AFLP analysis of genetic relationships in the genus Fosterella L.B. Smith (Pitcairnioideae, Bromeliaceae

Page 1

AFLP analysis of genetic relationships in thegenus Fosterella L.B. Smith (Pitcairnioideae,Bromeliaceae)

Martina Rex, Kerstin Patzolt, Katharina Schulte, Georg Zizka, Roberto Vasquez,Pierre L. Ibisch, and Kurt Weising

Abstract: The neotropical genus Fosterella L.B. Smith (Pitcairnioideae, Bromeliaceae) comprises about 30 species, with acentre of diversity in semiarid to humid habitats of the Andean slopes and valleys of Bolivia. Morphologic differentiationof species is difficult because of a paucity of diagnostic characters, and little is known about the infrageneric phylogeny.Here, we present the results of an amplified fragment length polymorphism (AFLP) analysis of 77 Fosterella specimens,covering 18 recognized species and 9 as-yet undescribed morphospecies. Eight primer combinations produced 310 bands,which were scored as presence/absence characters. Neighbour-joining tree reconstruction revealed 12 clusters (A–L) withvarious levels of support. Well-supported species groups were also recovered by a principal coordinates analysis. With fewexceptions, morphologically defined species boundaries were confirmed by the molecular data. Phylogenetic relationshipsbetween species groups remained ambiguous, however, because of short internal branch lengths. The AFLP data werecomplemented by a survey of the leaf anatomy of 19 Fosterella species. Species concepts and assemblages are discussedin the context of molecular, morphologic, anatomic, ecologic, and biogeographic data. The data suggest that accidentallong-distance dispersal and founder events have been important for Fosterella speciation.

Key words: Fosterella, Bromeliaceae, Pitcairnioideae, AFLP, molecular phylogeny, leaf anatomy, biogeography.

Resume : Le genre neotropical Fosterella L.B. Smith (Pitcairnioideae, Bromeliaceae) compte environ 30 especes dont lecentre de diversite est situe dans les habitats semi-arides a humides des flancs et vallees andeens en Bolivie. La differen-ciation morphologique des especes est difficile en raison du faible nombre de caracteres diagnostiques et peu de chosessont connues quant a la phylogenie infragenerique. Les auteurs presentent ici les resultats de l’analyse du polymorphismede longueur des fragments amplifies (AFLP) sur 77 specimens de Fosterella couvrant 18 especes reconnues et 9 morpho-especes inedites. Huit combinaisons d’amorces ont produit 310 bandes qui ont ete notees comme etant presentes ou absen-tes. La reconstruction d’arbres ‘neighbor-joining’ a revele 12 groupes (A–L) supportes a des degres divers. Des groupesd’especes bien supportes ont aussi ete obtenus par analyse de coordonnees principales. A quelques exceptions pres, lesfrontieres entre les especes definies sur des bases morphologiques ont ete confirmees suite aux analyses moleculaires. Lesrelations phylogenetiques entre les groupes d’especes demeuraient cependant ambigues en raison de la faible longueur desbranches internes. Les donnees AFLP ont ete completees par un examen de l’anatomie des feuilles chez 19 especes dugenre Fosterella. Les concepts d’espece et de groupes d’especes sont discutes dans le contexte de donnees moleculaires,morphologiques, anatomiques, ecologiques et biogeographiques. Il est suggere que la dispersion accidentelle sur de longuesdistances et l’effet fondateur ont ete importants au cours de la speciation chez le genre Fosterella.

Mots cles : Fosterella, Bromeliaceae, Pitcairnioideae, AFLP, phylogenie moleculaire, anatomie des feuilles, biogeographie.

[Traduit par la Redaction]

Introduction

Bromeliaceae (Poales) are one of the largest neotropicalfamilies of flowering plants, comprising more than 2 600

species in 56 genera (Smith and Till 1998). Bromeliads arerenowned for their high ecological plasticity, allowing themto occupy a great variety of terrestrial and epiphytic habitats(e.g., arid coastal plains, humid montane forests, and high

Received 7 July 2006. Accepted 6 October 2006. Published on the NRC Research Press Web site at http://genome.nrc.ca on 14 March2007.

Corresponding Editor: B. Golding.

M. Rex and K. Weising.2 Plant Molecular Systematics, Institute of Biology, Dept. of Sciences, University of Kassel, Heinrich-Plett-Str.40, D-34132 Kassel, Germany.K. Patzolt, K. Schulte, and G. Zizka. Research Institute Senckenberg and J.W. Goethe-University, Frankfurt am Main, Germany.R. Vasquez. Sociedad Boliviana de Botanica, Casilla 3822, Santa Cruz, Bolivia.P.L. Ibisch.1 Fundacion Amigos de la Naturaleza (FAN), Dept. de Ciencias, Santa Cruz de la Sierra, Bolivia.

1Present address: Faculty of Forestry, University of Applied Sciences, Eberswalde, Germany.2Corresponding author (e-mail: [email protected]).

90

Genome 50: 90–105 (2007) doi:10.1139/G06-141 # 2007 NRC Canada

Page 2

Andean savannahs). Their great ecologic versatility appearsto be linked to several key innovations, including unique fo-liar trichomes capable of water absorption, succulence, foliarimpoundment, and crassulacean acid metabolism photosyn-thesis (reviewed by Benzing 2000). In spite of the speciesrichness of bromeliads and their abundance in neotropicalhabitats, updated revisions are lacking for most of the gen-era, and our understanding of speciation in this group ispoor.

The genus Fosterella comprises small to medium-sized(up to 1 m in the flowering stage) meso- to xerophytic ter-restrial bromeliad species. The usually stemless (acaules-cent) plants carry rosulate leaves. Flowers areinconspicuous and mostly whitish; fruits are capsules thatrelease large numbers of minute, appendaged seeds at ma-turity. Fosterella species are distinguished from other, pre-sumably adjacent, genera by naked petals, basifixed anthers(which are coiled at anthesis), and inner filaments adnate tothe petals. The genus is distributed from the Peruvian Andesand the Brazilian Amazon in the north to Paraguay andnorthern Argentina in the south, with a centre of diversityin arid and semihumid habitats of the northeast Andeanslopes of Bolivia (Ibisch et al. 2002). Many of its speciesare characterized by only small distribution ranges. Adisjunct occurrence in Central America is marked byF. micrantha. Fosterella has traditionally been assignedto the subfamily Pitcairnioideae of Bromeliaceae, mainlybecause of its fruit characters. In the most recent mono-graph of this subfamily, Smith and Downs (1974) recog-nized 13 Fosterella species. Since then, additional taxahave been formally described, raising the species numberto 30 (Rauh 1979, 1987; Luther 1981, 1997; Smith andRead 1992; Ibisch et al. 1997, 1999, 2002; Kessler etal. 1999). Little is known so far about phylogenetic rela-tionships within the genus.

Recent molecular studies, based on chloroplast DNA se-quences, have demonstrated that Pitcairnioideae, in their tra-ditional circumscription, are paraphyletic, and actuallyrepresent 6 independent evolutionary lineages (Terry et al.1997; Horres et al. 2000, 2007; Crayn et al. 2004; Givnishet al. 2004, 2007). Fosterella remained with Pitcairnioideaes. str. and proved to be monophyletic in all molecular stud-ies. According to the best-resolved chloroplast trees avail-able, its closest relatives are Deuterocohnia, Dyckia, andEncholirium (Crayn et al. 2004; Givnish et al. 2004). Sev-eral reasons account for the lack of phylogenetic informationat the infrageneric level. First, morphological differentiationof species is notoriously difficult, and the few diagnosticcharacters are only available in the flowering stage. Second,many of the Fosterella species are rare, endemic, and (or)display narrow ecologic amplitudes. Some species areknown only from type collections (Kromer et al. 1999;Ibisch et al. 2002). Third, Fosterella collections are poorlyrepresented in herbaria and botanical gardens. Fourth, sev-eral undescribed taxa are awaiting taxonomic treatment, andthe discovery of additional species is likely.

In a first attempt to shed light on the infrageneric phylog-eny of Fosterella, a random amplified polymorphic DNA(RAPD) analysis of 26 accessions from 12 species was per-formed (Rex 2001; Ibisch et al. 2002). Neighbour-joining(NJ) analysis of the RAPD dataset identified a set of species

groups, most notably the so-called penduliflora alliance(F. penduliflora, F. latifolia, F. chiquitana), which was sup-ported by a 99% bootstrap value. However, there were alsoinstances where specimens did not group according to theirspecies assignment, or occupied a more isolated position.Most species were represented by single accessions, andtherefore the general validity of conclusions from thisRAPD investigation suffered from limited sampling. The po-tential of various noncoding chloroplast DNA sequences forresolving relationships within Fosterella was also tested.Unfortunately, infrageneric chloroplast (cp)DNA sequencevariation proved to be relatively low, as it is in other generaof the presumably young crown group of Bromeliaceae (e.g.,Horres et al. 2000). Attempts to amplify the nuclear riboso-mal internal transcribed spacer (ITS) region with PCR, usingstandard universal primers, were unsuccessful. GeneratingITS sequence data from Bromeliaceae is known to be diffi-cult (Barfuss et al. 2004; Tuthill and Brown 2004) and, toour knowledge, no published ITS phylogenies of Bromelia-ceae currently exist.

