United we stand, divided we fall: a meta-analysis
of
experiments
on clonal integration and its relationship
to
invasiveness
Yao-Bin
Song·
Fei-Hai
Yu
. Lidewij H.
Keser·
Wayne
Dawson·
Markus
Fischer·
Ming
Dong·
Mark
van
Kleunen
Abstract
Many ecosystems are dominated
by
clonal
plants. Among the most distinctive characteristics
of
clonal
plants
is
their potential for clonal integration (i.e. the
translocation
of
resources between interconnected ramets),
suggesting that integration may
playa
role
in
their success.
However, a general synthesis
of
effects
of
clonal integra-
tion on plant performance
is
lacking. We conducted a
meta-analysis on the effec
ts
of
clonal integration on bio-
mass production and asexual reproduction
of
the whole
clone, the recipient part (i.e. the part
of
a clone that imports
resources) and the donor part (i.e. the part
of
a clone that
Communicated by Laura Gough.
Electronic
supplementary
material
The
online version
of
this
article (doi: 10. 1007/s00442-012-2430-9) contains supplementary
material, which is available to authorized users.
Y.-B. Song ·
M.
Dong (
[gI
)
College
of
Life and Environmental Sciences, Hangzhou Normal
University, Hangzhou 310036, Zhejiang, China
e-mail: dongmingchina@126.com
Y.
-
B.
Song·
F.-H. Yu . M. Dong
State Key Laboratory
of
Vegetation and Environmental Change,
Institute
of
Botany, Chinese Academy
of
Sciences,
100083 Beijing, China
F.
-
H.
Yll
(
[gI
)
College
of
Nature Conservation, Beijing Forestry University,
100093 Beijing, China
e-mail : feihaiYll@bjfu.edll.cn
L.
H.
Keser .
M.
Fischer
Institute
of
Plant Sciences, University
of
Bern,
Altenbergrain 2 1, 3013 Bern, Switzerland
L.
H.
Keser . W.
Dawson·
M.
van Klellnen
Department
of
Biology, University
of
Konstanz,
Universitatsstrasse
10
, 78457 Konstanz, Germany
exports resources). The final dataset contained 389 effect
sizes from 84 studies covering 57 taxa. Overall, clonal
integration increased performance
of
recipient parts with-
out decreasing that
of
donor parts, and thus increased
perfoi·mance
of
whole clones. Among the studies and taxa
considered, the benefits
of
clonal integration did not differ
between two types
of
experimental approaches, between
stoloniferous and rhizomatous growth forms, between
directions
of
resource translocation (from younger to older
ramet or vice versa), or among types
of
translocated
resources (water, nutrients and carbohydrates). Clonal taxa
with larger benefits
of
integration on whole-clone perfor-
mance were not more invasive globally, but taxa
in
which
recipient parts
in
unfavorable patches benefited more from
integration were. Our results demonstrate general perfor-
mance benefits
of
clonal integration, at least in the short
term, and suggest that clonal integration contributes to the
success
of
clonal plants.
Keywords
Clonal plants · Environmental heterogeneity·
Global compendium
of
weeds . Invasiveness .
Physiological integration
Introduction
Plants with asexual reproduction (i.e. clonal plants) occur
in
many different taxonomic groups (Klimes et
a!.
1997), and
are dominant
in
many natural and man-made ecosystems
(Prach a
nd
Pysek 1994; Klimes et
a!.
1997). Moreover,
among invasive species
in
many alien floras, clonal plant
species such as
Alternanthera philoxeroides (Mart.) Gri-
seb.,
Eichhornia crassipes (Mart.) Solms, Spartina anglica
C.E. Hubbard, and Solidago canadensis L. are among the
most invasive (Pysek 1997; Lowe et
a!.
2000; Liu et
a!.
First publ. in: Oecologia ; 171 (2013), 2. - S. 317-327
Konstanzer Online-Publikations-System (KOPS)
URL: http://nbn-resolving.de/urn:nbn:de:bsz:352-202365
318
2006). Therefore,
an
important question
in
ecology and
evolution
is
what makes clonal plants so successful.
