Phylogeography of Arabidopsis halleri (Brassicaceae) in mountain regions of Central Europe inferred from cpDNA variation and ecological niche modelling

The study species

Arabidopsis halleri (L.) O’Kane & Al-Shehbaz is a perennial, self-incompatible and highly outcrossing (Llaurens et al., 2008), stoloniferous herb with a highly disjunctive distribution between Europe and the Far East. It occurs in mountain and upland environments on slopes, forest margins, rocky crevices and river banks from 200 to 2,200 m a.s.l. In Europe, it is widely distributed in the Alps, Carpathians, Sudetes and Dinaric Alps (Jalas & Suominen, 1994). Its distribution also covers some upland regions north from the Alps (including the Harz Mountains) and the Western Carpathians (Jalas & Suominen, 1994). The species is highly variable in leaf morphology, flower colour and traits connected with the development of stolons. At least three subspecies are quite distinct in terms of morphological variation (Al-Shehbaz & O’Kane, 2002): A. halleri subsp. halleri and A. halleri subsp. ovirensis occur in Europe, whereas A. halleri subsp. gemmifera occurs in the Far East. According to other authors (Kolník & Marhold, 2006) the species can be alternatively divided into five distinct morphological subspecies, with four occurring in Europe. Only diploid (2n = 2X = 16) individuals have so far been reported from throughout the distribution range (Al-Shehbaz & O’Kane, 2002; Kolník & Marhold, 2006). Other species from the genus Arabidopsis also co-occur within the investigated area (A.thaliana, A.arenosa, A.neglecta, A. lyrata). A recent study has shown that reticulation and hybridization among lineages that might have transferred cpDNA types from one lineage into the other is unlikely (Koch & Matschinger, 2007). The same authors have pointed out that cpDNA type diversity may predate separation of the main evolutionary lineages within the genus (Koch & Matschinger, 2007).

Sampling and DNA extraction

We sampled 1,281 individuals from 52 populations (Table 1) scattered across the species range in seven geographic regions of Central and Eastern Europe (see Fig. S1). Twenty-five populations included in the present study have been already described in previous studies (Pauwels et al., 2005; Pauwels et al., 2012). In the present study, sampling was extended to improve sampling representativeness in the area of the Sudetes as well as the Carpathians. All the locations sampled belonged to the native species range (Jalas & Suominen, 1994). In each locality, individual samples were taken from plants separated by at least three metres to avoid sampling clones (Van Rossum et al., 2004). Sample size generally reflected population size and was almost exhaustive in small populations. Permits for plant sampling were obtained from the following national parks: Krkonossky narodni park (CZ, permit no. 01004/2009), Karkonoski Park Narodowy (PL permits no. 24/2007, 2/2008)/, Tatrzanski Park Narodowy (PL, permit no. NB-056/119/07), Bieszczadzki Park Narodowy (PL, permits no. 54/06, 40/07) and Karpatskij biosfernij Zapovidnik (UA, permit no. 01110607/07). In other cases sampling was done outside protected areas.

Collected leaf tissue was immediately dried in silica gel prior to molecular analysis. DNA was extracted from 10 to 15 mg of dry material using NucleoSpin^® 96 Plant (Macherey-Nagel), and PCR amplifications were performed on 1:20 dilutions.

Genetic analysis

Data analysis

Since cpDNA is a non-recombining molecule, alleles observed at all twelve loci were combined into cpDNA haplotypes (chlorotypes, Table 3). Chlorotype nomenclature is fully consistent with our previous study (Pauwels et al., 2005).

Chlorotype diversity

A total of 12 cpDNA haplotypes were found in the investigated populations (Table 4). All neighboring chlorotypes were linked by a single mutation (Fig. 1), except for G and H (separated by two mutations). Thus, the MST topology did not allow the division of chlorotypes into clearly demarcated groups separated by more than one mutation.

Chlorotypes did not show equal frequencies in overall sampling (Table 4). The most widely represented was chlorotype J, present in 26.4% of the samples analyzed. Chlorotypes E, F and G had a share of 15.47%, 12.19% and 12.11% respectively. The share of the remaining chlorotypes in the overall sampling was significantly lower than 10%.