It has been argued that amplified fragment length poly-morphisms (AFLPs) (Vos et al. 1995) are the method ofchoice for analyzing relationships between closely relatedtaxa, where DNA sequencing reveals hardly any variation(Hodkinson et al. 2000; Despres et al. 2003). AFLPs are ap-plicable to all organisms without a priori sequence informa-tion, represent a reasonably random sample of the totalgenome, and often show good reproducibility with the appli-cation of stringent PCR conditions (Hansen et al. 1999). Inthe current investigation, we used AFLP profiling to findout whether Fosterella species, recognized on the basis ofthe few morphologic characters available, also represent ge-netically distinct units; to address genetic relationshipsamong Fosterella species; and to identify the taxonomic af-finities of specimens with yet-unknown species designation.Genetic relationships suggested by the molecular data arediscussed in the context of morphology, leaf anatomy, ecol-ogy, and biogeography.

Material and methods

Plant materialA total of 77 Fosterella accessions, representing 18 of

the 30 recognized Fosterella species, were included in theAFLP analysis (Table 1). Nine accessions that do not corre-spond morphologically to any described species were provi-sionally assigned to morphospecies F. spec. 1 to 9. Thirty-five accessions from 19 Fosterella species were subjectedto a complementary leaf anatomical study (Table 1). Allbut 2 specimens (F. micrantha from Mexico and F. spec. 8from Brazil) were collected in Bolivia, or represent materialof Bolivian species taken from living collections elsewhere.For molecular studies, leaves were used fresh or werestored frozen until DNA extraction. Attempts to quick-drythe moderately succulent leaves by treating them with silicagel were unsuccessful and resulted in degraded DNA prep-arations. Voucher specimens and duplicates have been de-posited in various herbaria and living collections. Detailson collection numbers of voucher specimens, accessionnumbers of living material, and collection sites are givenin Table 1.

Rex et al. 91

# 2007 NRC Canada

Page 3

Table 1. Origins and accession numbers of the plant material used in the current study.

Species* Accession No./collector (herbarium voucher)DNANo.

Anatomyslide No. Origin and locale

F. albicans(Grisebach)L.B.Smith

FAN RV 3796I/ Vasquez 3796I (VASQ, SEL, USZ) 8b 869 Bolivia, Santa Cruz, Florida

FAN PI 98.0204/ Ibisch 98.0204 (GOET, LPB, SEL, USZ,FR)

64a — Bolivia, Santa Cruz, Florida

FAN RV 4185/Vasquez 4185 (SEL, VASQ) 54a — Bolivia, Cochabamba, AyopayaFAN RV 4626b/Vasquez (GOET, LPB, SEL, VASQ) 94c — Bolivia, Cochabamba, ChapareFAN RV 3617/ Rex & Schulte RS151002–2 (FR) 62a — BoliviaFAN RV 3317/Vasquez 3317 (GOET, LPB, FR) 51a 870 Bolivia, Cochabamba, ChapareFAN RV 4023/Vasquez 4023 (SEL, VASQ, FR) 32a — Bolivia, Tarija, Arcre

F. caulescensRauh

HEID 103532/Rauh 40579a (HEID) 3a 786 Germany, Botanical GardenHeidelberg

BGO 85-33-0157-30/- — 937 Germany, Botanical GardenOsnabruck

FAN RV 4654b/Vasquez 4654b (VASQ) 112a — Bolivia, La Paz, Nor YungasF. chiquitana

Ibisch,Vasquez, &Gross

FAN RV 3762/Vasquez 3762 (VASQ, FR) 34a 871 Bolivia, Santa Cruz, Nuflo deChaves

FAN CN 2221/Nowicki 2221 (LPB) 20a — Bolivia, Santa Cruz, Nuflo deChavez

FAN PI 98.0125/Ibisch 98.0125 (FR, LPB) 11c 893 Bolivia, Santa Cruz, Nuflo deChavez

FAN RV 4685/Vasquez 4685 (VASQ) 120a — Bolivia, Santa Cruz, GuarayosFAN RS 301002-3/ Vasquez, Osinaga, Rex & Schulte

301002-3 (FR)118a — Bolivia, Santa Cruz, Guarayos

FAN RS 301002-1/Vasquez, Osinaga, Rex & Schulte301002-1 (LPB, FR)

— 936 Bolivia, Santa Cruz, Guarayos

F. cotacajensisKessler, Ibisch& Gross

FAN MK 9620b/Kromer 9620 (SEL, LPB) 76d 872 Bolivia, Cochabamba, Ayopaya

FAN IV 6377/Schulte 180901-20 (FR) — 876F. elata, H.

LutherFAN RV 4642/Vasquez 4642 (SEL, VASQ, FR) 119a — Bolivia, La Paz, Nor Yungas

FAN RS 251002-3/Vasquez, Osinaga, Rex & Schulte251002-3 (FR, SEL)

107a — Bolivia, La Paz, Nor Yungas

FAN RV 4655/Vasquez 4655 (VASQ, FR) 122b — Bolivia, La Paz, CaranaviFAN RS 261002-4/Vasquez, Osinaga, Rex & Schulte

261002-4 (FR)113a — Bolivia, La Paz, Nor Yungas

MSBG 1981-0059/SEL 63851 (SEL) 82a — Bolivia, La Paz, MunecasMSBG 1983-0004/SEL 63845 (SEL) 81a — Bolivia, La Paz, Nor YungasFAN CN 2061/- 60a — Bolivia, close to Rio La Paz,FAN RV 3633a/Vasquez 3633 (LPB, USZ, SEL, FR) 16c — Bolivia, La Paz, Sur YungasFAN RV 3654/Vasquez 3654 (VASQ, USZ, FR) 15a 616 Bolivia, La Paz, Nor YungasFAN RV 4177/Vasquez 4177 (VASQ, FR) 31a — Bolivia, Cochabamba, Ayopaya

F. floridensisIbisch, Vasquez& Gross

FAN RS 111002-11/Vasquez, Osinaga, Rex & Schulte111002-11 (FR)

103a 875 Bolivia, Santa Cruz, Florida

FAN PI 02.0001/Rex & Schulte 151002-23 (FR) 67b 874 Bolivia, Santa Cruz, FloridaFAN RV 3796/Vasquez 3796b (SEL, USZ, FR) — 906 Bolivia, Santa Cruz, Florida

F. gracilis(Rusby) L.B.Smith

FAN RS 281002-6/- 117a — Bolivia, Beni, Ballivian

FAN RS 281002-3/Vasquez, Osinaga, Rex & Schulte281002-3

— 945 Bolivia, Beni, Ballivian

FAN RS 281002-1/Vasquez, Osinaga, Rex & Schulte281002-1

— 947 Bolivia, Beni, Ballivian

92 Genome Vol. 50, 2007

# 2007 NRC Canada

Page 4

Table 1 (continued).

Species* Accession No./collector (herbarium voucher)DNANo.

Anatomyslide No. Origin and locale

F. graminea(L.B.Smith)L.B. Smith

FAN RM 216/Muller 216 (SEL) 71a — Bolivia, La Paz, Larecaja

BGO 115-19-83-83/- — 950 Germany, Botanical GardenOsnabruck

FAN RM 22/Muller 22 (LPB, USZ) 70a — Bolivia, La Paz, LarecajaF. heterophylla

RauhFAN RV 3661/Rex & Schulte 151002-24 (LPB, SEL, FR) 26a 894, 878 Bolivia, La Paz, Caranavi

FAN RS 241002-1/Vasquez, Osinaga, Rex & Schulte241002-1 (FR, SEL)

— 935 Bolivia, La Paz, Nor Yungas

FAN RS 241002-5/Vasquez, Osinaga, Rex & Schulte241002-5 (FR)

105a — Bolivia, La Paz, Nor Yungas

F. latifoliaIbisch, Vasquez& Gross

FAN PI 98.0098/Schulte 141002-7 (LPB, FR) 18c 599 Bolivia, Santa Cruz, Florida

FAN RS 111002-1/Vasquez, Osinaga, Rex & Schulte111002-1 (FR)

101a — Bolivia, Santa Cruz, Florida

FAN RV 3406/Schulte 141002-8 (FR) 35a 879 Germany, Greenhouse,University of Kassel

F. micrantha(Lindl.)L.B.Smith

HEID 103726/Schulte 171103-1 (FR) — 907 Germany, Botanical GardenHeidelberg

BGB 079-02-92-34/Schulte 101203-3 (FR) — 923 Germany, Botanical GardenBerlin-Dahlem

F. nowickiiIbisch, Vas-quez, & Gross

FAN RV 3636d/Schulte 141002-6 (FR) 12b 608 Bolivia, La Paz, Sur Yungas

FAN CN 2076 b/Nowicki 2076 (SEL,LPB, FR) 58c 880 Bolivia, La Paz, Sur YungasFAN CN 2060a/Nowicki 2060a (FR) 27b — Bolivia, Cochabamba, Hernando

SilesF. penduliflora

(C.H.Wright)L.B.Smith

FAN RV 2979/Vasquez 2979 (USZ, FR) 43a — Bolivia, Santa Cruz, Florida

FAN RV 3710/Vasquez (VASQ) 21a — Bolivia, Santa Cruz, CordilleraFAN RV 4051b/Vasquez (VASQ, FR) 46d — Germany, Greenhouse, Univer-

sity of KasselFAN RV 3624/Vasquez (VASQ, FR) 55a — Bolivia, Chuquisaca, InquisiviFAN RV 4003/Vasquez (LPB, SEL, USZ, VASQ) 45a — Bolivia, Tarija, Gran ChacoFAN RV 3724/Vasquez (FR) 57a — Bolivia, Santa Cruz, CordilleraFAN RV 3730/Vasquez 3730 (VASQ, FR) 17a 881 Bolivia, Cochabamba, Luis

CalvoF. rexiae Ibisch,

Vasquez &Gross

FAN RV 3666/Vasquez 3666 (USZ) 10a 602 Bolivia, La Paz, Caranavi

FAN RV 3673/Vasquez 3673 (LPB), SEL 9d 607, 883 Bolivia, La Paz, CaranaviF. spectabilis H.

