Dalibarda than in these four groups. In contrast, ITS 1 and ITS 2 divergence values in all these genera are markedly lower than those in Gentiana (5.0 to 48.9% in ITS 1 and 1.1 to 45.3% in ITS 2; Yuan, Küpfer, and Doyle, 1996). The percentage of aligned characters that are potentially phylogenetically informative is 20.0% in ITS 1 and 25.1% in ITS 2 in Rubus and Dalibarda, but 18% in Astragalus, Fouquieria, Gilia sect. Giliandra, and Viburnum. However, the numbers of potentially phylogenetically informative characters in Rubus plus Dalibarda ITS 1 and ITS 2 are almost identical to each other (Table 3).

Previous molecular systematic studies of Rubus
Three previous molecular studies, two of which were explicitly phylogenetic, focused on three economically important subgenera: Anoplobatus, flowering raspberries; Idaeobatus, raspberries; and Rubus, blackberries, and are generally congruent with our ITS results except in the placement of R. parviflorus (see below). These studies used chloroplast DNA (cpDNA) restriction fragment length polymorphism (Waugh et al., 1990), random amplified polymorphic DNA (RAPD) markers (Graham and McNicol, 1995), and ndhF sequences (Howarth, Gardner, and Morden, 1997) and each included 24 taxa and no more than 14 species. Our study is the first molecular phylogenetic study of Rubus subgenera based on a large taxonomic sample.

Congruence between ITS-based phylogeny and traditional Rubus classification
All ingroup species excluded from clade A (Figs. 2, 3) have been classified outside Rubus by some workers (see Bailey, 1941–1945). ITS data do not preclude exclusion of these species from Rubus given the lack of resolution and weak support for basal nodes. However, high bootstrap and decay values on the branch connecting Rubus and Dalibarda with the outgroups suggest a close relationship. Although one might recognize basal species in Fig. 2 as distinct from Rubus based on ITS data, we prefer to treat them (not including Dalibarda repens) as congeneric on morphological grounds. All ingroup species sampled have two ovules (the outgroups have one), and with the exception of Dalibarda, have fleshy aggregates of drupelets.

More evidence is necessary to determine the phylogenetic position of Dalibarda repens. This species is the fifth branch in the tree from the ingroup/outgroup node and thus appears to be nested within Rubus, although bootstrap support is <50% and decay values are only 1 or 2 for the first four nodes within the Rubus plus Dalibarda clade. Hence, Dalibarda could be the sister genus to Rubus or part of an unresolved basal complex. If Dalibarda is constrained as sister to Rubus, minimum-length trees are only three steps longer than those in Fig. 2. Dalibarda repens has been excluded from Rubus due to dry fruits, apetalous flowers, and reduced carpel number (5–10), but some Rubus species, such as R. pedatus, have only 3–6 carpels (Bailey, 1941–1945).

ITS data do not support monophyly of Rubus or of any subgenus for which we sampled more than one species except subg. Orobatus (Figs. 2, 3). In the sections that follow, we discuss this incongruence between our ITS phylogeny and Focke's (1910, 1911, 1914) classification for each subgenus.

Subg. Anoplobatus
ITS data clearly show that subg. Anoplobatus is not monophyletic; when it is constrained to be so, 15 steps are added to minimum-length trees in Fig. 2. Instead, the five species we sampled are divided into a New World group of four species that is strongly excluded from clade A and Japanese R. trifidus, which is allied to certain subg. Idaeobatus species.

