Conservation Genetics

, Volume 9, Issue 1, pp 85–95

Genetic consequences of past climate and human impact on eastern Mediterranean Cedrus libani forests. Implications for their conservation

Authors

    • INRA—UR629—Unité des Recherches Forestières Méditerranéennes
  • F. Lefèvre
    • INRA—UR629—Unité des Recherches Forestières Méditerranéennes
  • G. G. Vendramin
    • CNR—Istituto di Genetica Vegetale
  • A. Ambert
    • INRA—UR629—Unité des Recherches Forestières Méditerranéennes
  • C. Régnier
    • INRA—UR629—Unité des Recherches Forestières Méditerranéennes
  • M. Bariteau
    • INRA—UR629—Unité des Recherches Forestières Méditerranéennes
Research Article

DOI: 10.1007/s10592-007-9310-6

Cite this article as:
Fady, B., Lefèvre, F., Vendramin, G.G. et al. Conserv Genet (2008) 9: 85. doi:10.1007/s10592-007-9310-6

Summary

This study demonstrates the impact of natural factors and human activities on biodiversity at gene level on a keystone Mediterranean forest ecosystem species. We monitored the within and among population gene diversity of Cedrus libani, a forest tree species of the Eastern Mediterranean mountains. We used paternally inherited chloroplast microsatellites (57 haplotypes) and bi-parentally inherited isozymes (12 loci) to estimate allelic richness, heterozygosity, and differentiation in 18 natural and 1 planted populations from Turkey and Lebanon. We showed that there is a phylogeographic structure in C. libani, and that forests from Lebanon and Turkey constitute two genetically isolated groups which probably arose from distinct refugia after the last Quaternary glacial cycle. We found extensive gene flow and relatively low differentiation in Turkey, as well as little evidence of genetic drift within populations. However, one population we analyzed, which was planted more than 20 centuries ago, and is isolated from core populations in Turkey, demonstrated extremely low genetic diversity and deserves high conservation priority. In contrast, we found low gene flow, high differentiation and severe cases of genetic drift in Lebanon. As forests there are the remnants of millennia-long extensive deforestation, all deserve high conservation priority.

Keywords

Chloroplast microsatelliteGene flowGenetic driftHabitat fragmentationIsozyme

Introduction

Cedrus libani A. Rich has a long history of resource depletion over historical times. It is currently found in the mountains of Turkey, Syria and Lebanon from 1,400 to 2,200 m above sea level (Quézel and Médail 2003). Its wood has been greatly appreciated since ancient times for its durability, density, color, timber and insecticide properties. It was for example a choice timber for royal sepultures in the Middle East (Rogers and Kaya 2006). Its natural habitat in the mountains of Lebanon has been under considerable human impact over centuries. Intensive logging for ship building and construction, and land-clearing for agriculture were recorded as early as the 3rd millennium B.C. (Talhouk et al. 2001). Vanishing forests were already mentioned during the 1st century B.C. and as they continued to do so over historical times, it is now estimated that the current 2,000 ha of patchy cedar forests found in Lebanon are the remnants of more than 500,000 ha of post glacial forest (Alptekin et al. 1997). In Turkey, cedar forests cover almost 110,000 ha and occur primarily in the Taurus mountains, which steep slopes have somewhat sheltered its forests from over-exploitation and extirpation (Boydak 2003).

Fossil pollen indicates that cedars were present during the last Quaternary glacial cycle in Turkey (Zeist et al. 1975 and http://www.ngdc.noaa.gov/cgi-bin/paleo/webmapper.cgi). As there is a high genetic differentiation between C. libani from Turkey and C. libani from Lebanon (Scaltsoyiannes 1999, Bou Dagher-Kharrat 2001), it is highly probable that an independent cedar refugium was also present in Lebanon during the last Ice Age. Cedrus libani has the highest within-population genetic diversity of all cedars (Scaltsoyiannes 1999, Bou Dagher-Kharrat 2001), as do most of the eastern populations of circum-Mediterranean conifers (Fady 2005).

The expected genetic consequences of habitat fragmentation are a reduction of within-population and an increase of among-population diversity (Young et al. 1996). Because of demographic bottlenecks and as an effect of genetic drift, rare alleles tend to disappear, hence reducing allelic richness and, in the longer run, heterozygosity (Luikart et al. 1998). Because populations become isolated and gene flow reduced or non-existent, genetic drift tends to increase and, along with it, among-population diversity, although exceptions can be found. In tree species for example, very long distance gene flow is possible and, because of their long juvenile phase (Austerlitz et al. 2000), unusually high within- and low among-population genetic diversity is observed (Hamrick et al. 1992). As a consequence, trees may be somewhat genetically sheltered from demographic bottlenecks and habitat fragmentation, unless these events are extreme and leave less than a few tens of trees per population (Robledo-Arnuncio et al. 2004). Severe climate changes, for example such as those of the Quaternary glacial cycles (Fady 2005) as well as land-clearing or logging (Robledo-Arnuncio et al. 2004) are among such extreme events.