LutherMSBG 1995–0415/(SEL) 87a — USA, Florida, Marie Selby

Botanical GardensBGB 290-08-00-84/Leuenberger 41388 (Mus. Bot. Berol.,

Gartenherbar)— 949 Germany, Botanical Garden

Berlin-DahlemF. vasquezii

Gross & IbischFAN PI 98.0117/- 23b — Bolivia, Santa Cruz, Velasco

FAN PI 98.116/Schulte 020603-8 (FR) Schulte 020603-8(FR)

63a 897, 887 Bolivia, Santa Cruz, Velasco

F. villosula(Harms)L.B.Smith

FAN ‘‘H. Justiniano s.n.’’/H. Justiniano s.n.(VASQ), 59a — Bolivia, Santa Cruz, Florida

FAN RV 2994/Vasquez 2994 (USZ, FR) 39a 889 Bolivia, Santa Cruz, A. IbanezFAN PI 98.0173/Schulte 151002–7 (USZ, FR) 25a 618, 888 Bolivia, Santa Cruz, A. IbanezFAN PI 02.0002/Ibisch 02.0002 (FR) 66a — Bolivia, Santa Cruz, Florida

Rex et al. 93

# 2007 NRC Canada

Page 5

DNA isolationTotal genomic DNA, from 100 to 150 mg of fresh or fro-

zen leaves of individual plants, was isolated, either with aDNeasy Plant Mini Kit (Qiagen, Hilden, Germany) or withvarious modifications of the cetyl trimethylammonium bro-mide (CTAB) procedure (Weising et al. 2005). Final DNAconcentrations were determined electrophoretically, usingknown amounts of �-DNA as standards.

AFLP assaysAFLP analyses were performed in accordance with the

protocols described by Debener and Mattiesch (1999) andBanfer et al. (2004), with some modifications. In short,250 ng of genomic DNA were double-digested in a finalvolume of 30 mL at 37 8C with HindIII and MseI overnight,and ligated to HindIII and MseI adapters in the same reac-

tion. Two consecutive PCR amplifications with primers,containing either 1 (+1) or 3 (+3) selective nucleotides attheir 3’ ends (Table 2), were then carried out in a BiometraT-Gradient Cycler (Biometra, Gottingen, Germany) or aBio-Rad i-Cycler (Bio-Rad Laboratories, Munich, Ger-many).

In the first, preselective PCR, 5 mL of the restriction-liga-tion assay were combined with 2.5 ng/mL of unlabeledMseI+1 primer, 1.5 ng/mL of unlabeled HindIII+1 primer,1.5 mmol/L MgCl2, 0.2 mmol/L of each deoxynucleosidetriphosphate (dNTP), 20 mmol/L Tris–HCl (pH 8.0),50 mmol/L KCl, and 1 U Taq DNA polymerase (Invitrogen,Karlsruhe, Germany). Assays were adjusted to 50 mL withdistilled water and subjected to 20 cycles of amplification,each consisting of 94 8C for 30 s, 60 8C for 30 s, and72 8C for 60 s. Final extension was at 72 8C for 7 min. Se-

Table 1 (concluded).

Species* Accession No./collector (herbarium voucher)DNANo.

Anatomyslide No. Origin and locale

FAN RV 4623/Vasquez 4623 (VASQ, FR) 104a — Bolivia, Cochabamba, ChapareFAN RV 3318/Vasquez 3318 (VASQ) 24c — Bolivia, Cochabamba, ChapareFAN RV 4650a/Vasquez 4650a (VASQ, FR) 126a — Bolivia, La Paz, Nor YungasFAN RV 3322/Vasquez 3322 (VASQ) 40a — Bolivia, Cochabamba, Chapare

F. weberbaueri(Mez)L.B.Smith

FAN RV 3570/Vasquez 3570 (VASQ, SEL, LPB, USZ,FR)

48a — Bolivia, Cochabamba, Chapare

FAN RM 217*2/(SEL) 95a 896 Bolivia, La Paz, LarecajaFAN RV 4656/Schulte 020603–4 (FR) 121a 895 Bolivia, La Paz, CaranaviFAN RS 261002-18/Vasquez, Osinaga, Rex & Schulte

261002-18 (FR)114a — Bolivia, La Paz, Nor Yungas

F. cf. weber-baueri

FAN RV 4650/- 122b — Bolivia, La Paz, Caranavi

F. weddelliana(Brongniart exBaker)L.B.Smith

FAN CN 2034/Nowicki 2034 (FR) 37a — Bolivia

FAN RV 3612/Vasquez 3612 (VASQ, FR) 13a 611 Bolivia, La Paz, InquisiviFAN RV 3620/Vasquez 3620 (SEL, VASQ) 56a — Bolivia, La Paz, InquisiviFAN RV 3627/Vasquez (VASQ, FR) 36a — Bolivia, La Paz, Inquisivi

F. spec 1 FAN IV 6387/Vargas 6387 (FR) 30a — Bolivia, Cochabamba, AyopayaF. spec 2 FAN RS 261002-10/Vasquez, Osinaga, Rex & Schulte

261002-10 (FR)111a — Bolivia, La Paz, Nor Yungas

F. spec. 2 FAN RV 3652/Vasquez 3652 (VASQ, FR) 29a — Bolivia, La Paz, Nor YungasF. spec. 3 FAN TK 1398b/Kromer 1398b (USZ) 28d 948, 898 Bolivia, La Paz, CaranaviF. spec. 4 FAN RV 3729a/Vasquez 3729a (VASQ) 128a — Bolivia, Chuquisaca, Luis CalvoF. spec. 5 MSBG 1980-0854/- 88a — USA, Florida, Marie Selby

Botanical GardensF. spec. 6 FAN RS 271002-1/Vasquez, Osinaga, Rex & Schulte

261002-10 (FR)116a — Bolivia, Beni, Ballivian

F. spec. 7 FAN RV 3674a/Vasquez 3674a (SEL, VASQ, FR) 49b — Bolivia, La Paz, CaranaviF. spec. 8 J. B. Fernandes da Silva s.n. (cult. E. Leme 5078/HB); 129a — Brazil, Para, ItaitubaF. spec. 9 MSBG 1980-1733/ 79a — USA, Florida, Marie Selby

Botanical Gardens

Note: BGB, Botanical Garden Berlin-Dahlem; BGO, Botanical Garden Osnabruck, Germany; FAN, living collections of the Fundacion Amigos de laNaturaleza, Santa Cruz, Bolivia; FR, Forschungsinstitut Senckenburg, Frankfurt/Main, Germany; GOET, Herbarium of the Georg-August-University, Gottin-gen, Germany; HB, Herbarium Bradeanum, Rio de Janeiro, Brazil; HEID, Botanical Garden of the University of Heidelberg, Germany; LPB, Herbario Na-cional de Bolivia, La Paz, Bolivia; MSBG and SEL, Marie Selby Botanical Garden, Sarasota, US; USZ, Universdad Autonoma Gabriel Rene Moreno, SantaCruz, Bolivia; VASQ, Herbarium Vasquezianum.*Fosterella species are arranged alphabetically. Nomenclature follows Smith and Downs (1974), and for later described species Ibisch et al. (1997,

1999, 2002), Kessler et al. (1999), Luther (1981, 1997), Rauh (1979, 1987), and Smith and Read (1992). Abbreviations of author names follow Brummitand Powell (1992). Fosterella spec. 1–9 are as yet unnamed morphospecies.

94 Genome Vol. 50, 2007

# 2007 NRC Canada

Page 6

lective PCRs were then performed with 2.5 mL of a 1:20diluted preamplified DNA as a template, and various combi-nations of an unlabeled MseI+3 primer (3 ng/mL) (Roth,Karlsruhe, Germany) and a fluorescence-labeled HindIII+3primer (either 3 ng/mL of IRDye700 or 3.6 ng/mL of IR-Dye800-labeled primer) (MWG Biotech, Ebersberg,Germany) (Table 2). Each 30 mL assay also contained2 mmol/L MgCl2, 0.2 mmol/L of each dNTP, 0.3% TritonX-100, 20 mmol/L Tris–HCl (pH 8.0), 50 mmol/L KCl, and0.25 U Taq DNA polymerase (Invitrogen). A touchdownprotocol was followed for 47 cycles, each consisting of94 8C for 30 s, 658 to 56 8C for 30 s, and 72 8C for 60 s.Starting at 65 8C, the annealing temperature was reduced by0.8 8C per cycle during the first 11 cycles, and then left con-stant at 56 8C. Final extension was at 72 8C for 7 min.

Final products of the selective PCR were mixed with anequal volume of loading buffer, containing 98% (v/v) forma-mide, and denatured at 80 8C for 5 min. Samples (0.5 mL) ofeach reaction were electrophoresed on denaturing 41 cm �0.2 mm polyacrylamide gels (6% Sequagel XR (NationalDiagnostics, Atlanta, Ga.) in 1æ TBE buffer (Sambrook andRussell 2001)), using an automated sequencer (LI-COR4200 IR2; LI-COR Inc., Lincoln, Nebr.). Fragment mobilitywas measured with real-time laser scanning. Gel imageswere stored electronically for further analysis.

