Flora 224 (2016) 42–49
Contents lists available at ScienceDirect
Flora
journal homepage: www.elsevier.com/locate/flora
Fragmentation and environmental constraints influence genetic
diversity and germination of Stipa pennata in natural steppes
a,∗ a,b a,c a
Steffen Heinicke , Isabell Hensen , Christoph Rosche , Dennis Hanselmann ,
d e b,f
Polina D. Gudkova , Marina M. Silanteva , Karsten Wesche
a
Institute of Geobotany and Botanical Garden, Martin-Luther University of Halle-Wittenberg, Am Kirchtor 1, 06108 Halle (Saale), Germany
b
German Centre for Integrative Biodiversity Research (iDiv) Halle-Jena-Leipzig, Deutscher Platz 5e, 04103 Leipzig, Germany
c
Department of Botany, Faculty of Science, Charles University in Prague, Benátská 2, CZ-128 01 Prague, Czech Republic
d
Institute of Biology, Tomsk State University, Lenina Avenue 36, 634050 Tomsk, Russia
e
Institute of Botany, Altai State University, Prospekt Lenina 61, 656 049 Barnaul, Russia
f
Botany Department, Senckenberg Museum of Natural History Görlitz, PO Box 300 154, D-02806, Görlitz, Germany
a r t i c l e i n f o a b s t r a c t
Article history: Human impact and fragmentation often have negative effects on plant population sizes. This can lead
Received 24 November 2015
to declining genetic diversity due to restricted gene flow and genetic bottlenecks, and eventually result
Received in revised form 5 June 2016
in reduced reproductive fitness. Environmental conditions can also influence the genetic structure of
Accepted 6 June 2016
populations and directly affect their reproduction success.
Edited by W. Durka
For Stipa pennata, the key species of largely natural steppes in southern Siberia, using AFLP we tested
Available online 18 June 2016
whether genetic variability and germination are negatively influenced by fragmentation, and assessed
the influence of local environmental conditions. Genetic diversity was moderately high (mean percent-
Keywords:
Adaptation age of polymorphic bands = 38.4%), with high genetic differentiation occurring between populations
AFLP ( ST = 0.547). Genetic variation was mainly partitioned (41.8%) between two distinct grassland types.
Kulunda-steppe Isolation negatively affected genetic diversity, highlighting that fragmentation had an impact on genetic
Population-genetic structure structure. Higher mean precipitation negatively influenced population size, population density and
Precipitation genetic diversity. The speed of seed germination was correlated positively with population size and neg-
Vegetation
atively with vegetation cover, while we found no evidence for negative effects of low genetic diversity on
percentage of seed germination. The presence of different genetic groups shows that populations have
adapted to a range of environments. Germination speed also differed between groups, as a consequence
of maternal effects or of adaption to certain environmental conditions.
Our results show that fragmentation can have potentially strong effects even in natural grasslands. We
recommend that any future restoration schemes take the observed pronounced genetic differentiation
into account.
© 2016 Elsevier GmbH. All rights reserved.
1. Introduction to reduced reproductive fitness (Ellstrand and Elam, 1993; Hensen
and Oberprieler, 2005; Leimu et al., 2006). However, the rele-
In a growing number of threatened species, human medi- vance of such population-wide genetic processes to persistence
ated fragmentation significantly constrains population sizes and and conservation in endangered populations remains a topic of
increases among-population isolation due to restricted gene flow much debate (Vernesi et al., 2008). Biotic and abiotic factors can
(Allendorf et al., 2012; Eckert et al., 2008; Hoffmann and Willi, also have an impact on genetic diversity in populations (Hamasha
2008). Fluctuations, colonisation events and genetic bottlenecks et al., 2013; Huebner et al., 2009; Wang et al., 2006), mainly as
(Castric and Bernatchez, 2003) may result in an increased risk of a result of adaptation to environmental conditions. For example,
inbreeding, genetic drift and accumulation of deleterious muta- increased drought stress led to increased genetic diversity in Stipa
tions (Dudash and Fenster, 2000; Frankham et al., 2002; Young krylovii (Zhao et al., 2006) and Hordeum spontaneum (Nevo et al.,
et al., 1996), while diminishing genetic diversity is known to lead 1998), while similar habitats can host genetically similar popu-
lations (Hamasha et al., 2013; Zhao et al., 2004). A given species’
adaptive potential to particular environmental conditions should
∗ depend heavily on the level of within-population genetic diver-
Corresponding author.
sity, which can in-turn have a direct bearing on the reproduction
E-mail address: steffen [email protected] (S. Heinicke).
http://dx.doi.org/10.1016/j.flora.2016.06.003
0367-2530/© 2016 Elsevier GmbH. All rights reserved.