The success
of
clonal plants may be due
to
distinctive
clonal life-history traits (Tamm et al. 2002; van Kleunen
et al. 2002). One
of
these traits
is
clonal integration, i.e. the
ability to share resources such as water, carbohydrates and
mineral nutrients between interconnected clone parts (ra-
mets or groups
of
ramets; de Kroon et al. 1996; Hutchings
and Wijesinghe 1997; Herben and Suzuki 2002). Clonal
integration may be advantageous because it allows older
ramets
to
support developing younger ramets, and because
it allows support
of
clone parts growing in low-resource
patches (Alpert 1999; Herben 2004; Wang et
al.
2009).
Clonal integration should particularly benefit the recipient
ramets
of
the clone, and this could be at a cost
to
the donor
ramets (de Kroon et al. 1996; Wang et al. 2009).
If
the
benefits
to
the recipient ramets exceed the costs to the
donor ramets, clonal integration will increase the fitness
of
the whole clone (i.e. the genetic individual; van Kleunen
et al. 2000), which may give
an
advantage to clonal species
over non-clonal species. However, it
is
not clear whether
there are any patterns
in
the strength
of
the effects
of
integration among clonal plant species. Moreover, it
is
not
known whether
ef
fects
of
clonal integration are related
to
the success
of
the clonal plant species.
Studies on the effects
of
clonal integration have used two
major approaches (de Kroon et al. 1996; lonsdottir and
Watson 1997). Some studies looked at the effects
of
severing
the physical connection between different parts
of
a clone
(e.g., Alpert 1999; van Kleunen et al. 2000; Yu et
al.
2008).
This severing approach might overestimate the effects
of
clonal integration, because severing might cause physiologi-
cal stress and may make the plants more vulnerable
to
path-
ogen infections (Jonsdottir and Watson 1997). Therefore,
other studies left the connections intact but exposed the con-
ne
cted clone parts either to homogeneous or to heterogeneous
conditions (Stuefer et al. 1994; He et al. 20
II)
. This homo-
geneous-heterogeneous approach, however, may underesti-
mate the effects
of
clonal integration, because the implicit
assumption that
th
ere
is
no resource translocation under
homogeneous conditions might not hold (e.g., Wang et al.
2009). So far, it has not been tested whether measured effects
of
clonal integration depend on the experimental approach.
Effects
of
clonal integration may differ between species
with different clonal forms, i.e. between stoloniferous and
rhizomatous species. Previous studies found
th
at stolons
(i
.e. aboveground creeping stems) and rhizomes (i.e.
belowground creeping stems) have partly different func-
tions (Dong and de Kroon 1994). For instance, rhizomes
can store a larger amount
of
reserves such as carbohydrates
and proteins, and usually persist longer
in
the field than
stolons (Dong et a
l.
1996; Suzuki a
nd
Stuefer 1999).
Moreover, rhizomes can take up nutrients directly from the
soil and transport
th
em
to
the ramets (Brooker
et
al. 1999).
Furthermore, some studies found that rhizomes are less
plastic than stolons
in
response
to
changes
in
resource
availability (Dong and de Kroon 1994; de Kroon and
Hutchings 1995),
and-if
this
is
a general
pattt
~
rn
-
rhizo
matous species may be more reliant on clonal int
egrat
ion as
a means
of
coping with environmental heterogeneity
th
~
\Il
stoloniferous species. For these reasons, one
might
expect
that rhizomatous species have a higher capacity for clonal
integration than stoloniferous species.
Effects
of
clonal integration may also depend on the
potential direction
of
resource translocation, i.e. whether it
is
from developmentally older
to
developmentally younger
rame
ts
(acropetal) or from developmentally younger
to
developmentally older ramets (basipetal) (Marshall et al.
1990; lonsdottir and Watson 1997). Studies using labeled
resources have shown that acropetal translocation is usually
more common than basipetal translocation within a plant
(Marshall et al. 1990; de Kroon et a
l.