Data analysisAFLP banding patterns were recorded as the presence (1)

or absence (0) of a band at a particular position, with thehelp of GenImageIR version 4.02 software (ScanalyticsInc., Fairfax, Va.). The raw data were edited manually, aftervisual inspection. Faint or fuzzy bands were generallyignored. Scored fragment sizes ranged from 40 to 575 nu-cleotides. Standard lanes carrying identical samples wererun on each gel, which facilitated the combination of data-sets from 2 or more independent runs.

For tree construction, the binary character matrix wasconverted into a distance matrix based on the Nei and Li(1979) ( = Dice) index of similarity, using NTSYSpc 2.10p(Rohlf 2000) or Treecon 1.3b software (Van de Peer and DeWachter 1994). Phenograms were generated either by un-weighted pair group method with arithmetic mean(UPGMA) cluster analysis (Sneath and Sokal 1973) or bythe neighbour-joining (NJ) algorithm (Saitou and Nei 1987),using the same program packages. An unweighted maximum

parsimony analysis of the AFLP character matrix was con-ducted using PAUP* 4.0b10 (Swofford 2002). Characterchanges were interpreted under ACCTRAN optimizationon, random addition, tree bisection-reconnection swappingoptions, and Maxtrees set to increase without limits.Statistical support for individual furcations in both the NJand maximum parsimony trees was evaluated withbootstrapping (Felsenstein 1985). Bootstrap analyses used1000 replicates. The complemented programs INTERVALDATA, DCENTER, and EIGEN implemented in theNTSYSpc 2.10p software package (Rohlf 2000) were usedto perform a principle coordinates analysis.

Leaf blade anatomyMature leaves were used fresh, or fixed in formalin/acetic

acid/alcohol (FAA) and stored in 70% ethanol. Completetransverse-hand sections of the leaf blade (about halfwaybetween the leaf tip and base) were prepared from unstainedfresh material. Microtome sections were obtained frommaterial either embedded in Paraplast and stained withsafranine-astrablue or embedded in HEMA (Igersheim andCichocki 1996) and stained with toluidine. The sections wereinvestigated and documented with light microscopy (LeicaDialux 22), a digital camera system (Leica DC 300), andcamera lucida drawings. For measuring and analyzing thesections, IM 1000 software (Leica) was applied.

Morphology and biogeographyThe assessment of morphologic, ecologic, and biogeo-

graphic characters was derived from a database comprisingmore than 360 specimens of all known Fosterella species.The database includes data from the literature (Smith andDowns 1974; Rauh 1979, 1987; Luther 1981, 1997; Smithand Read 1992; Ibisch et al. 1997, 1999, 2002; Kessler etal. 1999), information obtained from the study of cultivatedplants (mainly in the Living Collection of the FundacionAmigos de la Naturaleza, Santa Cruz, Bolivia), and personalobservations that were made during numerous field trips toall ecoregions of Bolivia (Pierre L. Ibisch, 1993–2003).The classification of the Bolivian ecoregions follows thatdescribed by Ibisch et al. (2004).

Results

AFLP dataIn pilot experiments, 54 combinations of labeled HindIII-

primer and unlabeled MseI-primer with +3 selective baseseach were screened with a small set of template DNAs.Well-resolved banding patterns were obtained with 13 ofthese combinations, and 8 primer pairs were eventually re-tained to analyse the entire set of 112 samples (Table 2). Atotal of 77 template DNA samples produced distinct AFLPprofiles with all primer sets, and were included in the phylo-genetic analysis (Table 1). Unfortunately, we were not ableto generate consistent AFLP patterns from F. micrantha,which is the only Fosterella species present in CentralAmerica. The reproducibility of banding patterns was exem-plarily tested with 2 primer combinations and 61 taxa. Num-bers of unequivocally scorable band positions varied from30 to 52, depending on the primer pair. Of 310 band posi-tions totally scored, 96% were variable (Table 2). AFLP fin-

Table 2. Primer combinations used for selectiveamplification.

Primer combinationTotalbands

Polymorphicbands (%)

Mse+CAA/Hind+ACC 52 94Mse+CAC/Hind+ACC 34 94Mse+CAC/Hind+AGC 32 97Mse+CAA/Hind+ACA 36 94Mse+CAA/Hind+AAG 42 100Mse+AAG/Hind+AAC 32 94Mse+AGC/Hind+ACG 30 100Mse+AAG/Hind+ACA 52 96Total 310 96

Rex et al. 95

# 2007 NRC Canada

Page 7

gerprints generated by the primer combination Mse+CAAand Hind+AAG from a subset of samples are exemplarilyshown in Fig. 1.

Phenetic analysesPairwise Nei and Li (1979) similarity values ranged from

0.39 to 0.95 for all samples, and generally exceeded 0.66 atthe species level. The single plant of F. spectabilis and 1 ofthe 2 investigated samples of F. graminea exhibited thelargest distances from the other specimens and were arbitra-rily chosen to root the trees. Defining alternative outgroupsor using the midpoint rooting option had no major influenceon tree topology (not shown). Both the UPGMA and NJ treereceived cophenetic correlation coefficients > 0.83, and ex-hibited almost identical branching patterns. Only the NJ treeis shown here (Fig. 2). It is subdivided into 12 clusters thatreceive various levels of bootstrap support (BS). The moder-ately well-supported cluster A (73% BS) comprises allspecimens of F. elata. Clusters B and C each unite 2 con-specific accessions from closely adjacent localities (i.e.,F. vasquezii (87% BS) and F. floridensis (100% BS), re-spectively). Cluster D (87% BS) harbours 2 accessions ofF. heterophylla and 2 accessions of the unclassified F. spec.2. The monophyly of these 2 species also receives good sup-port. Cluster E is again moderately well-supported (69% BS)

and contains all specimens of F. nowickii andF. weddelliana, with the clonotype specimen of F. cotaca-jensis in a basal position. The monospecific cluster F (58%BS) is formed by the 2 accessions of F. caulescens. F. re-xiae, together with the unclassified F. spec. 3, makes upcluster G (73% BS). All but 1 of the investigated accessionsof F. albicans, plus the unclassified F. spec. 4, constitute theunsupported cluster H. Species groups A–H are united in alarge (but unsupported) supercluster, with F. graminea(RM216), F. gracilis, and F. albicans (RV4023) forming abasal grade.

The weakly defined cluster I contains a second F. grami-nea (RM22) and the unclassified F. spec. 5 and F. spec. 6,whereas the well-supported cluster J (98% BS) comprises allaccessions of F. villosula. Internal bootstrap values withincluster J are also relatively high, suggesting considerable sub-division within F. villosula. Cluster K (72% BS) comprisesall specimens of F. penduliflora, F. latifolia, and F. chiqui-tana, plus the unclassified F. spec. 7. The samples from these4 apparently closely related species are intermingled witheach other. Cluster L (57% BS) harbours all sampled plantsof F. weberbaueri, together with an unclassified species (F.spec. 8), which was collected in Brazil.

With few exceptions, relationships between the clustersoutlined above remain obscure. Clusters D and E are sister

Fig. 1. Amplified fragment length polymorphism (AFLP) profiles generated from a subset of the investigated Fosterella specimens afterselective amplification with a combination of unlabeled Mse+CAA and IRDye700-labeled Hind+AAG primers. PCR products were sepa-rated on a 6% denaturing polyacrylamide gel in an automated Li-COR sequencer.

96 Genome Vol. 50, 2007

# 2007 NRC Canada

Page 8

Fig. 2. Neighbour-joining tree of 77 Fosterella accessions based on 310 band positions obtained with 8 AFLP primer pair combinations.The Nei and Li (1979) index of similarity was used to generate the distance matrix used for tree calculation. Clusters A to L are referred toin the text. Symbols behind letters A to L are used in the principle coordinates analysis shown in Fig. 3. Numbers above the branches arebootstrap values obtained from 1000 replicates. The similarity scale is indicated by a horizontal bar.

Rex et al. 97

# 2007 NRC Canada

Page 9

groups, as are J and K, but each with very low support (BS57% and 50%, respectively). The latter relationship was alsoobserved in our preliminary RAPD study, where these 2clusters were referred to as villosula and pendulifloragroups, respectively (Ibisch et al. 2002). Finally, clusters I,J, and K are united in a larger cluster that receives 53%bootstrap support.

Phylogenetic analysisThe original binary character matrix was also subjected to

a phylogenetic analysis using maximum parsimony as an op-timality criterion, as was also done in other AFLP studies(e.g., Kardolus et al. 1998; Koopman et al. 2001; Despres etal. 2003). Defining F. spectabilis as an outgroup, PAUP*found 4 trees with 2491 steps, a consistency index of 0.12and a retention index of 0.5164 (not shown). The low con-sistency index value indicates a high level of homoplasy, aswould be expected with the application of AFLPs to a largedataset (see Despres et al. 2003 for a discussion). The strictconsensus tree displays a basal polytomy of various clades,the species compositions of which are generally congruentwith the clusters defined above in the NJ tree. The followingclades (clusters) receive bootstrap supports above 50% inthe maximum parsimony tree: A (56%), B (81%), C (99%),D (57%), E (65%), J (63%), K (97%), and L (53%). Otherclades, and relationships between clades, are not resolved.

Principal coordinates analysisRelationships among Fosterella specimens investigated by

AFLP are further illustrated in the results of a principal co-ordinates analysis (Fig. 3). The first 3 coordinates togetheraccount for 52% (i.e., 28.4%, 12.2%, and 11.4%) of the totalvariance. Essentially, the same groups are revealed as in theNJ tree, but resolution varies between the individual groups.Whereas clusters A, B, D, J, K, and L are each relativelywell resolved in the 3-dimensional principle coordinatesplot, other groups show considerable overlap.