ITS data indicate that New World members of subg. Anoplobatus are monophyletic and divided into two clades (Figs. 2, 3). One clade includes R. odoratus (eastern North America) and R. parviflorus (western North America), and the other contains R. deliciosus (southwestern North America) and R. trilobus (Mexico and Guatemala). Rydberg (1913) treated both species pairs as separate genera: Rubacer (Rubus odoratus and R. parviflorus) and Oreobatus (Rubus trilobus and R. deliciosus). Rydberg segregated these two genera from Rubus on the basis of different styles, stigmas, receptacles, stem armature, bark, and leaf morphology. Waugh et al. (1990), who sampled three species of subg. Anoplobatus, found R. odoratus and R. deliciosus to be sister taxa, but R. parviflorus was basal in a clade with three Asian subg. Idaeobatus species. Howarth, Gardner, and Morden (1997) also found that R. parviflorus nested within subg. Idaeobatus, although they sampled only one species of subg. Anoplobatus. These results conflict with ITS data, which strongly unite R. odoratus and R. parviflorus, apart from Asian subg. Idaeobatus species. The position of R. parviflorus in trees of Howarth, Gardner, and Morden (1997) is spurious because of rooting. When Fallugia is used as outgroup in the ndhF analysis, R. parviflorus is sister to the remaining Rubus species sampled (representing subgenera Dalibardastrum, Idaeobatus, Malachobatus and Rubus), in agreement with our ITS topology (C. Morden, personal communication, University of Hawai`i). Morphologically, R. parviflorus, together with other subg. Anoplobatus species, differs from most subg. Idaeobatus species in its unarmed stems and simple leaves with adnate stipules. Cluster analysis of RAPD markers (Graham and McNicol, 1995) placed R. deliciosus at the base of their Rubus phenogram, in agreement with ITS data.

Rubus trifidus strongly nests within clade A near certain Asian species of subg. Idaeobatus (Figs. 2, 3). Satomi and Naruhashi (1971) considered R. trifidus seeds to be identical to the R. idaeus type and noted that R. odoratus seeds are distinct from all Japanese species sampled. Naruhashi (1980) transferred R. trifidus to subg. Idaeobatus near R. crataegifolius, an alliance supported by ITS data.

Subg. Chamaebatus
ITS data indicate that the two species we sampled of this subgenus, R. nivalis (northwestern North America) and R. pectinellus (eastern Asia), occur within clade A (Figs. 2, 3) but may not be closely related. When monophyly of subg. Chamaebatus is forced, six steps are added to minimum-length trees in Fig. 2. Subgenus Chamaebatus contains five simple-leaved, prickly stemmed, prostrate species (Focke, 1910, 1914); the three members we did not sample occur in Mexico, the Himalayas, and eastern Asia. Rubus nivalis appears related to clade B and the subg. Orobatus clade. Rubus pectinellus is weakly united with R. nepalensis (Fig. 2) or sister to clade F when R. nepalensis is excluded (Fig. 3). Rubus nivalis and R. pectinellus are similar morphologically but differ in ploidy level: R. nivalis is diploid; R. pectinellus is hexaploid (Thompson, 1997).

Subg. Chamaemorus
Relationships of this monotypic subgenus are not fully resolved by ITS beyond excluding it from clade A (Figs. 2, 3). All species outside clade A have simple leaves and unarmed stems with the exception of R. pedatus, which has five-foliate leaves, and R. lasiococcus, which has simple or ternate leaves. Furthermore, nonclade A species are diploid (the triploid count of R. deliciosus is considered aberrant by Thompson, 1997) except for octaploid R. chamaemorus, the circumpolar cloudberry or baked-apple-berry, which also differs from other species excluded from clade A in its dioecy. The four other dioecious species we sampled, R. parvus, R. australis, and R. moorei of subg. Lampobatus and R. ursinus of subg. Rubus (not shown), all nest within clade A.

Subg. Comaropsis
Rubus geoides, the single species of subg. Comaropsis that we sampled, is strongly nested in clade B (Figs. 2, 3). This species was once placed in genus Dalibarda but later united with another southern South American species in subg. Comaropsis (Focke, 1910). Except for its prickly petioles, Rubus geoides morphologically resembles the unarmed R. gunnianus of subg. Dalibarda, which is related based on ITS data.

