Changing Ecological Opportunities Facilitated the Explosive ......endemism (ca. 3400 species, 75% endemic; Morat et al. 2012; Munzinger et al. 2016). The singular geological and climatic

HAL Id: hal-02361586https://hal.umontpellier.fr/hal-02361586

Submitted on 13 Nov 2019

HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, estdestinée au dépôt et à la diffusion de documentsscientifiques de niveau recherche, publiés ou non,émanant des établissements d’enseignement et derecherche français ou étrangers, des laboratoirespublics ou privés.

Changing Ecological Opportunities Facilitated theExplosive Diversification of New Caledonian Oxera

(Lamiaceae)Laure Barrabé, Sébastien Lavergne, Giliane Karnadi-Abdelkader, Bryan

Drew, Philippe Birnbaum, Gildas Gateble

To cite this version:Laure Barrabé, Sébastien Lavergne, Giliane Karnadi-Abdelkader, Bryan Drew, Philippe Birnbaum,et al.. Changing Ecological Opportunities Facilitated the Explosive Diversification of New CaledonianOxera (Lamiaceae). Systematic Biology, Oxford University Press (OUP), 2019, 68 (3), pp.460 - 481.�10.1093/sysbio/syy070�. �hal-02361586�

https://hal.umontpellier.fr/hal-02361586https://hal.archives-ouvertes.fr

Copyedited by: TP MANUSCRIPT CATEGORY: Systematic Biology

[16:56 11/4/2019 Sysbio-OP-SYSB180071.tex] Page: 460 460–481

Syst. Biol. 68(3):460–481, 2019© The Author(s) 2018. Published by Oxford University Press on behalf of the Society of Systematic Biologists. This is an Open Access article distributed under the terms of the CreativeCommons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, providedthe original work is properly cited.DOI:10.1093/sysbio/syy070Advance Access publication October 26, 2018

Changing Ecological Opportunities Facilitated the Explosive Diversification ofNew Caledonian Oxera (Lamiaceae)

LAURE BARRABÉ1,2, SÉBASTIEN LAVERGNE3, GILIANE KARNADI-ABDELKADER1, BRYAN T. DREW4,PHILIPPE BIRNBAUM1,5, AND GILDAS GÂTEBLÉ1,∗

1Institut Agronomique néo-Calédonien (IAC), Equipes ARBOREAL and SOLVEG, BP 711,Mont-Dore 98810, New Caledonia; 2Endemia, Plant Red List Authority, 7 rue Pierre Artigue, Nouméa 98800, New Caledonia;

3Laboratoire d’Ecologie Alpine, CNRS - Université Grenoble Alpes, UMR 5553,Grenoble F-38000, France; 4Department of Biology, University of Nebraska-Kearney, Kearney, NE 68849, USA; and 5UMR AMAP,

Université de Montpellier, CIRAD, CNRS, INRA, IRD, Montpellier 34398, France∗Correspondence to be sent to: Institut Agronomique néo-Calédonien (IAC),

Equipe ARBOREAL, BP 711, 98810 Mont-Dore, New Caledonia;E-mail: [email protected]

Received 09 July 2018; reviews returned 04 October 2018; accepted 23 October 2018Associate Editor: Vincent Savolainen

Abstract.—Phylogenies recurrently demonstrate that oceanic island systems have been home to rapid clade diversificationand adaptive radiations. The existence of adaptive radiations posits a central role of natural selection causing ecologicaldivergence and speciation, and some plant radiations have been highlighted as paradigmatic examples of such radiations.However, neutral processes may also drive speciation during clade radiations, with ecological divergence occurringfollowing speciation. Here, we document an exceptionally rapid and unique radiation of Lamiaceae within the NewCaledonian biodiversity hotspot. Specifically, we investigated various biological, ecological, and geographical drivers ofspecies diversification within the genus Oxera. We found that Oxera underwent an initial process of rapid cladogenesislikely triggered by a dramatic period of aridity during the early Pliocene. This early diversification of Oxera was associatedwith an important phase of ecological diversification triggered by significant shifts of pollination syndromes, dispersalmodes, and life forms. Finally, recent diversification of Oxera appears to have been further driven by the interplay ofallopatry and habitat shifts likely related to climatic oscillations. This suggests that Oxera could be regarded as an adaptiveradiation at an early evolutionary stage that has been obscured by more recent joint habitat diversification and neutralgeographical processes. Diversification within Oxera has perhaps been triggered by varied ecological and biological driversacting in a leapfrog pattern, but geographic processes may have been an equally important driver. We suspect that strictlyadaptive radiations may be rare in plants and that most events of rapid clade diversification may have involved a mixtureof geographical and ecological divergence. [Adaptive radiation, allopatry, leapfrog radiation, Lamiaceae, New Caledonia,niche shifts, Oxera.]

Oceanic islands are widely regarded as laborator-ies of evolution mainly owing to their isolated pastbiogeographical history (Losos and Ricklefs 2009). Theirgeographical isolation limits dispersal events from out-side, resulting in ecological niches being filled by speciesdiversification rather than colonization. Consequently,islands are expected to have sheltered many adaptiveradiations, where single lineages have diversified rapidlyinto several distinct niches, resulting in disproportion-ately high phenotypic diversity (Schluter 2000) whencompared with their continental relatives (Silvertownet al. 2005). This phenomenon can be particularlyapparent on large and high-elevation islands that exhibitdiverse and sharp environmental gradients (Whittakeret al. 2008). Although adaptive radiations may notnecessarily result from an acceleration of speciationrates (see Givnish 2015 for a discussion), all adaptiveradiations are spurred by ecological opportunities; thatis, the release of vacant niches due to unused resources,species extinctions or the acquisition of new traits (Losos2010). The existence of adaptive radiations thus posits acentral role of natural selection causing ecological diver-gence and speciation, and some plant radiations havebeen highlighted repeatedly as paradigmatic examplesof this scenario, such as the silversword alliance andlobeliads in Hawaii (Carlquist et al. 2003; Givnish et al.

2009), the Macaronesian Aeonium and Echium (Jorgensenand Olesen 2001), and Veronica (Hebe) in New Zealand(Wagstaff and Garnock-Jones 1998).

However, adaptive radiations sensu stricto appear tobe quite uncommon in the plant kingdom, with theaforementioned well-documented cases probably rep-resenting rare exceptions. In fact, many radiations likelyoccur, at least in part, in a non-adaptive way (Givnish1997, Rundell and Price 2009). Under such a scenario,speciation is not primarily driven by ecological diver-gence, but mainly by neutral divergence in allopatry.Ecological differentiation may occur following non-ecological speciation, through random trait divergence,adaptation to different climatic regimes or habitats, ordue to trait divergence favoring local coexistence ifsecondary sympatry occurs (e.g., Tobias et al. 2014).While molecular phylogenies will often provide littlepower to discriminate between different scenarios, itremains possible to assess the relative importance ofecological differentiation and geographic divergence inrapid species diversifications. Considering that theor-etical evidence suggests that most rapid evolutionaryradiations may have occurred through a mixture ofecological speciation and neutral divergence in allopatry(Aguilée et al. 2012, 2011), we can legitimately wonder

460

Dow

nloaded from https://academ

ic.oup.com/sysbio/article-abstract/68/3/460/5145071 by C

IRAD

Centre de C

ooperation Internationale en Recherche Agronom

ique pour le Developpem

ent user on 12 Novem

ber 2019

http://creativecommons.org/licenses/by/4.0/



2019 BARRABÉ ET AL.—THE RADIATION OF NEW CALEDONIAN OXERA 461

whether purely adaptive radiations have ever occurredin plants.

