Combining Data from DNA Sequences and Morphology for a

55

Systematic Botany (2002), 27(1): pp. 55–73q Copyright 2002 by the American Society of Plant Taxonomists

Combining Data from DNA Sequences and Morphology for a Phylogenyof Moringaceae (Brassicales)

MARK E. OLSON

Missouri Botanical Garden, PO Box 299, St Louis, Missouri 63166-0299;Current address: Instituto de Biologıa, Universidad Nacional Autonoma de Mexico, Circuito Exterior s/n,

Ciudad Universitaria, Copilco, Coyoacan A. P. 70-367, C. P. 04510, Mexico, D. F. Mexico

Communicating Editor: Thomas Lammers

ABSTRACT. The Old World dry tropical family Moringaceae is remarkable for the great diversity of habit and floralmorphology found within its only genus, Moringa. To infer the phylogenetic relationships of all 13 species, parsimonyanalyses of morphological data and DNA sequences from a low-copy nuclear region (PEPC), a chloroplast region (trnG),and a tandemly-repeated nuclear region present in high copy number (ITS) were conducted of each data set separately andcombined. Characters from studies of ontogeny substantially enhanced the resolution of the morphological data set. TheIncongruence Length Difference test indicated the congruence of all data sets, as did Templeton tests comparing the singletree resulting from the combined analysis in the context of the individual data sets. This tree is presented as the preferredtopology, in which the four bottle trees appear in a basal paraphyletic assemblage, with the three species of slender trees(including the economically important M. oleifera) forming a clade that is sister to a clade of the six species of tuberousshrubs and trees of northeast Africa. Moringa is currently divided into three sections, but because of the basal grade, itcannot be divided into useful monophyletic infrageneric taxa. The phylogeny-based informal terms ‘‘bottle tree grade’’,‘‘slender tree clade’’, and ‘‘tuberous clade’’ are suggested as alternatives. Relationships within Moringa were found to belargely congruent with a previous study of wood anatomy.

With just 13 species, Moringa Adans. (the only genusof the family Moringaceae) is for its size one of themost phenotypically varied groups of angiosperms.Ranging from huge ‘‘bottle trees’’ to tiny tuberousshrubs, and spanning the range from radial to bilateralfloral symmetry (Fig. 1), the small number of speciesin the genus makes it useful for investigating the di-versification of plant form. One species, M. oleifera(species authors are listed in Appendix 1), is cultivatedthroughout the tropics as the source of nutritious leafand fruit vegetables, high-quality seed oil, pharmaco-logically active compounds, and water clarificationagents (Oliveiera et al. 1999; Ghasi et al. 2000; Kalogoand Verstraete 2000; Saleem and Meinwald 2000; Jahn2001). The other twelve species all have local uses, butonly M. oleifera has been the subject of applied researchor breeding. To provide a framework for basic and ap-plied studies, I use DNA sequence data from one chlo-roplast and two nuclear loci, in addition to morpho-logical data, to construct a phylogenetic hypothesis forthe Moringaceae.

The Moringaceae and sister family Caricaceae arepart of the ‘‘mustard-oil plants’’ clade (Rodman et al.1998; the Brassicales of APG 1998), along with suchfamilies as Brassicaceae, Capparaceae, and Tropaeola-ceae. Moringa is found in the seasonally dry tropics ofAfrica, Asia, and Madagascar (map, Fig. 1K). The pa-chycaul species with massive, water-storing trunks andfleshy roots and actinomorphic flowers (Fig. 1A-C; seealso Table 1 for life form categories) occur in Africaand Madagascar (one species is apparently extinct inthe wild, Olson and Razafimandimbison 2000). The re-mainder of the family have bilaterally symmetrical

flowers, including the three species of slender trees(Fig. 1D-F). This habit class, which includes the eco-nomically important M. oleifera, is characterized by amore conventional trunk and tough, fibrous roots, andis principally south and southwest Asian. The drytropical habitats in Kenya, Ethiopia, and Somalia sup-port the highest number of Moringa species, and theentirety of what I call the ‘‘tuberous’’ species (Fig. 1E-J). These species are small trees or shrubs of varyinghabits, but always with fleshy, water-storing roots. Asis true of many taxa of the dry tropics, several speciesof this group are very poorly-known. Moringa arboreahas been seen twice by scientists and is known onlyfrom a single remote canyon on the Kenya-Ethiopiaborder. Moringa pygmaea is known only with certaintyfrom the type collection in northern Somalia, thougha sterile collection from the northern coast of Somalia(listed in Appendix 1) may represent this species.Verdcourt (1985) cites a probable undescribed specieson the Kenya-Somalia border; subsequent fieldworkhas shown this to be M. longituba (Olson 712 in Ap-pendix 1). Despite the great morphological diversity inthe family, the monophyly of Moringa is supported bynumerous distinctive synapomorphies such as gumducts in the pith and monothecal, bisporangiate an-thers.

Differing species groups within Moringa have beenproposed depending on the characters studied by theauthor, including leaf and floral morphology (Engler1902; Verdcourt 1985), palynology (Ferguson 1985),and wood anatomy (Olson and Carlquist 2001). Cur-rent sectional classification (Verdcourt 1985) is basedon floral morphology (Table 1; Fig. 1) and divides the

56 [Volume 27SYSTEMATIC BOTANY

FIG. 1. Morphological diversity and range of Moringa. A. Bottle tree habit. B-C. Bottle tree flowers: radially symmetrical, allpetals and sepals with an equal degree of flexion. D. Slender tree habit. E-F. Slender tree flowers: bilaterally symmetrical. G.Sarcorhizal trees (M. arborea, M. ruspoliana) have slender trunks and fleshy, brittle, tuberous roots; flowers similar to those inFigs. 1E, 1F. H. Tuberous shrubs of northeast Africa (M. borziana, M. longituba, M. pygmaea, M. rivae) have slender stems thatare often shed in times of severe drought, with massive, very soft tubers underground. Flowers similar to Figs. 1E, 1F, withthe exception of M. longituba (habit shown in Fig. 1I), which has bilaterally symmetrical flowers with a long tubular hypanthium(Fig. 1J). K. Range of Moringa showing India at upper right, part of Arabia at upper left, Madagascar at bottom center, andpart of Africa at left.

2002] 57OLSON: MORINGA PHYLOGENY

TAB

LE

1.Sp

ecie

sG

roup

sin

Mor

inga

.

Spec

ies

cate

gori

zed

byse

ctio

n(V

erd

cour

t,19

85)

Spec

ies

cate

gori

zed

byha

bit

(Ols

onan

dC

arlq

uist

,200

1)Sp

ecie

sca

tego

rize

dby

woo

dan

atom

y(O

lson

and

Car

lqui

st,2

001)

Gro

ups

show

nin

phyl

ogen

etic

tree

s

Don

ald

soni

a:flo

wer

sra

dia

llysy

mm

etri

-ca

lor

near

lyso

.M.d

rouh

ardi

i,M

.hild

e-br

andt

ii,M

.ova

lifol

ia,M

.ste

nope

tala

Bot

tletr

ees

(Cal

led

bott

leor

tank

tree

sbe

caus

eof

thei

rbl

oate

dtr

unks

that

are

fille

dw

ith

wat

er).

Equ

ival

ent

insp

ecie

sto

Sect

ion

Don

ald

soni

a

Stem

and

root

xyle

mha

veba

nds

ofco

n-flu

ent

para

trac

heal

pare

nchy

ma

alte

r-na

ting

wit

hba

nds

oflib

rifo

rmfib

ers,

som

eof

whi

chm

aybe

wid

ean

dpa

-re

nchy

ma-

like

(suc

hfib

ers

rare

inM

.dr

ouha

rdii)

.Equ

ival

ent

insp

ecie

sto

Sec-

tion

Don

ald

soni

a

Bot

tletr

ees:

para

phyl

etic

orpo

lyph

ylet

ic

Mor

inga

:flow

ers

bila

tera

llysy

mm

etri

cal

wit

hsh

ort

hypa

nthi

um.M

.arb

orea

,M.

borz

iana

,M.c

onca

nens

is,M

.ole

ifera

,M.

pere

grin

a,M

.pyg

mae

a,M

.riv

ae,M

.rus

-po

liana

Slen

der

tree

s:T

runk

ssl

ende

r,w

ith

thic

k,to

ugh

bark

and

toug

hro

ots.

M.c

onca

-ne

nsis

,M.o

leife

ra,M

.per

egri

na

Stem

xyle

mw

ith

libri

form

fiber

sof

vary

-in

gsh

apes

and

size

s,bu

tno

tfo

rmin

gse

ason

alba

nds.

Roo

txy

lem

wit

hal

ter-

nati

ngba

nds

oflib

rifo

rmfib

ers

and

ax-

ial

pare

nchy

ma;

thes

epa

renc

hym

aba

nds

are

not

wid

erth

anba

nds

offib

-be

rs.E

quiv

alen

tin

spec

ies

toth

esl

en-

der

tree

s

Slen

der

tree

s:m

onop

hyle

ticin

all

anal

y-se

s

Dys

mor

inga

:flow

ers

bila

tera

llysy

mm

et-

rica

lw

ith

long

hypa

nthi

um.M

.lon

gitu

-ba

Sarc

orhi

zal

tree

s:Sl

ende

rtr

unks

wit

ha

netw

ork

ofve

ryth

ick,

flesh

y,br

ittle

root

s.M

.arb

orea

,M.r

uspo

liana

Tub

erou

ssh

rubs

ofN

EA

fric

a:Sl

ende

r,of

ten

deci

duou

ssh

oots

,wit

hhu

getu

-be

rous

root

s.M

.bor

zian

a,M

.lon

gitu

ba,

M.p

ygm

aea,

M.r

ivae

Stem

xyle

mw

ith

narr

owlib

rifo

rmfib

ers

som

etim

esva

ryin

gin

size

wit

hse

ason

;so

met

imes

wit

hba

nds

ofpa

ratr

ache

alax

ial

pare

nchy

ma.