S. Heinicke et al. / Flora 224 (2016) 42–49 43
Fig. 1. Global distribution range of Stipa pennata and location of the study populations within the Kulunda steppe.
and long-term survival of the population (Bauert et al., 1998; Brook et al., 1965–1992). Its populations have become increasingly frag-
et al., 2002; Potvin and Tousignant, 1996). mented throughout their European range (Fig. 1), resulting in
Germination is a key stage in reproduction, and is influenced reduced levels of genetic diversity (Durka et al., 2012; Hensen
by factors including genetic diversity and/or environmental con- et al., 2010). For S. pennata, high genetic differentiation in its Cen-
ditions (Baskin and Baskin, 1998; Gasque and García-Fayos, 2003). tral European range edge is associated with decreasing population
Germination behaviour and seed viability are both dependent on sizes and related to populations’ degree of isolation (Wagner et al.,
species’ capacity to adapt to changing environmental conditions 2012). Moreover, Stipa species are known to develop cleistogamous
(Fenner and Thompson, 2005; Ronnenberg et al., 2008). As such, flowers, especially under drier conditions or in more arid regions
widely-distributed species in particular show local differentiation (Brown, 1952; Ronnenberg et al., 2011), and cleistogamous flow-
in germination behaviour (Fenner and Thompson, 2005). Seed ger- ers of Stipa leucotricha produce a higher ratio of viable seeds than
mination is also limited by the environmental conditions required chasmogamous flowers (Call and Spoonts, 1989).
to break dormancy (Baskin and Baskin, 2004), and unfavourable Working along a regional climate gradient, we tested (1)
climatic conditions can negatively influence germination in certain whether fragmentation of natural steppes in recent decades has
plants and lead to diminishing population sizes in the long-term, had a negative effect on genetic diversity and on germination, as
which in turn can lead to reduced genetic diversity (Montesinos a proxy for reproductive fitness. Furthermore, we examined (2)
et al., 2009). whether genetic structure as well as germination differs among dif-
A large share of studies on genetics and germination of grass- ferent steppe habitats. We additionally investigated (3) whether
land species originates from European grasslands that have been the regional aridity gradient spanning from the wetter north-
fragmented for long periods of time and are largely secondary eastern to the drier south-western part of Siberia is associated with
in nature. Genetic structures in novel and peripheral ranges may increasing genetic diversity within populations and, consequently,
differ much from those in the natural range (Durka et al., 2012; increased germination.
Hensen et al., 2010; Wagner et al., 2012), and effects of frag-
mentation may thus also differ (Hampe and Petit, 2005). Many 2. Materials and methods
European grassland species have native ranges in temperate grass-
lands, such as steppes and prairies, which cover approximately 8% 2.1. Study species
of the global terrestrial surface (White et al., 2000). The steppes
of Eurasia are among the world’s largest continuous land biomes. The original description of Stipa pennata by Linnaeus (1753)
Gradients in aridity, groundwater availability and edaphic factors was ambiguous, resulting in different species, such as Stipa joannis
can cause both small- and large-scale differentiation among natural (Celakovskˇ y),´ occasionally being included under the name. After
steppe species (Boonman and Mikhalev, 2005; Box, 2002; Dieterich, Freitag (1985) had determined a new lectotype, S. pennata was
2000). Steppes have, however, been subject to large scale conver- accepted as a separate taxon.
sion (Henwood, 1998), which has resulted in severe fragmentation Stipa pennata is a typical grass species of steppes, meadow
throughout their distribution range. The Kulunda steppe of south- steppes or forest steppes (Nosova, 1975). It is perennial and
2
western Siberia covers an area of approximately 80,000 km , and tussock-forming and is commonly described as tetraploid (2n = 44;
it is a striking example of an ecosystem that has undergone wide- Krasnikov, 1991; Sheidai et al., 2006), which is probably the result
ranging land use change. Due to the rising demand for agricultural of a hybridization event (Johnson, 1945). Lemmas typically have
crops, approximately 80% of the natural steppes have been trans- an up to 30 cm long bigeniculate awn, which is covered in feath-
formed for cultivation (Hoekstra et al., 2005; Meinel, 2002; Wein, ery hairs. Caryopses (hereafter referred to as seeds) are dispersed
1985). As a consequence, native species have become increas- either by wind or via animal skin or fur.
ingly threatened, with many being included in the national Red-list
(Aleksandrova et al., 2006; Solotov, 2009).