The New Caledonian biodiversity hotspot is a remotearchipelago in the southwest Pacific, composed of anold-, large-, and high-elevation main island (= GrandeTerre, ca. 37 Ma, ca. 16,600 km2, 1600 m max. elevation)exhibiting exceptional levels of species richness andendemism (ca. 3400 species, 75% endemic; Morat et al.2012; Munzinger et al. 2016). The singular geologicaland climatic New Caledonian history led to the imple-mentation of complex and abrupt environmental gradi-ents, resulting in a mosaic of highly distinct habitats.The archipelago topography is especially complex onGrande Terre, with many relatively steep valleys andhigh summits (Bonvallot 2012) presenting both physicalbarriers to dispersal and strong selection gradients. Thearchipelago also has a suite of diverse bedrock types,in particular volcano-sedimentary, metamorphic andultramafic (metal-rich soils with chemical and physicalproperties that constrain plant growth). These latterbedrock types, which act as strong environmental filtersand exhibit a patchy distribution, have probably playeda key role in plant speciation and flora evolution (Pillonet al. 2010). In addition, Pliocene and Pleistocene climaticfluctuations considerably affected the dynamic of NewCaledonian biotas, leading to the origination of newhabitats such as the unique shrubby sclerophyllousvegetation (i.e., maquis; Jaffré 1980), during periodsof forest contraction (Hope and Pask 1998; Stevenson2004; Stevenson and Hope 2005). This likely triggeredspeciation in palms (Pintaud et al. 2001), and hascontributed to the persistence of old angiosperm lin-eages in forest refugia (Pouteau et al. 2015). All theaforementioned characteristics have contributed to thetremendous plant diversity of New Caledonia (Jaffré1993), and have also driven numerous plant radiations(e.g., Psychotria; Pycnandra; Munzinger et al. 2016).Though plant radiations are increasingly highlighted inthe archipelago through molecular investigations, fewstudies have focused on the relative effect of driversimplied in their in situ diversification (e.g., Barrabé et al.2014; Pillon et al. 2014; Paun et al. 2016).

The old age, geographic isolation, and geologic andtopographic complexity of New Caledonia suggest thatadaptive radiations may be common among many gen-era endemic to New Caledonia. However, few of thesegenera exhibit the joint signatures of morphological andecological divergence expected from adaptive radiations,and rather seem to be relictual lineages (Pillon et al.2017). The woody genus Diospyros (Ebenaceae) was thefirst purported clear case of an adaptive radiation ofa plant group within New Caledonia due to its wideecological diversity (Paun et al. 2016), although theradiation was not related to obvious morphologicaland/or physiological differences. The tree genus Geissois(Cunoniaceae) was described as a cryptic adaptiveradiation, in which species exhibit notable differences inleaf element composition likely linked to the occupationof varied soils (Pillon et al. 2014), but does not fully

satisfy the criteria of an adaptive radiation as outlinedby Givnish (2015), as no physiological adaptation has yetbeen highlighted. It thus seems that New Caledonia hasharbored few, if any, adaptive radiations in a strict sense(sensu Givnish 2015 and Schluter 2000). This has beenexplained as a consequence of the relatively old age of theisland, the prevalence of woody species, the paucity ofpotential pollinator species, and/or a reduced number ofecological opportunities as compared with other islandsystems (Pillon et al. 2017).

In this work, we report on a molecular systematicinvestigation of the genus Oxera (Lamiaceae), whichhas been hypothesized to have diversified in NewCaledonia through adaptive radiation (Pillon et al.2014). The genus is composed of 33 species endemicto New Caledonia (Gâteblé unpublished data, availableat http://endemia.nc/flore/fiche588), and was recentlyenlarged to include five other Australasian and Pacificspecies (Barrabé et al. 2015). It was demonstrated thatthe entire New Caledonian clade originated from asingle colonization and constitutes the tenth largest plantradiation in the archipelago, which is puzzling giventhat Lamiaceae are an under-represented family withinthe flora of New Caledonia (Pillon et al. 2010). Distinctsubclades were partially circumscribed within Oxeraand can arguably be considered as several independentmicro-radiations within the genus (Barrabé et al. 2015).Oxera exhibits strongly divergent morphology in termsof life form, flowers and fruits, and occupies a vastdiversity of distinct habitats. In relation to this remark-able morphological diversity, we hypothesize that shiftsof functional, reproductive (pollination and dispersal),and environmental niches may have played a key roleduring the diversification of Oxera. This role can betwo-fold: 1) niche shifts may have triggered speciesdifferentiation and speciation, and 2) certain niche typesmay have accelerated the rate of genetic divergence andspeciation.

The dazzling morphological and ecological diversityof New Caledonian Oxera could reflect adaptationsto different ecological factors (see Givnish 2010 for adiscussion of life form adaptations). First, the great arrayof life forms observed in Oxera could be associated withdifferent environment preferences, as observed in thesilversword alliance (Carpenter et al. 2003), and thismay have created conditions that facilitated parapatricisolation and speciation. Robust lianas mainly growin closed rainforests with leaves usually reaching thecanopy, while slender lianas are mostly encounteredin open sclerophyllous vegetation and forest edges;monocaulous trees (i.e., single-stemmed) are mainlyconfined to rainforest understories; and finally, shrubspecies are often able to establish in both closed andmore open vegetation (Gâteblé, pers. obs.; see Fig. 1a andb). Such contrasting life forms clearly reflect functionalstrategies regarding growth under different light condi-tions (Givnish 1995; Santiago and Wright 2007; Selayaand Anten 2008).

Dow



IRAD

Centre de C




ber 2019

http://endemia.nc/flore/fiche588



462 SYSTEMATIC BIOLOGY VOL. 68

FIGURE 1. Biological and ecological data sets of Oxera. a) Illustration of habitat preferences and biological syndromes, from top to bottom:habitat preferences (UM = ultramafic soils; NUM = non-ultramafic soils), pollination niche, and dispersal niche. b) Illustration of life forms.c) Niche syndromes inferred from the Hill and Smith principal component method, according to the first two axes of multivariate analyses.

Second, most species of Oxera are confined to narrowenvironmental conditions within New Caledonia. Theyoccur either in rainforests, dry forests or maquis, or occurat different elevations and/or on different bedrock types(limestone, ultramafic, and volcano-sedimentary rocks).Although some species of Oxera are clearly specialized

toward particular habitats or have very limited geo-graphic ranges in narrow valleys or on remote summits,others are encountered throughout the archipelago dueto wider environmental tolerances. These contrastinghabitat and geographic patterns may be hypothesizedas the signature of parapatric or allopatric speciation

Dow



IRAD

Centre de C




ber 2019




processes, and provide an interesting basis for testingwhether habitat or geography (or both) have been majordrivers of clade diversification.

Third, the floral diversity of Oxera was previouslylinked to distinct flower pollinators (de Kok 1998) basedupon field observations of animal visits (Table 1).

These pollinator preferences could potentially bedrivers of evolution of reproductive isolation withinOxera. Honey-eaters (Meliphagidae) and white-eyes(Zosteropidae) have been observed visiting Oxera spe-cies displaying showy and curved flowers, profuselyproducing nectar, and bearing arched anthers typicallyextending out of the corolla (Table 1). Long-tubular whiteand/or green bilabiate flowers, sometimes releasing astrong fragrance, have been observed to be visited bybutterflies and moths (Pieridae and Sphingidae; Table 1;de Kok 1997; Haxaire and Salesne 2016). Despite the lackof recorded visitor observation, the flower attributes ofthe Australasian O. splendida, which bears large, flared,white flowers that open nocturnally and release a strongfragrance, are very similar to that of Fagraea, renownedfor its bat-pollination (Momose et al. 1998).

Fourth, the large variety of fruit morphology exhibitedby Oxera suggests different dispersal agents, althoughthese mechanisms have not been fully identified in thefield (Table 1). We assume that the different behaviorsand flight abilities of animal dispersers would haveaffected the spatial scale of gene flow, and thus produceddifferent rates of genetic divergence between Oxeralineages, due to varied dispersal abilities and habitatselection of dispersers (Givnish 2010, Theim et al. 2014).The largest white fruits encountered in Australasiaare consumed by bats and cassowaries (Richards 1990;Hutson et al. 1992; Cooper and Cooper 2004). Otherlarge colored fruits (ca. >2 cm) could be exclusivelyeaten by imperial pigeons (Ducula, Columbidae), whichis the only bird able to swallow and effectively dispersesuch large fruits (Meehan et al. 2002; Carpenter et al.2003; Cibois et al. 2017; P. Bachy pers. comm. forO. palmatinervia). It is noteworthy that the sedentarybehavior of pigeons confines them mostly to forestunderstories (Colombo 2008; Wotton and Kelly 2012).Smaller fruits (ca.



464 SYSTEMATIC BIOLOGY VOL. 68T

ABL

E1.