Chlorotypes E and J were the most widespread geographically (Fig. 2), with their occurrence established respectively in 6 and 5 out of the 8 investigated geographical regions. Most of the haplotypes occupying tips and terminal branches of the MST were more localized geographically. Chlorotypes B, C and H have been found only in the Bohemian Forest, chlorotype D in the Harz Mts., while chlorotype H only in the Eastern Carpathians (Fig. 2). Chlorotype G was found almost exclusively in the Eastern Carpathians, while chlorotype F showed a predominant occurrence in the Western Carpathians as well as in the geographically close region of the Northern Carpathian Foreland.

Genetic diversity indices: chlorotypic richness (A_Sc) and chlorotype diversity index (H_Sd) were calculated for each investigated population. They varied broadly from 1 to 4.983 for A_Sc and from 0 to 0.893 for H_Sd (Table 4). We also examined the geographical pattern of variation in chlorotypic richness. As shown in Fig. 3, populations with a high level of genetic diversity were co-located in the Bavarian Forest, the Harz Mts. and in the Western Carpathians (Tatra Mts.).

Also we compared genetic diversity indices between different geographical regions (Table 5). As expected, geographical regions differed substantially in terms of genetic diversity. Three regions: Western Carpathians, Harz Mts. and Bohemian Forest were found to be most diverse. The lowest genetic diversity was found in the Alps and in the Sudetes.

Genetic structure

The clustering approach employed in the present study (spherical k-means clustering—SKMC) enabled us to study the structure present in our dataset on several levels. We examined a broad spectrum of different k values from k = 2 to k = 25. The results of SKMC from k = 2 to k = 10 were plotted on the map (Fig. S2). Clearly, the populations from Eastern Carpathians formed one stable cluster (present in all the k levels), that differed from all the other populations. A subdivision of the populations into 6 clusters (Fig. 4) was statistically optimal (Fig. 5A) and had a high support in AMOVA (Fig. 5B). We carried out separate AMOVA analyses to test the distribution of genetic variance within and between groups identified by SKMC. Results of SKMC evidenced the presence of some level of genetic admixture in almost all studied geographical regions except in the Harz Mts., the Alps, and the Eastern Carpathians.

AMOVA performed without grouping populations showed that 79.16% of the total genetic variation was found between populations (Table 6). When assuming six groups of populations (according to k-means clustering), as much as 65.8% of the total variation was observed between groups of populations, whereas 15.82% was found among populations within groups (Table 6). These percentages of genetic variation were 42.88% and 37.90%, respectively, when assuming seven groups (characterised according to the geographical regions sampled; Table 6). Separate AMOVAs performed within each investigated geographical region revealed the highest percentage of among-population variation in the Sudetes and the Bohemian Forest: 85.66% and 74.57%, respectively (Table 6). The lowest values of among-population variation was found in the Alps, where we recorded the presence of just one chlorotype (0%; Table 6), and in the Harz Mts. (18.55%; Table 6). All the remaining geographic regions showed intermediate level of among-population variation.

Distribution modelling

Model performance was assessed using two different statistics: True Skill Statistic (TSS) and Area under Receiver Operating Characteristic (ROC). All models performed well and had TSS >0.7 and ROC >0.88. The performance of two models CTA and FDA was weaker than the performance of the remaining methods, but still of acceptable quality.

Two (CCSM4, MPI ESM) out of the three climatic models suggest that the species survived the LGM (ca. 22 kyr BP) only in the southernmost part of the Carpathians (Fig. 6). The Dinaric Alps and Balkan Mts. were also among the potential areas for species survival in southern Europe during the LGM (Fig. 6). Vast areas north from the Alps can be also considered as suitable for the survival of A. halleri (Fig. 6).

We also analysed the potential distribution of A. halleri during the Holocene climatic optimum (ca. 6 kyr BP). Our models showed that recolonization must have advanced very slowly in the Carpathians, when compared with areas north of the Alps. All the models showed that conditions facilitating the spread of A. halleri occurred much earlier in between the Western Carpathians, the Sudetes and areas north from the Alps, while suitable areas in the Eastern Carpathians were at first much more restricted (Fig. 6).