Subg. Cylactis
ITS data show that this subgenus of 14 species is clearly polyphyletic (Figs. 2, 3); forcing its monophyly yields minimum-length trees 28 steps longer than those in Fig. 2. We sampled one species from each of three of Focke's (1910, 1914) series and two species from the fourth. ITS data, including a synapomorphy in the 5.8S gene, indicate that R. arcticus (series Arctici) and R. pubescens (series Saxatiles) are closely related. The R. arcticus + R. pubescens clade is part of a multichotomy (Fig. 2) or sister to R. humulifolius (subg. Cylactis, series Humulifolii) + clade D (Fig. 3).

Rubus saxatilis (subg. Cylactis, series Saxatiles) is sister to R. idaeus in clade E, and differs from Swedish R. idaeus at only two sites, one of which is polymorphic and includes the R. idaeus residue. The alternative residue is an autapomorphy. Rubus saxatilis may be a tetraploid derivative of R. idaeus, but it lacks the primary character for distinguishing raspberries: dehiscence of the fruits without the receptacle.

The fifth sampled member of subg. Cylactis, R. minusculus, is tightly linked (100% bootstrap and a decay value of 18) with R. rosifolius of subg. Idaeobatus. These Asian diploids also share two gaps, pinnately compound leaves, and weak prickles. Satomi and Naruhashi (1971) and Naruhashi (1980) recognized the close relationship of these two species, which are considered synonymous in the Missouri Botanical Garden`s TROPICOS database (based on the Flora of China checklist). Constraining monophyly of subg. Cylactis excluding R. minusculus adds only six steps to minimum-length trees. Thus, our ITS results suggest that R. minusculus should be removed from subg. Cylactis.

Polyphyly of subg. Cylactis (Figs. 2, 3) is consistent with its morphological heterogeneity. Leaf type ranges from simple in R. humulifolius, to ternate in R. arcticus, R. pubescens, and R. saxatilis, and to pinnately compound in R. minusculus. Stem armature varies from prickly in Rubus humulifolius, R. saxatilis and R. minusculus, to unarmed in R. arcticus and R. pubescens.

Subg. Dalibarda
This subgenus of five species (Focke, 1910, 1914), four of which we sampled, is not monophyletic based on our ITS data (Figs. 2, 3). Three of the species, western North American R. lasiococcus, western North American-eastern Asian R. pedatus, and the eastern North American endemic R. dalibarda (= Dalibarda repens), are excluded from clade A, but do not form a monophyletic group. Rubus pedatus is sister to all Rubus species sampled plus genus Dalibarda, but bootstrap support is <50% and the decay value is only 2. Placement of R. lasiococcus (Bailey, 1941–1945) and R. pedatus (Bailey, 1941–1945; Naruhashi, 1980; Lu, 1983) in subg. Cylactis instead of subg. Dalibarda as proposed by Focke (1910), strongly conflicts with ITS data. Rubus pedatus has been placed in the genus Dalibarda and also in the genus Comaropsis (see Bailey, 1941–1945). A relationship with genus Dalibarda is supported here, but not with R. geoides of subg. Comaropsis. Dalibarda repens tenuously nests within Rubus and might instead be the sister of Rubus (as previously discussed).

Tasmanian R. gunnianus, the fourth species of subg. Dalibarda that we sampled, occurs in clade B with species of subgenera Comaropsis and Lampobatus from the extreme Southern Hemisphere. Separation of R. gunnianus from other subg. Dalibarda species is justified based on bootstrap and decay values (Fig. 2) and addition of 25 steps to minimum-length trees when monophyly of subg. Dalibarda is forced.

Subg. Dalibardastrum
Both sampled species of this subgenus nest in different lineages of clade A (Fig. 2). Chinese R. tricolor is part of a trichotomy with R. assamensis and R. tephrodes of subg. Malachobatus. This weakly supported group combines divergent morphologies: R. tricolor has bristly, prostrate stems and R. assamensis and R. tephrodes have prickles and/or bristles and upright stems.