Insi

tudi

rect

obse

rvat

ions

offlo

wer

visi

tors

and

frui

tcon

sum

ers

ofO

xera

Flow

ers

Frui

ts

Subc

lade

Spec

ies

Flow

ervi

sito

r

Flow

erty

pe(m

ulti

vari

ate

anal

yses

)Fr

uit

cons

umer

Frui

ttyp

e(m

ul-

tiva

riat

ean

alys

es)

Max

imum

mer

i-ca

rple

ngth

Ref

eren

ces

Aus

tral

asia

nO

xera

Oxe

rasp

lend

ida

Act

inom

orph

icC

asso

war

ies,

flyin

gfo

xes,

rats

Pota

to62

.5R

icha

rds

(199

0),H

utso

net

al.

(199

2),C

oope

ran

dC

oope

r(2

004)

Paci

ficO

xera

Oxe

raam

icor

umH

oney

eate

r(M

yzom

ela

card

inal

is,G

ymno

myz

avi

ridi

s)

Act

inom

orph

icC

olum

bida

e(D

ucul

apa

cific

a,D

idun

culu

sst

rigi

rost

ris)

,fly

ing

foxe

s

Rou

nded

41.7

Belc

her

and

Sibs

on(1

982)

,de

Kok

(199

7),B

eich

le(1

987)

,Tra

il(1

994)

,Pill

on(p

ers.

obs.

)

Paci

ficO

xera

Oxe

rale

hunt

eiH

oney

eate

rA

ctin

omor

phic

Rou

nded

30.5

deK

ok(1

997)

“bal

adic

a”su

bcla

deO

xera

aff.

nuda

Padd

le40

.6“b

alad

ica”

subc

lade

Oxe

raba

ladi

caH

oney

eate

r(G

lyci

fohi

aun

dula

ta),

Zos

tero

pida

e(Z

oste

rops

sp.)

Gul

let

Padd

le26

.3de

Kok

(199

7),F

leur

ot(p

ers.

obs.

)“b

alad

ica”

subc

lade

Oxe

rado

ubet

iae

sp.n

ov.in

ed.

Thor

ny40

.6

“bal

adic

a”su

bcla

deO

xera

garo

ensi

ssp

.nov

.ined

.“b

alad

ica”

subc

lade

Oxe

raou

nem

oae

sp.n

ov.in

ed.

Thor

ny30

.6

“bal

adic

a”su

bcla

deO

xera

papi

neau

isp

.nov

.ined

.Pa

ddle

25.9

“bal

adic

a”su

bcla

deO

xera

sess

ilifo

liaTh

orny

40.4

“bal

adic

a”su

bcla

deO

xera

vanu

atue

nsis

Bird

sG

ulle

tPa

ddle

35.0

deK

ok(1

997)

“ner

iifol

ia”

subc

lade

Oxe

rane

riifo

liaPi

erid

ae(A

naph

aeis

java

,E

urem

ahe

cabe

),Sp

hing

idae

(Gna

thot

hlib

user

otus

,Hip

poti

onsc

rofa

,H

ippo

tion

cele

rio,

Psi

logr

amm

alif

uens

is,

Com

psul

yxco

cher

eaui

,etc

.)

Bila

biat

eR

ound

ed14

.8de

Kok

(199

7),G

âteb

léan

dO

urie

t(pe

rs.o

bs.)

“ner

iifol

ia”

subc

lade

Oxe

raov

ata

Pier

idae

(Ana

phae

isja

va),

Sphi

ngid

ae(G

nath

othl

ibus

sale

snei

)

Bila

biat

eR

ound

ed15

.5H

axai

rean

dSa

lesn

e(2

016)

,de

Kok

574

(in

herb

.)“n

eriif

olia

”su

bcla

deO

xera

soro

ria

Rou

nded

13.5

“obl

ongi

folia

”su

bcla

deO

xera

cras

sifo

liaR

ound

ed18

.1

“obl

ongi

folia

”su

bcla

deO

xera

glan

dulo

saR

ound

ed11

.4

“obl

ongi

folia

”su

bcla

deO

xera

oblo

ngifo

liaBi

labi

ate

Stur

nida

e(A

plon

isst

riat

us)

Rou

nded

14.6

deK

ok56

8(i

nhe

rb.)

“obl

ongi

folia

”su

bcla

deO

xera

oreo

phila

Lepi

dopt

era

Bila

biat

eR

ound

ed11

.9de

Kok

(199

7)O

xera

mor

iere

ilin

eage

Oxe

ram

orie

rei

Hon

eyea

ter

(Gly

cifo

hia

undu

lata

)G

ulle

tR

ound

ed16

.2de

Kok

(199

7)

(Con

tinu

ed)

Dow



IRAD

Centre de C




ber 2019



2019 BARRABÉ ET AL.—THE RADIATION OF NEW CALEDONIAN OXERA 465T

ABL

E1.

(Con

tinu

ed)

Flow

ers

Frui

ts

Subc

lade

Spec

ies

Flow

ervi

sito

rFl

ower

type

(mul

tiva

riat

ean

alys

es)

Frui

tco

nsum

erFr

uitt

ype

(mul

tiva

ri-

ate

anal

yses

)

Max

imum

mer

i-ca

rple

ngth

Ref

eren

ces

“pul

chel

la”

subc

lade

Oxe

raaf

f.ba

lans

aeR

ound

ed12

.2

“pul

chel

la”

subc

lade

Oxe

raba

lans

aeBi

rds

Gul

let

Stur

nida

e(A

plon

isst

riat

us)

Rou

nded

8.4

deK

ok(1

997)

,Des

mou

lins

(per

s.ob

s.)

“pul

chel

la”

subc

lade

Oxe

rabr

evic

alyx

Hon

eyea

ter

(Gly

cifo

hia

undu

lata

)G

ulle

tSt

urni

dae

(Apl

onis

stri

atus

)

Rou

nded

7.9

deK

ok(1

997)

,Des

mou

lins

(per

s.ob

s.)

“pul

chel

la”

subc

lade

Oxe

rapu

lche

llaH

oney

eate

r(M

yzol

ema

cale

doni

ca),

Zos

tero

pida

e(Z

oste

rops

sp.)

Gul

let

Stur

nida

e(A

plon

isst

riat

us)

Rou

nded

11.3

Des

mou

lins

(per

s.ob

s.),

Fleu

rot

(per

s.ob

s.)

“rob

usta

”su

bcla

deO

xera

cori

acea

Hon

eyea

ter

(Gly

cifo

hia

undu

lata

)G

ulle

tR

ound

ed40

.8de

Kok

(199

7)“r

obus

ta”

subc

lade

Oxe

raco

rona

taR

ound

ed29

.9“r

obus

ta”

subc

lade

Oxe

ralo

ngifo

liaH

oney

eate

r(L

ichm

era

inca

na)

Gul

let

Rou

nded

37.2

deK

ok(1

997)

“rob

usta

”su

bcla

deO

xera

palm

atin

ervi

aH

oney

eate

r(G

lyci

fohi

aun

dula

ta,P

hile

mon

diem

enen

sis)

Gul

let

Col

umbi

dae

(Dre

pano

ptila

holo

seri

cea,

Duc

ula

Gol

iath

)

Rou

nded

34.0

deK

ok(1

997)

,Bac

hy(p

ers.

obs.

),D

esm

oulin

s(p

ers.

obs.

)

“rob

usta

”su

bcla

deO

xera

robu

sta

Rou

nded

35.7

“sub

vert

icill

ata”

subc

lade

Oxe

raau

reoc

alyx

sp.n

ov.in

ed.

Padd

le40

.7

“sub

vert

icill

ata”

subc

lade

Oxe

ram

eryt

ifolia

Padd

le32

.6

“sub

vert

icill

ata”

subc

lade

Oxe

rasu

bver

tici

llata

Hon

eyea

ter

(Phi

lem

ondi

emen

ensi

s)G

ulle

tPa

ddle

30.9

deK

ok(1

997)

“sub

vert

icill

ata”

subc

lade

Oxe

rati

wae

ana

sp.n

ov.in

ed.

Padd

le31

.9

“sul

fure

a”su

bcla

deO

xera

gmel

inoi

des

Rou

nded

13.8

“sul

fure

a”su

bcla

deO

xera

mic

roca

lyx

Hon

eyea

ter

(Gly

cifo

hia

undu

lata

),Z

oste

ropi

dae

(Zos

tero

psxa

ntho

chro

a)

Gul

let

Rou

nded

13.5

deK

ok(1

997)

“sul

fure

a”su

bcla

deO

xera

ouam

enie

nsis

sp.n

ov.in

ed.

