Narrow endemics to Mediterranean islands: Moderate genetic diversity but narrow climatic niche of the ancient, critically endangered Naufraga (Apiaceae)

Abstract

Narrow endemics constitute the cornerstone of Mediterranean plant diversity. Naufraga balearica (Apiaceae) is a critically endangered, extremely narrow endemic plant from the western Mediterranean island of Majorca. Because the species belongs to a monotypic genus, N. balearica was hypothesized to be a palaeoendemism. Here we conducted phylogenetic dating, population genetic and climatic niche analyses in order to understand the evolutionary history and conservation perspectives of this flagship species. Phylogenetic dating analysis of nuclear and plastid DNA sequences revealed a late Miocene to early Pliocene divergence between Naufraga and its sister genus Apium, supporting the palaeoendemic status of the former. Amplified fragment length polymorphism (AFLP) markers and plastid DNA sequences of the five Naufraga populations revealed moderate genetic diversity. This diversity is in line with that of other palaeoendemisms from western Mediterranean islands, as revealed by a comparison with 22 other narrow endemic species from this region. Despite the fact that all Naufraga populations are located at a maximum distance of 10 km in a straight line, a strikingly strong population differentiation was found for AFLP markers, which is explained by long-term isolation likely related to short-range pollination and dispersal strategies of the species. While the species is not genetically impoverished, species distribution modelling and microclimatic monitoring revealed that narrow ecological requirements underlie the current extreme rarity of Naufraga and may jeopardize its long-term survival. Our results indicate that a multidisciplinary approach provides powerful tools to develop conservation strategies for evolutionarily singular lineages.

Full text

Perspectives
in
Plant
Ecology,
Evolution
and
Systematics
16
(2014)
190–202
Contents
lists
available
at
ScienceDirect
Perspectives
in
Plant
Ecology,
Evolution
and
Systematics
jou
rn
al
hom
epage:
www.elsevier.com/locate/ppees
Research
article
Narrow
endemics
to
Mediterranean
islands:
Moderate
genetic
diversity
but
narrow
climatic
niche
of
the
ancient,
critically
endangered
Naufraga
(Apiaceae)
Mario
Fernández-Mazuecosa,∗,
Pedro
Jiménez-Mejíasa,1,
Xavier
Rotllan-Puigb,
Pablo
Vargasa
aReal
Jardín
Botánico
(RJB-CSIC),
Plaza
de
Murillo
2,
28014
Madrid,
Spain
bInstitut
Mediterrani
d‘Estudis
Avanc¸
ats
(CSIC-UIB),
Miquel
Marqués
21,
07190
Esporles,
Mallorca,
Balearic
Islands,
Spain
a
r
t
i
c
l
e
i
n
f
o
Article
history:
Received
18
November
2013
Received
in
revised
form
18
March
2014
Accepted
4
May
2014
Available
online
16
May
2014
Keywords:
Palaeoendemism
Amplified
fragment
length
polymorphism
Plastid
DNA
haplotypes
Species
distribution
modelling
Balearic
Islands
Apiaceae
a
b
s
t
r
a
c
t
Narrow
endemics
constitute
the
cornerstone
of
Mediterranean
plant
diversity.
Naufraga
balearica
(Api-
aceae)
is
a
critically
endangered,
extremely
narrow
endemic
plant
from
the
western
Mediterranean
island
of
Majorca.
Because
the
species
belongs
to
a
monotypic
genus,
N.
balearica
was
hypothesized
to
be
a
palaeoendemism.
Here
we
conducted
phylogenetic
dating,
population
genetic
and
climatic
niche
analy-
ses
in
order
to
understand
the
evolutionary
history
and
conservation
perspectives
of
this
flagship
species.
Phylogenetic
dating
analysis
of
nuclear
and
plastid
DNA
sequences
revealed
a
late
Miocene
to
early
Pliocene
divergence
between
Naufraga
and
its
sister
genus
Apium,
supporting
the
palaeoendemic
sta-
tus
of
the
former.
Amplified
fragment
length
polymorphism
(AFLP)
markers
and
plastid
DNA
sequences
of
the
five
Naufraga
populations
revealed
moderate
genetic
diversity.
This
diversity
is
in
line
with
that
of
other
palaeoendemisms
from
western
Mediterranean
islands,
as
revealed
by
a
comparison
with
22
other
narrow
endemic
species
from
this
region.
Despite
the
fact
that
all
Naufraga
populations
are
located
at
a
maximum
distance
of
10
km
in
a
straight
line,
a
strikingly
strong
population
differentiation
was
found
for
AFLP
markers,
which
is
explained
by
long-term
isolation
likely
related
to
short-range
pollination
and
dispersal
strategies
of
the
species.
While
the
species
is
not
genetically
impoverished,
species
distribu-
tion
modelling
and
microclimatic
monitoring
revealed
that
narrow
ecological
requirements
underlie
the
current
extreme
rarity
of
Naufraga
and
may
jeopardize
its
long-term
survival.
Our
results
indicate
that
a
multidisciplinary
approach
provides
powerful
tools
to
develop
conservation
strategies
for
evolutionarily
singular
lineages.
©
2014
Geobotanisches
Institut
ETH,
Stiftung
Ruebel.
Published
by
Elsevier
GmbH.
All
rights
reserved.