It is in this context of high expected genetic diversity that we used chloroplast microsatellites and isozymes to identify the potential effects of human impact vs. natural evolutionary processes on the genetic diversity of C. libani from Turkey and Lebanon. Chloroplast microsatellites (cpSSRs) are paternally inherited in Cedrus (Fady et al. 2003), as in most conifers, and thus more sensitive to genetic drift than bi-parentally inherited markers. They generally exhibit high within-population genetic variability, and their among-population diversity is similar or slightly more structured than that of nuclear markers (Petit et al. 2005). In this study, cpSSRs were used to describe within and among population genetic diversity in C. libani. At the time of the study, isozymes were the only bi-parentally inherited and co-dominant genetic markers available in this species. Their mutation rate is lower than that of cpSSRs. They were used on a sub-set of the cpSSR population sample to provide complementary information on within and among population genetic structure. We reasoned that if similar geographic patterns occurred using both kinds of makers, then long term evolutionary processes rather then recent human impact would be responsible for the observed genetic diversity structure. Conversely, we expected to identify those populations in our sample with a diversity and an evolutionary potential most at risk when cpSSR (and not necessarily isozyme) diversity was low.

Material and methods

Plant material for cpSSR analysis

A total of 19 C. libani populations was used in our study (Table  1 and Fig.  1): 14 populations from Turkey and 5 from Lebanon. Our plant material represents a significant part of the ecological diversity of C. libani, i.e., the western and central Taurus mountain ranges of Turkey and the Mount Lebanon ranges of Lebanon (Fig.  1). Our sample also includes a population from the anti-Taurus range in Turkey and an isolated (naturalized plantation) forest from the Pontic mountains in northern Turkey. Only missing from our sample is the small Amanos mountain range, from which no collection could be made for security reasons at time of collection. Each study population was made of 16 randomly selected samples. Needles were collected from trees planted in common-garden tests located in southern France. Trees in these experimental sites originate from a commercial seed collection. This type of seed collection is usually made from at least 30 seed-trees per population, each tree separated by at least 30 m from one another. We expect that our ex situ sampling strategy is a good surrogate for sampling natural populations, as: (i) seeds were collected during a mast seed year, (ii) mortality was low at nursery and field stages, (iii) our genetic markers are supposed to be neutral and not affected by the selection processes that may have occurred in the nursery and field test.
Table 1

Description of Cedrus libani populations used and within population diversity estimates from cpSSR and isozyme data

Population code

Population name

Country of origin

Latitude (N)

Longitude (E)

Stand surface (ha)

Natural regeneration

Sample size

cpSSR diversity estimates

Isozyme diversity estimates

cpSSR

isozyme

A

Rh

Ph

H

P

Ra

Hexp

Fis

AINZ

Ain Zhalta

Lebanon

33°39′

35°43′

40

high

16

5

3.66

0.0

0.61

BARO

Barouk

Lebanon

33°36′

35°41′

40

high

13

86

12

7.82

33.3

0.97

0.90

2.2

0.25

ns

HADE

Hadeth El Jebbe

Lebanon

34°14′

35°55′

200

low

15

41

13

7.84

7.6

0.98

0.70

1.9

0.27

−0.13

JABA

Jabal Kammouah

Lebanon

34°30′

36°13′

30

low

16

11

7.19

18.2

0.96

MASS

Masser Chouff

Lebanon

33°34′

35°41′

high

16

8

4.74

12.5

0.70

Overall for populations from Lebanon

  