Roo

tse

cond

ary

xy-

lem

alm

ost

enti

rely

axia

lpa

renc

hym

a,w

ith

orw

itho

utlib

rifo

rmfib

ers.

M.a

r-bo

rea,

M.b

orzi

ana,

M.p

ygm

aea,

M.r

ivae

Stem

woo

dw

ith

narr

owlib

rifo

rmfib

ers

som

etim

esva

ryin

gin

size

wit

hse

ason

wit

hout

band

sof

para

trac

heal

axia

lpa

-re

nchy

ma.

Roo

tse

cond

ary

xyle

mal

-m

ost

enti

rely

axia

lpa

renc

hym

a,w

ith

orw

itho

utlib

rifo

rmfib

ers

M.l

ongi

tuba

,M

.rus

polia

na

‘‘riv

aecl

ade’

’:eq

uiva

lent

toth

isw

ood

anat

omy

cate

gory

‘‘tub

erou

s’’c

lade

:equ

ival

ent

toth

ese

woo

dan

atom

yca

tego

ries

red

-flow

ered

clad

e:eq

uiva

lent

toth

isw

ood

anat

omy

cate

gory


genus into three sections: Donaldsonia (radial symme-try), Moringa (bilateral symmetry with a short hypan-thium), and Dysmoringa (bilateral symmetry with along hypanthium). Species groups based on habit andwood anatomy are contrasted in Table 1. Verdcourt(1985) provides the only phylogenetic hypothesis forthe family to date (Fig. 2), based on his study of her-barium specimens. His branching diagram lends itselfto interpretation in cladistic terms, and can be read tohypothesize a monophyletic bottle tree clade (corre-sponding to Section Donaldsonia) as the sister group tothe rest of the family, with a clade of slender treesforming the sister group to a clade of the tuberousspecies. Section Moringa (all of the species but Donald-sonia and M. longituba) is thus hypothesized to be par-aphyletic because of the position of M. longituba, thesole member of Section Dysmoringa. How these vari-ous, sometimes conflicting groupings compare to a cla-distic reconstruction based on data from various sourc-es is examined here.

To construct a phylogenetic hypothesis of Moringa,three molecular loci were selected showing interspe-cific variation in Moringa: one low-copy nuclear locus(PEPC), one chloroplast locus (trnG), and a tandemly-repeated nuclear region present in high copy number(ITS). The enzyme phosphoenolpyruvate carboxylase(PEPC; Enzyme Comission code 4.1.1.31) is of majorimportance in CO2 fixation in C4 and CAM plants andis implicated in anaplerotic carbon metabolism in C3plants (Latzko and Kelly 1983). The genes coding forthis enzyme appear to be present in low copy numbersin small multigene families (e.g., Panstruga et al. 1995found ;3 copies in Solanum tuberosum). Most plantPEPC genes that have been examined are characterizedby nine introns; sequences of the ca. 450 bp-long fourthintron have been used in our laboratory for phylogenyreconstruction in Solanaceae, Chenopodiaceae, and Ta-maricaceae. The tRNA gene for the amino acid glycineis located in the large single-copy region of the chlo-roplast genome. Hamilton et al. (1999) report popula-tion-level variation in the locus amplified by the primerpair trnG-trnS in Corythophora alta (Lecythidaceae). Thetandemly-repeated internal transcribed spacer region(ITS) of the 18s–26s nuclear ribosomal DNA has beenused to reconstruct phylogenies at the interspecific lev-el in numerous groups of plants, e.g., Asteraceae (Bald-win 1992; Bayer et al. 1996), Loasaceae (Moody et al.2001), and Rosaceae (Lee and Wen 2001).

In addition, a data set based on morphological char-acters was assembled to examine the effectiveness ofcharacters drawn from the study of ontogeny. Exam-ining immature ontogenetic stages can reveal addi-tional phylogenetic characters not identifiable in ma-ture organs (e.g., characteristics of cotyledons). Suchcharacters can be considered ‘‘instantaneous’’ in thatonly one ontogenetic stage is required for their obser-

vation. A second category of characters are ontogenetictransformations (e.g., leaves lobed -. leaves entire,which require knowledge of more than one ontoge-netic stage). Such characters are expected to improvephylogenetic resolution (de Queiroz 1985).

METHODS

Taxon Sampling for Molecular Data Sets. Taxa selected andgene regions sequenced are summarized in Appendix 1, alongwith voucher, locality, and species author information. At least onesample of each species of Moringa was available. In the case ofwide-ranging species, multiple samples from distributional ex-tremes were collected (e.g., M. longituba, M. rivae, and M. peregrina;a sample of ‘‘wild type’’ M. oleifera and a cultivar, PKM, were alsoincluded). Members of all four genera of Caricaceae (Carica L.,Cylicomorpha Urban, Jacaratia A. DC., and Jarilla Rusby) were se-lected as an outgroup.

Tissue Collection and DNA Extraction. Most tissue was col-lected in the field or from cultivated specimens and immediatelydried in silica gel. Voucher specimens for these collections are list-ed in Appendix 1. Leaves were usually used, but in some casesthe plants were leafless and stem tissue was prepared by separat-ing the bark from the xylem cylinder, removing the phellem, anddrying the remaining bark layers in silica gel. Tissue was groundvery finely in a mortar with a small amount of sterilized silicasand. DNA was extracted from ground tissue using the protocolof Edwards et al. (1994), followed by two 700ml 24:1 chloroform:isoamyl alcohol extractions (suggestion of Scott Hodges, pers.comm.). Material was available from the two known specimens ofMoringa pygmaea, but extraction using the Edwards et al. (1994)protocol, a CTAB protocol (Doyle and Doyle 1987), and QiagenDNeasy Plant Mini Kit recovered only degraded DNA.

PCR Amplification and Sequencing. The fourth intron of PEPCwas amplified using the primers PPCX4F and PPCX5R. PPCX4F(sequence 59 ACTCCACAGGATGAGATGAG) binds to the 4th exonand promotes extension across 4th intron; PPCX5R (sequence: 59GCGCCATCATTCTAGCCAA) binds to the 5th exon and promotesextension back across 4th intron. These primers were designedand kindly provided by John Gaskin. The trnG and ITS regionswere amplified with the primers of Hamilton 1999 (trnG) and Bay-er et al. 1996 (ITS21,22,23,24). The PCR thermal cycling profileconsisted of a 90 second denaturation at 948C followed by 30 cyclesof 948C for 50 seconds, 558C for 70 seconds, and 728C for 90 sec-onds. After these cycles, the samples were subjected to a finalextension at 728C for 3 minutes, and 308C for 1 minute. Each re-action contained a final concentration of 2.5 mM MgCl2, 10 mMTris HCl (pH 9.0), 50 mM KCl, 0.2 mM of each dNTP, 0.2 mM ofeach primer and 0.5 U/ml taq polymerase. Each PCR consisted offive separate 22.5 ml reactions which were combined for purifica-tion. PCR products were separated on agarose gels, purified witha Qiaquick Gel Extraction Kit (Qiagen), and quantified usingGibCo Low DNA Mass Ladder. Most sequencing reactions usedApplied Biosystems Incorporated Big Dye terminators and wererun on an Applied Biosystems Incorporated Model 373 or Model377 Prism DNA Automated Sequencer. GenBank accession num-bers are listed for all sequences in Appendix 1. Sequences werealigned by eye using the Se-Al Sequence Alignment Editor v1.0alpha 1 (Rambaut 1996–1998). Alignments are available from theTreeBASE database.

Morphological Characters. A data matrix of 28 morphologicalcharacters was constructed from herbarium, pickled, and livingmaterial of all 13 species of Moringa. Cylicomorpha was selected asan outgroup based on morphological and anatomical similarity toMoringa (Carlquist 1998; Olson and Carlquist 2001), and Badillo’s(1971) comments suggesting that, of all Caricaceae, Cylicomorphabears the largest proportion of plesiomorphic character states. Thecharacters and their coding are given in Appendix 2; vouchers arelisted in Appendix 1. For microscopy, most samples were collectedfrom living plants and preserved in 50–70% aqueous ethanol.Methods used to study wood and roots are detailed in Olson and


FIG. 2. Phylogeny based on Verdcourt’s (1985) ‘‘possible phylogeny based on the guess that the original ancestral specieswere trees with regular hypogynous flowers having hairy ovaries and winged seeds’’, constructed from similarities observedbetween herbarium specimens. Verdcourt envisioned three major groups (top to bottom): tuberous trees and shrubs of northeastAfrica (defined by the node marked ‘‘glabrous ovary, NE Africa, larger pollen grains’’), that is sister to a slender tree clade,and a bottle tree clade that is sister to the rest of the family.


TABLE 2. ILD Test pairwise comparisons.