Carpathian populations of A. halleri

Our data suggest a clear differentiation among populations from western and eastern part of the Carpathians. Those population groups differ in terms of chlorotype composition and frequencies. The same differentiation also appears clearly in the SKMC analysis since the western and eastern populations were grouped in different clusters from k = 2. This pattern of genetic variation seems to follow the division between the eastern and western part of the Carpathians which was first recognized by Wołoszczak (1896) and was established on the basis of floristic data. The nature of the barrier between the western and eastern part of Carpathians has been the subject of many studies employing different methodologies from floristic (Pax, 1898; Jasiewicz, 1965) to cytologic (Mráz & Szelag, 2004) and genetic (Mráz et al., 2007; Ronikier, Cieślak & Korbecka, 2008a; Těšitel et al., 2009). It has been hypothesized that specific climatic and orographic conditions of the westernmost part of Bieszczady Mts. (also known as Bukovske Vrchy Mountains) are among the main factors influencing the genetic landscape of this part of Carpathians (Domin, 1940). It seems that results of our study may suggest that differentiation between western and eastern part of Carpathians may be a historical phenomenon connected with recolonisation of the area by plants that survived in different refugia. In this case we do not need to postulate the presence of a specific barrier responsible for the existence of genetic discontinuity between western and eastern part of Carpathians since this phenomenon could be also explained by colonisation from two different directions together with the presence of gene flow between the two groups as was evidenced by our study (i.e., the presence of haplotype P in western and eastern Carpathians).

The Eastern Carpathian populations are characterized by the presence of haplotype G, at very high frequencies. This haplotype was considered as ancestral by Pauwels et al. (2005). The presence of the ancestral haplotype G at high frequencies has also been recorded in populations located south of the Alps as well as in the south-eastern part of this mountain range (Pauwels et al., 2012). Our genetic data showing low genetic diversity in the Eastern Carpathians suggest relatively recent recolonisation of this area by A. halleri as was also indicated by the modelling experiments. This is surprising, when we take into account that several studies have demonstrated (with high probablity) that a glacial refugium existed in the Eastern Carpathians (Willis & Van Andel, 2004; Tollefsrud et al., 2008). The low genetic diversity in the Eastern Carpathians was, however, also confirmed by Tollefsrud et al. (2008) in Picea abies. These authors suggested a genetic bottleneck as a result of substantial decrease in population size as a reason of this phenomenon. The same scenario cannot be ruled out for A. halleri, although modelling experiments carried out during the present study suggest postglacial recolonisation as a more plausible explanation of low genetic diversity in the Eastern Carpathians.

The Western Carpathian populations are characterized by high levels of genetic diversity and the presence of a private haplotype F at high frequencies. Both premises suggest the possibility of in situ glacial survival of the species in the Western Carpathians.

Survival of plant species in the Western Carpathians during the LGM has been a focal point of many studies employing different methodologies. These works have shown that the existence of a Western Carpathian refugium is quite probable for many plant species, including mountain plants. Macrofossil charcoal fragments found in Kraków, just about 30 km in a straight line from population PL8 and about 40 km from PL 7, indicate full glacial presence of Pinus, Larix and Abies from 26 to 28 kyr BP—during the coldest period of LGM that spanned between ca. 36–16 kyr BP (Damblon, Haesaerts & Van der Plicht, 1996; Willis & Van Andel, 2004; Musil, 2003). Dates from humic soil further down in the sequence are even earlier, and indicate the presence of trees as early as 36 kyr BP (Willis & Van Andel, 2004). There is also taxonomic evidence supporting the hypothesis of longstanding survival of different plant species in the Western Carpathians. Saxifraga wahlenbergii Ball. and Delphinium oxysepalum Borb. et Pax are good examples here. These two species are endemic to the Western Carpathians and occupy isolated systematic positions, what suggests that they are both of Tertiary age (Mirek & Piekos-Mirkowa, 1992). The survival of common yew (Taxus baccata), was also documented in charcoal for Moravany in Slovakia with a radiocarbon date of ca. 18 kyr BP (Lityńska-Zając, 1995). There is also a large body of phylogeographic evidence that indicates the existence of a major northern refugium for a variety of animal taxa in the area around the Carpathians, with some lineages predating LGM (Provan & Bennett, 2008).