29

6.25

37.9

0.84

0.80

2.1

0.258

−0.04

ABAN

Abanoz

Turkey

36°20′

32°56′

244

16

68

13

7.38

7.7

0.94

0.90

2.4

0.29

ns

ARMU

Armut Alani

Turkey

37°50′

31°18′

100

16

9

6.15

0.0

0.91

ARPA

Arpacik

Turkey

37°49′

29°14′

249

high

16

34

11

6.84

9.1

0.93

0.80

1.9

0.26

ns

ARSL

Arslankoy

Turkey

37°00′

34°14′

153

high

14

10

6.97

10.0

0.95

AYKI

Aykiricay

Turkey

36°27′

30°10′

94

16

35

9

5.66

11.1

0.85

0.80

2.0

0.26

ns

CATA *

Catalan

Turkey

40°47′

36°34′

290

low

12

3

2.34

0.0

0.32

DIRM

Dirmil

Turkey

37°08′

29°32′

40

16

8

5.50

12.5

0.87

GOKY

Gokyurt

Turkey

37°40′

32°02′

631

low

15

11

7.24

0.0

0.96

KAPI

Kapidag

Turkey

38°05′

30°42′

300

16

34

12

7.46

25.0

0.97

0.80

2.1

0.23

ns

KARA

Karacay

Turkey

36°24′

29°26′

142

16

38

8

5.53

12.5

0.88

0.80

2.1

0.24

ns

KONA

Konak

Turkey

37°17′

29°04′

107

low

16

28

6

4.37

0.0

0.77

0.70

1.8

0.26

ns

SEVI

Sevindik

Turkey

36°32′

29°46′

68

high

16

49

10

6.07

20.0

0.87

0.90

2.4

0.29

0.18

SULT

Sultandagi

Turkey

38°32′

31°08′

290

high

16

41

11

7.05

18.2

0.95

0.90

2.2

0.24

0.14

YGOK

Y. Gokdere

Turkey

37°44′

30°49′

151

16

26

11

7.05

9.1

0.95

0.80

2.0

0.24

ns

Overall for populations from Turkey

  

46

6.12

54.4

0.87

 

Overall for natural (excluding Catalan) populations from Turkey

  

46

6.41

54.4

0.89

0.82

2.1

0.258

0.11

Overall for all natural (excluding Catalan) C. libani populations

  

57

6.36

100

0.86

0.82

2.0

0.258

0.07

* This population is a plantation forest (see text). For cpSSRs: A = number of haplotypes; Rh = haplotypic richness after rarefaction to the smallest population size; Ph = proportion of private haplotypes, in % of A; H = gene diversity. For isozymes: P = proportion of polymorphic loci, Ra = allelic richness after rarefaction to the smallest population size; Hexp = expected gene diversity under Hardy Weinberg equilibrium, Fis = heterozygote excess or deficiency (only significant (P < 0.05) values are listed). “–" denotes non available data

https://static-content.springer.com/image/art%3A10.1007%2Fs10592-007-9310-6/MediaObjects/10592_2007_9310_Fig1_HTML.gif
Fig. 1

Schematic natural distribution of Cedrus libani (in light grey, from Alptekin et al.1997) and location of populations studied in Turkey and Lebanon (population code names are as in Table 1)

Plant material for isozyme analysis

We chose to use seed tissue to score isozyme polymorphism as scoring using needle tissue was very poor under our laboratory conditions. Isozyme polymorphism was estimated from the seeds collected in natural populations that remained in stock after seedlings were grown and planted in the common garden experiments. Only a subset of the original natural C. libani population collection was available: 10 populations from Turkey and 2 from Lebanon (Table 1). Each population was made of a bulk collection of 26–86 seeds.

DNA extraction and identification of cpSSRs

Total genomic DNA was extracted using the QIAGEN® DNeasy kit and 50 μg of frozen needles ground in a 2  ×  96 well-plate grinder. PCR amplifications were carried out using a PTC100 (MJResearch, Inc.) thermal cycler. Each reaction contained 0.2 mM of each dNTPs, 2.5 mM of MgCl2, 0.2 μM of each primer, 10 × reaction buffer (Pharmacia), approx. 25 ng of template DNA and 1 unit of Pharmacia Taq polymerase in a total volume of 25 μL. PCR profile was as follows: 5 min denaturation at 95°C followed by 25 cycles of 1 min denaturation at 94°C, 1 min annealing at 55°C and 1 min extension at 72°C, with a final extension step of 72°C for 8 min. After amplification, PCR products were mixed to a loading buffer (98% formamide, 10 mM EDTA pH 8.0, 0.1% bromophenol blue, 0.1% xylene cyanol and 10 mM NaOH), heated 5 min at 95°C and then set on ice. The seven cpSSR primers that were shown to amplify in Cedrus (Fady et al. 2003) were tested for polymorphism using one sample from each of the 19 study populations. One cpSSR fragment per polymorphic primer was sequenced (Amersham ALF Express) in C. libani, as well as in two related species C. atlantica and C. deodara, to verify the presence and the mutation pattern of the microsatellite motifs. These three samples were later used as control in all electrophoreses. Fragments were electrophoretically separated on a 6% polyacrylamide gel and stained using silver nitrate (Rajora et al. 2000). Each PCR product was genotyped twice on different gels.

Isozyme extraction and identification

Isozymes were extracted from both embryo and megagametophyte of germinating seeds, following the procedure of Fady and Conkle (1992). Isozyme polymorphism was scored using starch gel electrophoresis and specific activity staining following the procedure of Fallour et al. (2001). Ten enzyme systems, out of the 15 known to amplify in C. atlantica (Fallour et al. 2001), and showing clear banding patterns in one sample for each of all 19 populations analyzed, were used. These enzyme systems were: GDH (E.C. 1.4.1.3), GOT (E.C. 2.6.1.1), IDH (E.C. 1.1.1.42), LAP (E.C. 3.4.11.1), MDH (E.C. 1.1.1.37), MNR (E.C. 1.6.99.2), 6-PGD (E.C. 1.1.1.44), PGI (E.C. 5.3.1.9), PGM (E.C. 2.7.5.1), SkDH (E.C. 1.1.1.25).