Introduction
The
Mediterranean
Basin,
with
c.
25,000
plant
species
(Quézel,
1985),
constitutes
one
of
the
world’s
major
biodiversity
hotspots
(Myers
et
al.,
2000).
Around
60%
of
plant
species
endemic
to
the
Mediterranean
region
are
narrow
endemics,
i.e.
species
whose
distribution
is
restricted
to
a
single,
small
area
(Thompson,
2005).
Hence,
narrow
endemics
(both
palaeo-
and
neoendemics)
∗Corresponding
author
at:
Current
address:
Department
of
Plant
Sciences,
Uni-
versity
of
Cambridge,
Downing
Street,
Cambridge
CB2
3EA,
UK.
Tel.:
+44(0)1223333934.
E-mail
address:
mfmazuecos@rjb.csic.es
(M.
Fernández-Mazuecos).
1Current
address:
School
of
Biological
Sciences,
Washington
State
University,
Pullman
99164
WA,
USA.
are
considered
to
constitute
the
cornerstone
of
Mediterranean
plant
diversity
(Thompson,
2005).
Some
of
these
species,
termed
‘extremely
narrow
endemics’
(ENEs),
are
known
from
one
or
very
few
populations
(≤5)
and
display
very
small
census
sizes
(≤500
individuals)
(López-Pujol
et
al.,
2013).
ENEs
are
of
particular
con-
servation
concern
because
of
the
high
extinction
risk
associated
to
rarity
(O‘Grady
et
al.,
2004).
In
addition,
ENEs
usually
display
low
levels
of
genetic
diversity
(Gitzendanner
and
Soltis,
2000;
López-
Pujol
et
al.,
2013),
which
may
limit
their
evolutionary
viability.
This
genetic
impoverishment
is
variously
viewed
as
either
a
cause
or
consequence
of
rarity
(Gitzendanner
and
Soltis,
2000).
Indeed,
the
causes
of
species
rarity
have
long
been
discussed,
and
a
combina-
tion
of
ecological,
historical
and
genetic
factors
is
generally
invoked
to
account
for
it
(Kruckeberg
and
Rabinowitz,
1985).
Naufraga
balearica
(Fig.
1A)
is
an
extremely
narrow
endemic
plant
from
the
northern
coast
of
the
western
Mediterranean
island
http://dx.doi.org/10.1016/j.ppees.2014.05.003
1433-8319/©
2014
Geobotanisches
Institut
ETH,
Stiftung
Ruebel.
Published
by
Elsevier
GmbH.
All
rights
reserved.

M.
Fernández-Mazuecos
et
al.
/
Perspectives
in
Plant
Ecology,
Evolution
and
Systematics
16
(2014)
190–202
191
Fig.
1.
(A)
Specimens
of
Naufraga
balearica
with
flowers
and
fruits,
Cap
de
Catalunya
population
(photograph
by
P.
Vargas).
(B)
Phylogenetic
dating
analysis
of
the
tribe
Apieae
based
on
combined
nuclear
(ITS)
and
plastid
(rps16)
DNA
sequences.
The
maximum
clade
credibility
tree
produced
by
a
relaxed
molecular-clock
analysis
in
BEAST
is
shown.
Outgroup
taxa
have
been
pruned
for
clarity.
Values
above
branches
indicate
Bayesian
posterior
probabilities
(PP).
Node
bars
represent
the
95%
highest
posterior
density
intervals
for
the
divergence
time
estimates
of
clades
with
PP
=
1.
of
Majorca
(Balearic
Islands,
Spain)
(Rosselló,
2010).
It
is
listed
as
critically
endangered
(CR)
in
the
IUCN
Red
List
(Moreno,
2011)
and
the
Red
List
of
Spanish
vascular
flora
(Moreno,
2008).
The
species
was
first
described
from
a
single
locality
(Coves
Blanques)
discov-
ered
in
1962
(Constance
and
Cannon,
1967;
Duvigneaud,
1970).
Additional
populations
were
found
decades
later
in
the
Formen-
tor
Peninsula,
not
far
from
the
locus
classicus
(Bibiloni
and
Soler,
2002).
One
locality
(Finucchiaghia)
was
also
reported
in
Corsica
in
1981,
but
became
extinct
shortly
after
(Gamisans
et
al.,
1996;
Fridlender,
2001).
Genetic
analysis
of
cultivated
Corsican
plants
indicated
a
very
close
relationship
with
the
Coves
Blanques
popu-
lation
(but
not
with
other
populations
from
Majorca),
which
casted
doubt
on
the
spontaneity
of
N.
balearica
in
Corsica
(Fridlender
and
Boisselier-Dubayle,
2000).
It
has
long
been
hypothesized
that
N.
balearica
is
a
palaeoen-
demism
(Duvigneaud,
1970).
Palaeoendemics
have
historically
been
defined
as
relics
of
earlier
floras
which
have
survived
in
a
limited
portion
of
their
past
territory
(Wulff,
1943;
Favarger
and
Contandriopoulos,
1961).
They
are
systematically
isolated
as
a
result
of
their
early
divergence,
and
they
did
not
necessarily
originate
in
the
area
they
currently
occupy.
On
the
contrary,
neoen-
demics
originated
recently
in
a
given
region
and
have
not
yet
spread
beyond
it.
They
are
closely
related
to
other
species,
frequently
in
the
same
region.