Leaf blade anatomyOne group of 5 species (F. villosula, F. micrantha,

F. penduliflora, F. latifolia, and F. chiquitana), referred toas leaf anatomy group 1, is characterized by a set of com-mon leaf anatomical characters (exemplified in Figs. 4a–4c): cell walls of the adaxial epidermis are uniform andhardly thickened; sclerotic hypodermis is adaxially absent(Fig. 4b); chlorenchyma consists of isodiametric cells only,and is sharply demarcated from the adaxial water storagetissue; vascular bundles (first order) are adaxially not adja-cent to water storage tissue and are embedded in chloren-chyma (Fig. 4c); aerenchyma is loosely packed; abaxialwater storage tissue forms very large cushion-like groups(Fig. 4c); and the abaxial epidermis is planar. With the ex-ception of F. micrantha, where no AFLP data are available,the species of leaf anatomy group 1 fully correspond to clus-ters J and K of the AFLP tree (see Fig. 2).

The remainder of the investigated species are anatomi-cally quite heterogeneous. However, a group of 3 species(F. nowickii, F. cotacajensis, F. weddelliana (leaf anatomygroup 2, exemplified in Figs. 4d–4f)) share a number ofcharacters, as follows: cell walls of the adaxial epidermisare uniform and slightly thickened; sclerotic hypodermis is

adaxially present (Fig. 4e); chlorenchyma consists of isodia-metric and palisade-like cells; chlorenchyma is not sharplydemarcated from the adaxial water storage tissue but isseparated from the latter by an area of poorly differentiatedcells; vascular bundles (first order) are adaxially adjacent towater storage tissue (Fig. 4f); aerenchyma is more denselypacked; abaxial water storage tissue forms small tomedium-sized cushion-like groups (Fig. 4f); and abaxialepidermis has slight depressions. The species of this groupare also closely related to each other in the AFLP tree,where they form cluster E (see Fig. 2).

To sum up, the diacritic leaf anatomical characters are:the shape and thickening of cells of the adaxial epidermis;the absence/presence of adaxial sclerotic hypodermis; theform of cells of the chlorenchyma; the demarcation of chlor-enchyma from the adaxial water storage tissue; the positionof vascular bundles within the mesophyll; and the form andsize of abaxial water storage tissue.

Morphology, ecology, and biogeographyTable 3 summarizes the distributional patterns and some

selected characters of Fosterella species that were includedin the AFLP sampling. Several groups and individual spe-cies can be discerned by morphology. One group comprisespurely Andean species, with a caulescent-branched habit,serrate leaves, and petals that are recoiled like watch springs(and that remain so even after anthesis) (Fig. 5, right panel).This group contains 4 species: F. cotacajensis, F. nowickii,F. weddelliana, and F. caulescens. It is morphologically re-lated to some Andean acaulescent species with serrate leavesand recoiled petals, such as F. rexiae and F. albicans. An-other group is formed by several species with more-or-lessbroad-leaved acaulescent flat rosettes, entire leaves, andlily-like petals, such as F. chiquitana, F. latifolia,F. penduliflora, and F. villosula (Fig. 5, left panel). Speciesthat are morphologically clearly distinct are F. floridensis,which are narrow-leaved, acaulescent, and have a more-or-less upright rosette and straight petals, and F. spectabilis,which have rather long and straight reddish-orange petals.All other Fosterella species have whitish or cream-colouredpetals (yellow in F. gracilis). F. spec. 8 is remarkable be-cause of its disjunct occurrence in the northern Amazon,but shows morphologic affinities to F. weberbaueri. TheFosterella species with the most disjunct distribution, F. mi-crantha from Central America, is morphologically similar toF. villosula. With 1 exception (i.e., the inter-Andean groupwith F. cotacajensis and others), morphologic groups are re-flected by neither geographic distribution nor by habitat andecology.

Discussion

Whereas the monophyly and generic circumscription ofFosterella have never been questioned, species boundarieshave been difficult to define, and little is known about infra-generic relationships. The current AFLP study generallysupports species concepts that have been based on morpho-logic characters, habitat preferences, and geographic distri-bution (Ibisch et al. 1997, 1999, 2002). Thus, with theexception of F. graminea, F. albicans, and the species ofthe ‘‘penduliflora’’ alliance, all investigated accessions

98 Genome Vol. 50, 2007

# 2007 NRC Canada

Page 10

formed groups that were in full accordance with their spe-cies designation. Moreover, the AFLP data also define anumber of small species groups with various levels of statis-tical support, the coherence of which is discussed below.

F. elata, F. vasquezii, and F. floridensisThe analyzed specimens of F. elata, F. vasquezii, and

F. floridensis constitute clusters A, B, and C, respectively,which together form a monophylum in the NJ tree (Fig. 2).Their relative proximity in the tree seems to indicate a closerelationship of the 3 species, but this remains elusive be-cause of the lack of bootstrap support. F. elata and F. vas-quezii share a suite of morphologic similarities, especiallywith regard to flower and inflorescence characters (Table 3),but their habitats and geographic distribution differ consider-ably. F. vasquezii grows on Precambrian rocks of the Capa-rus mountains in northeastern Bolivia. These sites areinfluenced by the spray of waterfalls and might have beenrather humid, even during arid climatic periods. It is there-fore possible that F. vasquezii or its direct ancestors colon-ized this presumably very old habitat a long time ago. Incontrast, F. elata lives in the understory of the Andean Yun-gas cloud forests in northwestern Bolivia. These habitats arerelatively young, because the Andes have risen rather re-cently, in the Tertiary period. F. floridensis, known onlyfrom a very small range at the Andean knee, is clearly mor-phologically and ecologically distinct from both F. elata andF. vasquezii. It colonizes relatively raw soils, and withstandsa dry climate, in which there is little precipitation for severalmonths. It is also one of the few species with straight petalsthat are neither permanently nor temporarily recoiled or re-curved.

F. heterophylla, F. spec. 2, F. cotacajensis, F. nowickii,and F. weddelliana

This purely Andean group of 5 species is characterized by

flowers with permanently recoiled petals. Cluster D isformed by F. heterophylla and the 2 accessions of F. spec.2. These 2 species, known only from the Andean Yungascloud forests, are both stemless and possess entire leaves.They are obviously closely related. Cluster E comprises thestem-building and branched species F. cotacajensis,F. nowickii, and F. weddelliana, which also share the sametype of leaf blade anatomy (Figs. 4d–4f), have serrateleaves, and grow in relatively dry inter-Andean valleys.F. cotacajensis, found in the understory of a completely de-ciduous dry forest, withstands a harsh semiarid climate.F. nowickii and F. weddelliana grow at lower altitudes,where they prefer sites with a more semihumid climate. Theleaves of this subgroup’s species are very succulent andstiff, and the slow-growing and branched plants can possiblyreach considerable ages. Thus, despite their weakly sup-ported sister group status in the AFLP tree, the species ofclusters D and E are differentiated both morphologicallyand ecologically from each other.

F. rexiae, F. caulescens, F. spec. 3, F. spec. 4, andF. albicans

The 2 accessions each of F. rexiae and F. caulescensform groups that correspond to their species assignment,although with variable bootstrap support. The unclassifiedF. spec. 3 shows some affinity to F. rexiae. The F. albicansaccessions form only a loose and unsupported group, whichalso includes F. spec. 4, and their taxonomic status mightneed some revision. The species combined in clusters F, G,and H are also morphologically heterogeneous. All except F.spec. 4 are characterized by recoiled petals, and all exceptF. spec. 3 and F. spec. 4 have serrate leaves. Only 1 speciesis caulescent (F. caulescens). F. rexiae, F. caulescens, andF. spec. 3 are from the northern Yungas Andean rain forest(Department of La Paz), where F. spec. 4 and F. albicans donot occur. In accordance with its genetic inhomogeneity,

Fig. 3. Principal coordinates analysis of AFLP data from Fosterella, based on the same dataset that was used to construct the neighbour-joining (NJ) tree. Axes are scaled with artificial units. Specimens belonging to a particular cluster in the NJ tree (A to L in Fig. 2) share thesame symbol. Well-resolved species groups are highlighted. For details see text. Specimens marked with an open circle are not assigned toany subclade.

Rex et al. 99

# 2007 NRC Canada

Page 11

F. albicans also shows a wide distributional range (fromnorthern Argentina to central Bolivia) and grows in varioushabitats. It is a typical species of the Tucuman-Bolivian for-est ecoregion, but is also present in the montane rainforestsof the Yungas ecoregion. F. spec. 4, until now, has beenknown from only 1 collection in the Tucuman-Bolivian for-est.

F. graminea, F. spec. 5, and F. spec. 6F. graminea, a species from the sub-Andean rain forests

of northern Bolivia, is characterized by narrow serrateleaves and recoiled petals. The 2 F. graminea specimensstudied here do not group together, showing, as F. albicansdoes, some necessity of further taxonomic revision. Thesamples of F. spec. 5 and F. spec. 6 group together, but aregenetically quite distant from F. graminea and from anyother species. They need to be studied more closely.

F. villosulaThe 8 accessions that are morphologically recognized as

Fig. 4. Cross sections of Fosterella leaves. (a–c) Leaf anatomy group 1: (a) F. chiquitana; (b) F. villosula, adaxially; (c) F. micrantha,abaxially; (d–f), leaf anatomy group 2: (d) F. cotacajensis; (e) F. cotacajensis, adaxially; (f) F. nowickii, abaxially. A, air lacunae; C,chlorenchyma; E, epidermis; S, sclerotic hypodermis; WT, water storage tissue. For details, see text.