2.2. Study area
Here, we focus on the feather grass Stipa pennata, which is a
characteristic biogeographical element of temperate western and
The Kulunda steppe represents the north-eastern distribution
middle Eurasian steppes (Lavrenko and Karamysheva, 1993 Meusel
range of S. pennata (Fig. 1). The region has been influenced by agri-
44 S. Heinicke et al. / Flora 224 (2016) 42–49
culture since the 1740s, when arable farming was introduced in tions F1–F9) contained all populations that had a grass cover of at
the north and subsequently spread southwards. Steppe conver- least 50% (mainly Stipa and Festuca species). The meadow steppe
sion accelerated in the 1950s (Meinel, 2002; Silanteva, personal group (Table 1, Fig. 1, populations M1–M12) in contrast, was char-
communication), while in the 1990s low grain prices resulted in acterized by lower grass cover and a higher proportion of forbs.
large-scale fallowing of former fields, especially in the south. Stipa Population sizes and densities were estimated by counting the tus-
2
species can, however, require up to 10 years before the first indi- socks (per m , then extrapolated), and ranged between 140 and
viduals become established and begin to recolonize former fields 250,000 in M4 and F5, respectively (Table 1). Seeds were collected
(Boonman and Mikhalev, 2005). from six sites of the feather grass steppe group and five sites of the
The Kulunda steppe is characterised by declining precipita- meadow steppe group, and pooled from about 20 sampled indi-
tion towards the south (Bergmann and Frühauf, 2011). In the viduals across a given population. The total annual precipitation at
wetter northeast (annual precipitation 350–500 mm), degraded each site was obtained from the WorldClim model (Hijmans et al.,
chernozems underlie extensive forests which, travelling south- 2005). Populations were classified based on their degree of isola-
wards, give way to forest steppes with mosaics of steppe elements tion, which was used as a proxy for anthropogenic intervention. A
interspersed with groups of trees (Boonman and Mikhalev, 2005). high degree of isolation was assumed for populations surrounded
Further south, along a decreasing precipitation gradient, highly by barriers, such as forests or cultivated fields, and with no possi-
diverse meadow steppes overlying black chernozems become more bility for establishment of new individuals in the adjacent area.
dominant (Ronginskaja, 1963; Solotov, 2009), followed by typical Isolation was classified as medium if barriers were present but
steppes, which are considered the main habitat of Stipa species potential establishment sites existed in the adjacent area, such
and are defined as “feather grass steppe” (Boonman and Mikhalev, as roadsides, while isolation was considered low if there were no
2005). apparent barriers (Table 1).
Due to the high level of agricultural conversion of the Kulunda
region, natural steppe vegetation now is almost exclusively
restricted to areas of less interest to agriculture, i.e. roadsides, field 2.4. AFLP genotyping and analyses
edges, lake and river terraces and small refugia (Lapschina, 1992;
Solotov, 2009). We employed AFLP markers to facilitate direct comparisons
with previous studies on genetic variation of Stipa species. Extrac-
tion of DNA and AFLP analysis followed the protocol of Hensen
2.3. Sampling et al. (2012) with the following primer combinations being
employed: AAG (FAM)-CCA, AGC (HEX)-CCA and AGC (HEX)-CAA
TM
In summer 2012, we collected leaves from 21 S. pennata popu- (Wagner et al., 2012). We analyzed samples using a MegaBACE
lations across the Kulunda steppe, with minimum and maximum 1000 sequencer (Amersham Bioscience, UK) and visualized peaks
distances between populations equalling six and 332 km, respec- between 60 and 400 bp in length with the program MegaBACE
tively (Fig. 1). A population was defined as a group of individuals Fragment Profiler (Amersham Bioscience, 2003). For populations
separated by at least 1 km from another group. For each popula- M5 and M9, only 11 samples per population were used for further
tion, the 12 sampled individuals were distributed equally across analysis due to the limited quality of the band profiles.