Phylogenetic
analyses
based
on
nuclear
internal
tran-
scribed
spacer
(ITS)
sequences
have
revealed
that
Naufraga
is
sister
to
the
genus
Apium
(c.
20
spp.),
and
divergence
between
the
two
genera
has
been
dated
back
to
>4.8
Ma
(Spalik
et
al.,
2010;
Banasiak
et
al.,
2013).
These
results
are
congruent
with
a
palaeoendemic
status
of
Naufraga.
However,
additional
evidence
(from
phyloge-
netics,
population
genetics
and
ecology)
is
required
before
firm
conclusions
can
be
drawn.
A
previous
population
genetic
analysis
based
on
random
ampli-
fied
polymorphic
DNA
(RAPD)
markers
found
strong
genetic
differentiation
between
populations
of
N.
balearica
(Fridlender
and
Boisselier-Dubayle,
2000).
However,
the
population
sampling
in
that
study
was
unsatisfactory
(number
of
localities
included)
and
the
reliability
of
RAPD
markers
has
been
questioned
because
of
their
low
reproducibility
(Newton
et
al.,
1999).
More
reliable
markers,
together
with
a
deeper
sampling
of
individuals
and
popu-
lations,
are
therefore
needed
for
a
reliable
genetic
characterization
of
N.
balearica
populations.
Here,
phylogenetic
and
population
genetic
analyses
(based
on
nuclear
and
plastid
DNA
markers)
and
species
distribution
mod-
elling
were
conducted
to
achieve
the
following
objectives:
(1)
to
estimate
the
divergence
time
of
N.
balearica;
(2)
to
disclose
the
genetic
diversity
and
spatial
genetic
structure
of
extant
popula-
tions;
and
(3)
to
characterize
the
climatic
niche
of
the
species
both
at
the
macro-
and
microclimatic
scales.
Our
results
were
further
integrated
with
previous
ecological
and
life
history
data
in
order
to
understand
the
evolutionary
history
and
conservation
perspec-
tives
of
the
species.
Our
working
hypothesis
in
this
study
was
that
the
current
rarity
of
N.
balearica
can
be
explained
by
its
narrow
ecological
requirements,
together
with
its
short-range
pollination
and
dispersal
strategies.
Materials
and
methods
Study
species
Naufraga
balearica
Constance
&
Cannon
(Apiaceae,
Apioideae,
Apieae)
(hereafter
Naufraga)
is
a
perennial
herb
(Fig.
1A)
inhabit-
ing
shady,
humid
sites
on
calcareous
coastal
cliffs,
25–250
m
above
sea
level
(Fridlender,
2001;
Bibiloni
and
Soler,
2002).
Its
distribu-
tion
range
encompasses
a
short
(c.
15
km)
stretch
of
the
northern
Majorcan
coast
(Fig.
2A).
It
is
a
xenogamous,
ant-pollinated
species
(Cursach
and
Rita,
2012).
Flowering
starts
in
April
and
ends
in
August,
and
fruiting
occurs
from
June
to
September
(Rosselló,
2010).
Barochory
seems
to
be
its
only
mode
of
seed
dispersal
(Fridlender,
2001;
Moragues,
2005).
Vegetative
reproduction
by
stolons
is
frequent
(Moragues,
2005).
A
diploid
chromosome
num-
ber
of
2n
=
20,
with
0–2
accessory
chromosomes,
has
been
reported
(Castro
and
Rossello,
2005).
Demographic
analyses
have
shown
that
populations
are
declining
(Cursach
et
al.,
2013;
Cursach
and
Rita,
2013)
and
that
seedling
survival
is
low
(Cursach
and
Rita,
2012).
Major
threats
include
interspecific
competition,
changing
climatic
conditions,
soil
erosion,
umbel
predation,
collection
and
fires
(Rosselló,
2010;
Cursach
and
Rita,
2012;
Cursach
et
al.,
2013;
Cursach
and
Rita,
2013).
A
certain
degree
of
herbivore
pressure
is
thought
to
benefit
Naufraga
populations
by
reducing
interspecific
competition
(Cursach
et
al.,
2013).
Phylogenetic
dating
Spalik
et
al.
(2010)
and
Banasiak
et
al.
(2013)
obtained
the
first
estimates
for
the
divergence
time
of
Naufraga
based
on
broad-
scale
analyses
of
ITS
sequences
of
subfamily
Apioideae.
In
order
to
obtain
a
more
precise
estimate,
a
deeper
sampling
of
closely
related
genera
and
additional
DNA
markers
are
required.
To
this
end,
we
conducted
a
dating
analysis
using
sequence
matrices
from
a
separate
phylogenetic
study
of
the
tribe
Apieae
(Jiménez-Mejías
and
Vargas,
under
review).
Forty-four
nuclear
ITS
sequences
and
44

192
M.
Fernández-Mazuecos
et
al.
/
Perspectives
in
Plant
Ecology,
Evolution
and
Systematics
16
(2014)
190–202
Fig.
2.
Analysis
of
ptDNA
(rpl32-trnLUAG/ycf6-psbM)
haplotypes
of
Naufraga
balearica.
The
three
haplotypes
are
represented
as
different
shades
of
grey
in
both
A
and
C.
(A)
Geographical
distribution
of
haplotypes
across
the
five
sampled
popu-
lations.