Statistics

Variation at each of the amplified chloroplast DNA region is referred to as size variant. The combination of size variants from different amplified cpDNA regions defines a haplotype. Variants at isozyme loci are referred to as alleles, which together combine into genotypes. Mendelian inheritance of alleles was assumed to be that observed by Fallour et al. (2001) for C. atlantica. Both monomorphic and polymorphic isozyme loci were used to generate population diversity indices in C. libani. To account for the risk of not detecting rarely occurring haplotypes because of sample size limitations, genetic diversity estimates and confidence intervals generated were weighted by sample size.

Within population

The number of different haplotypes per population (A) was counted directly. To eliminate the possible effect of sample size on within population diversity, haplotypic richness (Rh) was estimated using the rarefaction method of El Mousadik and Petit (1996), where the rarefied sample size corresponds to the smallest population size of the dataset, calculated as follows (software used: Fstat, http://www.unil.ch/izea/softwares/fstat.html):
$$ R_{h}\,=\,\sum_{i\,=\,1}^n {\left[ {1-\frac{\left( {\begin{array}{l} {N-N_{i}} \\ n \end{array} } \right)}{\left( {\begin{array}{l} N \\ n \\ \end{array} } \right)}} \right]}, $$
where n is the rarefied sample size, and Ni is the frequency of haplotype i within a collection of N haplotypes. Significance of allelic richness differences between groups of populations was tested using 1,000 permutations. The proportion of polymorphic isozyme loci (P) was counted directly. Isozyme allelic richness (Ra) was estimated in the same way as for Rh, but rarefied sample size was 2n and genotype sample size was 2N (diploid genome).
Genetic diversity within population was evaluated using Nei’s unbiased (1973) diversity index, H, calculated as follows:
$$ H=\frac{n}{n-1}(1 - \sum_{k\,=\,1}^n p_{i}^{2}) $$
where p is the relative frequency of the ith haplotype (or allele), k the number of haplotypes (or loci) and n the sample size within a population. The significance of differences in haplotypic diversity among groups of populations was performed by permuting populations within groups 1,000 times (software used: Fstat). In the case of isozymes, the expected heterozygosity Hexp was estimated in the same way as H, although p is the relative frequency of the ith allele, k the number of loci and n must be replaced by 2n.
Departures from Hardy-Weinberg expectations were estimated using isozymes and comparing expected (Hexp) and observed (Hobs) heterozygote frequencies within populations, as follows (software used: Fstat):
$$ F_{is} = 1 - \frac{H_{\rm obs}}{H_{\rm exp}} $$

Among populations

For both isozymes and cpSSRs, differentiation among groups of populations was calculated using Gst (Nei 1973), as follows:
$$ G_{st}=\frac{H_{t}-H_{s}}{H_{t}} $$
were Ht is the total diversity and Hs the gene diversity within populations (software used: Fstat). We then calculated pair-wise differentiation values (calculated as for Gst) among all population pairs and tested their significance by permuting individuals within populations 1,000 times (software used: Fstat). Haplotype and genotype frequencies were also used separately to calculate pair-wise genetic distances (Nei 1972) among all population pairs (software used: GenAlEx, http://www.anu.edu.au/BoZo/GenAlEx/), as follows:
$$ D = -\ln \frac{\left( {\sum\limits_m {\sum\limits_i P_{1mi} \ast P_{2mi} } } \right)}{\sqrt{\sum\limits_m {\sum\limits_i {P^{2}_{1mi}} } }\sqrt{\sum\limits_m {\sum\limits_i {P^{2}_{2mi}} } }} ,$$
where m is the number of loci, P1mi is the frequency of allele (haplotype) i at locus m in population 1.

Both cpSSR and isozyme pair-wise measures were used to construct phylogenetic trees using the Saitou and Nei (1987) neighbor joining method (software used: Phylip, http://www.evolution.genetics.washington.edu/phylip.html). Possible correlation between genetic and geographic (Euclidean) distances was tested using the Mantel approach permuting the data 2,000 times (software used: Fstat).

For cpSSRs, assuming a stepwise mutation model, differentiation was also estimated taking into account genetic distances between haplotypes (Pons and Petit 1996), i.e., differences in the total number of repeats between all haplotypes, denoted Rst and estimated as the sum of the squared differences in number of repeats (software used: Fstat). The significance of information provided by Rst vs. Gst, i.e., the fact that ordered haplotypes could be significantly more geographically structured than unordered ones, was also tested by randomly permuting 1,000 times haplotype identity within populations (software used: cpSSR, http://www.pierroton.inra.fr/genetics/labo/Software/CpSSR/). Finally, measuring the amount of variation due to differentiation among populations and groups of populations (Turkey and Lebanon) was done using an Analysis of Molecular Variance approach permuting individuals within populations 1,000 times (software used: GenAlEx).