Rou

nded

12.7

“sul

fure

a”su

bcla

deO

xera

panc

heri

Rou

nded

13.5

“sul

fure

a”su

bcla

deO

xera

rugo

saR

ound

ed18

.4“s

ulfu

rea”

subc

lade

Oxe

rasu

lfure

aH

oney

eate

r(L

ichm

era

inca

na,M

yzom

ela

cale

doni

ca,G

lyci

fohi

aun

dula

ta)

Gul

let

Stur

nida

e(A

plon

isst

riat

us)

Rou

nded

11.7

deK

ok(1

997)

,Des

mou

lins

(per

s.ob

s.),

Gât

eblé

(per

s.ob

s.)

Dow



IRAD

Centre de C




ber 2019




protocols described in Barrabé et al. (2015). All accessionsused in this study are listed in the SupplementaryAppendix S1a available on Dryad.

For divergence time estimation, we also used two addi-tional alignments composed of four DNA plastid regions(matK, rps16, trnL-F, and trnL-trnF; SupplementaryAppendix S2a available on Dryad) from two datasets, with taxon sampling that spanned the Lamialesand Lamiaceae. The Lamiales data set included: 1)a subset of the Oxera data set (i.e., five species;Supplementary Appendix S1c available on Dryad);and 2) 177 representatives of various lineages withinLamiales, whose sequences were downloaded fromGenBank (Supplementary Appendix S1c available onDryad). The Lamiaceae data set was thus composedof 1) the aforementioned Oxera subset and 2) 59 taxarepresenting all major Lamiaceae lineages (downloadedlikewise from GenBank; Supplementary Appendix S1bavailable on Dryad). This large-scale sampling allowedus to incorporate Lamiales and Lamiaceae fossils ascalibration points (see below).

All DNA alignment matrices used in this study areavailable on Dryad.

Phylogenetic InferenceFor the three molecular data sets, phylogenetic infer-

ences were conducted using Bayesian Markov ChainMonte Carlo (MCMC) as implemented in MrBayesv3.2.1 (Ronquist et al. 2012) based on single-locus andconcatenated data sets. Best-fit nucleotide sequence evol-ution models were identified using jModelTest (Posada2008) based on the Akaike criterion (more details onsettings are provided in Supplementary Appendix S2aavailable on Dryad). The combined data sets werepartitioned to allow each locus to possess specific modelparameters (Nylander et al. 2004). The methodologicalapproach of the Bayesian MCMC analyses followedthat described in Barrabé et al. (2015) and the detailsof parameter settings are provided in SupplementaryAppendix S2b available on Dryad. All analyses were runon the Cyber Infrastructure for Phylogenetic Researchcluster (CIPRES; http://www.phylo.org/; last accessed18 October 2017). The post-burn-in trees resulting fromthe MCMC stationary phase were used to constructa majority-rule consensus tree and calculate Bayesianposterior probabilities (PPs); clades were consideredwell-supported when PP values were ≥0.95). The threeresulting consensus trees are available in TreeBASE (ID23353) and Dryad.

In order to check for phylogenetic incongruencebetween nuclear and plastid loci, we also ran twophylogenetic reconstructions separately for all six con-catenated nuclear regions and all six concatenatedplastid regions, using the same MrBayes parametersettings and models of nucleotide evolution as above.These two analyses were only applied to the Oxera dataset. We then assessed whether there was any supported(PP ≥0.95) incongruence between two trees.

Fossil Dating and Divergence Time EstimationWe performed a two-step approach to estimate the

temporal evolution of Oxera. This allowed us to test theimpact of different dating calibrations and taxonomicsampling on divergence time estimates. In the first step,we performed a fossil calibration on the Lamiales andLamiaceae data sets using fossils described in a recentLamiaceae molecular dating study (Yao et al. 2016).We incorporated three Lamiaceae (Melissa, Ocimum,and Stachys) and three Lamiales fossils (Bignoniaceae,Acanthaceae, and Oleaceae) for dating the phylogenyinferred with the Lamiales data set, and two Lamiaceaefossils (Melissa and Ocimum) for dating the phylogenyinferred with the Lamiaceae data set (for further detailson the design and parameters of these calibrations, seeSupplementary Appendix S2c, available on Dryad andexplanations provided inYao et al. 2016). For root calib-rations we used divergence times estimated in Magallónet al. (2015): 1) the divergence between Lamiales andGentianales—Solanales for the Lamiales data set and2) the divergence between Plocospermataceae and otherLamiales for the Lamiaceae data set. In the second step,the divergence time estimate between Rotheca (sp. Wen9487) and other Lamiaceae recovered during the first step(i.e., the dating of the Lamiaceae data set) was used as asecondary root calibration point (see Results section) forthe Oxera data set.

For the three molecular data sets, molecular diver-gence times were estimated using the Bayesian MCMCapproach implemented in BEAST v.1.8.0 (Drummondand Rambaut 2007). DNA regions were combined andpartitions were set as in the above Bayesian MCMCanalyses (Supplementary Appendix S2c available onDryad). An uncorrelated relaxed molecular clock modelwas selected following a lognormal distribution, andthe Birth–Death process implemented for the tree prior.Other details on analyses, calibration and parametersettings are provided in Supplementary Appendix S2cavailable on Dryad. The post burn-in trees were summar-ized, and Bayesian PPs, median height (= age estimate),and the 95% highest posterior density heights interval ofeach node (95% HPD) assessed, using a Maximum CladeCredibility target tree (named as MCCT tree for the Oxeradata set) in Treeannotator v.1.8.0 (Drummond and Ram-baut 2007). For the Oxera data set a subset of 100 datedtrees (RD trees hereafter) were also randomly sampledfrom the stationary phase to integrate phylogeneticuncertainty in some of the following analyses (seebelow). All divergence time analyses were conductedusing CIPRES. The three resulting Maximum CladeCredibility trees are available in TreeBASE (ID 23353)and Dryad.

Comparative Analyses of Species DiversificationTo obtain an overall view of the Oxera diversification,

we first generated a lineage through time diagram,based on a pruned version of the MCCT tree includingonly Oxera species and a single sample per species

Dow



IRAD

Centre de C




ber 2019

https://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttp://www.phylo.org/https://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-data




(pMCCT tree), and also on the RD trees (pruned as inthe pMCCT tree). We also calculated net diversificationrates, following the conservative approaches implemen-ted in Pillon (2012) and Barrabé et al. (2014), to allowcomparisons with other oceanic insular plant radiations,using median crown ages and their 95% HPD estimatedfrom the molecular dating, and two extreme values forrelative extinction (null and equal to 0.9). Net diversific-ation rates were likewise estimated per unit of area andlog(area).

To assess more precisely the rate and tempo of speciesdiversification in Oxera through time, we carried outdiversification analyses with maximum likelihood (ML)and Bayesian modeling approaches. The ML analyseswere conducted in the TreePar R package (Stadler 2011),using an optimization algorithm to infer past rates of spe-ciation and extinction and their temporal variation. Wefitted various diversification models: a pure birth model,a constant birth–death model, several more complexmodels allowing one to several rate shifts through time(up to 10 shifts), and finally several diversity-dependentmodels. All models were fitted on the pMCCT tree (seeSupplementary Appendix S2d, available on Dryad forfurther details on parameter settings) and on the RDtrees to account for phylogenetic uncertainty. The best-fitdiversification model was identified using the correctedAkaike criterion, and a likelihood-ratio test was used forcomparing nested pure birth, birth–death, and rate shiftmodels.