Pie
charts
represent
haplotype
frequencies,
obtained
by
sequencing
19–20
individuals
per
population.
(B)
Fifty
per
cent
majority-rule
consensus
tree
of
the
Bayesian
phylogenetic
analysis
of
ptDNA
haplotypes;
numbers
above
branches
are
Bayesian
posterior
probabilities.
(C)
Statistical
parsimony
network
of
ptDNA
hap-
lotypes;
lines
represent
single
nucleotide
substitutions.
The
rectangle
(haplotype
A)
indicates
the
haplotype
suggested
as
ancestral
by
the
TCS
software.
Oval
and
rectangle
sizes
are
proportional
to
the
number
of
sequences
(N)
obtained
for
each
haplotype.
plastid
DNA
(ptDNA)
rps16
sequences
from
33
Apieae
and
11
out-
group
species
(one
individual
per
species;
Table
S1)
were
analyzed
using
the
relaxed
molecular
clock
approach
implemented
in
BEAST
1.7.5
(Drummond
et
al.,
2006;
Drummond
and
Rambaut,
2007).
Since
no
fossils
of
Apieae
appropriate
for
calibration
are
known
to
date
(Martínez-Millán,
2010;
Banasiak
et
al.,
2013),
we
imple-
mented
a
secondary
calibration
based
on
the
result
of
Banasiak
et
al.’s
(2013)
dating
analysis
of
Apioideae.
A
lognormal
distribu-
tion
was
fitted
to
the
posterior
distribution
of
ages
(in
million
years)
for
the
crown
age
of
Apieae
obtained
by
Banasiak
et
al.
(2013)
using
the
R
package
MASS
(Ł.
Banasiak,
pers.
comm.).
The
obtained
dis-
tribution,
with
log(mean)
=
2.724
and
log(stdev)
=
0.138,
was
then
used
as
a
prior
to
calibrate
the
crown
age
of
Apieae
in
our
dating
analysis.
Models
of
nucleotide
substitution
were
selected
for
each
DNA
region
under
the
Akaike
information
criterion
(AIC)
in
jModel-
Test
0.1
(Guindon
and
Gascuel,
2003;
Posada,
2008).
A
birth-death
process
(Gernhard,
2008)
was
employed
as
tree
prior.
The
substitu-
tion
rate
variation
was
modelled
using
an
uncorrelated
lognormal
distribution.
Based
on
previous
estimates
for
herbaceous
plants,
uniform
prior
distributions
were
set
for
the
substitution
rates,
with
ranges
5
×
10−4–5
×
10−2substitutions
per
site
per
Myr
for
ITS,
and
1
×
10−4–1
×
10−2substitutions
per
site
per
Myr
for
ptDNA
(see
Blanco-Pastor
et
al.,
2012
for
details).
For
each
dataset,
four
MCMC
analyses
with
10
million
generations
each
and
a
sample
frequency
of
1000
were
run
through
the
CIPRES
Science
Gateway
(Miller
et
al.,
2010).
Parameter
analysis
in
Tracer
1.5
(Rambaut
and
Drummond,
2007)
showed
adequate
chain
length,
with
effective
sample
size
(ESS)
values
above
300.
Chains
were
combined
using
LogCombiner
1.7.5,
after
discarding
the
first
10%
of
sampled
generations
as
burn-
in.
Trees
were
summarized
in
a
maximum
clade
credibility
(MCC)
tree
obtained
in
TreeAnotator
1.7.5
and
visualized
in
FigTree
1.3.1
(http://tree.bio.ed.ac.uk/software/figtree/).
Population
genetics
Sampling
strategy
and
DNA
isolation
Plant
materials
were
sampled
from
100
individuals
in
the
five
known
populations
of
Naufraga
covering
its
entire
known
distribu-
tion
(Bibiloni
and
Soler,
2002)
(Fig.
2A;
Table
1).
As
a
result,
twenty
individuals
were
sampled
per
population.
Given
the
stoloniferous
character
of
the
species,
care
was
taken
to
sample
distinct
individ-
uals.
All
plant
material
was
collected
in
the
field
and
dried
in
silica
gel.
Based
on
previous
phylogenetic
results
(Spalik
et
al.,
2010),
three
individuals
of
different
species
of
Apium
(A.
graveolens,
A.
prostratum
and
A.
panul)
were
additionally
sampled
to
be
used
as
the
outgroup
in
phylogenetic
and
phylogeographic
analyses
(see
below).
Total
genomic
DNA
was
extracted
using
the
DNeasy
Plant
Mini
Kit
(Qiagen
Inc.,
California)
following
the
manufacturer’s
rec-
ommended
protocols.
ptDNA
sequence
variation
Procedures
used
for
ptDNA
amplification
and
sequencing
fol-
lowed
Fernández-Mazuecos
and
Vargas
(2011).
First,
a
pilot
study
using
five
individuals
(one
per
population)
of
Naufraga
was
per-
formed
to
find
consistently
amplified
and
variable
ptDNA
regions.
We
tested
eight
regions
previously
used
in
phylogeographic
anal-
yses:
rpl32-trnLUAG,
trnQ-rps16,
trnS-trnG,
trnH-psbA,
petL-psbE,
psbJ-petA,
ycf6-psbM
and
atpI-atpH
(Hamilton,
1999;
Shaw
et
al.,
2005,
2007;
Hollingsworth
et
al.,
2009).