Results

Characterization of marker polymorphism

Polymorphic primer pairs (Pt63718, Pt15169, Pt71936) among and within the three Cedrus species tested were the same as the ones detected by Fady et al. (2003) on a smaller sample size. Within C. libani, locus Pt63718 had six variants (most frequent variant had 100 base pairs), locus Pt15169 had four variants (most frequent variant had 126 base pairs) and locus Pt71936 had five variants (most frequent variant had 142 base pairs). Sequenced fragments from the three cpSSR regions confirmed the presence of a microsatellite motif and indicated that variation among samples was due to stepwise changes in the number of repeats of the microsatellite motifs (Table  2). No heteroplasmy was observed at any of the three amplified sites in all samples tested. The most commonly displayed haplotype combinations were those with variants of intermediate sizes, as expected in a stepwise mutation process.
Table 2

Sequence alignment of three cpDNA fragments containing the microsatellite motifs analyzed, from three different samples used as standards in all electrophoresis gels

https://static-content.springer.com/image/art%3A10.1007%2Fs10592-007-9310-6/MediaObjects/10592_2007_9310_Figa_HTML.gif

Primer zones are underlined and microsatellite motifs are colored in gray. Fragment lengths (in base pairs) appear at the end of the stretch. Samples are as follows: CA: C. atlantica, CL: C. libani, CD: C. deodara

The 10 enzyme systems yielded 12 scorable loci. Polymorphism within loci was identical to that of C. atlantica (Fallour et al. 2001), except that in C. libani, locus Idh was monomorphic and locus 6Pgd was polymorphic (2 alleles).

Within population diversity

Size variants at the three polymorphic cpSSRs combined into 55 different haplotypes out of 120 mathematically possible combinations. Most haplotypes (48 out of 55, i.e., 87%) had a frequency of less than 5% over all individuals, while only one (i.e., 2%) had a frequency over 10%. Mean within-population haplotypic richness (R) was 6.36, and was lowest in Çatalan, the planted population of Turkey (R = 2.34). Mean haplotypic richness of populations from Lebanon was not significantly different from that of populations from Turkey (Table  1). Mean haplotypic diversity of C. libani populations was H = 0.86, and was not significantly different in populations of Lebanon and Turkey. Haplotypic diversity was over 0.75 in all but three populations (Table  1), and was lowest in the planted C. libani population Çatalan of Turkey (H = 0.32).

Mean isozyme allelic richness of C. libani was R = 2.0, and not significantly different in Lebanon and Turkey. Mean gene diversity expected under Hardy Weinberg equilibrium was Hexp = 0.26, and was identical in Lebanon and Turkey (Table  1).

Departure from Hardy Weinberg equilibrium

There was a significant overall heterozygote deficiency in C.  libani (Fis = 0.07, P < 0.05). Mean Fis of populations from Turkey was significant (Fis = 0.11, P < 0.01) although it was not in populations from Lebanon (Fis = −0.04, P > 0.05). Two (Turkish) populations had a significant (P < 0.05) homozygote excess, although one (Lebanese) population had significant heterozygote excess (Table  1).

Geographic structure of gene diversity

In populations from Lebanon, 38% (11 out of 29) of all haplotypes were found to be specific to Lebanon. In populations from Turkey, 61% (28 out of 46) of all haplotypes were found to be specific to Turkey. The remaining, most common, 16 haplotypes (29% of all C. libani haplotypes) were shared among the two different geographic groups.

Pair-wise differentiation values (Table  3) showed that the planted Çatalan population was significantly different from all other populations. Within the Turkish regional group and excluding the non-natural population Çatalan, 65 population-pair values out of 78 (83%) were non significant. In the Lebanese regional group, only three population-pair values out of 10 (30%) were non significant. And, out of 65 population-pair values generated from the comparison of natural Turkish-Lebanese population pairs (excluding population Çatalan), 18 (28%) were non significant. The Analysis of Molecular Variance (also excluding population Çatalan) revealed that, although most of the haplotypic variation observed was among individuals (77%, P < 0.001), variation among populations was mostly due to differences between the two regional groups of populations from Turkey and Lebanon (18% of total variance, P < 0.001) while the remaining 5% of variation was among populations within regions.
Table 3

Differentiation between all C. libani population pairs, expressed as pair-wise Gst values

Populations

Lebanon

Turkey

 

BARO

HADE

JABA

MASS

ABAN

ARMU

ARPA

ARSL

AYKI

CATA *

DIRM

GOKY

KAPI

KARA

KONA

SEVI

SULT

YGOK

AINZ

0.163

0.131

0.184

0.256

0.215

0.242

0.229

0.207

0.270

0.523

0.256

0.193

0.200

0.258

0.312

0.262

0.185

0.220

BARO

0.000

0.004

0.152

0.003

0.043

0.007

0.018

0.078

0.340

0.074

0.000

0.014

0.051

0.121

0.060

0.029

0.022

HADE

0.040

0.000

0.146

0.005

0.023

0.001

0.013

0.053

0.333

0.044

0.020

0.009

0.024

0.064

0.015

0.000

0.006

JABA

  