Pairwise comparison p

PEPC 1 trnGPEPC 1 ITSPEPC 1 morphologytrnG 1 ITStrnG 1 morphologyITS 1 morphology

0.4280.10810.0840.3310.016

Carlquist 2001. For sectioning, leaves and flowers were passedthrough a dehydration series from 70% aqueous ethanol to 95%,through three changes of absolute ethanol, ending with threechanges of tertiary butyl alcohol, with the sample being allowedto remain in each solution at least overnight. Samples were em-bedded in paraffin, sectioned on a rotary microtome at 13 mm andstained in a series corresponding to Northen’s modification of Fos-ter’s ferric chloride-tannic acid staining series (Johansen 1940),with ferric aluminum sulfate substituted for ferric chloride. Forscanning electron microscope (SEM) observations, leaf and floraldissections were dehydrated to absolute ethanol, critical-pointdried, and mounted on aluminum stubs. The samples were sput-ter-coated on a Polaron E-5000 and observed with a Hitachi S-450SEM at 20 kilovolts.

Analysis. Phylogenetic analyses employed PAUP* 4.0b8 (Swof-ford 2001). Searches were heuristic with the parsimony optimalitycriterion, unweighted and unordered characters, starting treesfound via 1000 random additions, TBR branch swapping, theCOLLAPSE and STEEPEST DESCENT options off, MULTREESand ACCTRAN options in effect. Gaps were interpreted as miss-ing data and ignored. Bootstrap values were derived from 1000replicates of a full heuristic search. AutoDecay (Eriksson 1998) wasused to calculate decay indices (Bremer support).

‘‘Total evidence’’ Analysis. A combined data set was construct-ed to include all three molecular data sets and the morphologicaldata for each species. Three species are represented by more thanone sample: M. oleifera (PEPC, trnG 5 Stanley, s.n.; ITS 5 Olson,s.n.) M. ovalifolia (PEPC and trnG 5 Olson 718, ITS 5 Olson, s.n.);M. rivae (PEPC and trnG 5 Olson 677, ITS 5 Olson 701). The otherspecies were represented by the following samples: Cylicomorphaparviflora C. Kayombo 1296; M. drouhardii Olson 679; M. hildebrand-tii Olson ‘‘2’’; M. peregrina Danin, s.n.; M. borziana Olson 678; M.longituba Olson 708; M. ruspoliana Olson 702. While not ideal, it issometimes justifiable to allow different individuals to contributesequence data from different regions, especially if the species inquestion represent monophyletic lineages. Moringa hildebrandtiiwas absent from the trnG data set and these characters were codedas missing for this species.

Congruence testing. There is ample evidence that different to-pologies may be recovered from phylogenetic data sets of differentorigin (e.g., chloroplast and nuclear regions, Dumolin-Lapegue,Kremer, and Petit 1999; morphological and molecular data sets,Larson 1994). The Incongruence Length Difference test (ILD; Farriset al. 1994; the Partition Homogeneity Test option of PAUP*) andthe Templeton test (Templeton 1983; Larson 1994; Mason-Gamerand Kellogg 1996; Johnson and Soltis 1998) were used to assessthe level of congruence between the data sets.

ILD TEST. For this test, each possible pair of individual datasets are combined to form a single data set. Thus the four datasets used in this study were combined to form six data sets com-posed of two partitions each. These pairs are listed in Table 2. Onethousand Partition Homogeneity Test replicates were run, usingthe same heuristic search settings as those of the parsimony anal-yses of the individual data sets. The null hypothesis that the sumof the treelengths from the random partitions should not be sta-tistically significantly different from the sum of the treelengthsproduced by the original partitions are rejected in cases that dis-played a P value ,0.01 (as suggested by Johnson and Soltis 1998).

TEMPLETON TEST. The Templeton test compares the partition-ing of a data set (the ‘‘test data’’ ; terminology of Mason-Gamer

and Kellogg 1996) onto two trees, one resulting from an analysisof that data set (the ‘‘test tree’’) and a tree resulting from an anal-ysis of another data set (the ‘‘rival tree’’). The null hypothesis forthis test is that both the test tree and the rival tree represent sta-tistically equivalent, if not equally parsimonious, explanations ofa given data set, as assessed by an application of a Wilcoxonsigned-ranks test (Templeton 1983; Felsenstein 1985; Larson 1994).

The ‘‘total-evidence’’ data set was used for Templeton tests. Mor-inga pygmaea was absent from these comparisons because it waspresent only in the morphology data set. Similarly, Caricaceae se-quences were not included in these tests because they were notused in the PEPC and trnG analyses. Topologies resulting fromindividual data set analyses were compared to each other, and thetopology resulting from the total-evidence analysis was used as arival tree with respect to the individual data sets. Polytomies inrival trees were resolved by constraining a parsimony analysis ofthe test data set with the rival tree to find the resolution(s) ofpolytomies most compatible with the test data. This approachavoids inflating the rival tree length with polytomies, which arehere assumed to represent a lack of resolving power of the dataset rather than multiple simultaneous divergences (choice of rivaltrees is discussed by Mason-Gamer and Kellogg 1996; Cunning-ham 1997; Graham et al. 1998). The two-tailed critical values cal-culated using PAUP* 4.0b8 (Swofford 2000) were halved to com-pare them to the one-tailed values in Table 30 of Rohlf and Sokal(1981). In addition, the null hypothesis was not rejected in com-parisons had four characters or less that differed in length on thetwo trees (i.e., in cases where N,5).

RESULTS

Abbreviations and Figure Notes. Bootstrap values(generally those greater than 50%) and decay indicesare shown in the Figures below branches, and branchlengths are shown above. The following abbreviationsare used in the text and figure legends; TL 5 treelength in number of steps; CI 5 consistency index(Kluge and Farris 1969); RI 5 retention index (Farris1989); RC 5 rescaled consistency index (Farris 1989).The species groups referred to as the bottle trees (M.drouhardii, M. hildebrandtii, M. ovalifolia, M. stenopetala),slender trees (M. concanensis, M. oleifera, M. peregrina),and the tuberous shrubs and trees of northeast Africa(the ‘‘tuberous clade’’: M. arborea, M. borziana, M. lon-gituba, M. pygmaea, M. rivae, M. ruspoliana) are labeledon all trees. Within this latter group, the ’’rivae group’’(consisting of M. rivae and the three morphologicallysimilar species M. arborea, M. borziana, and M. pygmaea)and Section Dysmoringa (consisting only of M. longi-tuba) are also identified on the total-evidence tree (Fig.6). Indel typology follows Golenberg et al. (1993) asmodified by Hoot and Douglas (1998), where Type 1aindels are simple repeats or deletions of the same nu-cleotide, Type 1b indels are repeated motifs of two ormore bases that include more than one nucleotide, andall other indels are referred to the Type II category.

PEPC. The sequence used corresponds approximate-ly to bases 799–1297 of the Arabidopsis thaliana phos-phoenolpyruvate carboxylase sequence (GenBank ac-cession AF071788, Paterson, K. M. and H. G. Nimmo,unpubl.). Bases before position 798 and past position1297 were excluded from phylogenetic analyses be-cause they were missing for some species. Because of


alignment ambiguity between the ingroup and out-group sequences, the outgroup sequences were exclud-ed and the trees were rooted with Moringa drouhardii.This species is strongly supported as the sister taxonto the rest of the family in the ITS and morphologicalanalyses, and shares many morphological featureswith the Caricaceae (Olson and Carlquist 2001). Threepercent of the cells were coded as missing, mostly dueto missing sequence for M. hildebrandtii Olson ‘‘W’’and M. peregrina Danin s.n. Four Type 1a indels of 1bpeach also contribute to the cells coded as missing(alignment deposited as TreeBASE M1027). Of the 494bp used in the analysis, 39 were phylogenetically in-formative (8% of the total characters). Two most-par-simonious trees were recovered of 69 steps (CI 5 0.95;RI 5 0.97; RC 5 0.92). The strict consensus of thesetrees is shown in Fig. 3A. The trees differed in theirarrangement within the rivae group (a M. arborea—M.rivae clade sister to M. borziana vs. M. arborea as thesister group to a M. borziana—M. rivae clade).

trnG. The first ca. 144 bases align to the tRNA (Gly)gene of Sinapis alba (Liere and Link 1994) before ex-tending into the adjacent noncoding spacer. Beyondapproximately the 350th position, the sequences werecharacterized by alternating poly-A and poly-T tractsand a region of more than 100bp in length that distin-guished the rivae group species but was unalignable tothe rest of the family or the outgroup. These and sub-sequent bases were excluded from the analysis. As inthe PEPC analysis, M. drouhardii was used to root thetree. Of the 4.5% of cells coded as missing, most ofthese are accounted for by three Type 1a indels andthree Type 1b indels 1–7 bp in length. One region ofType 1a indels varied from 6–16 A residues and wasbounded by Gs on each side. This area was alignedsuch that the Gs aligned and gaps were introduced tocreate contiguous A blocks adjacent to the 59 G (align-ment deposited as TreeBASE M1026). Of the 335 char-acters used in the analysis, 20 were phylogeneticallyinformative (6% of the total characters). Four most-par-simonious trees were recovered of 30 steps (CI 5 0.87;RI 5 0.96; RC 5 0.83). The strict consensus of thesetrees is shown in Fig. 3B. The trees differed in theirarrangements within the slender tree clade and rivaegroup species 1 M. ruspoliana.