The Bayesian diversification analyses were computedusing BAMM (Rabosky et al. 2014). This allows mod-elling complex dynamics of speciation and extinctionon phylogenetic trees using a reversible jump MCMCalgorithm. We pruned the MCCT tree in the samemanner as the pMCCT tree, but retained species belong-ing to the Clerodendrum clade to assess early variationin Oxera diversification rates. The sampling fractionwas completed with an estimate of species richness,especially for the Clerodendrum clade (retrieved from theKew Checklist website: http://wcsp.science.kew.org;last accessed 18 October 2017). As we expected two rateshifts to have taken place during the evolution of Oxeraand its relatives (considering the large sizes of the generaClerodendrum and Oxera as compared with the othersmall genera), the prior on the number of diversificationshifts was set to 2. All other priors were estimated inthe BAMMtools R package (Rabosky et al. 2014), withthe default settings used for all other parameters. TheMCMC was run for 50×106 generations and sampledevery 10,000 generations. The analysis was conductedwith three independent Markov chains. The first 10%generations of each run were discarded after checking forchain convergence and adequate MCMC sampling. Theremaining generations were summarized to generate the95% credibility interval for rate shift configurations, themarginal shift probability tree (where branch lengthsare proportional to the probability that a shift occurredon a given branch), the best rate shift configurationwith the highest maximum a posteriori probability,

the phylorate plot (showing variation of mean diver-sification rates with a colored gradient) and finallythe rates-through time plots (only built for the Oxerafocal group).

Biological, Ecological, and Geographic Range Data setsTo identify and disentangle suspected drivers of

species diversification in Oxera, we built a data setby compiling data on five major life traits: 1) flowermorphology, 2) fruit morphology, 3) life forms, 4)geographical occurrences, and 5) environmental pref-erences. The flower and fruit morphology (i.e., organdimensions, colors, shapes, and textures) were com-piled from herbarium and fieldwork observations onorgan dimensions, colors, shapes, and textures. Thosetraits were coded as 13 discrete and three continuouscharacters for flower morphology, and three discreteand three continuous traits for fruits (SupplementaryAppendix S3a, d, and e available on Dryad). The lifeform data set was composed of a single discrete traitcorresponding to the four life forms usually ascribed toOxera, namely monocaulous, robust liana, slender liana,and shrub/small tree (Supplementary Appendix S3aand b available on Dryad). Note that no functionalor physiological traits related to light requirements ornutrient use strategy were available for Oxera.

The geographical distributions of New Caledonianspecies were retrieved from the databases of the herb-aria of Nouméa (NOU, VIROT), the Museum nationald’Histoire naturelle of Paris (P, SONNERAT), the Uni-versity of Zurich (Z), and additional field observations.The geographical coordinates of herbarium specimenrecords were then databased, error-corrected for dis-tribution, and finally incorporated into a GeographicalInformation System (GIS) under QGIS 2.8.1. For Aus-tralasian and Pacific species distributions were assessedfrom de Kok and Mabberley (1999b) and manually addedinto the same GIS layer.

The environmental data set was composed of threevariables that best explain plant distributions acrossNew Caledonia (Jaffré 1993; Morat 1993; SupplementaryAppendix S3a and c available on Dryad): geologicalsubstrates, vegetation types, and elevation. Species’ rocktypes were coded as occurring on ultramafic rocks,on volcano-sedimentary and metamorphic rocks, orubiquists. These geological attributes were determinedby coupling geographical occurrences with a layerof geological substrate provided by the Direction del’Industrie, des Mines et de l’Energie de la Nouvelle-Calédonie (New Caledonia). Species’ vegetation typeswere coded as occurring in wet forests, in sclerophyl-lous vegetation types (i.e., dry forests and/or maquisvegetation), and ubiquists, based on herbarium andfield observations (Gâteblé, pers. obs.). Species’ elevationranges were extracted from the GIS layer generated bya digital elevation model (resolution of 10 m) providedby the Direction des Infrastructures de la Topographieet des Transports Terrestres (New Caledonia), and

Dow



IRAD

Centre de C




ber 2019

https://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttp://wcsp.science.kew.orghttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-data




then averaged across geographical occurrences of eachspecies.

To identify ecological syndromes among Oxera spe-cies, we performed multivariate analyses on the envir-onmental, flower, and fruit data sets. As we used bothcontinuous and discrete traits, we conducted the PCAanalyses by using the Hill and Smith (1976) principalcomponent method under the ade4 R package (Chesselet al. 2004). This analysis allowed clustering speciesamong pollinator syndromes, dispersal syndromes, andhabitat preferences, which we subsequently treated asnew discrete traits according to the species clustering(i.e., the niche data sets; Supplementary Appendix S3favailable on Dryad). We posit that the ecological syn-drome of a species reflects the ecological niche it occupies(Johnson 2010, Pauw 2013). No data on species growthand architecture were available, we therefore, couldnot evaluate life forms in a similar manner; this traitwas treated as a discrete variable without any step ofordination analysis.

Comparative Analyses of Niche EvolutionTo assess the mode and tempo of evolution of each

niche during the Oxera radiation (i.e., of pollinationand dispersal syndromes or habitat preferences), wemeasured their respective amount of phylogenetic signalby estimating � and � Pagel statistics (Pagel 1999). Valuesof � and � close to 1 indicate a strong phylogeneticsignal and gradual trait evolution (i.e., according to amodel of Brownian motion), and deviation from thisexpectation provides insight into temporal patterns oftrait evolution. Values of � close to 0 depicts punctualevolution, where trait divergence occurs independentlyfrom branch lengths in the phylogeny. Values of � closeto 0 reveal a low phylogenetic signal, where most traitchange occurred late in evolutionary history and closelyrelated species thus share very little trait similarity.For the habitat, pollination, and dispersal niches � and� statistics were estimated using the “fitContinuous”and “phylosig” functions of the geiger and phytools Rpackages, respectively (Harmon et al. 2008; Revell 2012).Estimations were performed on the RD trees, and byaveraging estimated values for each of the first n axesof each PCA explaining 90% of the cumulative variance.For life form, each statistic was simply averaged throughthe RD trees using the “fitdiscrete” function (geiger) andthe single discrete trait.

To reconstruct the evolutionary history of habitatpreference, pollination, dispersal niches, and life form,we conducted ancestral state reconstruction analyses.We first assessed ancestral niche states using the “ace”function in the ape R package (Paradis et al. 2004) appliedto the pMCCT tree, and the niche and life form data sets.We fitted three ML evolution models: the “ER” (withequal state transition rates), “SYM” (symmetrical), and“ARD” (with all rates unequal) models, identifying thebest-fitting model using comparisons performed on thecorrected Akaike criterion, and likelihood ratio test. We

then estimated their respective absolute shift numberand timing by performing stochastic character mappings(Huelsenbeck et al. 2003), using the “make.simmap”function in phytools. We launched 500 simulations usingan estimated prior distribution on the root node, thebest-fit ML model recovered in the previous analysesand applied first to the pMCCT tree, and then to theRD trees to obtain most likely envelopes of past shiftnumber for habitat preference, pollination, dispersalniches and life-forms. For 11 of the 100 RD trees, weencountered intractable optimization issues when usingthe “make.simmap” function. We subsequently removedthe 11 trees from these analyses and used only 89 RDtrees. Absolute shift timings were extracted using acustom R-script from all simulations and summarized byplotting the median number of state changes in differenttime bins of a 0.5 Myr time-grid.

To investigate whether a particular niche state (pol-lination, dispersal, or habitat) or life-form would haveaffected clade diversification rates of Oxera, we con-ducted multi-state trait-based analyses that estimatedsimultaneously trait-dependent speciation, extinction,and transition rates on a phylogeny and characterdistribution (Fitzjohn et al. 2009). These were performedusing the “make.musse” function under the diversitreeR package (Fitzjohn et al. 2009), applied to the pMCCTtree, and the niche and life form data sets. We first fitted17 models from a null model (rates independent of traitstates) to a full model (all rate components dependentof trait states; Supplementary Appendix S4 available onDryad), conserving the best-fitting using the correctedAkaike criterion. We performed subsequent Bayesiananalyses through a MCMC, run for 10,000 generations,seeding them with rate estimates recovered from theprevious ML analyses as priors, and discarding thefirst 1000 generations as burn-in. PP distributions of allparameters were summarized using the diversitree R-package. We also fitted the best model and compared itto the second best model across all RD trees to check thatmodel selection was robust to phylogenetic uncertainty.

Geographical Evolution versus Habitat EvolutionTo investigate the relative roles of geographical and

environmental divergence during the diversification ofOxera, we computed age-range correlations (Fitzpatrickand Turelli 2006; Warren et al. 2008) where metricsof range overlaps, range asymmetries, and habitatdistances computed between species pairs are comparedwith their estimated time of divergence (from thephylogeny). These metrics have been recurrently usedto diagnose different speciation and post-speciationscenarios (Anacker and Strauss 2014; Grossenbacheret al. 2014).