The other subg. Dalibardastrum species we sampled, Himalayan R. nepalensis, is weakly allied to R. pectinellus (Fig. 2) and may be of hybrid origin (see Possible hybrid taxa below). Graham and McNicol (1995) showed that R. nepalensis clustered with R. coreanus Miq., an eastern Asian species of subg. Idaeobatus, but R. nepalensis has trifoliate leaves and weak, bristly, prostrate stems, while R. coreanus has 5–7 leaflets and stout, prickly, upright stems (Ohwi, 1965). Forcing the monophyly of subg. Dalibardastrum adds only three steps to the shortest trees. The strict consensus topology with subg. Dalibardastrum (not shown) constrained to be monophyletic is identical to Fig. 3 except that R. nepalensis and R. tricolor are sister species, and uncertain support for relationships of R. nepalensis and R. tricolor leave unresolved the status of subg. Dalibardastrum.

Subg. Idaeobatus
Our ITS phylogeny indicates that subg. Idaeobatus is polyphyletic with three apparent lineages (Figs. 2, 3), a finding consistent with its occurrence on six continents. Thus, the raspberry fruits separating from the receptacle may have evolved at least three times. Waugh et al. (1990) and Howarth, Gardner, and Morden (1997) also found subg. Idaeobatus to be polyphyletic, forming at least two distinct groups.

Subgenus Idaeobatus species R. crataegifolius and R. rosifolius are members of clade D (Figs. 2, 3), which has uniform seed morphology (Satomi and Naruhashi, 1971; R. rosifolius seeds were not studied), but is otherwise diverse morphologically. Rubus rosifolius and R. minusculus of subg. Cylactis have pinnately compound leaves and weak prickles. The remaining two clade D species, on the other hand, have simple leaves, and only R. crataegifolius has prickles. When monophyly is imposed on subg. Idaeobatus, 31 steps are added to minimum-length ITS trees in Fig. 2. However, exclusion of R. rosifolius and R. crataegifolius from the constraint tree, reduces the monophyly cost to 14 and six, respectively. Thus, separation of R. rosifolius and R. crataegifolius from other subg. Idaeobatus species is supported.

Rubus hawaiensis (subg. Idaeobatus) is sister to the remaining taxa in clade A (Figs. 2, 3), but bootstrap and decay support are low. Rubus hawaiensis was classified by Focke (1911, 1914) in sect. Spectabiles with three other Hawaiian species and their hypothesized North American continental relative, R. spectabilis Pursh (not sampled). Only two native species, R. hawaiensis and R. macraei, are currently recognized in Hawai`i (Wagner, Herbst, and Sohmer, 1990). Neither ITS nor ndhF (Howarth, Gardner, and Morden, 1997) support a close relationship of these two species, suggesting that two colonization events (both from western North America; see below) are responsible for Hawaiian Rubus. Yet, minimum-length ITS trees only three steps longer than Fig. 2 are obtained if R. hawaiensis and R. macraei are constrained to be sister taxa.

Subg. Lampobatus
ITS data indicate that the four species we sampled of this subgenus, R. australis and R. parvus (both of New Zealand), R. moorei (Australia), and R. alpinus (West Indies and Central–South America), are divided into two groups (Figs. 2, 3). The first three species were originally placed in sect. Micranthobatus (Focke, 1894) and unite with Tasmanian R. gunnianus of subg. Dalibarda and R. geoides of subg. Comaropsis to form the well-supported clade B. Members of this clade are diverse morphologically, including species with simple leaves and unarmed stems (R. parvus), species with simple or ternate leaves and prickly petioles (R. geoides), and species with compound leaves and prickly stems (R. moorei). Rubus gunnianus and R. moorei are apparently more closely related to each other than either is to the New Zealand species R. parvus and R. australis.

Rubus alpinus, initially put in sect. Lampobatus with several Mexican/West Indian species plus one Himalayan species (Focke, 1894), is more closely related to species of subg. Rubus, in agreement with Rydberg (1913) and Kalkman (1987). ITS data therefore suggest that R. alpinus should be removed from subg. Lampobatus and that R. geoides and R. gunnianus should be included.