ITS. Several tracts of outgroup sequences were ex-cluded from the analysis because they were so di-verged as to preclude unambiguous alignment (at po-sitions 68–170; 243–259; 450–469; 478–490; 517–530;603–669). Of the cells in this alignment, 15.63% werescored as missing, partly because of missing sequencewithin the 5.8s gene for five species. Indels were morecommon in the spacer regions than in the coding 5.8sgene: the ITS-1 spacer had eight indels of Type 1a, andtwo Type 1b indels; the ITS-2 spacer had seven Type1a indels and two Type 1b indels; the 5.8s gene had

just one indel, of Type 1a (alignment deposited asTreeBASE M1028). Of the 705 characters used in theanalysis, 228 were phylogenetically informative (32%of the total characters). Eighteen most-parsimonioustrees were recovered of 431 steps (CI 5 0.81; RI 5 0.90;RC 5 0.73). The strict consensus of these trees isshown in Fig. 4. Most of the differences between themost-parsimonious trees were different arrangementsof the rivae group species and the multiple samples ofM. longituba. The bottle tree M. drouhardii is the sistertaxon to the rest of the family, making the bottle treespoly-, or perhaps para-, phyletic. The slender tree andtuberous clades are recovered with high support frombootstrap and decay indices.

Morphological Characters. The 28 characters inthis data set and their coding are enumerated in Ap-pendix 2. In this data matrix, 2.04% of the cells werescored as missing. Ten most-parsimonious trees wererecovered of 54 steps (CI 5 0.61; RI 5 0.73; RC 5 0.45).These trees differ in their arrangments of species with-in the slender trees and the tuberous clade. The reso-lution recovered from the bootstrap analysis is shownin Fig. 5. The bottle trees occur in this tree in a para-phyletic grade at the base of the tree, with the slendertree and tuberous clades sister to each other with mod-erate support. Nonmolecular characters provide sup-port at all levels of relationship, e.g., M. arborea and M.rivae are paired by the presence of crystalliferous ty-loses, a unique situation in the family (character 7; seelower tree in Fig. 5). Likewise, M. ruspoliana and M.longituba are united by entirely lacking paratracheal ax-ial parenchyma in their shoots (character 6).

To examine the effectiveness of characters basedsolely on mature morphologies, the nine charactersthat derive from ontogenetic studies (marked by anasterisk in Appendix 2) were removed. The 4 most-parsimonious trees from this analysis (40 steps; CI 50.75; RI 5 0.86; RC 5 0.65) showed less resolution,particularly at the base of the tree. This tree is alsoshown in Fig. 5.

Comparison of Individual Data Set Analyses. Thegroupings based on life form classes show differingpatterns of phylogenetic status: 1) The slender treesform a clade in all analyses with strong support exceptfor the trnG analysis. Although the monophyly of theslender trees is clear, no pattern of relationship amongthe three species that form the clade emerges in theseanalyses. 2) The tuberous clade appears strongly sup-ported in all analyses. In the PEPC analysis, two majordivisions appear within the tuberous clade: the rivaegroup, and a clade consisting of the two red-floweredspecies M. longituba, and M. ruspoliana. In the ITS andtrnG analyses, M. ruspoliana pairs with the rivae group.In all analyses, little resolution was found within therivae group. 3) The bottle trees are para- or poly- phy-letic in all analyses, with some members of this class


FIG. 3A-B. PEPC and trnG analyses. A. Strict consensus of the 2 most-parsimonious trees recovered in the analysis of PEPCsequence data (TL 69, CI 0.95, RI 0.97, RC 0.92). B. Strict consensus of the 4 most-parsimonious trees recovered in the analysisof trnG data (TL 30, CI 0.87, RI 0.96, RC 0.83).

grouping weakly with the tuberous clade. In the ITSand morphological analyses, M. drouhardii is well-sup-ported as the sister taxon to the rest of the family.

Total Evidence Analysis. Of the 1563 charactersused in the analysis, 155 were phylogenetically infor-mative (11% of the total characters). In the analysis ofthe ITS data set described above, 32% of the characterswere phylogenetically informative. However, of thecharacters in the ITS partition of the total-evidence

data set, the proportion of phylogenetically-informa-tive characters was reduced to 11%. This reduction invariation is due to the removal of multiple samples ofspecies such as M. longituba, of which six were includ-ed in the ITS analysis compared to the single sampleincluded in the total evidence analysis. The percent-ages of phylogenetically informative characters con-tributed by the PEPC, trnG, and morphological datasets were comparable to those found in the individual


FIG. 4. Strict consensus of the 18 most-parsimonious trees recovered in the analysis of ITS data (TL 431, CI 0.81, RI 0.90,RC 0.73).

data set analyses. A single most-parsimonious tree of432 steps was recovered (CI 5 0.86; RI 5 0.85; RC 50.73; this tree is shown in Fig. 6). The relationship be-tween the major groups in Moringa is resolved withbetter support in the total evidence analysis relative tothe individual analyses. The bottle trees are paraphy-letic in this analysis, with robustly-supported slendertree and tuberous clades. Moringa ruspoliana pairs with81% bootstrap and a decay index of 3 with the otherred-flowered species, M. longituba.

Congruence Testing. ILD TEST. Pairwise compar-isons indicate a generally high degree of congruencebetween the data sets, with P . 0.01 in all cases; all Pvalues were greater than 0.05 except for the ITS 1 mor-phology comparison. Test statistics are summarized inTab 2.

TEMPLETON TEST. In just less than half (46%) ofthe comparisons between trees derived from analysesof the individual data sets, the null hypothesis wasrejected, indicating that the test trees had a signifi-


FIG. 5. Top left tree: Consensus tree from bootstrap analysis of morphological data (TL 54, CI 0.61, RI 0.73, RC 0.45). Topright tree: Strict consensus of the 4 most-parsimonious trees recovered in the analysis of the morphological data with the 9ontogenetic characters removed (TL 40, CI 0.75, RI 0.86, RC 0.65). Lower tree: same as top left tree with unambiguous changesindicated on branches. The numbers correspond to the characters as numbered in App. 2. Unless indicated, change is to thederived state (0 to 1).


FIG. 6. Single most-parsimonious tree recovered in the ‘‘total evidence’’ analysis of all molecular, morphological, anatomical,and ontogenetic data, with M. hildebrandtii trnG characters (not sequenced for this species) coded as missing. (TL 432, CI 0.86,RI 0.85, RC 0.73). The position of M. pygmaea is shown in dashed lines based on the morphological analysis.

cantly more parsimonious partitioning of the test datathan did the rival trees. For example, the PEPC testdata set was mutually congruent with both the trnGand morphology data sets, but the null hypothesis wasrejected in all comparisons with ITS rival trees. In thereciprocal test, the PEPC rival trees were compatiblewith the ITS test data set. The results for all compar-isons, with number of characters differing in lengthbetween the two topologies and critical values, are giv-en in Table 3. In contrast to the variable results of Tem-pleton tests involving trees from individual data setanalyses, none of the comparisons of the total-evidencerival tree in the context of individual data sets resultedin a rejection of the null hypothesis. In other words,the tree derived from a pooling of all of the data rep-resents a topology that is compatible with each indi-vidual data set.

Population Sampling. Nine of the 13 describedMoringa species were represented by more than onesample in at least one data set (data set and supportfor the node subtending all the samples of the species.50% indicated in parentheses): M. concanensis (ITS),M. drouhardii (ITS, 100%, 44 decay), M. hildebrandtii(PEPC, 91%, 2 decay), M. longituba, (PEPC, 85%, 2 de-cay; trnG, 57%, 1 decay; ITS, 77%, 1 decay), M. oleifera

(ITS, 53%, 1 decay), M. ovalifolia (trnG), M. peregrina(trnG, 59%, 1 decay; ITS, 72%, 1 decay), M. rivae (ITS),and M. ruspoliana (trnG, 60%, 1 decay; ITS, 63%, 1 de-cay). None of the analyses rejected the monophyly ofany of these species, though support was particularlyweak for the pairings of the multiple samples of M.concanensis, M. oleifera, M. ovalifolia and M. rivae.

DISCUSSION

Congruence Testing and a Phylogeny of Moringaceae.Congruence of the data sets is indicated by the ILD testand that none of the individual data sets was able toreject the total evidence topology in the Templeton testcomparisons. I assume that this is because the total-ev-idence rival topology faithfully represents the phyloge-netic signal present in each of the data sets. Therefore,I choose the total evidence tree (Fig. 6) as the best skel-eton for a phylogenetic hypothesis of the Moringaceae.Moringa pygmaea was included only in the morpholog-ical analysis where it appears as the sister species toM. borziana. The best estimate of its position in theMoringa phylogeny is therefore sister to M. borziana,and it is shown in this position with dashed lines inFig. 6.


TABLE 3. Templeton test values. N 5 number of characters of different lengths between test tree and rival tree.