1996;
D'
Hertefeldt
and lonsdottir 1999). The latter most likely reflects that
clonal integration in many plants primarily provides support
to establishing daughter ramets. However, this does not
necessarily mean that effects
of
clonal integration on plant
performance will be larger when
re
sources are translocated
acropetally (Stuefer et al. 1994). Moreover, the effect
of
integration could depend on the resource (i.e. water, nutri-
ents or carbohydrates) that
is
being translocated (van
Kleunen and Stuefer 1999). So far, it has not been tested
whether the effects
of
integration depend on
the
kind
of
resources and the direction
of
resource translocation.
The success
of
clonal plants
is
not only apparent from
their high frequencies
in
native floras but also
in
the fact that
many invasive alien plants, particularly
in
(semi-)natural
areas, are clonal, and that many
of
the most invasive plants
are clonal (Pysek 1997; Lowe et al. 2000; Liu et al. 2006).
This suggests that clonal life-
hi
story traits may
playa
key
role
in
plant invasiveness (Yu et al. 2009; Aguilera et al.
2010; Xu et al. 2010). Although this hypothesis has been
posed repeatedly, there have so far been no explicit tests.
If
clonal integration is usually advantageous and results
in
larger size and more offspring, one might expect that spe-
cies with higher levels
of
integration are more successful
than species with lower levels
of
integration, and thus are
also more likely to have become invasive.
We conducted a meta-analysis
of
existing experimental
studies on
th
e effects
of
clonal integration to address the
following specific questions.
(I)
Are the effects
of
clonal
integration generally positive? (2) Are effects
of
il1tegra-
tion estimated
in
experiments using a severing treatment
larger than when estimated using comparisons between
plants growing
in
heterogeneous a
nd
homogeneous envi-
ronments? (3) Are effects
of
integration larger for rhizo-
. matous plants than for stoloniferous plants? (4) Are the
effects
of
integration larger for acropetal translocation than
for basipetal translocation? (5) Do effects
of
integration
depend on the kind
of
resource being translocated? And (6)
are species with larger benefits
of
integration more invasive
at a global scale?
In
this meta-analysis, we only considered
the short-term benefits and costs associated with clonal
integration. We could not address the long-term benefits
and
costs-for
example, due
to
possible trade-offs between
sharing resources among ramets and storage (e.g., Poor
et
al.
2005)-because
most experimental studies
of
clonal
integration lasted only a few months.
Materials and methods
Selection
of
studies on clonal integration
To collect studies that quantified the effects of clonal inte-
gration,
we
searched lSI Web
of
Science (http://apps.
isiknowledge.com) on 20 April 2011, using the keyword
combinations 'clon* integration' and 'physiolog* integra-
tion'. We also added studies listed in the references of the
articles we obtained, including unpublished data
in
theses
and non-English articles that did not appear in lSI Web
of
Science. In total,
we
found
158
studies covering
98
taxa.
We used the following criteria to select among these
158
studies the experimental ones that were suitable for meta-
analysis.
(1) We only included studies that reported means,
sample sizes and standard errors or standard deviations for
treatment and control, because these data are required
to
calculate effect sizes
of
clonal integration and their vari-
ances; and (2)
we
only included studies that reported
effects
of
integration on biomass production and/or asexual
reproduction (i.e. number
of
ramets). These are common
measures
of
fitness
in
clonal plants (Pan and Price 200 I).
We did not include data
on
effects
of
integration on sexual
reproduction because these studies were too scarce. The
final dataset contained 389 cases from 84 studies covering
57 taxa (see Electronic supplementary material, ESM S I).
Data extraction and calculation
of
effect sizes
We extracted data on biomass and asexual reproduction for
the whole clone, the potential recipient part, and/or the
potential donor part. Where possible, we distinguished
between whether the potential recipient and donor parts
were the developmentally older or younger parts
of
the
clone. When extracting biomass data, our first choice was
to
use data
on
total biomass.
If
instead
of
total biomass the
biomasses
of
different plant parts (e.g., root mass and mass
of leaves) were reported,
we
extracted these data, and later
pooled the effect sizes
of
each
of
these parts per species
and study (see below).
319
We extracted means, sample sizes and standard devia-
tions or standard errors for biomass and asexual repro-
duction
of
the whole clone, and the potential recipient and
donor parts. We extracted the data directly from tables and
the main text or from graphs using the program ImageTool
(http://ddsdx.uthscsa.edu/diglitdesc.html).