All
regions
were
consis-
tently
amplified
and
sequenced.
The
two
regions
(rpl32-trnLUAG
and
ycf6-psbM)
that
yielded
nucleotide
variation
in
the
pilot
study
were
then
sequenced
for
all
sampled
individuals
of
Naufraga
and
outgroup.
In
all
cases,
the
same
standard
primers
were
employed
for
amplification
and
sequencing.
Sequences
were
assembled
in
Geneious
Pro
(Drummond
et
al.,
2010).
Sequences
of
each
DNA
region
were
separately
aligned
using
MAFFT
6
(Katoh
et
al.,
2002)
with
default
parameters,
and
further
adjustments
were
made
by
visual
inspection.
The
three
ptDNA
regions
were
concatenated
in
a
single
matrix.
All
new
sequences
were
deposited
in
the
GenBank
database
(see
Table
S2
for
accession
numbers).
The
concatenated
dataset
was
analyzed
using
the
statistical
par-
simony
algorithm
(Templeton
et
al.,
1992),
as
implemented
in
TCS
1.21
(Clement
et
al.,
2000),
in
order
to
infer
genealogical
rela-
tionships
among
haplotypes
and
phylogeographic
patterns.
The
maximum
number
of
differences
resulting
from
single
substitut-
ions
among
haplotypes
was
calculated
with
95%
confidence
limits,
treating
gaps
as
missing
data.
Relationships
among
haplotypes
were
additionally
assessed
using
Bayesian
inference.
Models
of
nucleotide
substitution
Table
1
Geographic
location,
average
AFLP
gene
diversity
±
standard
deviation,
DW
index
values
and
BAPS
clustering
of
Naufraga
balearica
populations.
Population
Label
Coordinates
Gene
diversity
DW
BAPS
clustering
Coves
Blanques
COV
39.933◦N
3.055◦E
0.1995
±
0.2016
45.15
I
Coll
de
la
Creueta COL
39.930◦N
3.109◦E
0.1149
±
0.1763
30.73
II
Orelles
de
l’Ase
ORE
39.944◦N
3.139◦E
0.0531
±
0.1300
19.98
III
Les
Fonts
Salades
FON
39.955◦N
3.160◦E
0.2233
±
0.2002
48.57
IV
Cap
de
Catalunya
CAP
39.957◦N
3.171◦E
0.1970
±
0.1879
40.57
IV
Total
0.2686
±
0.1621

M.
Fernández-Mazuecos
et
al.
/
Perspectives
in
Plant
Ecology,
Evolution
and
Systematics
16
(2014)
190–202
193
(GTR
for
both
DNA
regions)
were
selected
under
the
AIC
in
jModelTest
0.1.
The
Bayesian
analysis
was
performed
in
MrBayes
v3.1.2
(Ronquist
and
Huelsenbeck,
2003)
using
two
searches
with
10
million
generations
each
and
a
sample
frequency
of
1000.
The
two
regions
were
partitioned,
and
substitution
models
were
unlinked
across
partitions.
A
fifty-percent
majority
rule
consen-
sus
tree
with
Bayesian
posterior
probabilities
(PP)
of
clades
was
calculated
after
removing
the
first
10%
generations
as
burn-in.
Amplified
fragment
length
polymorphism
The
genetic
diversity
and
spatial
genetic
structure
of
Naufraga
populations
were
analyzed
using
AFLPs
(amplified
fragment
length
polymophisms;
Vos
et
al.,
1995).
Laboratory
procedures
followed
Gaudeul
et
al.
(2000).
First,
we
performed
a
pilot
study
in
which
32
primer
combinations
were
screened
in
six
samples
(one
individ-
ual
per
population,
plus
one
replicate).
We
chose
the
four
primer
combinations
that
yielded
the
highest
numbers
of
reproducible
fragments
and
informative
characters:
EcoRI
ACT
(FAM)
–
MseI
CTG,
EcoRI
AGG
(VIC)
–
MseI
CTA,
EcoRI
AGA
(FAM)
–
MseI
CTA,
and
EcoRI
AGG
(VIC)
–
MseI
CTA.
The
100
sampled
individuals
were
then
analyzed
using
the
four
chosen
primer
combinations.
Selective
PCR
products
were
run
on
a
capillary
sequencer
(ABI
PRISM
3700;
Applied
Biosystems,
Foster
City,
CA,
USA)
with
the
internal
size
standard
GeneScan
500
LIZ
(Applied
Biosystems).
Data
collection
and
fragment
sizing
were
performed
using
the
software
GeneMap-
per
v3.7
(Applied
Biosystems).
Fragments
in
the
range
50–500
bp
were
automatically
scored
and
manually
revised.
The
results
were
exported
as
a
presence/absence
(1/0)
matrix.
Reproducibility
was
estimated
based
on
nine
replicated
samples
(9%
of
the
sampling)
as
the
average
proportion
of
correctly
replicated
bands
(Bonin
et
al.,
2004).
Markers
with
low
reproducibility
were
excluded,
resulting
in
a
final
error
of
1.9%.
Linked
alleles
were
removed
from
the
matrix.
Phenotype
diversity
was
evaluated
using
the
R
script
AFLP-
dat
ver.