Test dataand tree

Rivaltree N P

Test dataand tree Rival tree N P

PPC1PPC1PPC2PPC2PPC1PPC1PPC2PPC2PPC 1–2trnG 1–4trnG 1–4trnG 1–4

trnG 1trnG 2trnG 1trnG 2ITS 1ITS 2ITS 1ITS 2morph 1–2PPCITS 1–2morph 1–4

533579972242

0.17970.08330.08330.17970.0082*0.0196*0.0196*0.0082*10.15730.0455*0.1573

morph 4morph 5, 7–8morph 6morph 9morph 10morph 1–3morph 4morph 5, 7–8morph 6morph 9morph 10morph 1–3

trnG 1trnG 1trnG 1trnG 1trnG 1trnG 2trnG 2trnG 2trnG 2trnG 2trnG 2ITS

101011121310128

13101111

0.16550.0578*0.23620.19670.25970.0578*0.19670.0339*0.25970.16550.23620.0018*

ITS 1ITS 2ITS 1ITS 2ITS 1ITS 2morph 1–3morph 4morph 5morph 6–7morph 8–10morph 1–3

PPCPPCtrnGtrnGmorph 1–3morph 1–3PPCPPCPPCPPCPPCtrnG 1

5119

117

132375

8

0.17970.13170.0209*0.0075*0.0588*0.0522*10.56370.70550.6547same topology0.0339*

morph 4morph 5, 7–8morph 6, 9morph 10trnG 1–4ITS 1–2morph 6morph 1–3morph 4, 10morph 5, 7–8morph 9

ITSITSITSITStotaltotaltotaltotaltotaltotaltotal

161315172

11725

3

0.0221*0.0030*0.0144*0.0184*0.15730.13170.705510.6547same topology0.5637

Ontogenetic Characters. Ontogenetic studies pro-vided characters that greatly improved the resolutionof the morphological analysis (cf. trees resulting fromanalyses with and without ontogenetic characters inFig. 5). Many of these are simply characters drawnfrom immature stages, e.g., germination type (charac-ter 15). The three basal species have cotyledons thatemerge from the seed, whereas in the rest of the familythe cotyledons remain within the seed coat. Character14, seasonal persistence of the shoot at different onto-genetic stages, can be considered a non-instantaneouscharacter in that it is necessary to observe more thanone ontogenetic stage to determine the state present ineach species. Another such character, the pattern of an-ther orientation in ontogeny (character 24), unites theslender trees and tuberous clades. Examination of thischaracter substantially altered homology interpreta-tions that were based only on mature flowers. The ra-dially symmetrical flowers of the pachycaul speciesseemed clearly to be symplesiomorphic with the acti-nomorphic flowers of Caricaceae. However, examina-tion of their ontogeny showed bilateral symmetry tobe a property of all species of Moringa (Olson, unpubl.data; see also Appendix 2).

Phylogeny and the Distribution of Moringa. Twomonophyletic groups in Moringa show clear geograph-ical associations (highlighted on the total evidence treein Fig. 7). Within these groups, there are pairs of sisterspecies that share morphological similarities and ap-pear to have largely allopatric geographical distribu-tions. One of these clades, the slender trees, is nearly

restricted to Asia, and its species occur in a brokenband from Arabia to Bangladesh. Moringa peregrina oc-curs from the Dead Sea to southern Arabia and north-ern Somalia. The pair M. concanensis and M. oleifera arevery similar in habit, leaves, and flowers but differmost conspicuously in bark morphology. Moringa con-canensis is widely distributed from Pakistan to Bang-ladesh and along the length of peninsular India, butM. oleifera is apparently native to dry lowlands ofnorthern India and has never been recorded co-occur-ring with M. concanensis (though recent documentationthat M. oleifera still occurs in the wild is scant).

The tuberous group forms a monophyletic cladewith three species pairs, all of which are restricted tothe Horn of Africa. The members of the red-floweredspecies pair, M. longituba and M. ruspoliana co-occur inthe same general region only in extreme northeastKenya and central northern Somalia. Otherwise, theirdistributions are exclusive, with M. longituba having amore southerly range extending nearly 300 km farthersouth into Kenya, from Moyale to Wajir, and reachingnorthern Somalia via an arc roughly south of the So-malia-Ethiopia border. In contrast, the northern andsouthern extremes of the distribution of M. ruspoliana,where it co-occurs with M. longituba, are connected byan arc extending through southeastern Ethiopia alongthe line roughly delimited by the northwestern edgeof the hot Ogaden lowlands. A member of another spe-cies pair, M. borziana, is well-documented from south-eastern Kenya and southern Somalia, always fromwithin 200 km of the coast. Its sister species, M. pyg-


FIG. 7. Geographic ranges of Moringa species and phylogeny. Broad regions are shaded in gray and labeled at left. Morespecific ranges are cited at right. The small gray boxes at right highlight largely allopatric species pairs within the tuberousclade.

maea, may replace M. borziana in similar situations innorthern Somalia, but exploration of this area is stilltoo limited for definitive conclusions. Likewise, verylittle is known about the geographical distributions ofboth members of the species pair M. rivae—M. arborea,but the species do appear to have differing habitatpreferences, with M. arborea exploiting the dynamichabitat of canyon bottoms and M. rivae preferring dri-er, more exposed sites.

In addition to forming a non-monophyletic group,the bottle trees show a less coherent pattern of distri-bution than the slender tree and tuberous clades. Threeof the four species making up the basal grade in thefamily have austral distributions with M. drouhardiiand M. hildebrandtii being restricted to Madagascarand M. ovalifolia reaching from central southern Na-mibia to southwestern Angola. Moringa stenopetala isfound well to the north in the western Horn of Africajust to the west of the area occupied by the tuberousgroup.

Previous Classifications of Moringa. That a genusof just 13 species should be divided into infragenerictaxa is a reflection of its remarkable morphological di-versity. In 1902, Engler placed seven of the eight spe-cies then known into Section (Eu)Moringa, and created

Section Dysmoringa to emphasize the unique flowers ofM. longituba (Fig. 1J). Verdcourt (1958, 1985) noted themorphological cohesiveness of the four phylogeneti-cally basal species and transferred them to SectionDonaldsonia. However, both Donaldsonia and Moringaemerge in phylogenetic reconstructions as paraphylet-ic, and the single species of Section Dysmoringa is em-bedded within the Section Moringa clade. Because ofthe paraphyletic assemblage at the base of the family,there is little gain to communication in dividing Mor-inga into monophyletic groupings (e.g., creating a sec-tion for each of the four species of the basal grade). Itherefore recommend that the sections should be dis-mantled. In keeping with the intent of previous au-thors to provide a means of designating distinctivegroups within the genus, the informal terms ‘‘bottletree grade,’’ ‘‘slender tree clade,’’ and ‘‘tuberous’’clade, which is divided into the ’’rivae clade’’ and the‘‘red-flowered clade’’ seem suitable for this purpose(these groups are highlighted on the tree in Fig. 6).

When groupings based on life form (Table 1) aresuperimposed on the phylogeny, the bottle trees ap-pear paraphyletic, and slender trees monophyletic; thesarcorhizal trees (M. arborea and M. ruspoliana) arepolyphyletic within the tuberous shrubs. In constrast,


the species groups delimited by wood anatomy (Table1) are entirely congruent with the major groupings re-covered in phylogenetic analyses. Particularly note-worthy is the confirmation of the pairing of M. longi-tuba and M. ruspoliana, a relationship not proposed be-fore wood anatomy studies. Likewise, the grouping ofthe tuberous species into a clade mirrors the conclu-sions of Ferguson (1985) based on pollen size variation.Verdcourt’s (1985) phylogeny (Fig. 2), developed fromoften fragmentary herbarium specimens and withoutbenefit of an outgroup, is very similar to the one fa-vored here. It differs from the tree in Fig. 6 only indepicting a monophyletic bottle tree clade and M. rus-poliana sister to the tuberous clade rather than to M.longituba.

Verdcourt (1985) noted that ‘‘adequate materials fora really thorough monograph were no more availabletoday than they were 25 years ago when the idea wasfirst conceived by Mr. J. B. Gillett.’’ The statement isequally true now. As they were 40 years ago, collec-tions are most seriously lacking from the tuberousclade of the Horn of Africa. In particular, there havebeen almost no collections from the key region ofsoutheastern Ethiopia during this time. Human conflictin the Horn, the heart of Moringa species diversity, hasproven a significant deterrent to exploration. The re-moteness of the localities of most of these species andthe difficulty of access to large areas of this dry trop-ical region are further challenges to exploration. Bettersampling within the tuberous clade should clarify phy-logenetic relationships within the rivae clade, provideadditional samples of M. pygmaea, and will almost cer-tainly reveal undescribed species.

ACKNOWLEDGEMENTS. Grassie, Barbara; ‘s’avuto dir? My con-tinued gratitude for the collaboration of David Odee in Nairobiand Joseph Machua in the field. Gilfrid Powys was an inspirationin getting to the field and sky. Field work would have been im-possible without Ambia A. Osman and Mohammed, Abdiaziz‘‘Jack’’ Bashir, Halima Abdi Mohammed and Ahmad Salat Omar,Geoffrey Muluvi, Hassan A. Sheikh, Shahina Ghazanfar, MartinFisher, Sylvain Razafimandimbison, V. Amalan Stanley, Fr. K. M.Mathew, and Herta Kolberg. DNA lab work at Washington Uni-versity was possible thanks to the graciousness of Barbara Schaal.Thank you to Peter Raven, Mick Richardson, and Allan Larson,for their time and endless support. James Rodman and BernardVerdcourt provided much encouragement and ideas. Joe and Mir-ella Olson have been amazingly patient. Sherwin Carlquist is awealth of assistance, encouragement, and inspiration. Peter Ste-vens, Elizbeth Kellogg, Mike Dyer, Jessica Ingram, and Mike Veithgave help and suggestions. Chuck Hanson, Avinoam Danin, TomVanDevender, Burl Mostul, David Orr, Nathan Wong, and WinnieSingeo generously provided material. Jason Bradford, Ana LucıaCaicedo Samper, John Gaskin, Paula Kover, Simon Malcomber, Al-lison J. Miller, Ken Olsen, and Jason Rauscher made lab work apleasure. Sara Hoot and an anonymous reviewer generously im-proved the manuscript. Field and lab work were supported bygrant # 6141–98 from the Committee for Research and Explorationof the National Geographic Society, United States National ScienceFoundation Doctoral Dissertation Improvement Award DEB-9801128, and the Andrew Mellon Foundation.