Subg. Malachobatus
This primarily Asian subgenus, in which we sampled four species, is not monophyletic based on ITS data (Figs. 2, 3). Three of the four species, Rubus assamensis, R. tephrodes, and R. lambertianus, are united with R. tricolor of subg. Dalibardastrum in the weakly supported clade F. Species of clade F are Asian, tetraploid, simple-leaved, and armed with prickles and/or bristles.

Rubus lineatus, the fourth sampled species of subg. Malachobatus, is part of a multichotomy in clade A (Fig. 2). When R. nepalensis is excluded from analysis, R. lineatus is sister to clade E (Fig. 3). However, if transversion-transition weighting is employed or gaps are included, R. lineatus is allied with clade F (not shown). Rubus lineatus differs from most subg. Malachobatus species in its palmately compound leaves and occasionally unarmed stems. Constraining subg. Malachobatus to be monophyletic requires only two additional steps. Thus, additional data are necessary to evaluate the monophyly of this subgenus.

Subg. Orobatus
The two Ecuadorian species we sampled, R. nubigenus and R. roseus, of this primarily South American subgenus form a strongly supported clade (Figs. 2, 3), including one synapomorphy in the 5.8S gene, and differ by only two transitions and a one-base deletion in R. nubigenus.

Subg. Rubus
Based on ITS data, sampled species of this subgenus form a well-supported clade (clade C; Figs. 2, 3) that includes R. alpinus of subg. Lampobatus but excludes R. ursinus (not shown). Section Rubus is not monophyletic because North American R. cuneifolius of sect. Rubus is more closely related to South American R. robustus of sect. Floribundi than to European R. ulmifolius of sect. Rubus. Previous molecular data show that subg. Rubus sect. Rubus species are closely related. Three sampled sect. Rubus species in Waugh et al. (1990) formed a monophyletic group and four species in Graham and McNicol (1995) also clustered together using RAPD markers. Approximately 1050 bp of ndhF sequence show no variability between R. argutus and R. cuneifolius of subg. Rubus sect. Rubus (Howarth, Gardner, and Morden, 1997).

Six nucleotide sites distinguish North American and European species of sect. Rubus, and five species (R. allegheniensis, R. divaricatus, R. nessensis, R. sapidus, and R. sulcatus) are polymorphic for at least one of these sites. Separate analysis of Rubus subg. Rubus species (excluding R. ursinus) with R. chamaemorus, R. odoratus, and Dalibarda repens as outgroups produces a highly unresolved strict consensus tree (not shown), but does recover a clade of New World species plus European R. nessensis. If, however, the five polymorphic taxa noted above are removed, BRANCH AND BOUND searches yield 20 most parsimonious trees of length 94 (CI = 0.820, RI = 0.893). The strict consensus tree (not shown) indicates that European subg. Rubus species form a weakly supported clade that is sister to a New World clade (79% bootstrap value) with South American R. robustus of sect. Floribundi sister to North American sect. Rubus species. These polymorphic species could be hybrids (three of the five are tetraploid) of recent origin. Alternatively, they could be ancient and ancestral species wherein concerted evolution has failed to homogenize ITS repeats (Campbell et al., 1997), and lineage sorting or biased gene conversion has resulted in distinct New World and European lineages.

Congruence between ITS data and nonmolecular features
Morphology
We mapped leaf type and stem armature, which are commonly used in Rubus classification (Focke, 1910, 1911, 1914; Bailey, 1941–1945), onto the Fig. 2 topology. We recognized only two leaf-type character states, simple and compound, and for simplicity did not distinguish among digitate, ternate, and pinnate leaves. The most parsimonious mapping of leaf type onto the ITS strict consensus tree (Fig. 4) requires eight evolutionary steps and is congruent with some clades, but not others. New World subg. Anoplobatus species (R. odoratus, R. parviflorus, R. deliciosus, and R. trilobus) and clade F species have simple leaves, and clade C and E species have compound leaves. However, clades B and D include simple-leaved and compound-leaved species. The presence of simple and ternate leaves in R. lasiococcus and other species also indicates that leaf type is phylogenetically plastic and therefore of limited value in Rubus at the subgeneric level.