0.167

0.038

0.044

0.046

0.034

0.088

0.263

0.076

0.031

0.026

0.083

0.130

0.087

0.038

0.030

MASS

   

0.141

0.195

0.183

0.127

0.206

0.472

0.178

0.094

0.096

0.209

0.260

0.213

0.168

0.171

ABAN

0.289

0.255

  

0.033

0.000

0.002

0.086

0.341

0.058

0.000

0.019

0.000

0.084

0.031

0.000

0.000

ARMU

     

0.022

0.006

0.062

0.325

0.008

0.020

0.004

0.007

0.000

0.008

0.008

0.013

ARPA

0.372

0.331

  

0.021

 

0.011

0.075

0.350

0.055

0.000

0.023

0.000

0.066

0.007

0.000

0.004

ARSL

       

0.023

0.335

0.000

0.000

0.000

0.009

0.046

0.037

0.000

0.012

AYKI

0.351

0.305

  

0.038

 

0.006

 

0.377

0.014

0.055

0.027

0.056

0.055

0.044

0.071

0.078

CATA *

         

0.378

0.335

0.329

0.380

0.431

0.384

0.341

0.302

DIRM

          

0.025

0.000

0.009

0.000

0.000

0.000

0.069

GOKY

           

0.000

0.007

0.074

0.000

0.000

0.000

KAPI

0.436

0.412

  

0.089

 

0.090

 

0.093

   

0.000

0.000

0.046

0.000

0.026

KARA

0.386

0.338

  

0.023

 

0.000

 

0.000

   

0.056

0.000

0.000

0.000

0.022

KONA

0.396

0.352

  

0.027

 

0.060

 

0.006

   

0.000

0.007

0.000

0.044

0.076

SEVI

0.343

0.302

  

0.033

 

0.012

 

0.000

   

0.062

0.000

0.005

0.015

0.019

SULT

0.398

0.359

  

0.050

 

0.046

 

0.008

   

0.043

0.005

0.000

0.000

0.003

YGOK

0.395

0.372

  

0.046

 

0.054

 

0.055

   

0.000

0.033

0.004

0.028

0.018

Above the diagonal, values computed from cpSSR haplotype frequencies. Below the diagonal, values computed from allozyme frequencies. Non significant values (P < 0.05 after 1,000 permutations) appear in bold type. Population codes are as in Table 1

We used both un-ordered (Gst) and ordered (Rst) differentiation measures to estimate the divergence among populations based on cpSSRs. Among population haplotypic variation of natural C. libani populations (population Çatalan excluded) was highly significant, and the difference between Gst and Rst was also highly significant (Table  4). If only the 11 populations that are in common between the cpSSR and isozyme analyses were considered, differentiation remained significant (Gst = 0.024, P < 0.001). Differentiation was also significant among the natural C. libani populations from Turkey, and again the difference between Gst and Rst was significant (Table 4). Finally, differentiation was also significant among populations from Lebanon, but the difference between Gst and Rst was not significant (Table 4). Fig. 2 (pair-wise Nei’s distance-based dendrogram) shows C. libani populations from Lebanon as a loose group, as opposed to natural populations from Turkey which form a more cohesive, and spatially separated, group.
Table 4

Estimates of differentiation among populations in Cedrus libani calculated from cpSSR data. Rst is a measure of differentiation that takes into account differences in the total number of repeats between haplotypes, although Gst. does not. A significant value in the Gst vs. Rst column indicates a significant phylogeographic structure

Population group

Gst

P value

Rst

P value

Gst vs. Rst (P-value)

Lebanon

0.123

<0.001

0.143

<0.0001

non significant

Turkey (excluding Catalan)

0.022

<0.005

0.109

<0.005

<0.01

All populations

0.067

<0.067

0.216

<0.0001

<0.001

https://static-content.springer.com/image/art%3A10.1007%2Fs10592-007-9310-6/MediaObjects/10592_2007_9310_Fig2_HTML.gif
Fig. 2

Phylogenetic tree built using Saitou and Nei (1987) neighbor joining method and Nei (1972) minimum genetic distance on cpSSR data. Distance range among Turkish populations is 0.035–0.141. It is 0.051–0.266 among Lebanese populations, and 0.063–0.355 between Turkish and Lebanese populations