Species geographical ranges were retrieved from thegeographical data set by creating for each species asingle convex polygon enclosing its locations using the“convex hull” tool under QGIS. For narrow endemicspecies, with one or two occurrences, we applied a

Dow



IRAD

Centre de C




ber 2019

https://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-data




buffer of 0.5 km around each location, and used arange of 3.141592 × 0.52 km2 for the former, and of6.283184 × 0.52 km2 for the latter. From these distributiondata, range overlaps, range asymmetries, and habitatdistances metrics were computed between all possiblepairs of species. Range overlap was defined as the areaoccupied by two species divided by the area of themore narrowly ranging species, ranging from 0 (noco-occurrence) to 1 (complete co-occurrence). Rangeasymmetry was calculated as the area of the wider-ranging species divided by that of the smaller (rangingfrom 1 to infinity). The habitat similarity between twospecies was assessed by calculating their Euclidiandistance (habitat distance) based on the eigenvectorvalues of the first n axes explaining 90% of the cumulatevariance retrieved from the environmental PCA (seeabove). These geographical variables (areas of all Oxeraspecies, range asymmetries and range overlaps of allOxera species pairs) are available on Dryad.

Age-range correlation analyses were performed usingan extended version of the “age.range.correlation” func-tion of the phyloclim R package (Heibl and Calenge2013; https://github.com/danlwarren/arc-extensions/blob/master/age.range.correlation.2.R; last accessed 18October 2017), applied to our pMCCT and the RDtrees. We launched 1000 iterations for the Monte Carloresampling procedure to create the null hypothesis ofno relationship between phylogenetic relatedness andrange overlap, range asymmetry, or habitat distance, andtested it. We calculated linear regressions, whose slopesand intercepts indicate speciation mode (allopatric vs.sympatric) or habitat evolution (conservatism vs. diver-gence) through time. For each metric a “super-p-value”was calculated, corresponding to the tree proportionacross the RD trees for which the linear regression wassignificant. We also assessed the frequency of rangeoverlaps across all random species pairs and across allsister species pairs using a kernel density plot.

Finally, we estimated the respective importance of geo-graphical versus habitat divergence and whether localco-existence is possible between closely related Oxeraspecies. To do so, we superimposed species geographicaloverlaps to their habitat preferences (summarized intodiagrams combining information on their geological,elevation, and vegetation attributes) within each Oxerasubclade (as delimited in Barrabé et al. 2015) to determinewhether closely related species could possibly occur insympatry.

RESULTS

Phylogenetic InferenceThe Bayesian MCMC analyses of the three species

sampling (Lamiales, Lamiaceae, and Oxera data sets)recovered the same following robust phylogenetic rela-tionships (Supplementary Appendices S5–S7 availableon Dryad). The sister genera Hosea and Oxera formed awell-supported clade that was sister to the Clerodendrum

clade (PP =1). Analyses based on the Oxera data setprovided more resolution and detail on the internalOxera relationships, which are congruent with the onesestablished in Barrabé et al. (2015) (SupplementaryAppendix S5 available on Dryad). The Australasian O.splendida was placed as sister to the New Caledonianradiation (PP =0.72), which included all New Cale-donian and nested Pacific subclades. We retrievedseven well-supported New Caledonian subclades (PP=1; Fig. 2a), namely the oblongifolia, neriifolia, pulchella,baladica, subverticillata, robusta, and sulfurea subclades.

A few supported incongruences (PP ≥0.95) wereobserved between the phylogenetic trees reconstructedseparately from the six nuclear loci and from the sixplastid loci (Supplementary Appendix S8 available onDryad), but these incongruencies were all located nearthe tips (i.e., between close species). This discordancedid not affect the robustness of the seven New Cale-donian subclades previously recovered. These minorincongruences underscore the importance of accountingfor phylogenetic uncertainty by using the 100 RD treesin most of our evolution and geographical analyses (seeMaterial and Methods section).

Molecular DatingThe three molecular BEAST dating analyses provided

congruent divergence times for Oxera and relatives(Fig. 2a, Supplementary Appendices S9–S11 availableon Dryad), which can thus be considered as a reliableappraisal of colonization dates and tempo of speciesdiversification. Oxera likely emerged during the lateMiocene, and the New Caledonian radiation sensu strictois estimated to have initiated during the early Pliocene.The median divergence time estimates for the stem nodeof Oxera were 11.46 Ma (95% HPD = 5.65–19.68), 11.2Ma (95% HPD = 5.61–19.06), and 12.02 Ma (95% HPD =6.21–19.16), as recovered with the Lamiales, Lamiaceae,and Oxera data sets, respectively (Fig. 2a, SupplementaryAppendices S9–S11 available on Dryad). For the crownnode, age estimates were 6.34 Ma (95% HPD = 2.89–11.48), 6.28 Ma (95% HPD = 3.07–10.73), and 5.39 Ma(95% HPD = 2.97–8.42), respectively. The Oxera data setprovided an estimate of 4.54 Ma (95% HPD = 2.7–6.92)for the crown node of the New Caledonian radiation(Fig. 2a).

Oxera DiversificationThe lineage through time plot highlighted a gradual

and exponential increase of the number of Oxera lineagesthrough time (Fig. 2a). The net conservative diversifica-tion rates are summarized in Tables 2 and 3. For the NewCaledonian radiation, these rates were 0.2 species/Myrfor an extinction of 0.9 and 1.04 species/Myr for anull extinction. With respect to per-unit-area and per-unit-log(area) they were estimated between 1.1×10−5and 5.6×10−5 species/Myr/km2, and between 0.02 and

Dow



IRAD

Centre de C




ber 2019

https://github.com/danlwarren/arc-extensions/blob/master/age.range.correlation.2.Rhttps://github.com/danlwarren/arc-extensions/blob/master/age.range.correlation.2.Rhttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-data




FIGURE 2. Evolutionary history of Oxera in New Caledonia, and of its habitat preferences, pollination and dispersal niches, and life forms. a)BEAST chronogram of the Oxera data set with superimposed lineage through time plot. Median age estimates and blue node bars (correspondingto the 95% highest posterior densities) are indicated for each node of interest. Bayesian PP from the BEAST analyses (on the right) and MCMCBayesian analyses (on the left) are indicated with an asterisk for each node of interest when >0.95, and with a hyphen when



2019 BARRABÉ ET AL.—THE RADIATION OF NEW CALEDONIAN OXERA 471T

ABL

E2.

Com

pari

son

ofne

tdiv

ersi

ficat

ion

rate

sof

Oxe

raw

ith

othe

rN

ewC

aled

onia

npl

antl

inea

ges/

radi

atio

ns

Fam

ilyR

adia

tion

Stem

age

(Ma)

Cro

wn

age

(Ma)

Num

ber

ofN

ewC

aled

onia

nsp

ecie

sre

trie

ved

from

Mun

zing

eret

al.(

2016

)

Spec

ies

num

ber

rest

rict

edto sc

lero

phyl

lve

geta

-ti

ons

Spec

ies

num

ber

able

togr

owth

insc

lero

phyl

lve

geta

-ti

ons

Prop

orti

onof

spec

ies

stri

ctly

grow

ing

insc

lero

phyl

lve

geta

-ti

ons

(%)

Prop

orti

onof

spec

ies

able

togr

owin

scle

roph

yll

vege

ta-

tion

s

Net

dive

rsifi

cati

onra

te(b

ased

onth

eeq

uati

onof

Mag

alló

nan

dSa

nder

son

(200

1))

for

anex

tinc

tion

null,

and

usin

gcr

own

ages

(in

spec

ies/

Myr

)

Wor

ksw

here

age

esti

mat

esha

vebe

enpu

blis

hed

Sapo

tace

aeP

lanc

hone

llacl

ade

III

1815

31

233

.366

.70.

03Sw

enso

net

al.(

2013

)

Prot

eace

aeB

eaup

rea

clad

eI

(“fil

ipes

”cl

ade)

78.2

28.3

71

514

.371

.40.

04H

eet

al.(

2016

)

Not

hofa

gace

aeN

otho

fagu

s26

.116

.45

00

00

0.06

Sauq

uete

tal.