View larger version (48K): [in this window] [in a new window]   Fig. 4. Topology of Fig. 2 tree with leaf type mapped. For simplicity no distinction was made between digitate, ternate, and pinnate leaves. Missing squares at branch tips indicate variability in this character.

View larger version (48K): [in this window] [in a new window]   Fig. 5. Topology of Fig. 2 tree with stem armature mapped. Missing squares at branch tips indicate variability in this character.

Ploidy level
Base chromosome number is phylogenetically informative among Rosaceae subfamilies (Morgan, Soltis, and Robertson, 1994), and ploidy is largely congruent with our Rubus ITS phylogeny (Fig. 6). Five or possibly six of the seven species excluded from clade A (Figs. 2, 3) with known chromosome counts are diploid, clade D species are diploid, clade F species are tetraploid, and members of the subg. Orobatus clade are hexaploid. In contrast, ploidy in clades C and E ranges from diploid to tetraploid.

View larger version (50K): [in this window] [in a new window]   Fig. 6. Topology of Fig. 2 tree with ploidy mapped. Ploidy of Waldsteinia is from W. ternata , not W. fragarioides , which we sampled for ITS. Ploidy of Geum represents the lowest chromosome count reported in the genus (tetraploid); G. urbanum , included in our study, is hexaploid. Chromosome counts for the three outgroups are from the Missouri Botanical Garden's index of plant chromosome number database. Missing squares at branch tips indicate variability in this character or data not available.

View larger version (52K): [in this window] [in a new window]   Fig. 7. Topology of Fig. 2 tree with biogeographic region mapped. "Other" represents noncoded regions or two or more coded regions. "NA-Europe" is North America, Mexico plus Guatemala, and/or Europe. "Southern Hemisphere" includes Australia, New Zealand, Tasmania, and southern South America.

Taxon jackknifing identified one species, R. nepalensis, whose removal markedly reduces the number of equally parsimonious trees recovered and increases resolution in clade A (see Results). This impact on tree topology could be explained by a hybrid origin of R. nepalensis (McDade, 1992; Campbell et al., 1997) or by homoplasy. We examined 21 potentially phylogenetically informative characters that might create the instability introduced by this taxon. For six homoplastic characters (the remaining 15 are largely autapomorphic for terminal clades), patterns of relationship of R. nepalensis are particularly complex. This species is least similar to clades B and D (Fig. 2) at the six homoplastic sites and shares five nucleotides with clade F, four with clade E, three with clade C, and two with R. arcticus + R. pubescens. Based on ITS sequence divergence, R. nepalensis is closest to R. idaeus, R. saxatilis, and R. tricolor. The first two species are sister taxa in clade E, and R. tricolor is in clade F.

Homoplasy in these six characters alone may be responsible for the reduction in tree number and increased resolution in clade A when R. nepalensis is excluded. We removed each of the six homoplastic characters individually and ran HEURISTIC searches in PAUP to determine their impact on tree number. Two characters had no effect, one increased tree number to 3068, two reduced tree number to 48 and 36, and character 68 in ITS 2 (Fig. 1) reduced tree number to 12 and tree length by nine.

In conclusion, our ITS results show that Focke's (1910, 1911, 1914) classification of Rubus contains mostly nonmonophyletic subgenera although several groups are strongly supported. Subgenus Rubus (including R. alpinus of subg. Lampobatus and excluding R. ursinus) forms clade C (Fig. 2), and extreme Southern Hemisphere species form clade B. Furthermore, it appears that R. minusculus is closely related to R. rosifolius and should be removed from subg. Cylactis, Tasmanian R. gunnianus should be allied with Australian and New Zealand subg. Lampobatus species, and R. trifidus is not a member of subg. Anoplobatus. Although our results do provide strong phylogenetic signal for some infrageneric clades in Rubus, we sampled only 56 of the ~750 Rubus species. Biogeographic and ploidy level variations are generally more consistent with ITS-based trees than leaf type and stem armature, which are highly homoplastic and of limited phylogenetic value among Rubus subgenera.