Although no allozyme was found to be population or region specific, isozymes confirmed the population structure found using cpSSRs. Differentiation among populations was highly significant (Gst = 0.152, P < 0.01). Differentiation among Turkish populations was globally significant (Gst = 0.024, P < 0.01), although 53% of all pair-wise differentiation values (19 out of 36) were non significant (not shown). Differentiation between the two Lebanese populations tested was not significant (Gst = 0.020, P > 0.05). Finally, all Turkish-Lebanese pair-wise differentiation values were significant (P < 0.01, not shown) as well as the differentiation between the two groups (Gst = 0.201, P < 0.01). Again, natural populations from Turkey form a rather cohesive group, significantly separated from populations from Lebanon (Fig. 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs10592-007-9310-6/MediaObjects/10592_2007_9310_Fig3_HTML.gif
Fig. 3

Phylogenetic tree built using Saitou and Nei (1987) neighbor joining method and Nei (1972) minimum genetic distance on isozyme data. Distance range among Turkish populations is 0.002–0.032. It is 0.015 between the two Lebanese populations, and 0.097–0.186 between Turkish and Lebanese populations

Using the Mantel test procedure with two different estimates of genetic structure (pairwise differentiation and Nei’s distance) and the two types of markers (cpSSRs and isozymes), there was no indication of isolation by distance when Turkish and Lebanese populations were analyzed separately. However, when all C.  libani populations were grouped together, matrices were no longer independent. Using Nei’s distance, for example, geographic distance explained 15% of cpSSR genetic variance (P < 0.005).

Discussion

As in most population studies involving tree species and using paternally or bi-parentally inherited genetic markers, most of the genetic diversity found in C. libani was present within population. Despite its long history of resource depletion, biodiversity at gene level was found to be among the largest values observed for conifers both at cpSSR (Petit et al. 2005) and isozyme loci (Hamrick et al. 1992). This was not unexpected, as Eastern Mediterranean conifers harbor higher levels of genetic variation than any other conifers, which is due to the fact that their populations during the late glacial maximum (18,000 years ago) were somewhat sheltered in favorable environments and did not suffer strong demographic and genetic bottlenecks (Fady 2005). However, this high genetic diversity was not evenly partitioned among C.  libani populations, which we interpreted as the result of natural and human-made environmental constraints over different time scales.

Phylogeographic patterns of diversity

The overall among-population genetic diversity found in natural C. libani populations is compatible with an isolation by distance model, and is mostly due to differentiation between our Lebanese and Turkish regional population groups. Both groups carry some specific haplotypes. Considering the rather low mutation rate of cpSSRs (Provan et al. 1999), and the even lower one of isozymes, considering also that post-glacial recolonization is expected to have generated rather low levels of population differentiation in trees (Austerlitz et al. 2000), the existence of two genetically isolated (groups of) glacial refugia during the last glacial cycle (which lasted approximately 100,000 years and ended 15,000 years ago (Hewitt 2000)) is a likely hypothesis for this range-wide spatial structure. This hypothesis could be further argued by analyzing the haplotypic diversity found in the C. libanianti-Taurus and Amanos populations of Turkey. However, recent studies have also demonstrated that adaptive traits also differ markedly between the Turkish and Lebanese groups, e.g., those related to water stress resistance (Ladjal et al. 2005). We thus tend to agree with floras (Davis 1965, Greuter et al. 1984) and with population genetic studies (Scaltsoyiannes 1999, Bou Dagher-Kharrat 2001) that recognize the existence of two C. libanitaxa, one in Lebanon and one in Turkey, which, we think, should be granted the taxonomic level of subspecies.

Comparing differentiation measures that do (Rst) and do not (Gst) take into account differences in number of repeats between cpSSR haplotypes, indicated that Lebanese populations probably emerged from a single refugium (no significant difference between Rst and Gst). By contrast, the highly significant difference between Rst and Gst within the Turkish regional group is indicative of a phylogeographic structure. Turkish populations probably emerged from several refugia. The fact that there is a significant overall (isozyme) homozygote excess in Turkey (and not in Lebanon) when populations are pooled together, despite seemingly effective contemporary gene flow (low overall cpSSR and isozyme differentiation in Turkey), does not contradict this hypothesis of a “multi-refugium” origin for C. libani in Turkey.

Human induced patterns of diversity

Populations from Turkey make a more genetically cohesive regional group than populations from Lebanon. This is a probable indication that gene flow is currently reduced in Lebanon, a likely consequence of human impact through deforestation. At equilibrium between gene flow and drift, it is expected that Gst estimates from paternally and bi-parentally inherited markers should be statistically identical (or slightly lower for bi-parentally inherited genes), because paternally and bi-parentally transmitted genes are carried both by pollen and seed (Petit et al. 2005). Experimental data have confirmed the theory in many conifer species (Petit et al. 2005). In our study, however, cpSSR Gst (0.024 using only populations that are shared between the cpSSR and isozyme studies) is much lower than isozyme Gst (0.152). Further, all pairwise Turkey–Lebanon isozyme Gst values, as well as the only available Lebanon–Lebanon value, are higher than their corresponding cpSSR Gst values, although the majority (21 out of 36) of within Turkey pairwise isozyme Gst values are not. Although gene flow and drift seem to have reached an equilibrium in Turkey, it is not the case in Lebanon. And as there are no significant ecological requirement differences between Turkish and Lebanese habitats, over exploitation in the form of logging and/or grazing is the likely explanation for the overall lower genetic diversity found in cedars from Lebanon.