Where
neces-
sary, we obtained data directly from the corresponding
authors
of
the papers. We also extracted from each paper
data
on
the duration
of
the experiment (ESM S
l).
Effect sizes are based
on
the difference between a 'treat-
ment' and a 'control'. To get positive signs
of
effect sizes for
benefits
of
integration, we considered treatments
in
which
resource translocation was possible or most likely
as
'treat-
ment', and treatments
in
which resource translocation was
prevented or unlikely
as
'control' .. This means that, for
studies that used the severing approach, the treatment with
intact connections (i.e. with resource translocation possible)
was used
as
'treatment' and the treatment with severed
connections (i.e. with resource translocation not possible)
as
'control' . For studies that left all connections intact but used
the heterogeneous versus homogeneous approach,
we
con-
sidered the heterogeneous treatment (i.e. resource translo-
cation
is
likely)
as
'treatment' and the homogeneous
treatment (i.e. resource translocation is not likely)
as
'con-
trol'. The latter approach could only be used to look at effects
of
integration for potential recipient clone parts (i.e. clone
parts exposed to relatively low local resource availability)
and donor clone parts (i.e. clone parts exposed to relatively
high local resource availability). Here,
we
only compared
clone parts
of
the same developmental stage and with the
same local conditions, where one part was connected
to
another clone part experiencing the same conditions (i.e.
homogeneous treatment) and the other one to a clone
part experiencing different conditions (i.e. heterogeneous
treatment).
We used pairs
of
'treatment' and 'control' as defined
above to calculate effect sizes (Hedges'
d) and their vari-
ances following Rosenberg et
al.
(2000). These calcula-
tions were performed with the software Meta Win,
v.2.1
(Rosenberg et al. 2000). When a study included separate
data for more than one genotype or environment, we first
calculated effect sizes and their variances for each pair
separately. To avoid pseudo-replication, we pooled effect
sizes and their variances
of
different genotypes or envi-
ronments per species and study by doing a separate meta-
analysis
on
these effect sizes (sensu Leimu et
al.
2006), and
used the pooled effect sizes
in
the final analyses.
Global invasion success
of
study species
As studies usually did not refer to the invasion success of
their study species, we used the Global Compendium of
Weeds (GCW; Randall 2002) to get estimates
of
global
320
invasiveness for our species. The GCW
is
a list
of
plant
species (over 28,000 taxa) that have been cited
in
primary
literature, floras, and government reports (a total
of
300
references) as weeds
in
a given location (Randall 2002).
The number
of
these references
in
the GCW that list a
species as being a weed has been used as a continuous
index
of
global invasiveness
in
recent studies (Pysek et al.
2009; Schlaepfer et al. 20
10;
Dawson et al. 20
II;
Jenkins
and Keller 20
II).
Although the accuracy
of
the GCW has
been criticized (Richardson and Rejmanek 2004), it
is
the
best available and most comprehensive database on global
invasiveness
of
plants (Pysek et al. 2009). Because some
global regions may have a greater bias
in
recording alien
invasive species, the number
of
references in the GCW
may not clearly indicate the global level
of
species inva-
siveness (Dawson et al. 20
II)
. Therefore, we also used
another proxy
of
invasiveness proposed by Dawson et al.
(2011), which
is
the number
of
global regions containing a
GCW reference. We
log+
I-transformed numbers
of
GCW
references and numbers
of
GCW regions prior to analysis.
Data analysis
All analyses were conducted with the software MetaWin,
v.2.1 (Rosenberg
et
al. 2000).
In
the final analyses, we
deleted one outlier with an extremely high effect size.
However, before doing so, we confirmed that the results
were qualitatively the same when including this outlier
in
the analyses. We used the random-effects model setting,
which assumes that the differences among studies are not
only due to sampling error but also due to true random
variation, as
is
the default for ecological data (Gurevitch
and Hedges 2001).