On the other hand, results of niche modelling conducted during the present study suggest postglacial recolonization of the area. In this scenario high genetic diversity in the area could be at least partially explained by relatively recent gene flow (occurring later than 6 kyr BP, according to our modelling experiments) from the Eastern Carpathians. The presence of a genetic admixture in Western Carpathian populations was also revealed by Pauwels et al. (2012) in A. halleri and by Těšitel et al. (2009) in Melampurum sylvaticum. It seems that the question of the existence of glacial refugium for A. halleri in the Western Carpathians should remain open taking into account that evidence are still not conclusive.

In the SKMC analysis, populations from the Western Carpathians were clustered together with populations from upland regions of southern Poland. This fact supports the hypothesis of a Western Carpathian origin of populations located north of the Western Carpathians. However, for the reasons stated above, it is very difficult to say when these populations were founded. We have shown that populations of A. halleri in western and eastern part of Carpathians form two different genetic groups. It might be hypothesized that these groups could be derived from two different refugia. Populations in the Eastern Carpathians might belong to the group that survived in a southern refugium (areas south of the Alps, Dinaric Alps and Balkan Mts.) as suggested by Pauwels et al. (2012).

Genetic differentiation between the Harz, the Bohemian Forest and the Alps

The situation in the areas north and north-east of the Alps is more complex than in the Carpathians. SKMC analysis showed that at least two groups of populations can be recognized regardless of the level of k: populations from the Harz Mountains and from the Alps. Populations from the Bohemian Forest were usually assigned to different clusters, forming very heterogeneous group. This heterogeneity was also evidenced in AMOVA.

Differentiation is also apparent between the regions of the Harz Mts. and the Bohemian Forest, with the first harbouring haplotype D, which is not present in the Bohemian Forest. In the latter region the occurrence of haplotype C can be observed, which, in turn, is present neither in the Harz, nor in the Alps or the Šumava. A relatively recent (postglacial) origin of this differentiation could be hypothesized as both haplotypes occupy external nodes of the MST tree.

It is not easy to explain this pattern of genetic differentiation. Some ideas might be provided by the results published by Tollefsrud et al. (2008). Investigating the genetic variation of Norway Spruce together with pollen data they established that one possible glacial refugium of the species might have extended from the northern slopes of the Alps up to the Šumava (the Bohemian Massif). It seems that A. halleri might have also survived in a vast area, and that its Pleistocene distribution covered not only the region mentioned above, but also extended northwards and westwards to the Ardennes and Hautes Fagnes (High Fens) in Belgium. So far one natural population from this area has been tested by Pauwels et al. (2008). This, nowadays isolated, population from Hautes Fagnes, harbouring a chlorotype with an extremely restricted geographical range (Pauwels et al., 2008), might be the trace of a past A. halleri distribution in Western Europe. The vast extent of possible refugial areas north of the Alps was also clearly evidenced by our modelling experiments. Other studies based on molecular methods also suggest the presence of a glacial refugium in the Central Europe (Reisch, Poschlod & Wingender, 2003; Koch, 2002; Rejzková et al., 2008).

Genetic variation in the closely related species Arabidopsis lyrata from the Harz, southern Germany and the Alps (Clauss & Mitchell-Olds, 2006), can give us some insights into a possible explanation of this pattern. Studies on genetic variation of A. lyrata (carried out on the basis of nuclear microsatellite loci) showed high within-population diversity throughout central Europe, accompanied by low regional differentiation and geographically widespread polymorphism. The authors hypothesized that (given the unlikeliness of gene flow) a common gene pool must have existed for central European populations (Koch & Matschinger, 2007). It should be noted that “central European” in this case is not a precise term and describes sites located approximately between the 10th and 16th eastern meridian. This area corresponds to the locations of the populations sampled by us north of the Alps in the the Bohemian Forest and the Harz. The same scenario is also probable for A. halleri, where haplotype E, present in all regions north of the Alps, can be interpreted as a testimony to a common gene pool in the past. Other haplotypes, with restricted geographical range, would be, under this model, local derivatives that evolved after climatic warming and fragmentation of a previously vast range.