LITERATURE CITED

APG (ANGIOSPERM PHYLOGENY GROUP). 1998. An ordinal classi-fication for the families of flowering plants. Annals of theMissouri Botanical Garden 85: 531–553.

BADILLO, V. M. 1971. Monografia de la familia Caricaceae. Maracay:Asociacion de Profesores.

BAYER, R. J., D. E. SOLTIS, and P. S. SOLTIS. 1996. Phylogenetic in-ferences in Antennaria (Asteraceae: Gnaphalieae: Cassiinae)based on sequences from nuclear ribosomal DNA internaltranscribed spacers (ITS). American Journal of Botany 83:516–527.

CARLQUIST, S. 1998. Wood and bark anatomy of Caricaceae: cor-relations with systematics and habit. International Associa-tion of Wood Anatomists Journal 19: 191–206.

CUNNINGHAM, C. W. 1997. Can three incongruence tests predictwhen data should be combined? Molecular Biology and Evo-lution 14: 733–740.

DE QUEIROZ, K. 1985. The ontogentic method for determiningcharacter polarity and its relevance to phylogenetic system-atics. Systematic Zoology 34: 280–299.

DOYLE, J. J. and J. L. DOYLE. 1987. A rapid isolation procedure forsmall quantities of fresh leaf tissue. Phytochemistry Bulletin19: 11–15.

DUMOLIN-LAPEGUE, S., A. KREMER, and R. J. PETIT. 1999. Are chlo-roplast and mitochondrial DNA variation species indepen-dent in oaks? Evolution 53: 1406–1413.

EDWARDS, K., C. JOHNSTONE, and C. THOMSON. 1994. A simplemethod of extraction. Nucleic Acid Research 19: 1349.

ENGLER, A. 1902. Contribuzioni alla conoscenza della floradell’Africa orientale. Annuario Reale dell’Istituto Botanico diRoma 9: 241–256.

ERIKSSON, T. 1998. AutoDecay ver.4.0 (program distributed by theauthor). Department of Botany, Stockholm University. Stock-holm.

FARRIS, J. S. 1989. The retention index and the rescaled consistencyindex. Cladistics 5: 417–419.

———, M. KAELLERSJO, A. G. KLUGE, and C. BULT. 1994. Testingsignificance of incongruence. Cladistics 10: 315–319.

FELSENSTEIN, J. 1985. Confidence limits on phylogenies with a mo-lecular clock. Systematic Zoology 34: 152–161.

FERGUSON, I. K. 1985. Pollen morphology of the Moringaceae. KewBulletin 40: 25–34.

GHASI, S., E. NWODOBO, and J. O. OFILI. 2000. Hypocholesterolemiceffects of crude extract of leaf of Moringa oleifera Lam in high-fat diet fed wistar rats. Journal of Ethnopharmacology 69: 21–25.

GOLENBERG, E. M., M. T. CLEGG, M. L. DURBIN, J. DOEBLY, and D.P. MA. 1993. Evolution of a noncoding region of the cholor-plast genome. Molecular Phylogenetics and Evolution 2: 52–64.

GRAHAM, S. W., J. R. KOHN, B. R. MORTON, J. E. ECKENWALDER,and S. C. H. BARRETT. 1998. Phylogenetic congruence anddiscordance among one morhpological and three moelculardata sets from Pontederiaceae. Systematic Biology 47: 545–567.

HAMILTON, M. B. 1999. Four primer pairs for the amplification ofchloroplast intergenic regions with intraspecific variation.Molecular Ecology 8: 513–525.

HOOT, S. B. and A. W. DOUGLAS. 1998. Phylogeny of the Protea-ceae based on atpB and atpB-rbcL intergenic spacer regionsequences. Australian Systematic Botany 11: 301–320.

JAHN, S. A. 2001. Drinking water from Chinese rivers: challengesof clarification. Journal of Water Supply Research and Tech-nology-Aqua 50: 15–27.

JOHANSEN, D. A. 1940. Plant microtechnique. New York: McGrawHill.

JOHNSON, L. A. and D. E. SOLTIS. 1998. Assessing incongruence:empirical examples from molecular data. Pp 297–343 in Mo-


lecular systematics of plants II. DNA sequencing, eds. D. E. Soltis,P. S. Soltis, and J. J. Doyle. Boston: Kluwer.

JUMELLE, M. H. 1930. Les Moringa de Madagascar. Annales duMusee Colonial deMarseille ser. 4. 8: 1–20.

KALOGO, Y. and W. VERSTRAETE. 2000. Technical feasibility of thetreatment of domestic wastewater by a CEPS-UASB system.Environmental Technology 21: 55–65.

KLUGE, A. G. and J. S. FARRIS. 1969. Quantitative phyletics and theevolution of anurans. Systematic Zoology 18: 1–32.

LARSON, A. 1994. The comparison of morphological and moleculardata in phylogenetic systematics. Pp. 371–390 in Molecularecology and evolution: Approaches and applications, eds. B. Schier-water, B. Streit, G. P. Wagner, and R. DeSalle. Basel, Switzer-land: Birkhauser Verlag.

LATZKO, E. and G. J. KELLY. 1983. The many-faceted function ofphosphoenolpyruvate carboxylase in C3 plants. PhysiologieVegetale 21: 805–815.

LEE, S. and J. WEN. 2001. A phylogenetic analysis of Prunus andthe Amygdaloideae (Rosaceae) using ITS sequences of nucle-ar ribosomal DNA. American Journal of Botany 88: 150–160.

LIERE, K. and G. LINK. 1994. Structure and expression character-istics of the chloroplast DNA region containing the split genefor tRNA(Gly) (UCC) from mustard (Sinapis alba L.) CurrentGenetics 26: 557–563.

MASON-GAMER, R. J. and E. KELLOGG. 1996. Testing for phyloge-netic conflict among molecular data sets in the Triticeae (Gra-mineae). Systematic Biology 45: 524–545.

MOODY, M. L., L. HUFFORD, D. E. SOLTIS, and P. E. SOLTIS. 2001.Phylogenetic relationships of Loasaceae subfamily Grono-vioideae inferred from matK and ITS sequence data. Amer-ican Journal of Botany 88: 326–336.

OLIVEIRA, J. T. A., S. B. SILVEIRA, I. M. VASCONCELOS, B. S. CA-VADA, and R. A. MOREIRA. 1999. Compositional and nutri-tional attributes of seeds from the multiple purpose tree Mor-inga oleifera Lamarck. Journal of the Science of Food and Ag-riculture 79: 815–820.

OLSON, M. E. and S. CARLQUIST. 2001. Stem and root anatomicalcorrelations with life form diversity, ecology, and systematicsin Moringa (Moringaceae). Botanical Journal of the LinneanSociety 135: 315–348.

——— and S. G. RAZAFIMANDIMBISON. 2000. Moringa hildebrandtii:a tree extinct in the wild but preserved by indigenous hor-ticultural practices in Madagascar. Adansonia ser. 3 22: 217–221.

PANSTRUGA, R., A. SEILER, H.-J. HIRSCH, and F. KREUZALER. 1995.Genomic structure of a phosphoenolpyruvate carboxylasegene from potato (Solanum tuberosum). Plant Physiology 109:1126.

RAMBAUT, A. 1996–1998. Se-Al Sequence Alignment Editor v1.0alpha 1. Obtained at http://evolve.zps.ox.ac.uk/software/Se-Al/Se-Al10a1.hqx.

RODMAN, J. E., P. A. SOLTIS, D. E. SOLTIS, K. J. SYTSMA, and K. G.KAROL. 1998. Parallel evolution of glucosinolate biosynthesisinferred from congruent nuclear and plastid gene phyloge-nies. American Journal of Botany 85: 997–1006.

ROHLF, F. J. and R. R. SOKAL. 1981. Statistical tables. W. H. Freeman,New York.

SALEEM, R. and J. MEINWALD. 2000. Synthesis of novel hypotensivearomatic thiocarbamate glycosides. Journal of the ChemicalSociety-Perkin Transactions 1: 391–394.

SANG, T. and Y. ZHONG. 2000. Testing hybridization hypothesesbased on incongruent gene trees. Systematic Biology 49: 422–434.

SIIKAMAKI, P. 1999. Developmental instability in hybrids bewtweenLychnis viscaria and Lychnis alpina (Caryophyllaceae). Ameri-can Journal of Botany 86: 1683–1686.

SWOFFORD, D. L. 2000. PAUP*. Phylogenetic Analysis Using Par-simony (*and Other Methods). Version 4. Sinauer Associates,Sunderland, Massachusetts.

TABERLET, P. L., F. GIELLY, F. PAUTOUT, and J. BOUVET. 1991. Uni-versal primers for amplification of noncoding regions ofcpDNA. Plant Molecular Biology 17: 1105–1109.

TEMPLETON, A. R. 1983. Phylogenetic inference from restrictionendonuclease cleavage site maps with particular reference tothe evolution of humans and the apes. Evolution 37: 221–244.

VERDCOURT, B. 1958. Moringaceae: a correction. Kew Bulletin 13:385.