2008
(Ehrich,
2006;
updated
in
24th
June
2010).
Clones
were
identified
as
those
phenotypes
that
differed
in
a
propor-
tion
of
bands
below
the
error
rate
(2%,
see
above).
We
used
Arlequin
v3.5
(Excoffier
and
Lischer,
2010)
to
evaluate
gene
diver-
sity
of
populations
according
to
Nei’s
formula
(Nei,
1987).
The
‘frequency-down-weighted
marker’
(DW)
value,
a
rarity
measure
(Schönswetter
and
Tribsch,
2005),
which
accounts
for
differences
in
sample
size
(Ehrich
et
al.,
2008),
was
also
calculated
using
AFLPdat.
Several
methods
were
employed
to
evaluate
the
genetic
structure
of
populations.
An
unrooted
neighbour-joining
(NJ)
tree
of
AFLP
genotypes
was
obtained
in
PAUP*
v4.0b10
(Swofford,
2002)
using
Nei-Li
distances
(Nei
and
Li,
1979).
Branch
support
was
evaluated
with
10,000
bootstrap
replicates.
A
principal
coordinate
analysis
(PCoA)
was
performed
in
the
package
GenAlEx
v6.5
(Peakall
and
Smouse,
2006)
using
Euclidean
distances
among
samples
(Huff
et
al.,
1993).
The
Bayesian
clustering
software
BAPS
v6.0
(Corander
et
al.,
2004)
was
also
used
to
estimate
population
structure.
A
genetic
mixture
analysis
was
conducted
at
the
individual
level,
with
an
upper
bound
to
the
number
of
populations
K
=
20.
An
admixture
analysis
was
then
conducted
with
default
parameters.
Analysis
of
molecular
variance
(AMOVA;
Excoffier
et
al.,
1992)
was
conducted
in
Arlequin
to
assess
genetic
differentiation
among
populations
and
BAPS
groups.
Finally,
a
Mantel
test
(Smouse
et
al.,
1986)
was
performed
in
GenAlEx
to
evaluate
isolation-by-distance,
i.e.
corre-
lation
between
genetic
and
geographic
distances.
Climatic
niche
Species
distribution
modelling
Species
distribution
modelling
(SDM)
was
performed
to
evaluate
the
potential
distribution
of
Naufraga
under
present
macroclimatic
conditions.
We
employed
the
maximum
entropy
algorithm,
as
implemented
in
Maxent
v3.3
(Phillips
et
al.,
2006),
because
it
is
appropriate
for
low
numbers
of
presence-only
data
and
its
good
predictive
performance
has
been
demonstrated
(Elith
et
al.,
2006;
Pearson
et
al.,
2007).
We
retrieved
a
set
of
19
bio-
climatic
variables
under
current
conditions
from
the
WorldClim
website
(www.worldclim.org;
Hijmans
et
al.,
2005).
Following
Fernández-Mazuecos
and
Vargas
(2013),
we
then
selected
a
set
of
seven
variables
that
are
uncorrelated
in
the
western
Mediterranean
(including
the
Balearic
Islands):
bio3
(isothermality),
bio4
(tem-
perature
seasonality),
bio5
(maximum
temperature
of
warmest
month),
bio6
(minimum
temperature
of
coldest
month),
bio13
(precipitation
of
wettest
month),
bio14
(precipitation
of
driest
month)
and
bio15
(precipitation
seasonality).
The
seven
variables
were
used
as
predictors
to
calibrate
the
distribution
model
in
Max-
ent.
Analyses
were
conducted
at
two
different
geographic
scales:
western
Mediterranean
region
(latitude
32–47◦N;
longitude
11◦W
to
19◦E),
and
Balearic
Islands
(latitude
38.5–40.3◦N;
longitude
1–4.5◦E).
In
the
occurrence
dataset,
we
included
precise
coordi-
nates
of
the
five
Balearic
populations
sampled
for
genetic
analyses.
Due
to
the
doubtful
spontaneity
of
the
extinct
Corsican
population
(Fridlender
and
Boisselier-Dubayle,
2000),
analyses
at
the
western
Mediterranean
scale
were
conducted
both
excluding
and
includ-
ing
this
locality.
The
predictive
power
of
Maxent
with
as
few
as
five
localities
was
demonstrated
by
Pearson
et
al.
(2007).
Given
the
low
number
of
presence
records,
we
did
not
split
localities
into
training
and
test
data
(Pearson
et
al.,
2007).
Jackknife
analyses
were
employed
to
evaluate
variable
contributions
to
the
models.
To
convert
continuous
suitability
values
to
discrete
presence/absence
(1/0)
values,
we
chose
the
‘minimum
training
presence’
threshold,
which
provides
a
conservative
estimate
of
suitable
areas
(Pearson
et
al.,
2007).
Microclimatic
niche
To
complement
the
modelling
results,
we
characterized
the
cli-
matic
niche
of
Naufraga
at
a
smaller
spatial
and
temporal
scale.
Temperature
and
relative
humidity
in
the
five
sampled
locali-
ties
were
monitored
in
the
course
of
one
year,
from
September
2012
to
August
2013.
We
used
ten
HOBO
U23
Pro
v2
data
loggers
(Onset
Computer
Corporation,
Bourne,
MA,
USA),
which
recorded
the
two
variables
at
30-min
intervals.