(201

2)Sa

pota

ceae

Pla

ncho

nella

clad

eI

15.3

12.0

54

03

075

0.06

Swen

son

etal

.(20

13)

Prot

eace

aeK

erm

adec

ia-

Sleu

mer

oden

dron

13.6

12.3

50

00

0.07

Sauq

uete

tal.

(200

9)

Myr

tace

aeM

etro

side

ros

clad

eB

21.2

511

51

320

600.

08Pa

pado

pulo

set

al.(

2011

)

Myr

tace

aeM

etro

side

ros

clad

eA

16.2

513

.67

47

57.1

100

0.09

Papa

dopu

los

etal

.(20

11)

Are

cace

aeC

linos

perm

a—C

ypho

kent

ia21

.510

.56

00

00

0.1

Bake

ran

dC

ouvr

eur

(201

2)

Myr

tace

aeM

etro

side

ros

clad

eC

29.1

8.25

52

440

800.

11Pa

pado

pulo

set

al.(

2011

)

Are

cace

aeC

ham

beyr

onia

—K

enti

opsi

s9

4.5

60

00

00.

12Ba

ker

and

Cou

vreu

r(2

012)

Sapo

tace

aeP

icho

nia

1910

.17

55

71.4

57.1

0.12

Swen

son

etal

.(20

13)

Sapo

tace

aeP

lanc

hone

llacl

ade

II33

17.4

187

738

.938

.90.

13Sw

enso

net

al.(

2013

)

Rub

iace

aeM

arga

rito

psis

8.47

5.21

41

425

100

0.13

Barr

abé

etal

.(20

14)

Rub

iace

aeP

sych

otri

acl

ade

NC

117

.06

4.78

44

410

010

00.

15Ba

rrab

éet

al.(

2014

)

Rut

acea

eO

xant

hera

max

5.8

5.8

55

510

010

00.

16Pf

eila

ndC

risp

(200

8)Sa

pota

ceae

Ple

iolu

ma

clad

eI

104.

95

02

040

0.19

Swen

son

etal

.(20

13)

Are

cace

aeB

urre

tiok

enti

a—C

ypho

phoe

nix

11.5

7.5

90

00

00.

2Ba

ker

and

Cou

vreu

r(2

012)

Sapo

tace

aeP

ycna

ndra

29.8

16.2

5717

2029

.835

.10.

21Sw

enso

net

al.(

2013

)A

rauc

aria

ceae

Ara

ucar

ia16

813

07

053

.80.

23K

rani

tzet

al.(

2014

)Lo

gani

acea

eG

enio

stom

a9

6.5

101

410

400.

25Fo

ster

etal

.(20

14)

Podo

carp

acea

eD

acry

dium

53.

055

03

060

0.3

Kep

pele

tal.

(201

1)Po

doca

rpac

eae

Podo

carp

us6.

94.

28

22

2525

0.33

Qui

roga

etal

.(20

16)

Sapo

tace

aeP

leio

lum

acl

ade

II7.

54.

28

68

7510

00.

33Sw

enso

net

al.(

2013

)

Ana

card

iace

aeE

uros

chin

us11

.85

3.7

70

20

28.6

0.34

Wee

kset

al.(

2014

)Eb

enac

eae

Dio

spyr

oscl

ade

III

9.1

7.2

258

1632

640.

35Tu

rner

etal

.(20

13)

Pand

anac

eae

NC

Pand

anus

subg

en.

Loph

osti

gma

8.5

6.5

242

28.

38.

30.

38G

alla

her

etal

.(20

15)

Eric

acea

eD

raco

phyl

lum

5.2

3.5

87

687

.575

0.4

Wag

staf

feta

l.(2

010)

Rub

iace

aeP

sych

otri

acl

ade

NC

27.

636.

978

3240

4151

.30.

53Ba

rrab

éet

al.(

2014

)

Lam

iace

aeN

CO

xera

5.39

4.54

335

1115

.229

.70.

62T

his

stud

yR

ubia

ceae

Thi

ollie

rea

6.5

313

1313

100

100

0.62

Man

nset

al.(

2012

)

Not

e:In

grey

,scl

erop

hyllo

uscl

ades

.

Dow



IRAD

Centre de C




ber 2019




TA

BLE

3.C

ompa

riso

nof

netd

iver

sific

atio

nra

tes

ofO

xera

wit

hot

her

rapi

dis

land

plan

tlin

eage

s/ra

diat

ions

Fam

ilyR

adia

tion

Arc

hipe

lago

Num

ber

ofsp

ecie

sA

rea

(km

2 )

Med

ian

crow

nag

e(9

5%H

PDin

Ma)

Net

dive

rsifi

cati

onra

tean

dus

ing

crow

nag

es(s

peci

es/M

yr)

Net

dive

rsifi

cati

onra

teat

aar

eaun

it(s

peci

es/

Myr

/km

2 )

Net

dive

rsifi

cati

onra

teat

alo

g(ar

ea)u

nit

(spe

-ci

es/M

yr/

log(

km2 )

)

Wor

ksw

here

age

esti

mat

esan

dsp

ecie

snu

mbe

rha

vebe

enpu

blis

hed

Exti

ncti

on=

0Ex

tinc

tion

=0.

9

Lam

iace

aeN

ewC

aled

o-ni

anO

xera

New

Cal

edon

ia33

18,6

002.

76.

920.

41–1

.04

0.2–

0.51

0.00

0011

–0.

0000

560.

02–0

.106

This

stud

y

Rub

iace

aeP

sych

otri

acl

ade

NC

2N

ewC

aled

onia

7818

,600

4.62

9.82

0.37

–0.7

90.

22–0

.46

0.00

0012

–0.

0000

430.

022–

0.08

1Ba

rrab

éet

al.

(201

4)A

ster

acea

eC

heilo

roph

usM

acar

ones

ia20

10,3

725.

9815

.35

0.15

–0.3

90.

07–0

.17

0.00

0006

–0.

0000

370.

007–

0.04

2V

ital

eset

al.

(201

4)Eu

phor

biac

eae

Eup

horb

iaH

awai

i16

16,6

440.

723.

970.

52–2

.89

0.22

–1.2

0.00

0013

–0.

0001

740.

022–

0.29

7Ya

nget

al.

(201

8)A

ster

acea

eB

iden

sH

awai

i19

16,6

441.

33.

10.

73–1

.73

0.32

–0.7

50.

0000

19–

0.00

0104

0.03

2–0.

178

Kno

peet

al.

(201

2)A

ster

acea

eE

chiu

mM

acar

ones

ia19

10,3

722.

73.

90.

58–0

.83

0.25

–0.3

60.

0000

24–

0.00

008

0.02

7–0.

09K

nope

etal

.(2

012)

Ges

neri

acea

eC

yrta

ndra

Haw

aii

5816

,644

5.2

0.65

0.36

0.00

0021

–0.

0000

390.

037–

0.06

7C

lark

etal

.(2

008,

2009

)C

ampa

nula

ceae

Lobe

liads

Haw

aii

126

16,6

4410

.49

16.7

10.

25–0

.39

0.15

–0.2

40.

0000

09–

0.00

0024

0.01

6–0.

041

Giv

nish

etal

.(2

009)

Cra

ssul

acea

eA

eoni

umM

acar

ones

ia63

10,3

7213

.25

17.1

50.

2–0.

260.

11–0

.15

0.00

0011

–0.

0000

250.

012–

0.02

8K

imet

al.(

2008

)Pl

anta

gina

ceae

Ver

onic

aN

ewZ

eala

nd12

026

8,68

05.

70.

720.

440.

0000

02–

0.00

0003

0.03

5–0.

057

Ogu

tcen

(201

6),

Wag

staf

feta

l.(2

002)

Dow



IRAD

Centre de C




ber 2019




rate shift configurations (with a cumulative PP of0.378; Supplementary Appendix S13 available on Dryad)showed significant rate shifts mainly located on the twoconsecutive branches leading to the New CaledonianOxera radiation, and those leading to most of the Clero-dendrum clade (Supplementary Appendix S13 availableon Dryad). Both shifts were depicted in the marginalshift probability tree, where these latter branches weresignificantly longer than most others, indicating a highprobability that these shifts occurred on their respectivebranches (Supplementary Appendix S12b available onDryad). The Oxera rates through time plots highlightedthat speciation rates remained constant (with a meanvalue of ca. 0.71 species/Myr). Extinction rates decreasedslightly at the radiation onset but remained constant after(mean value of ca. 0.31 species/Myr). The resulting netdiversification rates increased early in the radiation onsetand later remained quite constant (mean value of ca.0.4 species/Myr; Fig. 2b, Supplementary Appendix S12cavailable on Dryad). The ML diversification analysesselected the pure birth model as best-fitting our data forthe genus Oxera, with a net diversification rate of 0.575species/Myr (Supplementary Appendix S14 available onDryad). Accordingly, the pure birth model was the bestselected model on 93% of the RD trees.