100 Genome Vol. 50, 2007

# 2007 NRC Canada

Page 12

Tab

le3.

Dis

trib

utio

npa

ttern

s,ha

bita

ts,

ecor

egio

ns,

and

sele

cted

anat

omic

and

mor

phol

ogic

char

acte

rsof

Fos

tere

lla

spec

ies

incl

uded

inth

eA

FLP

sam

plin

g.

Mor

phol

ogic

and

anat

omic

char

acte

rsB

ioge

ogra

phy

Spec

ies*

CL

LSA

BM

PHPL

PSPC

Geo

grap

hic

dist

ribu

tion

RS

Eco

regi

on

Alt.

rang

e(m

abov

ese

ale

vel)

F.

elat

aA

abse

rbl

a7

rec

cB

:L

P(N

.&

S.Y

unga

s,C

aran

avi,

Inqu

isiv

i,M

une-

cas)

,C

B(A

yopa

ya)

smSA

A,

Y80

0–21

00

F.

vasq

uezi

iB

abse

rbl

a5

rec

wB

:SC

(Vel

asco

)le

BSC

A30

0–85

0F

.fl

orid

ensi

sC

abe

nla

14st

rw

B:

SC(F

lori

da)

leIA

D,

Y45

0–14

50F

.sp

ec.

1is

os

ebl

a>

3re

cw

B:

CB

(Ayo

paya

)n.

d.IA

D,

Y11

50.

F.

spec

.2

Dab

ebl

a4–

5re

cw

B:

LP

(N.

Yun

gas)

leY

900–

2000

F.

hete

roph

ylla

Ds

ecb

5re

cw

B:

LP

(N.

&S.

Yun

gas,

Car

anav

i)sm

Y10

00–2

200

F.

cota

caje

nsis

Eab

ser

cb8

rec

wB

:L

P(I

nqui

sivi

),C

B(A

yopa

ya)

smIA

D21

00–2

200

F.

now

icki

iE

abse

rcb

5re

cg

B:

LP

(S.

Yun

gas)

smIA

D,

Y90

0–14

00F

.w

edde

llia

naE

abse

rnl

a4

rec

wB

:L

P(I

nqui

sivi

)sm

IAD

,Y

1250

–180

0F

.ca

ules

cens

Fab

ser

cb7

rec

pB

:L

P(N

.Y

unga

s)sm

SAA

,Y

800–

1200

F.

rexi

aeG

abse

rnl

a5

rec

wB

:L

P(C

aran

avi)

leSA

A,

Y80

0–13

00F

.sp

ec.

3G

abe

nla

4–5

rec

wB

:L

P(C

aran

avi)

leY

1500

.F

.sp

ec.

4H

abe

bla

7–8

llw

B:

CH

(Lui

sC

alvo

)le

TB

1430

.F

.al

bica

nsH

abse

rnl

a5

rec

wA

:S

(Ora

n,B

arr.

del

R.B

lanc

o,Sa

nA

ndre

s,Q

uebr

.de

Ach

eral

);B

:SC

(Vel

asco

,Fl

orid

a,V

alle

Gra

nde)

,C

B(C

hapa

re,

Ayo

paya

,Sa

caba

,J.

Car

-ra

sco

T.,

S.C

inti,

J.M

endo

za),

CH

(Tom

ina)

,B

eni

(Mox

os),

LP

(Mur

illo,

Inqu

isiv

i,F.

Tam

ayo,

S.Y

unga

s),

Tar

ija(A

rce)

rwT

B,

Y20

0–21

00

F.

grac

ilis

iso

abe

bla

5–6

lly

B:

LP

(Lar

ecaj

a),

Ben

i(B

alliv

an)

smPA

A20

0–50

0F

.gr

amin

eaI

abse

rnl

a3

rec

wB

:L

P(L

arec

aja)

smSA

A,

Y70

0–12

00F

.sp

ec.

5I

n.d.

n.d

n.d.

n.d.

n.d.

n.d.

n.d.

n.d

n.d.

n.d.

F.

spec

.6

In.

d.e

bla

n.d.

n.d.

n.d.

B:

Ben

i(B

alliv

ian)

n.d

Y85

0.F

.vi

llos

ula

Js

ebl

a5–

7ll

wB

:C

B(C

hapa

re,

Inca

chac

a),

SC(A

.Ib

anez

,Fl

orid

a,N

uflo

deC

have

z),

LP

(F.

Tam

ayo,

Lar

ecaj

a,N

.Y

unga

s)

smPA

A,

SAA

,Y

,T

B40

0–16

00

F.

chiq

uita

naK

abe

bla

9–10

llw

B:

SC(N

uflo

deC

have

z,G

uara

yos)

smC

D20

0–50

0F

.pe

ndul

iflo

raK

se

bla

7–10

llw

P:H

uanu

co(T

ingo

Mar

ia);

A:

J(S

ierr

ade

Cal

ila-

gua,

Sier

rade

Sant

aB

arba

ra),

Sa(O

ran,

Arr

oyo

deT

arta

gal,

Rio

Bla

nco,

Cer

ros

deR

ioY

tau,

San

Pedr

ico,

San

Jose

);B

:SC

(Flo

rida

,G

uara

yos,

Cor

dille

ra,

A.

Iban

ez),

CB

(H.

Sile

s,L

.C

alvo

),T

arija

(Arc

e,G

.C

haco

,O

Con

nor)

,C

H(I

nqui

sivi

)

rwT

B,

CD

,M

C22

0–24

30

F.

lati

foli

aK

se

bla

4–5

llw

B:

SC(F

lori

da)

leT

B60

0–11

00F

.sp

ec.

7K

abe

bla

5–6

rec

wB

:L

P(C

aran

avi)

n.d.

Y83

0..F

.sp

ec.

8L

se

bla

6ll

wB

r:Pa

rale

MA

P<

200.

F.

web

erba

ueri

La

ebl

a4–

5tr

ecw

P:H

uanu

cu,

Juni

n;B

:C

B(C

hapa

re),

LP

(Car

anav

i,L

arec

aja,

N.

Yun

gas)

rwPA

A,

SAA

,Y

600–

1500

F.

spec

.9

iso

n.d.

n.d

n.d.

n.d.

n.d.

n.d.

n.d.

n.d

n.d.

n.d.

F.

spec

tabi

lis

iso

abe

bla

22–2

4st

rr

B:

SC(F

lori

da),

CH

(Tom

ina)

smT

B13

00–1

700

Rex et al. 101

# 2007 NRC Canada

Page 13

F. villosula form a well-supported group of acaulescentherbs with entire more-or-less broad leaves, lily-like nonre-coiled petals (Fig. 5, left panel), and a shared leaf anatomy(Figs. 4d–4f). Morphologic, leaf anatomical, and preliminarycpDNA sequence data suggest that F. micrantha from Cen-tral America also belongs to this group (see Figs. 4a–4c).The arising hypothesis is that F. micrantha is the product ofa relatively recent long-distance dispersal event from centralSouth America to the dry forests of El Salvador, westernGuatemala, and Mexico. F. villosula itself, as currently de-fined, possibly includes 2 distinct species. Some strikingmorphologic and ecologic intraspecific differences had beennoticed before (Ibisch et al. 2002), which have now beenconfirmed by the AFLP data. Thus, the specimens collectedclose to the type locality of F. villosula, in the highly humidrainforests of the Chapare region in central Bolivia, grouptogether, and so do the more southern specimens from semi-humid forests at the Andean knee near Santa Cruz. The mo-lecular and ecologic differences should be sufficient toestablish a new taxon for the southern specimens (e.g.,P. Ibisch & C. Nowicki 98.0173).

F. penduliflora, F. latifolia, F. chiquitana, and F. spec. 7This group of accessions is characterized by lily-like pet-

als and a common leaf blade anatomy that is shared withF. villosula and F. micrantha (Figs. 4d to 4f). Whereas thewhole group (i.e., cluster K in Fig. 2) is moderately sup-ported by AFLP data, its internal topology does not reflecttaxonomic boundaries. F. penduliflora is a widespread spe-cies, occurring from northern Argentina to central and east-ern Bolivia, in several ecoregions, such as the Tucuman-Bolivian forests, the montane Chaco, and the Chiquitanodry forest. Although seemingly a mesophytic herb, it with-stands considerable phases of drought. Under extremedrought, the plants can shed their leaves entirely, and thebulbous young shoots are able to endure several monthswithout precipitation. The species occurs in the understoryof Andean dry forests, and can survive on sun-exposed rockswith very limited water supply. F. chiquitana andF. latifolia were recently described as new taxa from thelowlands of the Chiquitano dry forest and from semihumidAndean forests, respectively (Ibisch et al. 1999). Their statusas separate species is not confirmed by the AFLP data, andmight require revision. If F. chiquitana and F. latifolia (andthe yet-unclassified F. spec. 7) are included inF. penduliflora, the latter taxon becomes even more variablethan previously thought. For the time being, we still con-sider F. latifolia to be a geographically and morphologicallydistinct taxon with a very small range.

F. weberbaueri and F. spec. 8The accessions combined in cluster L are acaulescent

herbs with entire leaves and only temporarily recoiled petalsthat become straight again after anthesis (see Ibisch et al.2002 for differentiation of F. weberbaueri from F. schido-sperma). They live in the pre- and sub-Andean rain forestsand even reach the montane Yungas forests. According tothe AFLP data, a specimen from northern La Paz (Nor Yun-gas province; acc. No. RS 261002-18) is clearly distinctfrom plants collected in the southern provinces of Caranavi,Larecaja, and Chapare. A potential morphologic differentia-T

able

3(c

oncl

uded

). Mor

phol

ogic

and

anat

omic

char

acte

rsB

ioge

ogra

phy

Spec

ies*

CL

LSA

BM

PHPL

PSPC

Geo

grap

hic

dist

ribu

tion

RS

Eco

regi

on

Alt.

rang

e(m

abov

ese

ale

vel)

Not

e:A

lt.,

altit

ude.