Two
loggers
were
placed
at
each
locality,
one
in
a
shady
slope
facing
north,
with
presence
of
Naufraga,
and
the
other
in
a
sunny
slope
facing
south,
in
which
Naufraga
was
absent.
Distance
between
loggers
in
the
same
local-
ity
ranged
between
75
and
310
m,
while
altitude
differences
ranged
between
2
and
130
m.
Resulting
data
were
processed
using
the
HOBOware
software.
Results
Phylogenetic
dating
The
phylogenetic
dating
analysis
of
nuclear
and
plastid
sequences
of
Apieae
(Fig.
1B)
recovered
a
sister
relationship
between
Naufraga
and
Apium
(PP
=
1),
and
estimated
a
divergence
time
between
the
two
genera
in
the
late
Miocene
to
early
Pliocene
(95%
highest
posterior
density
interval
3.5–9.0
Ma).
Population
genetics
ptDNA
sequence
variation
Two
ptDNA
regions,
rpl32-trnLUAG and
ycf6-psbM,
yielded
nucleotide
diversity
within
Naufraga
in
the
pilot
study
(one
nucleotide
substitution
each).
After
extensive
sequencing
of
the
two
regions
(Fig.
2A–C),
the
combined
matrix
consisted
of
1896
bp
from
96
individuals
of
Naufraga
and
three
of
Apium.
In
the
TCS

194
M.
Fernández-Mazuecos
et
al.
/
Perspectives
in
Plant
Ecology,
Evolution
and
Systematics
16
(2014)
190–202
Fig.
3.
Analysis
of
the
population
genetic
structure
of
Naufraga
balearica
based
on
AFLP
markers.
Colours
represent
genetic
clusters
obtained
in
BAPS:
red,
cluster
I;
yellow,
cluster
II;
blue,
cluster
III;
green,
cluster
IV.
Populations
are
labelled
as
in
Table
1.
(A)
Neighbour-joining
tree
of
AFLP
genotypes.
Numbers
above
branches
are
bootstrap
values
(in
percentage).
(B)
Principal
coordinate
analysis
(PCoA).
Values
for
the
first
three
axes
are
plotted,
and
percentage
of
variation
explained
by
each
axis
is
shown
in
brackets.
(C)
Genetic
clustering
of
individuals
based
on
Bayesian
analysis
in
BAPS.
The
result
of
the
admixture
analysis
is
shown.
Each
vertical
bar
represents
a
single
individual,
with
colours
representing
the
genetic
contributions
of
the
four
genetic
clusters
detected
in
the
mixture
analysis.
(For
interpretation
of
the
references
to
color
in
this
figure
legend,
the
reader
is
referred
to
the
web
version
of
this
article.)
analysis,
two
unconnected
networks
were
obtained:
one
formed
by
three
haplotypes
of
Naufraga,
and
the
other
formed
by
three
Apium
haplotypes
(one
per
species;
results
not
shown).
The
three
Naufraga
haplotypes
(A–C)
formed
a
network
with
no
loops
and
no
missing
haplotypes
(Fig.
2C).
The
three
haplotypes
constituted
a
mono-
phyletic
group
(PP
=
1)
in
the
Bayesian
phylogenetic
analysis,
but
relationships
between
them
were
unresolved
(Fig.
2B).
The
central
haplotype
A
was
the
most
widely
distributed
(Fig.
2A):
it
was
the
only
haplotype
found
in
western
populations
(COV,
COL),
and
it
was
also
found
in
one
eastern
population
(FON).
Haplotype
B
was
found
in
central-eastern
populations
(ORE,
FON).
Haplotype
C
was
exclu-
sively
found
in
the
easternmost
population
(CAP).
Remarkably,
all
haplotypic
variation
was
represented
in
the
two
easternmost
popu-
lations
(FON,
CAP),
separated
by
c.
1
km
but
displaying
no
shared
haplotypes
between
them
(Fig.
2A).
Amplified
fragment
length
polymorphism
The
final
AFLP
matrix
had
185
characters.
Gene
diversity
for
the
whole
species
was
0.27.
Only
one
sample
was
identified
as
a
putative
clone
considering
an
error
rate
of
2%.
The
highest
popu-
lation
diversity
and
rarity
(DW)
values
were
found
in
population
FON,
while
ORE
displayed
the
lowest
values
(Table
1).
The
NJ
tree
(Fig.
3A),
the
PCoA
(Fig.
3B)
and
the
BAPS
analysis
(Fig.
3C)
recovered
congruent
results.
All
three
analyses
revealed
four
distinct
genetic
clusters,
each
constituted
by
individuals
from
one
or
two
popu-
lations.
All
individuals
from
each
of
the
populations
belonged
to
one
of
the
clusters
detected
by
BAPS
(Fig.
3C).
Clusters
I,
II
and
III
were
formed
by
individuals
from
populations
COV,
COL
and
ORE
respectively,
while
cluster
IV
included
individuals
from
the
adjacent
populations
FON
and
CAP.
No
admixture
between
genetic
clusters
was
detected
by
BAPS
(Fig.
3C).
In
the
population-based
AMOVA
(Table
2),
similar
percentages
of
variation
were
found
within
(53.42%)
and
among
(46.58%)
populations.
When
analysing
BAPS
groups,
higher
variation
was
found
among
groups
(36.55%)
than
among
populations
within
group
IV
(11.99%)
(Table
2).