the entire population area and were at least 1 m apart from one Using these binary data matrices, polymorphic DNA bands were
another. Sites were described by conducting vegetation surveys, given scores of “0” (absence) or “1” (presence). To determine
which then enabled us to distinguish between two habitat groups the most reliable of the automatically scored loci, we followed
(Heinicke, 2014): The feather grass steppe group (Table 1, popula- the approach of Ley & Hardy (2013) and compared automatically
Table 1
Overview of the collected S. pennata populations. Pop: populations, F: feather grass steppe group, Populations M: meadow steppe group; Geographic coordinates in decimal
degrees (LAT: latitude; LONG: longitude); P: annual precipitation (mm); n: number of individuals included in the analysis; S: population size (estimated number of adult
2
individuals); D: population density (tussocks per m ); I: spatial isolation, H: highest, M: middle, L: lowest; He: expected heterozygosity; PPB: proportion of polymorphic
bands; G: final germination percentage (mean ± s.e.; n = 4); TI: Timson index (mean ± s.e.; n = 4).
Pop LAT LONG Region P n S D I He PPB G TI
F1 53.554 81.57 Kamenskij Rajon 388 12 50400 7 M 0.120 30.2 55 (±15) 17 (±5)
F2 53.434 78.955 Burlinskij Rajon 317 12 42000 6 L 0.208 79.7 – –
F3 53.228 79.788 Suetskij Rajon 323 12 24500 7 L 0.198 75.7 38 (±23) 15 (±9)
F4 53.108 79.832 Blagoweschtschenskij Rajon 318 12 30000 6 L 0.173 40.6 – –
F5 53.026 80.006 Blagoweschtschenskij Rajon 321 12 250000 10 H 0.100 28.7 35 (±19) 14 (±7)
F6 52.78 79.732 Blagoweschtschenskij Rajon 306 12 210000 7 H 0.104 30.7 38 (±14) 14 (±4)
F7 52.96 81.34 Sawjalowskij Rajon 385 12 195000 13 M 0.174 42.1 45 (±8) 6 (±2)
F8 53.249 83.941 Perwomajskij Rajon 440 12 30000 5 M 0.098 27.7 28 (±16) 14 (±7)
F9 52.646 82.152 Alejskij Rajon 449 12 2400 3 H 0.078 27.7 – –
M1 52.1 80.219 Woltschichinskij Rajon 319 12 9500 5 M 0.154 39.6 20 (±8) 8 (±3)
M2 52.776 82.658 Toptschichinskij Rajon 468 12 27000 3 M 0.109 30.7 40 (±12) 7 (±4)
M3 53.493 81.917 Schipunowskij Rajon 446 12 240 2 M 0.12 31.7 19 (±17) 19 (±7)
M4 52.815 81.752 Mamontowckij Rajon 421 12 140 1 H 0.098 30.7 18 (±7) 10 (±4)
M5 53.39 82.073 Schelabolichinskij Rajon 428 11 39000 3 M 0.114 31.2 – –
M6 53.351 81.614 Tjumenzewskij Rajon 399 12 51200 4 M 0.162 41.6 30 (±11) 12 (±5)
M7 52.872 81.612 Mamontowckij Rajon 410 12 11000 2 M 0.148 31.7 – –
M8 52.892 80.531 Blagoweschtschenskij Rajon 342 12 18000 3 M 0.191 63.4 – –
M9 52.206 81.635 Nowitschichinskij Rajon 434 11 16500 3 H 0.091 29.7 – –
M10 52.906 83.453 Kalmanskij Rajon 451 12 480 3 M 0.102 31.2 – –
M11 52.627 82.253 Altajskij Rajon 460 12 1500 2 M 0.103 31.2 – –
M12 52.627 82.344 Altajskij Rajon 467 12 1400 1 M 0.125 29.7 – –
Mean F 361 92700 7 0.139 42.6
Mean M 420 14663 3 0.126 35.2
Mean Total 395 48108 5 0.132 38.4
S. Heinicke et al. / Flora 224 (2016) 42–49 45
Table 2
Overview of the statistical models. PPB: percentage of polymorphic bands; TI: Timson’s index; LM: linear models; LMM: linear mixed.