To resolve phylogenetic relationships within Rubus, additional sampling of species, particularly those of large and disjunct subgenera, is needed. Because gene trees may not represent species trees (Doyle, 1992; Kellogg, Appels, and Mason-Gamer, 1996), more data are critical for confident determination of organismal relationships in Rubus. Levels of homoplasy in key morphological characters, such as leaf type and stem armature, encourage further use of molecular data. Given that weak support of several nodes in ITS-based trees is due largely to a limited number of characters, either a faster evolving or longer nuclear DNA region should be sought. Hybridization within and between Rubus subgenera and prevalence of polyploidy mandate use of a plastid gene. These data might provide a more robust phylogeny and allow one to address the possible origins of polyploid Rubus taxa.

	FOOTNOTES

 
¹ The authors thank the U.S. Department of Agriculture-Agricultural Research Service, National Clonal Germplasm Repository-Corvallis, Missouri Botanical Garden Herbarium, Harvard University Herbaria, Swedish Museum of Natural History, Morton Arboretum, Helsingborg Botanical Garden, and Royal Tasmanian Botanical Garden for plant samples; C. W. Morden for Rubus macraei and R. parviflorus DNA; H. O. Martensen for German Rubus samples; A. Oredsson for Swedish R. nessensis ; T. Eriksson for Swedish Rubus samples and for unpublished ITS sequences of Fallugia, Geum, Waldsteinia , and Rubus chamaemorus ; the University of Maine DNA sequencing facility; W. S. Judd and M. Woods for assistance in the field; and A. C. Dibble, T. Eriksson, C. W. Greene, I. L. Kornfield, L. C. Merrick, T. F. Vining, M. P. Widrlechner, and W. A. Wright for comments on the manuscript. This research was supported in part by University of Maine Association of Graduate Students and American Society of Plant Taxonomists Graduate Research grants to LAA and by a National Science Foundation grant (BSR 91-06226) to CSC.

	REFERENCES

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Aalders, L. E., and I. V. Hall.1966A cytotaxonomic survey of the native blackberries of Nova Scotia. Canadian Journal of Genetics and Cytology 8: 528–532. [ISI]

Alice, L. A., T. Eriksson, B. Eriksen, and C. S. Campbell.1997Intersubgeneric hybridization between a diploid raspberry, Rubus idaeus, and a tetraploid blackberry, R. caesius (Rosaceae). American Journal of Botany 84: 171 (Abstract).

Bailey, L. H.1941–45Species Batorum I-VIII. Gentes Herbarum 5: 1–918.

Baldwin, B. G.1992Phylogenetic utility of the internal transcribed spacers of nuclear ribosomal DNA in plants: An example from the Compositae. Molecular Phylogenetics and Evolution 1: 3–16. [CrossRef][Medline]

———, M. J. Sanderson, J. M. Porter, M. F. Wojciechowski, C. S. Campbell, and M. J. Donoghue.1995The ITS region of nuclear ribosomal DNA: a valuable source of evidence on angiosperm phylogeny. Annals of the Missouri Botanical Garden 82: 247–277. [CrossRef][ISI]

Brainerd, E., and A. K. Peitersen.1920Blackberries of New England—their classification. Vermont Agricultural Experiment Station Bulletin 217: 1–84.