The Bohemian Forest hosts a population with the highest genetic diversity in the whole sampled area. The question whether this high genetic diversity should be attributed to the existence of a glacial refugium in this region or the presence of a contact zone between eastern and western lineages remains open. We think that results cited above (Clauss & Mitchell-Olds, 2006; Tollefsrud et al., 2008) as well as data clearly suggesting the presence of a glacial refugium for Fagus in the southern part of Czech Republic (Magri et al., 2006) make the hypothesis of the existence of a glacial refugium in this area more probable. The fact that even nowadays the range of A. halleri and Norway Spruce in Central Europe is highly similar (Tollefsrud et al., 2008) also suggest this scenario. However, no palaeobotanical data are available for A. halleri. Therefore, such a hypothesis cannot be fully confirmed or rejected. It should be also noted that the area between the 10th and 16th eastern meridian has been identified as a contact zone for some plant species (Fjellheim et al., 2006; Daneck et al., 2011). It is clear that further detailed studies in this region, also including other taxa, are needed to clarify these findings more precisely.

Origin of A. halleri populations in the Sudetes

We have shown that A. halleri populations in Sudetes are characterized by a very low genetic variation, with most of the populations harbouring only one cpDNA haplotype. This finding is not surprising, given the evidence from geological research showing that this region was severely impacted during the glacial period. Is seems that at least one glacial maximum (48–43 kyr BP) had a devastating effect on the regional flora, mainly due to the close proximity of the continental ice sheet (Marks, 2005). Geological evidence show that even during LGM this region was affected by the presence of the mountain glaciers (Badura & Przybylski, 1998). Therefore, it has been postulated that in situ glacial refugia, supporting the remains of the autochthonous flora, did not exist in the Sudetes (Mitka et al., 2007). Quite a different scenario involving survival on nunataks and in peripheral refugia has been suggested for the Alps (Stehlik, 2000). It is noteworthy that even nowadays the climate of the highest parts of the Sudetes is cold enough to maintain the occurrence of tundra-like ecosystems (Soukupova et al., 1995) . It seems therefore, that A. halleri populations in the Sudetes could be of a recent (postglacial) origin. This is supported by the very low genetic variation found in Sudetes as well as the fact that despite dense sampling, we have not found any chlorotypes specific to this area. On the basis of the evidence from cpDNA variation, it could be hypothesized that populations from the Bohemian Forest could be a source of migrants that established new populations of the species in this region after LGM.

We have also found the presence of haplotype I in high-mountain populations of A. halleri in the Sudetes. The occurrence of this uncommon haplotype has also been recorded in high-mountain populations in the Western Carpathians as well as for one population in the Bohemian Forest. A similar pattern have been also recognized in Pulsatilla vernalis (Ronikier et al., 2008b). The most reasonable explanation for the observed pattern is the assumption that a contact between the Sudetic and Carpathian flora occurred in the past. Mitka et al. (2007) suggested that this contact could have occurred especially for high-mountain taxa, which could easily disperse within the open landscapes that were present between the Sudetes and the Carpathians during glacial maxima. This supposition is also supported by the floristic evidence showing that several high-mountain taxa such as Erigeron macrophyllus Herbich, Melampyrum herbichii Woł., Sesleria tatrae (Degen) Deyl and Thymus carpathicus Čelak. that are present in the Carpathians, occur also in high-mountain environments of the Sudetes (Pawłowski, 1969). The connections between the Carpathian and Sudetic populations of A. halleri surely require further studies.

Taxonomic concept of A. halleri in the Carpathians

Taxonomy of A. halleri is still much debated and to date three European subspecies have been recognised: subsp. halleri, subsp. tatrica (Pawł.) Kolník, and subsp. dacica (Heuff.) Kolník (Hohmann et al., 2014). This division is based on a morphological study of the Carpathian populations published by Kolník & Marhold (2006). Our results seem to question this taxonomic division that has already been cited by various authors (Clauss & Mitchell-Olds, 2006; Koch, Wernisch & Schmickl, 2008). We have shown that main genetic groups identified by cpDNA variation are not consistent with the division proposed by Kolník & Marhold (2006), nor with the distribution of the described taxa.