For each mean effect size, we calculated the bias-cor-
rected
95
% bootstrap confidence intervals based on 4,999
permutations (Adams et al. 1997). Using this method, a
mean effect size
is
significantly different from zero when
its 95
% confidence interval does not include zero. For the
recipient and the donor parts, we compared mean effect
sizes between different categories (i.e. experimental
approaches, clonal forms, directions
of
resource translo-
cation and kinds
of
resources being translocated). We first
used Chi-squared tests
to
assess whether heterogeneity
among effect sizes
(Qtotal) was significantly larger than the
expected sampling error. Then, we calculated heterogene-
ity
in
effect size between categories (Qb) and within groups
(Qw), and tested their significances with a randomization
test and a Chi-squared test, respectively (Rosenberg et al.
2000). A significant
Qb suggests that the categorical vari-
able explained a significant part
of
the heterogeneity, while
a significant
Qw
implies that there was still heterogeneity
in
effect sizes among studies not explained by the categorical
variable (Rosenberg et al. 2000).
For the whole clones, we could only use
data
based on
the severing approach. The resulting dataset allowed us
to
compare mean effect sizes between different clonal forms,
but not between directions
of
resource translocation or
kinds
of
resources being translocated.
To
assess whether
effect sizes
of
clonal integration were associated with
global invasiveness, we tested whether the slope
of
a
regression with global invasiveness as
an
explanatory
variable was significantly different from zero with a ran-
domization test (Rosenberg et al. 2000).
To examine whether there was evidence for a publica-
tion bias (i.e. whether significant findings have a greater
chance
to
be published than non-significant ones), we
inspected funnel and normal-quantile plots, performed
Spearman's rank-order correlation tests, and estimated fail-
safe numbers (Palmer 1999; Jennions and MI/lller 2002), as
implemented
in
MetaWin. The scatter plot
of
effect s
iz
e
versus sample size showed a funnel-shaped distribution
without any obvious underrepresentation
of
effect sizes
(ESM S2A, S3A). The normal quantile plot showed that
effect sizes were normally distributed (ESM
S2B;
Rosen-
berg et al. 2000). Moreover, effect sizes were not corre-
lated with sample sizes
(r
= 0.063, P = 0.214). Finally,
the fail-safe number (i.e. number
of
studies that would have
to
be added
to
change the results
of
the meta-analysis from
significant to non-significant) was 24,531, which was much
bigger than 1,955 (i.e.
5n +
10;
n being the number
of
cases in our dataset; Rosenthal 1979; Rosenberg et al.
2000). Together, these plots and statistical tests indicate
that there was no publication bias.
As
an
alternative to Hedges'
d,
we also calculated log-
response ratios. However, because the normal-quantile plot
showed that log-response ratios were not normally distributed
(ESM
S3B)
, we present the results
of
analyses
ba
sed on
Hedges'
d.
Nevertheless, the results based on log-response
ratios were qualitatively similar, and are presented in ESM
S4.
Because clonal growth characteristics and invasion success
are likely to be taxonomically biased, we also tested whether
effect sizes varied significantly among families. Indeed, effect
sizes
of
biomass varied significantly among families for
recipient parts and donor parts, and effect sizes
of
asexual
reproduction varied significantly among families for whole
clones (ESM
S5, S6). Therefore, to test whether the results are
consistent within families, we repeated all analyses for the two
families with > 10 studies (Poaceae and Rosaceae).
Results
Effects
of
integration on performance
of
whole clones
Mean effect sizes for biomass
of
the whole clone were
significantly greater than zero (Fig. I a), indicating that
clonal integration generally increased biomass
of
the whole
clone. However, clonal integration did not significantly
increase effect sizes for asexual reproduction
of
th
e whole
clone (Fig. I
b)
. There were
no
significa
nt
differences
in
effect sizes between stoloniferous and rhizomatous plants
(for biomass
Qb
= 0.1
2,
P = 0.750; for asexual repro-
duction
Qb
= 0.1
2,
P = 0.753; ESM S7). The results were
similar within the two largest families, Poaceae and Ro
s-
aceae (ESM S8, S9), which indicates that the results were
robust with
re
spect
to
taxonomy.