Mor

phol

ogic

alan

dan

atom

ical

char

acte

rs:

n.d.

,no

tde

term

ined

;C

L,

clus

ter;

iso,

isol

ated

posi

tion,

not

arra

nged

ina

clus

ter;

LSA

,le

afsc

ales

abax

iall

y(a

b,ab

unda

nt;

a,ab

sent

;s,

scar

ce);

BM

,bl

ade

mar

gin

(e,

entir

e,se

r,se

rrat

eat

the

base

);PH

,pl

ant

habi

t(b

la,

broa

d-le

aved

acau

lesc

ent

flat

rose

ttes;

nla,

narr

ow-l

eave

dac

aule

scen

t±

upri

ght

rose

ttes;

cb,

caul

esce

nt-b

ranc

hed)

;PL

,m

axim

alpe

tal

leng

th(m

m);

PS,

peta

lsh

ape

(rec

,re

coile

d,ev

enaf

ter

anth

esis

;ll,

lily-

like;

trec

,te

mpo

rari

lyre

coile

d,st

raig

htaf

ter

anth

esis

;st

r,st

raig

ht);

PC,

peta

lco

lor

(w,

whi

te;

r,re

ddis

hto

cora

l-or

ange

;c,

crea

man

dro

sesp

ots;

p,pa

legr

een,

then

slig

htly

redd

ish;

g,gr

eeni

sh;

y,ye

llow

).G

eogr

aphi

cdi

stri

butio

n(c

ount

ries

and

dist

rict

s):

B,

Bol

ivia

;P,

Peru

;B

r,B

razi

l;A

,A

rgen

tina;

J,Ju

juy;

LP,

La

Paz;

S,Sa

lta;

SC,

Sant

aC

ruz;

CB

,C

ocha

bam

ba;

CH

,C

huqu

isac

a.R

S,ra

nge

size

(le,

loca

lly

ende

mic

;sm

,sm

all;

rw,

rela

tivel

yw

ides

prea

d).

Eco

regi

ons:

BSC

A,

Sout

hwes

tA

maz

on/B

eni

and

Sant

aC

ruz

Am

azon

Fore

sts;

CD

,C

hiqu

itano

Dry

Fore

st;

IAD

,In

ter-

And

ean

Dry

Fore

sts;

MA

P,M

oist

Am

azon

Fore

sts

ofPa

ra;

PAA

,So

uthw

est

Am

azon

/Pre

-And

ean

Am

azon

Fore

sts;

SAA

,So

uthw

est

Am

azon

/Sub

-And

ean

Am

azon

Fore

sts;

TB

,T

ucum

an-B

oliv

ian

Fore

sts;

Y,

(Per

uvia

n-B

oliv

ian)

Yun

gas.

*Sp

ecie

sar

ear

rang

edac

cord

ing

toth

eir

occu

rren

ce(c

lust

ers

Ato

L)

inth

ene

ighb

our-

join

ing

tree

(see

Fig.

2).

102 Genome Vol. 50, 2007

# 2007 NRC Canada

Page 14

tion should be revised. The molecular data confirm that aspecimen that resembles F. weberbaueri but that was re-corded in northern Brazil (about 1 600 km away, state ofPara; acc. No. Leme 5078) is clearly distinct and probablymerits species status (F. spec. 8).

Species not assigned to any groupSeveral species and unclassified morphospecies have iso-

lated positions in the AFLP tree and are not yet assignableto any of the above species groups. One of these species isF. gracilis, which is remarkable for having yellowish flowersand for being the only species in northern Bolivia that occursat altitudes below 500 m (i.e., on rocks in pre-Andeanforests belonging to the Southwest Amazon ecore-gion). F. spectabilis exhibits the largest geneticdistance from all other taxa. This acaulescent herb,with entire leaves, is morphologically well differentiatedfrom all other species, having large, straight, reddish tocoral-orange petals. Other accessions that take isolatedpositions are F. spec. 1 (between clusters C and D), F.spec. 9, and 1 specimen of F. albicans (acc. No.RV4023).

ConclusionsRobert W. Read (personal communication) discovered an

important morphologic difference that divides Fosterellainto 2 subgroups: 1 characterized by strongly recoiled petalsthat stay so even after anthesis; and 1 with straight or some-what recurved petals that straighten out after anthesis(Fig. 5). He suggested that this characteristic could be valua-ble in establishing 2 subgenera, but never published hisideas. Although it has become clear that there are more sub-groups, the molecular data partly corroborate Read’s obser-vations and indicate that at least the group with unrecoiled,lily-like petals (i.e., the species of clusters J and K in Fig. 2)could be monophyletic. The AFLP data also support some

other relationships that were initially suggested by morpho-logic characters (e.g., a group of caulescent-branched plantswith serrate leaves from the inter-Andean valleys (F. cota-cajensis, F. nowickii, and F. weddelliana); and a group ofbroad-leaved acaulescent herbs with entire blades and lily-like flowers (F. penduliflora and associates)). Because ofshort internal nodes, the AFLP tree does not provide conclu-sive information about relationships between the species andspecies groups outlined above. Unresolved internal nodes as-sociated with long terminal branches are commonly encoun-tered in AFLP trees (e.g., Koopman et al. 2001; Banfer et al.2004), and limit the usefulness of the AFLP technique forphylogenetic analyses to a certain extent.

Our results also shed light on Fosterella evolution andbiogeography. The small but not dust-like Fosterella seedsdo not permit effective wind dispersal. As a consequence,many Fosterella species are confined to rather small ranges,representing local endemics. The same is true for other bro-meliad genera belonging to the Pitcairnioideae, such as Pit-cairnia, Puya, or Deuterocohnia. Obviously, taxa with moreefficient reproduction mechanisms tend to have less restrictedranges, such as the mostly bird- and mammal-dispersed Bro-melioideae and the wind-dispersed Tillandsioideae (Ibisch1996; Ibisch et al. 2001). Several species of Fosterella arefound in the South American lowlands, and others are foundin different habitats of the Andes. The AFLP trees neither re-vealed a consistent close relationship among all Andean spe-cies nor one among all lowland Fosterella species. The dataat hand do not give clear evidence establishing whether An-dean habitats where colonized from the Amazon lowlands orvice versa. Nevertheless, repeated colonization and founderevents have to be postulated in Fosterella. In the Bromelioi-deae, molecular data point toward sub-Andean taxa beingmore basal than lowland ones (Schulte et al. 2005). However,given the comparatively young age of the higher AndeanFosterella habitats and the fact that the lowland taxa areconfined to very old and ecologically stable habitats related

Fig. 5. Flower morphology types in Fosterella. Lily-like petals of F. penduliflora (left) and watch-spring-like recoiled petals of F. elata(right).

Rex et al. 103

# 2007 NRC Canada

Page 15

to different kinds of pre-Cambric rocks (e.g., F. chiquitana,F. hatschbachii, F. penduliflora, F. rojasii, F. windischii,F. vasquezii, and F. yuvinkae), a lowland origin can be hy-pothesized. Apparently, the genus also reached remote re-gions, such as Central America, by rare events of long-distance dispersal, perhaps through seeds attached to birds.

Further interdisciplinary studies on diversity, biogeogra-phy, and phylogeny of this genus seem promising. The re-sults will likely facilitate further taxonomic treatment byconfirming hypotheses about new taxa and the genetic dif-ferentiation according to geographic provenances.

AcknowledgementsThe authors acknowledge the support of the Deutsche

Forschungsgemeinschaft (DFG grants We 1830/5-1, IB 85/1-1 and ZI 557/6-1). Part of the plant material was kindlysupplied by the Palmengarten Frankfurt am Main, the Bota-nical Gardens of Heidelberg, Berlin and Osnabruck; theMarie Selby Botanical Garden of Sarasota, Fla., and by theliving collection of FAN, where Arturo Osinaga provided in-valuable help. We also thank Elton Leme for support, andChristoph Nowicki, Robert Muller, and Jule Peters for helpwith the data and tables.

ReferencesBanfer, G., Fiala, B., and Weising, K. 2004. AFLP analysis of phy-

logenetic relationships among myrmecophytic species of Macar-anga (Euphorbiaceae) and their allies. Plant Syst. Evol. 249:213–231. doi:10.1007/s00606-004-0219-4.

Barfuss, M.H.J., Samuel, M.R., and Till, W. 2004. Molecular phy-logeny in subfamily Tillandsioideae (Bromeliaceae) based on sixcpDNA markers: an update. J. Bromeliad Soc. 54: 8–16.

Benzing, D.H. 2000. Bromeliaceae: profile of an adaptive radiation.Cambridge University Press, Cambridge.

Brummit, R.K., and Powell, C.E. 1992. Authors of plant names.Royal Botanic Gardens, Kew. Whitstable Litho Ltd., Kent.

Crayn, D.M., Winter, K., and Smith, J.A.C. 2004. Multiple origins ofcrassulacean acid metabolism and the epiphytic habit in the Neo-tropical family Bromeliaceae. Proc. Natl. Acad. Sci. U.S.A. 101:3703–3708. doi:10.1073/pnas.0400366101. PMID:14982989.