The human-made population Çatalan of Turkey was significantly different from all other Turkish populations, both in terms of within (lowest gene diversity of all populations studied) and among-populations gene diversity. Scaltsoyiannes (1999) had also found such diversity patterns in his isoenzyme survey. This now acclimated forest was planted in the 3rd century B.C. as a gift from a local Pontus king to his Mediterranean-born wife (in Kayihan 2000). The RAPD analysis of 14 Turkish populations by Kayihan (2000) suggested that the parent population could be the Mediterranean population of Arslankoy. In our study, the most frequent haplotype of Çatalan was found to be rare in the data set and present at frequency lower than 13% in only three other populations, two of which were from Turkey but not on the Mediterranean (Armut Alani and Gokdere), and one was from Lebanon. The two rare haplotypes found in Çatalan were shared by eight Turkish and one Lebanese populations. The three Çatalan haplotypes were never found together in any single population of our data set. Thus, our data confirm that population Çatalan is the result of a very strong demographic bottleneck following seed transfer and plantation, and possibly originate from a foundation involving as little as three male pollinators. As one haplotype has invaded the current gene pool, it is probable that the male reproductive success of one of those founder trees was much higher than that of the others. Finally, our data does not make it possible to identify the source population for this founding event. The most likely hypotheses for the foundation of Çatalan is that it originates from one population in our dataset for which we were not able to detect all rare haplotypes, or alternatively, a seed source not represented in our sample.

Threats to the genetic diversity of Cedrus libani

In Turkey, genetic diversity was high in most natural populations and there was no overall indication of serious genetic threat. Heterozygotes were never in excess, although homozygotes were in two (Sevindik and Sultandagi) out of nine populations, which is somewhat unexpected in natural forest tree populations. Assuming our sampling does not include different subpopulations in these two populations, this could indicate that mating occurs among neighboring trees, that selfing is high, and/or that the populations are recent and composed of related trees, all of which hypotheses need to be confirmed by a fine scale analysis of genetic diversity. Population Çatalan, however, had a gene diversity 2/3 lower than the rest of the populations from Turkey. Founded as the result of seed transfer from very few seed trees, it has been continuously isolated from core populations, and can probably be now considered as a new ecotype. Population Çatalan thus deserves high conservation priority and may now face new adaptation challenges. In such bottlenecked and isolated populations, the correlation between neutral gene markers and adaptive genes can be expected to be strong (Le Corre and Kremer 2003). Hence, the low diversity measured at neutral genes in this study is likely to reflect a low genetic variance at adaptive genes. Regeneration at low elevation is scarce and adult trees have suffered recent dieback in Çatalan (Alptekin et al. 1997), which could indicate reduced adaptation there. In this case, restoring some low level of between population gene flow is advisable. To avoid affecting local adaptation significantly, only areas demonstrating dieback should be targeted, using seeds harvested in source populations where the common Çatalan haplotype is found.

In Lebanon, most populations are currently under threat, either because of low gene diversity, small size or low natural regeneration. The fact that even highly polymorphic genetic markers (cpSSRs) are able to detect a strong signal of low gene diversity in two populations is an indication of strong and long term disruption of the gene flow—drift equilibrium. Historical facts and our data suggest that human activity forms a likely candidate for this disruption, mostly through deforestation (Talhouk et al. 2001). All Lebanese populations deserve high conservation priority as they are the remnants of what was once a large continuous forest (Alptekin et al. 1997). However, not all forests have low genetic diversity, and making sure that natural regeneration can be achieved (and, alternatively, planting using local seed sources when it is not achieved) may be a perfectly suitable way to protect their biodiversity, such as in population Hadeth El Jebbe, for example. In populations Ain Zhalta and Masser Chouff where gene diversity has been seriously reduced, natural regeneration is high. However, introducing some level of gene flow from high diversity populations might reduce the risk of future maladaptation. Candidate source populations should either be chosen from a mix of all populations from Lebanon, as all arise from a common recent ancestor, or from populations displaying similar adaptive trait variability if differentiation has started to occur at this level.

Acknowledgements

We are grateful to D. Vauthier for collecting plant material and to B. Jouaud for laboratory assistance. This work was funded by the project FAIR CT95-0097 “Adaptation and selection of Mediterranean Pinus and Cedrus for sustainable afforestation of marginal lands" financed by the Commission of the European Communities.

Copyright information

© Springer Science+Business Media, Inc. 2007