———. 1985. A synopsis of the Moringaceae. Kew Bulletin 40: 1–23.


AP

PE

ND

IX1.

Spec

ies

sam

plin

g,lo

cise

quen

ced,

and

vouc

her

info

rmat

ion.

Car

icac

eae

dist

ribu

tion

note

sfr

omBa

dillo

,197

1.

Spec

ies

Col

lect

oran

d#

Loc

alit

yR

egio

nsse

quen

ced

and

Gen

Ban

kac

cess

ion

num

ber

Her

bari

aw

ith

vouc

hers

Car

ica

mic

roca

rpa

Jacq

uin

Wai

mea

Arb

oret

um90

p260

Cul

tivat

edpl

ant

from

Wai

mea

Arb

oret

um,H

alei

wa,

Haw

aii:

orig

inal

lyfr

omVe

nezu

ela:

Ara

gua:

Parq

ueN

acio

nalH

.Pitt

ier

PEPC

AF3

7860

4;tr

nGA

F378

623;

ITS

AF3

7857

8M

O

Cyl

icom

orph

apa

rvifl

ora

Urb

anK

ayom

bo12

96Ta

nzan

ia:I

ring

a:M

ufind

iD

istr

ict:

Lula

nda

Fore

stR

eser

vePE

PCA

F378

605;

trnG

AF3

7862

7;IT

SA

F378

579

MO

Mw

ango

ka38

7Ta

nzan

ia:T

anga

:058

0494

49S

3882

5934

0Etr

nGA

F378

626;

ITS

AF3

7858

0M

OJa

cara

tia

coru

mbe

nsis

Kun

tze

Wai

mea

Arb

oret

um84

p423

Cul

tiva

ted

plan

tfr

omW

aim

eaA

rbor

etum

,Hal

eiw

a,H

awai

i:re

port

edfr

omB

oliv

ia,P

arag

uay,

Arg

enti

natr

nGA

F378

622

MO

Ols

on,s

.n.

Cul

tiva

ted

plan

tfr

omA

rid

Lan

ds

Gre

enho

uses

,Tuc

son,

AZ

ITS

AF3

7857

5M

OJa

rilla

choc

ola

Stan

dle

yR

eina

99-9

62A

Mex

ico:

Sono

ra:M

unic

ipio

deYe

cora

:Cur

eatr

nGA

F378

624;

ITS

AF3

7857

7M

OJa

rilla

hete

roph

ylla

(Cer

v.ex

La

Lla

ve)

Rus

byO

lson

,s.n

.C

ulti

vate

dpl

ant

kind

lypr

ovid

edby

Sher

win

Car

lqui

st;n

ativ

eto

cent

ral

wes

tern

Mex

ico

trnG

AF3

7862

5;IT

SA

F378

576

MO

Mor

inga

arbo

rea

Verd

c.O

lson

and

Mac

hua

714

Ken

ya:N

orth

east

ern

Prov

ince

:Man

dera

Dis

tric

t:ca

.35

kmN

Wof

Rha

mu

onro

adto

Mal

kaM

ari,

insm

allr

ocky

cany

onPE

PCA

F378

616;

trnG

AF3

7863

8;IT

SA

F378

592

MO

,EA

,K

Mor

inga

borz

iana

Mat

tei

Ols

onan

dM

uluv

i67

8K

enya

:Coa

stPr

ovin

ce:T

aita

Dis

tric

t:SW

ofTs

avo

Eas

tN

a-tio

nal

Park

Voi

gate

PEPC

AF3

7861

7;tr

nGA

F378

639;

ITS

AF3

7859

3M

O,R

HT,

FT,K

Mor

inga

conc

anen

sis

Nim

mo

Ols

onan

dSt

anle

y70

0(i

ndiv

idua

l2un

less

spec

ified

)

Ind

ia:T

amil

Nad

u:U

dim

elpe

tD

ist.

Coi

mba

tore

Rd.

toPa

ra-

pad

dur

Dam

inPa

lni

Hill

sca

.200

kmW

SWof

Tir

uchi

ra-

palli

PEPC

AF3

7861

3;tr

nGA

F378

633;

ITS

AF3

7858

6,A

F378

587

MO

,EA

,FT,

K

Mor

inga

drou

hard

iiJu

m.

Ols

on67

9an

d‘‘A

m-

boas

ary’

’M

adag

asca

r:T

ulea

r:ne

arA

mbo

asar

yPE

PCA

F378

606,

AF3

7860

7,tr

nGA

F378

628;

ITS

AF3

7858

1,A

F378

582

MO

,K,T

AN

Mor

inga

hild

ebra

ndti

iEng

l.O

lson

‘‘E’’

and

‘‘W’’

Mad

agas

car:

Tul

ear:

cult

ivat

edin

villa

geof

Am

bohi

mah

ave-

lona

PEPC

AF3

7860

8;A

F378

609

MO

,EA

,FT,

K,

TAN

Ols

on‘‘2

’’C

ulti

vate

dpl

ant

prov

ided

byR

are

Plan

tR

esea

rch,

Port

land

,O

rego

nPE

PCA

F378

610;

ITS

AF3

7858

4M

O,K

,TA

N

Mor

inga

long

itub

aE

ngl.

Ols

onan

dM

achu

a70

4K

enya

:Nor

thea

ster

nPr

ovin

ce:M

ande

raD

istr

ict:

ca.2

0km

WN

Wof

Man

dera

near

loca

lity

ofFi

lqo

PEPC

AF3

7862

0;IT

SA

F378

598

MO

,K

Ols

onan

dM

achu

a70

8K

enya

:Nor

thea

ster

nPr

ovin

ce:W

ajir

Dis

tric

t:ca

.20

kmN

ofW

ajir

PEPC

AF3

7861

9;tr

nGA

F378

644;

ITS

AF3

7859

9M

O,E

A,F

T,K

Ols

onan

dM

achu

a70

9K

enya

:Nor

thea

ster

nPr

ovin

ce:W

ajir

Dis

tric

t:L

afal

eyvi

llage

,ca

.8km

Nof

Waj

irtr

nGA

F378

646;

ITS

AF3

7860

1M

O,E

A,F

T,K

Ols

onan

dM

achu

a71

0K

enya

:Nor

thea

ster

nPr

ovin

ce:W

ajir

Dis

tric

t:ca

.12

kmE

ofW

ajir

trnG

AF3

7864

3;IT

SA

F378

602

MO

,EA

,FT,

K

Ols

onan

dM

achu

a71

2K

enya

:Nor

thea

ster

nPr

ovin

ce:W

ajir

Dis

tric

t:ca

.30

kmE

ofW

ajir

(the

‘‘Mor

inga

sp.K

enya

’’of

Verd

cour

t,19

85)

trnG

AF3

7864

5;IT

SA

F378

603

MO

,EA

,FT,

K

Hor

woo

d,s.

n.So

mal

ia:N

ugal

:Las

Ano

d:n

ear

Las

Ano

dIT

SA

F378

600

MO

Mor

inga

olei

fera

Lam

.St

anle

ys.

n.In

dia

:Tam

ilN

adu:

Chr

ompe

t(a

neig

hbor

hood

ofC

henn

ai)

PEPC

AF3

7861

4;tr

nGA

F378

634;

ITS

AF3

7858

8M

O

Ols

on,s

.n.

Com

mer

cial

lyav

aila

ble

annu

alcu

ltiv

arca

lled

‘‘PK

M’’,

prov

id-

edby

V.A

mal

anSt

anle

y,C

henn

ai,I

ndia

ITS

AF3

7858

9M

O


AP

PE

ND

IX1.

Con

tinue

d.

Spec

ies

Col

lect

oran

d#

Loc

alit

yR

egio

nsse

quen

ced

and

Gen

Ban

kac

cess

ion

num

ber

Her

bari

aw

ith

vouc

hers

Mor

inga

oval

ifolia

Din

ter

and

A.B

erge

rO

lson

716

and

718

Nam

ibia

:Nam

ib-N

aukl

uft

Park

:Wof

Kui

seb

Pass

PEPC

AF3

7861

2;tr

nGA

F378

629,

AF3

7863

0M

O,W

IND

Ols

on,s

.n.

Nam

ibia

:Kar

ibib

:see

dpu

rcha

sed

from

B&

TW

orld

Seed

str

nGA

F378

631;

ITS

AF3

7858

3M

OM

orin

gape

regr

ina

(For

ssk.

)Fi

ori

Ols

on56

9O

man

:Dho

far:

inbr

oad

Wad

ibe

twee

nR

aysu

tan

dM

ughs

ayl

trnG

AF3

7863

7M

O,E

A,F

T,K

Ols

on,s

.n.

Seed

spu

rcha

sed

inC

airo

mar

ket,

prov

ided

byB

enSt

ern,

Uni

-ve

rsit

yof

Bra

dfo

rd,U

Ktr

nGA

F378

636;

ITS

AF3

7859

1M

O

Dan

in,s

.n.