Debener, T., and Mattiesch, L. 1999. Construction of a geneticlinkage map for roses using RAPD and AFLP markers. Theor.Appl. Genet. 99: 891–899. doi:10.1007/s001220051310.

Despres, L., Gielly, L., Redoutet, B., and Taberlet, P. 2003. UsingAFLP to resolve phylogenetic relationships in a morphologicallydiversified plant species complex when nuclear and chloroplastsequences fail to reveal variability. Mol. Phylogenet. Evol. 27:185–196. PMID:12695084.

Felsenstein, J. 1985. Confidence limits on phylogenies: an ap-proach using the bootstrap. Evolution, 39: 783–791. doi:10.2307/2408678.

Givnish, T.J., Millam, K.C., Evans, T.M., Hall, J.C., Pires, J.C.,Berry, P.E., and Sytsma, K.J. 2004. Ancient vicariance or recentlong-distance dispersal? Inferences about phylogeny and SouthAmerican-African disjunctions in Rapateaceae and Bromeliaceaebased on ndhF sequence data. Int. J. Plant Sci. 165: S35–S54.doi:10.1086/421067.

Givnish, T.J., Millam, K.C., Evans, T.M., Pires, J.C., Berry, P.E.,and Sytsma, K.J. 2007. Phylogeny, biogeography, and ecologicalevolution in Bromeliaceae: insights from ndhF sequences. InMonocots: comparative biology and evolution. Edited by J.T.

Columbus, E.A. Friar, C.W. Hamilton, J.M. Porter, L.M. Prince,and M.G. Simpson. In press.

Hansen, M., Kraft, T., Christiansson, M., and Nilsson, N.O. 1999.Evaluation of AFLP in Beta. Theor. Appl. Genet. 98: 845–852.doi:10.1007/s001220051143.

Hodkinson, T.R., Renvoize, S.A., Chonghaile, G.N., Stapleton,C.M.A., and Chase, M.W. 2000. A comparison of ITS nuclearrDNA sequence data and AFLP markers for phylogenetic studiesin Phyllostachys (Bambusoideae, Poaceae). J. Plant Res. 113:259–269. doi:10.1007/PL00013936.

Horres, R., Zizka, G., Kahl, G., and Weising, K. 2000. Molecularphylogenetics of Bromeliaceae: evidence from trnL(UAA) in-tron sequences of the chloroplast genome. Plant Biol. 2: 306–315. doi:10.1055/s-2000-3700.

Horres, R., Schulte, K., Weising, K., and Zizka, G. 2007. Systema-tics of Bromelioideae (Bromeliaceae) - molecular, anatomicaland morphological Studies. In Monocots: comparative biologyand evolution. Edited by J.T. Columbus, E.A. Friar, C.W. Ha-milton, J.M. Porter, L.M. Prince, and M.G. Simpson, M.G. Inpress.

Ibisch, P.L. 1996. Neotropische Epiphytendiversitat - das BeispielBolivien. Martina-Galunder-Verlag, Wiehl.

Ibisch, P.L., Gross, E., Rauer, G., and Rudolph, D. 1997. On thediversity and biogeography of the genus Fosterella L.B. Smith(Bromeliaceae) with the description of a new species from East-ern Bolivia. J. Bromel. Soc. 47: 211–217.

Ibisch, P.L., Vasquez, R., and Gross, E. 1999. More novelties ofFosterella L.B. Smith (Bromeliaceae, Pitcairnioideae) from Bo-livia. Rev. Soc. Boliv. Bot. 2: 117–132.

Ibisch, P.L., Nowicki, C., and Vasquez, R. 2001. Towards an un-derstanding of diversity patterns and conservation requirementsof the Bolivian Bromeliaceae. J. Bromel. Soc. 51: 99–113.

Ibisch, P.L., Vasquez, R., Gross, E., Kromer, T., and Rex, M. 2002.Novelties in Bolivian Fosterella (Bromeliaceae). Selbyana, 23:204–219.

Ibisch, P.L., Beck, S.G., Gerkmann, B., and Carretero, A. 2004.Ecoregions and ecosystems. In Biodiversity: the richness of Bo-livia. State of knowledge and conservation. Edited by P.L. Ibischand G. Merida. Ministerio de Desarrollo Sostenible y Planifica-cion/Editorial FAN, Santa Cruz. pp. 47–88.

Igersheim, A., and Cichocki, O. 1996. A simple method for micro-tome sectioning of prehistorical charcoal specimens, embeddedin 2-hydroxyethyl methacrylate (HEMA). Rev. Palaeobot. Paly-nol. 92: 389–393.

Kardolus, J.P., van Eck, H.J., and van den Berg, R.G. 1998. Thepotential of AFLPs in biosystematics: a first application in Sola-num taxonomy (Solanaceae). Plant Syst. Evol. 210: 87–103.doi:10.1007/BF00984729.

Kessler, M., Ibisch, P.L., and Gross, E. 1999. Fosterella cotacajen-sis, una nueva especie de Bromeliaceae de los valles secos andi-nos de Bolivia. Rev. Soc. Boliv. Bot. 2: 111–116.

Koopman, W.J.M., Zevenbergen, M.J., and van den Berg, R.G.2001. Species relationships in Lactuca s.l. (Lactuceae, Astera-ceae) inferred from AFLP fingerprints. Am. J. Bot. 88: 1881–1887.

Kromer, T., Kessler, M., Holst, B.K., Luther, H.E., Gouda, E.J.,Ibisch, P.L., Till, W., and Vasquez, R. 1999. Checklist of Boli-vian Bromeliaceae with notes on species distribution and levelsof endemism. Selbyana, 20: 201–223.

Luther, H.E. 1981. Miscellaneous new taxa of Bromeliaceae (I).Selbyana, 5: 310–311.

Luther, H.E. 1997. A showy new Fosterella from Bolivia. J. Bro-mel. Soc. 47: 118–119.

Nei, M., and Li, W.-H. 1979. Mathematical model for studying ge-

104 Genome Vol. 50, 2007

# 2007 NRC Canada

Page 16

netic variation in terms of restriction endonucleases. Proc. Natl.Acad. Sci. U.S.A. 76: 5269–5273. doi:10.1073/pnas.76.10.5269.PMID:291943.

Rauh, W. 1979. Bromelienstudien. I. Neue und wenig bekannte Ar-ten aus Peru und anderen Landern (9.Mitteilung). 4. Fosterella.Trop. Subtrop. Pflanzenwelt, 31: 23–29.

Rauh, W. 1987. Bromelienstudien. 4. Fosterella L.B. Smith. Trop.Subtrop. Pflanzenwelt, 60: 24–26.

Rex, M. 2001. Molekulare Analyse von Verwandtschaftsbeziehun-gen in ausgewahlten Gattungen der Pitcairnioideae (Bromelia-ceae). Diploma Thesis, University of Kassel.

Rohlf, F.J. 2000. NTSYSpc version 2.10p. Numerical taxonomyand multivariate analysis system. Exeter, New York.

Saitou, N., and Nei, M. 1987. The neighbour joining method: anew method for reconstructing phylogenetic trees. Mol. Biol.Evol. 4: 406–425. PMID:3447015.

Sambrook, J., and Russell, D.W. 2001. Molecular Cloning. A La-boratory Manual. 3rd ed. Cold Spring Harbor Laboratory Press,Cold Spring Harbor, New York.

Schulte, K., Horres, R., and Zizka, G. 2005. Molecular phylogenyof Bromelioideae and its implications on biogeography and theevolution of CAM in the family. Senckenb. Biol. 85: 113–125.

Smith, L.B., and Downs, R.J. 1974. 13. Fosterella. In Flora Neotro-pica, Monograph no. 14, Pitcairnioideae. Hafner Press, NewYork. pp. 199–209.

Smith, L.B., and Read, R.W. 1992. Flora Neotropica, MonographNo. 14 (1), suppl. no. 3. Fosterella. Bradea, 6: 134–140.

Smith, L.B., and Till, W. 1998 Bromeliaceae. In The Families andGenera of Vascular Plants. Edited by K. Kubitzki. Vol. 4.Springer, Berlin, Heidelberg, New York. pp. 74–99.

Sneath, P.H.A., and Sokal, R.O. 1973. Numerical Taxonomy. Free-man, San Francisco.

Swofford, D.L. 2002. PAUP*: Phylogenetic analysis using parsi-mony, (*and other methods) v. 4.0b10 [computer program]. Si-nauer Assoc., Sunderland.

Terry, R.G., Brown, G.K., and Olmstead, R.G. 1997. Examinationof subfamilial phylogeny in Bromeliaceae using comparative se-quencing of the plastid locus NDHF. Am. J. Bot. 84: 664–670.doi:10.2307/2445903.

Tuthill, D., and Brown, G.K. 2004. Bromeliad endophytes and theserendipity of science. J. Bromeliad Soc. 54: 28–32.

Van de Peer, Y., and De Wachter, R. 1994. TREECON for Win-dows: a software package for the construction and drawing ofevolutionary trees for the Microsoft Windows environment.Comput. Appl. Biosci. 10: 569–570. PMID:7828077.

Vos, P., Hogers, R., Bleeker, M., Reijans, M., Van de Lee, T.,Hornes, M., et al. 1995. AFLP: a new technique for DNA fin-gerprinting. Nucleic Acids Res. 23: 4407–4414. PMID:7501463.

Weising, K., Nybom, H., Wolff, K., and Meyer, W. 2005. DNAFingerprinting in Plants: Principles, Methods, and Applications.2nd ed. CRC Press, Taylor & Francis Group, Boca Raton, Fla.

Rex et al. 105

# 2007 NRC Canada