Dependent variables Statistical models Explanatory variables Random effects No. populations
PPB LM log (population size) – 21
log (TI) LMM log (population size), PPB population 11
PPB LM vegetation cover*precipitation – 21
log (TI) LMM vegetation cover*precipitation population 11
scored genotypes of replicates in SPAGeDi 1.4 software (Hardy and in darkness (Hensen, unpublished data; Ronnenberg et al., 2008).
Vekemans, 2002). Based on a dataset containing replicated individ- Twenty-five seeds each were placed in four Petri dishes per popu-
uals only (28.4% of 250 individuals), the broad-sense heritability of lation and incubated in a germination chamber under the following
◦ ◦
each marker was estimated as an FST value. Significance of repro- conditions: 20 C/10 C at day/night in a 12 h cycle (in accordance
ducibility for the FST values was tested by executing 1000 random with Hensen and Hoffmann, unpublished data) and kept constantly
permutations. Markers with FST > 0.5 and P < 0.05 were selected for moist with demineralized water. Seeds were checked 10 times over
further statistical analysis. The final dataset included 202 polymor- a 27 day period, with germinated seeds being removed as appro-
phic bands, with an error rate of 2.2%, which is within the expected priate. Final seed germination was determined in %. The modified
range of AFLP analysis (2–5%; Hansen et al., 1999). Timson Index (TI) was used to estimate germination speed, and was
Since different approaches exist for analysing AFLP data calculated as the cumulative sum of all germinated seeds divided
(Bonin et al., 2007), we determined key parameters follow- by the overall germination period (Pérez-Fernández et al., 2006).
ing: (1) Band-based approaches by counting polymorphic bands The Kruskal-Wallis test was also used to test whether the classi-
(PB), estimating the percentage of polymorphic bands (PPB) and fication of the degree of isolation had an impact on the reproductive
the Sørensen dissimilarity between individuals, and (2) Allele- fitness of S. pennata. We used the Mann-Whitney U Test to deter-
frequency approaches for estimating Neiıs´ gene diversity (He) and mine significant differences between the grassland types.
pairwise ST distance between populations, respectively. How-
ever, we were not able to precisely determine allele frequencies
2.6. Congruence between data on genetics, germination and
due to the dominance of the AFLP marker system and the poly-
environmental conditions
ploidy of S. pennata. As with previous studies on polyploid species
(e.g. Durka et al., 2012; Wagner et al., 2011), allele frequency was
A further Mantel test (9999 repetitions) was conducted with the
estimated as being equal to band frequency.
pairwise FST distance matrix and a dissimilarity matrix of the vege-
We estimated PB, PPB and He in accordance with Lynch and
tation data (Bray-Curtis dissimilarity) to examine whether genetic
Milligan (1994) using AFLP-SURV 1.0 (Vekemans et al., 2002). For
differentiation was associated with the plant community com-
further analyses, we consistently used the PPB as a parameter for
position. Correlations between population size, density, PPB and
genetic diversity, since its calculation is not based on any assump-
precipitation were calculated using Pearson’s r.
tion, in contrast to He (the calculation of which is based on allele
Effects of population size, precipitation and vegetation cover on
frequencies that can only be estimated with dominant markers
genetic diversity as well as on germination were tested with lin-
such as AFLP). Since the data of the tested genetic parameters
ear models and linear mixed effects models (Table 2) using the R
(He and PPB) between the grassland types were not normally
packages “MASS” (Venables and Ripley, 2002) and “nlme” (Pinheiro
distributed according to a Shapiro-Wilk test, we used the Mann-
et al., 2013). Further models were used to examine the relation-
Whitney U Test to assess the level of significance. Effects of degree
ship between genetic diversity and germination. Population sizes
of isolation on gene diversity where analysed with Kruskal-Wallis
and TI were log-transformed to meet assumptions of normality. In
tests.
order to develop the minimum adequate model, all non-significant
To calculate the partitioning of genetic variation within and
variables were incrementally excluded using a backward approach.
between populations, we performed Analyses of Molecular Vari-
Models of different complexity were compared using a Chi-square
ance (AMOVA, based on ST-distance, Excoffier et al., 1992) using
test for the LMM and an F-test for the LM, while parameter esti-
Arlequin 3.5 (Excoffier and Lischer, 2010). We included the two
mates were separately determined.
grassland types as an additional hierarchical level in the AMOVA.