The
result
of
the
Mantel
test
was
not
significant
(R2=
0.0598;
P
=
0.350).
Climatic
niche
Species
distribution
modelling
According
to
the
distribution
model
at
the
scale
of
the
west-
ern
Mediterranean
region
excluding
the
Corsican
locality
(Fig.
4A),
suitable
areas
for
Naufraga
include
northern
Majorca
(where
the
species
actually
occurs),
and
four
other
small
areas
dispersed
across
the
region:
the
adjacent
island
of
Minorca,
the
north-western
coast
of
Portugal,
the
coast
of
Kabylie
in
northern
Algeria,
and
three
islands
of
the
Italian
offshore
Tuscan
Archipelago.
No
suitability
was
inferred
in
the
Corsican
locality.
According
to
the
jackknife
test
(results
not
shown),
the
minimum
temperature
of
coldest
month
(bio6)
was
the
most
important
variable
for
the
model,
followed
by
the
temperature
seasonality
(bio4).
When
including
the
Corsi-
can
locality
in
the
model
(Fig.
S1),
the
potential
distribution
was
still
restricted
in
the
context
of
the
western
Mediterranean,
but
an
unrealistically
broad
potential
distribution
was
obtained
both
in
the
Balearic
Islands
and
Corsica.
Table
2
AMOVA
analyses
of
AFLP
genotypes
in
Naufraga
balearica
populations.
Grouping
compared
and
source
of
variation
d.f.
Sum
of
squares
Variance
components
Percentage
of
variation
Populations
Among
populations
4
1075.120
12.71029
46.58%
Within
populations
95
1384.550
14.57421
53.42%
BAPS
groups
Among
groups
3
992.645
10.35028
36.55%
Among
populations
within
groups
1
82.475
3.39504
11.99%
Within
populations
95
1384.550
14.57421
51.46%

M.
Fernández-Mazuecos
et
al.
/
Perspectives
in
Plant
Ecology,
Evolution
and
Systematics
16
(2014)
190–202
195
Fig.
4.
Maximum
entropy
distribution
models
of
Naufraga
balearica
fitted
to
current
climatic
conditions.
The
‘minimum
training
presence’
threshold
was
applied
to
the
logistic
output
of
Maxent
in
order
to
obtain
presence
(red)/absence
data.
(A)
Distribution
model
at
the
scale
of
the
western
Mediterranean
region.
Location
of
known
populations
(northern
Majorca)
is
indicated
by
a
small
box.
Other
suitable
areas
are
marked
with
arrows.
The
dotted
circle
marks
the
location
of
the
previously
reported
(but
now
extinct)
Corsican
population
(Gamisans
et
al.,
1996),
which
was
excluded
from
the
analysis
due
to
its
doubtful
spontaneity.
Notice
the
lack
of
habitat
suitability
for
this
station.
(B)
Distribution
model
at
the
scale
of
the
Balearic
Islands.
(For
interpretation
of
the
references
to
color
in
this
figure
legend,
the
reader
is
referred
to
the
web
version
of
this
article.)
In
the
distribution
model
at
the
scale
of
the
Balearic
Islands
(Fig.
4B),
the
inferred
potential
distribution
was
restricted
to
the
island
of
Majorca.
No
climatic
suitability
was
inferred
for
the
remaining
islands
of
the
Balearic
archipelago.
The
main
suitable
area
was
located
around
the
Pollenc¸
a
Bay
in
the
north
of
Majorca,
including
the
Formentor
and
Alcúdia
peninsulas.
In
addition,
a
very
small
suitable
area
was
inferred
in
one
the
highest
peaks
of
the
island,
in
the
Serra
de
Tramuntana
mountain
range
(Puig
de
Massanella,
1364
m).
According
to
the
jackknife
test
(results
not
shown),
the
precipitation
of
the
wettest
month
(bio13)
was
the
most
important
variable
for
the
model.
Microclimatic
niche
The
one-year
monitoring
of
temperature
and
relative
humidity
in
sunny
and
shady
sites
of
the
five
localities
is
summarized
in
Fig.
5
and
Table
3.
Sunny
sites
displayed
much
wider
diurnal
ranges
of
temperature
(particularly
in
spring-summer),
and
much
higher
maximum
temperatures
than
shady
sites.
Temperatures
never
fell
below
0◦C
at
shady
sites,
and
only
occasionally
at
sunny
sites.
Differences
in
the
diurnal
range
of
relative
humidity
were
also
found,
particularly
in
spring-summer,
when
lower
minimum
values
were
clearly
recorded
in
sunny
sites.
More
similar
patterns
of
temperature
and
humidity
variation
were
found
across
localities
at
shady
than
at
sunny
sites.
Discussion
Naufraga
balearica,
a
pre-Mediterranean
endemism
Continental
islands
in
the
central-western
Mediterranean
(Balearic
Islands,
Corsica,
Sardinia,
Sicily)
constitute
one
of
the
regional
hotspots
and
glacial
refugia
of
plant
diversity
within
the
large
Mediterranean
hotspot
(Médail
and
Quézel,
1997;
Médail
and
Diadema,
2009).
They
represent
both
a
reservoir
of
genetic
diversity
for
widespread
Mediterranean
plants
(Fineschi
et