Effects
of
integration on performance
of
potential
recipient parts
Mean effect sizes for biomass and asexual reproductions
of
the potential recipient parts were significantly greater than
zero (Fig. Ic, d), indicating
th
at recipient parts benefited
from integration. These benefits
of
clonal integration
to
recipient parts did not differ between acropetal and basip-
etal directions (for biomass
Qb
=
1.61
, P = 0.237; for
32 1
asexual reproduction
Qb
= 1.43, P = 0.533), between
stoloniferous and rhizomatous species (for biomass
Qb
= 0.20, P = 0.646; for asexual reproduction '
Qb
=
1
.25
, P = 0.290), between
th
e sever
in
g approach and
homogeneou
s-
heterogeneous approach (for biomass
Qb
=
2.
1
4,
P = 0.
14
3; for asexual reproduction
Qb
= 2.85,
P = 0.117),
or
among different resources be
in
g translocated
(i
.e.
carbohydrates, nutrie
nt
s and water; for
bioma
ss
Qb
=
5.24, P = 0.078; for asexual reproduction
Qb
= 3.95,
P = 0.188; ESM
S7).
There were some minor
dev
iations
from the overall pattern within
th
e two largest families, Po-
aceae and Rosaceae, but overall
th
e results were
quit
e robust
with respect to taxonomy (ESM
S8, S9).
Effects
of
integration on performance
of
potential
donor parts
Mean
ef
fect sizes for biomass and asexual reproduction
of
the donor parts were not sig
nifi
cantly different from zero
(Fig. Ie,
f)
, indicating that clonal integration usually did
Biomass
Asexual
reproduction
Fig. 1 Mean effect sizes
(Hedges'
d)
of
clonal
integration for biomass and
asexual reproduction
of
the
whole clone
(a, b), the potential
recipient
cl
one parts (c,
d)
and
the potential donor clone part
(e, O. The bars around the
means
denote bias-corrected
Overall (a)
if-O-f
1
K>-l
(41,21)
(16,
12)
(25
,9
)
(b)
~
(25,14)
95
% bootstrap confidence
intervals, a
nd
a mean effect size
is sig
nifi
cantly different from
zero when
it
s
95
% confidence
interval does n
ot
include zero.
The
first a
nd
seco
nd
numbe
rs
in
parentheses
are number
of
studies and number
of
species,
respectively
Rhizomatous
Stoloniferous
1 I--O--i
-3 -2 -1
023
Overall (c)
I
11-0-1
(69,31)
Basipetal
Acropetal
Rhizomatous
Stoloniferous
Severing
vs. non-severing
Heterogeneous
vs. homo.
Carbohydrates
Nutrients
Water
If-O-I
(20,
14)
: 1-0-1 (37, 22)
If-O-l
.1
1-0-1
1
1
1-0-1
11-0-1
1
(17, 13)
(52,
18)
(3
2,22
)
(37,
16)
1
1
I--O-l
(19,11)
.
1~
(15
,
8)
f--
~~_
I-7O-I
+-----'--
--r..:.
(
-,
12,
8)
-3 -2 -1 0
2 3
t:
(10,
8)
(15,
6)
f----.-
--r-t---
--.---.-
~
-3
-2
-1 0 2 3
I
(d) 1
1-0-1
1
f-j-O---j
1 1--01
1
1 1-0-1
11-0-1
~~
1
1
I-O--i
11-0--1
~
(45
,2
1)
(12,7)
(25, 16)
(10,8)
(35,13)
(25, 19)
(20,
8)
(9
,5
)
(12,
6)
(4
,3
)
~
-'--r--+--
.-~~
-3
-2
-1 0
2 3
Overall (e)
(82,
30)
(f)
(54, 19)
(31
, 14)
(17,9)
Basipetal
Acropetal
Rhizomatous
Stoloniferous
Severing
vs. non-severing
Heterogeneous
vs. homo.
Carbohydrates
Nutrients
Water
-3 -2 -1 0
(42,
19)
(26,
16)
(21
,
12)
(61
,
18)
(30
,19
)
(52
,2
1)
(16,8)
(21,
11)
(15,7)
2 3
-3
-2 -1 0
Effect
size
(12,6)
(42, 13)
(2
2,14
)
(3
2,14
)
(8,4)
(14
,6
)
(11
,4
)
2 3