Isra

el:S

outh

ern

Reg

ion:

En-

Ged

iPE

PCA

F378

615;

trnG

AF3

7863

5;IT

SA

F378

590

MO

Mor

inga

pygm

aea

Verd

c.N

ugen

t25

Som

alia

:Nor

ther

nre

gion

:13

kmE

ofQ

ard

hoai

rstr

ipno

nesu

cces

sful

EAG

love

ran

dG

illila

nd11

94So

mal

ia:N

orth

coas

tne

arB

erbe

rano

nesu

cces

sful

EA,K

Mor

inga

riva

eC

hiov

.O

lson

and

Pow

ys67

7K

enya

:Eas

tern

Prov

ince

:Mar

sabi

tD

istr

ict:

Esl

ope

ofB

aio

Mou

ntai

nPE

PCA

F378

618;

trnG

AF3

7864

0;IT

SA

F378

594

MO

,EA

,FT,

K

Ols

onan

dM

achu

a70

1K

enya

:Nor

thea

ster

nPr

ovin

ce:M

ande

raD

istr

ict:

ca.4

kmN

ofR

ham

uIT

SA

F378

595

MO

,EA

,FT,

K

Mor

inga

rusp

olia

naE

ngl.

Ols

onan

dM

achu

a70

2K

enya

:Nor

thea

ster

nPr

ovin

ce:M

ande

raD

istr

ict:

Rha

mu-

Dim

-tu

Div

isio

n:ar

ound

villa

geof

Yabi

cho

PEPC

AF3

7862

1;tr

nGA

F378

641;

ITS

AF3

7859

7M

O,E

A,F

T,K

Ols

onan

dM

achu

a70

3K

enya

:Nor

thea

ster

nPr

ovin

ce:M

ande

raD

istr

ict:

onK

enya

-So-

mal

iabo

rder

Sof

Daw

aR

iver

trnG

AF3

7864

2;IT

SA

F378

596

MO

,EA

,FT,

K

Mor

inga

sten

opet

ala

(Bak

erf.)

Cuf

odon

tis

Ols

on67

5K

enya

:Rif

tVa

lley

Prov

ince

:Bar

ingo

Dis

tric

t:Pa

rmal

okIs

land

,L

ake

Bar

ingo

PEPC

AF3

7861

1;tr

nGA

F378

632;

ITS

AF3

7858

5M

O,E

A,F

T,K


APPENDIX 2. Morphological characters and data set.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28

CylicomorphaM. drouhardiiM. hildebrandtiiM. ovalifoliaM. stenopetalaM. concanensisM. oleiferaM. peregrinaM. arboreaM. rivaeM. borzianaM. pygmaeaM. longitubaM. ruspoliana

00000110000000

00000000010011

00110000100000

00010011010000

0011010010000?

00000000000?11

00000000110000

11100100110101

00011110000000

00000111111111

00000111000000

00000000011110

00010111111111

00010111112221

00001111111111

00000111?00?0?

00011111111111

00111110111111

00010011111111

000?0000000011

00000000111?11

00000000110022

000000000

0/11100

01111222222222

00111111000000

?0?0?111000?00

00000111222222

01111222222222

Assignment of morphological character states. Characters marked with an asterisk are derived from ontogenetic studies.

A. Wood, bark, and root anatomical characters. The following characters are for the most part depicted and discussed in Carlquist(1998) and Olson and Carlquist (2001).

1. Callose plugs on sieve plates: Both in transections and longisections stained with safranin, large pink-staining plugs, presumablycallose, are conspicuous on the sieve plates of most species. 0 5 present, 1 5 absent.

2. Rhomboidal crystals in phloem rays. 0 5 absent, 1 5 present.3. Phloem ray sclereids. 0 5 absent, 1 5 present.4. Druses in cortical sclerenchyma. The thick-walled cells present in the outer bark of some species contain druses that are completely

immobilized within their massive walls. 0 5 absent, 1 5 present.5. Phelloderm sclereids. 0 5 absent, 1 5 present.6. Paratracheal axial parenchyma in shoots: 0 5 present, 1 5 absent.7. Druses in tyloses. 0 5 absent, 1 5 present.8. Druses in xylem rays. 0 5 present, 1 5 absent.9. Uniserate wings on multiseriate rays. 0 5 present, 1 5 absent.

10. Principal xylem cell type in stems. Some species are characterized by large amounts of paratracheal axial parenchyma and fewlibriform fibers, other species by the opposite condition. 0 5 paratracheal axial parenchyma, 1 5 libriform fibers.

11. Growth rings. Most species are characterized by seasonal change in xylem cell size and shape and often cell type. A few speciesshow little if any seasonal fluctuation in cell shape. 0 5 present, 1 5 absent.

12. Libriform fibers in root secondary xylem. 0 5 present, 1 5 absent.

B. Seed, seedling, and leaf characters.

13. *Ground tissue proliferation in epicotyl. All Moringa and all Caricaceae examined form swollen, often tuberous hypocotyls early inontogeny. In many species, this swelling of the lower part of the stem extends above the insertion of the cotyledons into the epicotyl.In other species, there is a strong differentiation into a bloated underground tuberous epicotyl/root and a slender abovegroundstem that usually survives only one season before dying back to the tuber. 0 5 epicotyl swollen, 1 5 epicotyl slender.

14. *Seasonal persistence of the shoot at different ontogenetic stages. In several species the shoot formed upon germination grows veryrapidly in height and, barring injury, eventually forms the main bole of the tree. These shoots may be considered permanent in thatthey are not shed naturally by the plant. In contrast, the juvenile plants of other species often persist through many seasons astubers that send up seasonal shoots when conditions are favorable and die back to the tuber during drought. A subset of the Moringaspecies with ephemeral juvenile shoots eventually form root systems that are sufficiently large to support permanent shoots, andthe plant begins to grow into a tree with a permanent trunk. Still other species maintain the characteristic of regularly dying backto the tuber in times of drought throughout the life of the plant. 0 5 juvenile and adult shoots permanent, 1 5 juvenile shootsephemeral, adult shoots permanent, 2 5 juvenile and adult shoots ephemeral.

15. *Germination phanerocotylar (cotyledons emerge from the seed coat) 5 0, 1 5 germination cryptocotylar (cotyledons remain in theseed coat).

16. *Leaves palmate 5 0, leaves pinnate 5 1. In the Caricaceae, most leaves are palmate, especially in Cylicomorpha and Jacaratia. InMoringa, adult leaves are pinnate. However, the juvenile leaves of many species are distinctly palmate, and the transition to pinnateleaves occurs over the first seven leaves produced by the seedling.

17. *Leaf margin of 1st leaves. The first leaves of some species are characterized by irregular margins with occasional lobes andindentations (here denoted ‘‘irregular margins’’). Other species have entire leaf margins. 0 5 irregular margins, 1 5 entire margins.

18. Seed wings. Most Moringa species are characterized by wide, hyaline wings that run the length of the seeds. Two species, and theCaricaceae, are characterized by wingless seeds with variously irregular surfaces. 0 5 wings absent, 1 5 wings present.

19. Leaftip epidermal layers. Some species have just one epidermal layer at the leaf margin, whereas others have several. Surprisingly,this character does not correlate with leaflet size. 0 5 one epidermal layer, 1 5 multiple epidermal layers.

20. Leaf trichomes. Whereas the young leaves of many species are covered with sometimes dense indumentum, the adult leaves of mostof these species are apparently glabrous. Character states were assessed both from examination of intact leaves and of leaf serialsections. 0 5 absent, 1 5 present.


21. Epidermal cell dimples. In some species, the cells surrounding the stomata show dimples or folds along the surfaces contacting theguard cells and perpendicular to them. 0 5 dimples absent, 1 5 dimples present.

C. Floral characters.

22. Flower color 1. Distribution of red pigmentation. Some species have localized maroon or pink patches, whereas others have densered pigmentation distributed throughout the perianth. 0 5 absent or along midvein, 1 5 distinctly aggregated at petal tips, 2 5throughout perianth.

23. Flower color 2. Yellow. Flowers cream, with or without pink or brown lines or blotches, or red 5 0, 1 5 bright yellow.24. *Anther orientation in ontogeny. The five anthers of all species studied initially point to the center of the flower. As anthesis

approaches, the filaments twist. In some species, three of the anthers come to face one way while the others come to face theopposing direction (here denoted 3/2 orientation). In other species, four anthers point the same direction and only one faces theopposing direction (here denoted 4/1 orientation). The Caricaceae appear to maintain centrally-pointing anthers throughout ontog-eny. 0 5 no change in orientation; 1 5 3/2, 2 5 4/1.

25. Ovary pubescence. The basal half of the ovary in some species is clothed in a dense coat of long unicellular trichomes that seem toform a barrier to the nectariferous lower part of the hypanthium. 0 5 ovary glabrous, 1 5 ovary pubescent.

26. *Carpel emergence. During floral ontogeny, the ovary emerges at the same time as the anthers in some species, but only well afterthe anther primordia are distinct in others. 0 5 ovary emergence coincident with anthers, 1 5 ovary emergence after anthers.

27. *Filament and staminode postgenital adhesion. The filaments and staminodes of all species are free upon differentiation from theprimordia. In some species, they remain so throughout all of ontogeny. In other species, the filaments and staminodes bringthemselves into contact in a semicircle and adhere to one another via an unknown substance, presumably a sticky secretion, visiblein sections.

28. *Flora symmetry in ontogeny. All species of Caricaceae have radially symmetrical flowers, and the species studied show thisarrangement from the earliest stages of development. All Moringa species studied exhibit bilateral symmetry very early in ontogeny.Some species have adult floral morphologies that are radially symmetrical or nearly so, whereas others have very clearly bilaterallysymmetrical flowers. 0 5 symmetrical throughout development, 1 5 bilateral to radial, 2 5 bilateral throughout development.

Documents

Combining Data from DNA Sequences and Morphology for a