Bremer, K.1988The limits of amino acid sequence data in angiosperm phylogenetic reconstruction. Evolution 42: 795–803. [CrossRef][ISI]

Brown, S. W.1943The origin and nature of variability in the Pacific coast blackberries (Rubus ursinus Cham. et Schlecht. and R. lemurum sp. nov.). American Journal of Botany 30: 686–697. [CrossRef][ISI]

Campbell, C. S., M. J. Donoghue, B. G. Baldwin, and M. F. Wojciechowski.1995Phylogenetic relationships in Maloideae (Rosaceae): evidence from sequences of the internal transcribed spacers of nuclear ribosomal DNA and its congruence with morphology. American Journal of Botany 82: 903–918. [CrossRef][ISI]

———, M. F. Wojciechowski, B. G. Baldwin, L. A. Alice, and M. J. Donoghue.1997Persistent nuclear ribosomal DNA sequence polymorphism in the Amelanchier agamic complex (Rosaceae). Molecular Biology and Evolution 14: 81–90. [Abstract]

Davis, H. A.1990Studies in Rubus. Castanea 55: 22–30.

Donoghue, M. J., R. G. Olmstead, J. F. Smith, and J. D. Palmer.1992Phylogenetic relationships of Dipsacales based on rbcL sequences. Annals of the Missouri Botanical Garden 79: 333–345.

Downie, S. R., and D. S. Katz-Downie.1996A molecular phylogeny of Apiaceae subfamily Apioideae: evidence from nuclear ribosomal DNA internal transcribed spacer sequences. American Journal of Botany 83: 234–251. [CrossRef][ISI]

Doyle, J. J.1992Gene trees and species trees: molecular systematics as one-character taxonomy. Systematic Botany 17: 144–163. [CrossRef][ISI]

———, and J. L. Doyle.1987A rapid DNA isolation procedure for small quantities of fresh leaf tissue. Phytochemical Bulletin 19: 11–15.

Eriksson, T., M. J. Donoghue, and M. Vretblad.In pressPhylogenetic analysis of Potentilla using nuclear ribosomal ITS sequences, and implications for the classification of Rosoideae (Rosaceae). Plant Systematics and Evolution.

———, and N. Wikstrom.1996AutoDecay version 3.0. Stockholm: distributed by the authors, Stockholm University, Sweden.

Felsenstein, J.1985Confidence limits on phylogenies: an approach using the bootstrap. Evolution 39: 783–791. [CrossRef][ISI]

Fernald, M. L.1950Gray's manual of botany, 8th ed., 818–864. Van Nostrand Reinhold, New York, NY.

Focke, W. O.1894Rosaceae. In A. Engler and K. Prantl [eds.], Die natürlichen pflanzenfamilien, vol. 3, part 3, 1–60. Leipzig.

———.1910Species Ruborum monographiae generis Rubi prodromus. Bibliotheca Botanica 17: 1–120.

———.1911Species Ruborum monographiae generis Rubi prodromus. Bibliotheca Botanica 17: 121–223.

———.1914Species Ruborum monographiae generic Rubi prodromus. Bibliotheca Botanica 17: 1–274.

Gleason, H. A., and A. Cronquist.1991Manual of the vascular plants of northeastern United States and Canada, 248–253. New York Botanical Garden, New York, NY.

Graham, J., and R. J. McNicol.1995An examination of the ability of RAPD markers to determine the relationships within and between Rubus species. Theoretical and Applied Genetics 90: 1128–1132. [ISI]

Gu, Y., C. M. Zhao, W. Jin, and W. L. Li.1993Rubus resources in Fujian and Hunan provinces. Acta Horticulturae 345: 117–125.

Gustafsson, A.1942The origin and properties of the European blackberry flora. Hereditas 28: 249–277. [ISI]

———.1943The genesis of the European blackberry flora. Lunds Universitets Årsskrift 39: 1–200.

Hillis, D. M., M. W. Allard, and M. M. Miyamoto.1993Analysis of DNA sequence data: phylogenetic inference. In E. A. Zimmer [ed.], Methods in enzymology 224, Molecular evolution: producing the biochemical data, 456–487. Academic Press, San Diego, CA.

———, and J. P. Huelsenbeck.1992Signal, noise and reliability in molecular phylogenetic analyses. Journal of Heredity 83: 189–195. [Abstract/Free Full Text]