Pairwise FST-values among populations and their significance
3. Results
(10,000 permutations) were determined using AFLP-SURV 1.0. Fur-
ther AMOVAs were calculated for three subsets of data comprising
3.1. Genetic diversity and differentiation
populations from the three levels of isolation. We tested for genetic
isolation by (geographical) distance with a Mantel test using the
Population PPB ranged between 28.7% and 79.7% (Table 1), while
R-package “ade4” (Dray and Dufour, 2007) and using a pairwise
Neiıs´ gene diversity (He) ranged between 0.091 and 0.208. Both
FST distance matrix along with a geographical distance matrix. Sig-
measures of genetic diversity were highly correlated (Pearsonıs´
nificance of the resulting standardized Mantel correlation rM was
r = 0.82, p < 0.001). Mean genetic diversity across all populations
tested with 9999 permutations. We obtained a dissimilarity matrix
was moderately high at PPB = 38.4% and He = 0.132. In addition,
of square-root transformed Sørensen coefficients between indi-
Kruskal-Wallis tests showed that an increase of the degree of isola-
viduals, which places emphasis on shared bands (Legendre and
2
tion resulted in reduced genetic diversity (PPB: = 9.28, p < 0.01,
Legendre, 1998). This was visualised by Principal Coordinate Anal-
2
Fig. 3; He: = 12.19, p < 0.005). Limited power of pair-wise post hoc
ysis (PCoA) with the package “vegan” (Oksanen et al., 2009) in R
non-parametric comparisons resulted in moderate initial p-values
3.1.2 (R Development Core Team, 2013).
(p ≈ 0.025 for all three tests), which did not pass the scrutiny of
Bonferroni-type (Holm) corrections.
2.5. Reproductive fitness: germination experiments According to the Mann-Whitney U Test, populations from the
two grassland types did not differ with respect to genetic diversity.
One hundred seeds per population were used for a germina- However, population sizes and densities were significantly smaller
◦
tion experiment following cold stratification at 5 C for 10 weeks in the meadow steppe group (Mann-Whitney U = 16; p < 0.01;
46 S. Heinicke et al. / Flora 224 (2016) 42–49
Fig. 2. Principal Coordinate Analysis (PCoA) for all AFLP data of S. pennata populations; Axes 1 and 2 explain 35.8% and 12% of the total variance, respectively. Genetic
data were analysed based on square-root transformed Sørensen dissimilarity (ellipses indicate the three main clustered groups). Different symbols and colours indicate
the different populations and their classification within the main habitat types (blue: meadow steppe group; orange: feather grass steppe group). (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version of this article.)
2
Fig. 3. Relationship between the percentage of polymorphic bands and the degree of isolation. Differences were overall significant (PPB: = 9.28, p < 0.01), while limited
statistical power resulted in non-significant pair-wise post hoc comparisons if Holm-Bonferroni-correction was applied.
Mann-Whitney U = 5; p < 0.001, respectively). In addition, we did ing to the first group, and M2, M3, M4, M5, M7, M9, M10, M11 and
not find a relationship between population size and PPB. According M12 to the second group. The less important second axis differ-
to the linear regression models, in wetter regions vegetation cover entiated a third group (Fig. 2), with populations F3 and F4. Along
−
had a negative impact on PPB (parameter estimate = 0.01, F = 7.00, axis 1, individuals of populations F2 and F7 were assigned between
p < 0.05). the first and third group. Differences along the second axis were
The hierarchical AMOVA (Table 3) indicated that the largest limited and largely caused by the presence/absence of two private
share of molecular variance was partitioned within populations alleles. The first and second group corresponded to populations
of S. pennata (43.7%), followed by variances among the two grass- from the feather grass steppe group and the meadow steppe group,
land types (41.8%). Variation among populations within grassland respectively.
types was comparatively small (14.5%). Populations were strongly Genetic structures were related to plant community compo-
differentiated overall (AMOVA, ST = 0.547, Table 3). All pairwise sition (rM = 0.32, p < 0.005), and genetically similar populations