Niche OrdinationThe multivariate analyses allowed the separation of

four, three, and five clearly distinct ecological syndromesfrom the species’ environmental, flower, and fruit datasets, respectively (Fig. 1c). The first four, nine, andsix axes of the PCA analyses (explaining 90 % of thecumulate variance), respectively, were retained for sub-sequent comparative analyses. The three major habitatsoccupied by Oxera were forests on non-ultramafic rocks,forests on ultramafic rocks, sclerophyllous vegetation,and a fourth class was defined for ubiquist species.The three pollination syndromes characterized specieswith actinomorphic, gullet, and bilabiate floral morphseach adapted to bat, bird, moth/butterfly pollination,respectively. For dispersal syndromes the PCA axismarkedly separated species with small fruits ≤2 cm (onthe left) from those with large fruits >2 cm (on the right;Fig. 1c). The five dispersal syndromes discriminatedspecies with potato-like (consumed by cassowaries andbats), paddle-like (by Columbidae), small rounded (byColumbidae and/or small birds such as Sturnidae), largerounded (by Columbidae), and thorny fruit morphs(presumably by large geckos).

Niche EvolutionOur estimates of both Pagel statistics showed dis-

tinct evolutionary patterns (Fig. 2c; SupplementaryAppendix S15 available on Dryad). We found lowvalues of �, indicating low phylogenetic signal (i.e., arecent accelerated evolution), for the habitat preference(Fig. 2c), and high phylogenetic signal for the dispersal

niche and life form, suggesting that these niche charac-teristics have diversified early. For the pollination niche,estimates of � were more spread out, with a lower 5%quantile of 0.2, a higher 95% quantile of 1.01, and amedian value of 0.99; this also indicated a relatively highphylogenetic signal (Fig. 2c). Estimates of the � statisticwere close to 0 for the habitat preference (indicating apunctual evolution), and close to 1 for the dispersal nicheand life form, respectively (indicating more gradualevolution). For the pollination niche, � scaled between0.35 (5% quantile) and 0.99 (95% quantile), with a medianvalue of 0.69.

The best-fit evolutionary model identified in ourancestral state reconstruction analyses was the “ER”model for each niche/trait data set (SupplementaryAppendix S16 available on Dryad). The reconstructedancestral habitat preference of Oxera was ambiguous:forests on both ultramafic and non-ultramafic rockswere together reconstructed with a high likelihood(forested vegetation was recovered as ancestral forthis node with cumulated likelihoods of both foresttypes of ca. 0.57). However, the ancestral pollination,dispersal, and life form were inferred with much lessambiguity; the Oxera ancestor was very likely a robustliana with gullet flowers producing small rounded fruits(Supplementary Appendix S17 available on Dryad).Analyses of stochastic character mapping allowed us toestimate that very few pollination, dispersal, and lifeform shifts occurred during the evolution of Oxera (upto two shifts per time bin of 0.5 Myr; Fig. 2b), andthat most of these trait shifts occurred between 2 Maand 0.5 Ma for pollination, and following 3 Ma and3.5 Ma for dispersal and life form, respectively. Habitatshifts occurred between 9 Ma and present, but from 4Ma onward their number increased markedly, finallyreaching 17 shifts between 0.5 Ma and the present.

The best-fitting model of trait-dependent speciesdiversification identified for habitats had speciationand transition rates all equal, a null extinction rate(Supplementary Appendix S4 available on Dryad),and no discernible effect of any habitat type. For thepollination syndromes, dispersal types and life forms,the best-fitting model was that with a null extinction,transition rates all equal, and a particular characterstate in which speciation rates were significantly dif-ferent from other states (Supplementary Appendix S4available on Dryad). These results were very robustto phylogenetic uncertainty, as 100% of the RD treesconsistently yielded the same best models. Slender lianashad on average significantly higher speciation ratescomparing to other life forms, actinomorphic flowers,and potato-like fruits had on average significantly lowerspeciation rates comparing to other floral and fruit types(Supplementary Appendix S18 available on Dryad).

Geographical EvolutionThe range overlap between all species pairs was

generally low (mean value of 0.187, and a standard

Dow



IRAD

Centre de C




ber 2019

https://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-datahttps://academic.oup.com/sysbio/article-lookup/doi/10.1093/sysbio/syy070#supplementary-data




−0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2

01

23

45

range overlap

spec

ies

pair

dens

ity

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●● ●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

● ● ● ● ●●

●●

●●

●●

●●

● ●● ● ● ● ●

●●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●● ● ● ● ●

●●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●●

● ● ● ● ●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●●

●●

●●

●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ●● ● ●

● ●● ●

● ●●

●●

●●

●●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●●

●● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●

● ● ● ● ● ● ● ●●

●●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●●

●●

●●

●● ● ● ● ● ● ● ● ●

all species pairssister species pairs

FIGURE 3. The Kernel density plot of range overlap through allpossible species pairs (in black) and all possible sister species pairs (ingrey) in New Caledonian Oxera.

deviation of 0.357), suggesting a high level of allopatrybetween all species. The kernel density plot highlightedtwo range overlap frequency peaks, the highest locatedfor a range overlap of 0 and the shortest to 1 (Fig. 3). Agerange correlation analyses showed that the slope of rangeasymmetry through time was significantly positive forlinear regressions (all P-values




FIGURE 4. Species coexistence within Oxera, indicated by subclade, with stars indicating likely coexistence; mapped on the BEAST maximumcredibility clade chronogram from the Oxera data set. a) Geographical overlap between all species of each subclade. b) Ecological ranges ofall species indicated per subclade, plotted according to their altitudinal ranges and their habitat preferences (UM = ultramafic soils, NUM =non-ultramafic soils). Species names are indicated with the following abbreviations, listed according to subclade: robusta subclade (cori = O.coriacea, coro = O. coronata, lon = O. longifolia, pal = O. palmatinervia, rob = O. robusta), subverticillata subclade (aur = O. aureocalyx sp. nov. ined.,mer = O. merytifolia, sub = O. subverticillata, tiw = O. tiwaeana sp. nov. ined.), sulfurea subclade (gme = O. gmelinoides, mic = O. microcalyx, oua =O. ouameniensis sp. nov. ined., pan = O. pancheri, rug = O. rugosa, sul = O. sulfurea), baladica subclade (bal = O. baladica, anu = O. aff. nuda, dou =O. doubetiae sp. nov. ined., gar = O. garoensis sp. nov. ined., oun = O. ounemoae sp. nov. ined., pap = O. papineaui sp. nov. ined., ses = O. sessilifolia,van = O. vanuatuensis), neriifolia subclade (neri = O. neriifolia, ova = O. ovata, sor = O. sororia), pulchella subclade (bal = O. balansae, balV = O.aff.balansae, bre = O. brevicalyx, pul = O. pulchella), and oblongifolia subclade (cra = O. crassifolia, gla = O. glandulosa, obl = O. oblongifolia, ore = O.oreophila).

vegetation), especially since the Miocene (Donoghue andEdwards 2014). It is important to note that some ofthese novel habitats exhibit a patchy spatial distribution,which has likely increased genetic divergence. Thisostensibly has generated vacant niches, created ecolo-gical opportunities, increased spatial divergence and

thus triggered speciation in many Australasian lineages,as highlighted in Australia for Banksia, Eucalyptus, andAllocasuarina (Crisp et al. 2004; Gallagher et al. 2003). Theearly rapid diversification of Oxera suggests that similarclimatic and physical events fostered diversificationwithin Oxera shortly after its arrival into New Caledonia.

Dow



IRAD

Centre de C



e

Documents

Changing Ecological Opportunities Facilitated the Explosive ......endemism (ca. 3400 species, 75% endemic; Morat et al. 2012; Munzinger et al. 2016). The singular geological and climatic