Flora 208 (2013) 453–463 Contents lists available at ScienceDirect Flora journal homepage: www.elsevier.com/locate/flora Phylogeography of the heathers Erica arborea and E. trimera in the afro-alpine ‘sky islands’ inferred from AFLPs and plastid DNA sequences Abel Gizaw a,b,∗ , Mulugeta Kebede a,b,1 , Sileshi Nemomissa a , Dorothee Ehrich b,2 , Biructawit Bekele a,b,3 , Virginia Mirré b , Magnus Popp b , Christian Brochmann b a b Department of Plant Biology and Biodiversity Management, College of Natural Sciences, Addis Ababa University, P.O. Box 3434, Addis Ababa, Ethiopia National Centre for Biosystematics, Natural History Museum, University of Oslo, P.O. Box 1172, Blindern, NO-0318 Oslo, Norway a r t i c l e i n f o Article history: Received 8 September 2012 Accepted 8 June 2013 Available online 2 August 2013 Keywords: Tree heather Afro-alpine Pleistocene refugia Montane forest bridge Ethiopia a b s t r a c t The ericaceous vegetation zone of the unique and highly fragmented afro-alpine environment in the eastern African high mountains is typically dominated by the heather Erica arborea, often in combination with its close relative E. trimera. Both species are shrubs or small trees with tiny seeds, potentially capable of dispersal by wind over long distances. While E. arborea is widely distributed in Africa, the Middle East and Europe, E. trimera is endemic to the afro-alpine region where it is restricted to higher altitudes than E. arborea. We used Amplified Fragment Length Polymorphisms (AFLPs) and variation in non-coding plastid DNA sequences to test whether these two morphologically and ecologically very similar species display similar phylogeographic patterns in the afro-alpine region. We predict that the more high-altitudinal E. trimera shows more distinct genetic structuring than E. arborea, because dispersal of the latter may have been facilitated by formation of interglacial forest bridges between mountains. Based on extensive field sampling in most of the high mountains of Ethiopia and East Africa, we show that the two species are clearly distinct at AFLP and plastid DNA loci. Both showed low levels of overall AFLP diversity, suggesting bottlenecking in small refugial populations during unfavourable climatic periods. However, their genetic structuring and inferred phylogeographic histories were conspicuously different. The more highaltitudinal E. trimera consisted of three to four distinct AFLP groups, which also had different plastid DNA haplotypes and different geographic distributions, suggesting long-term restriction to several refugia (at least one in Ethiopia and two in East Africa). In contrast, E. arborea showed little geographic structuring at AFLP loci and only a single, widespread plastid DNA haplotype, which may suggest recent colonization of the entire study area from a single source population, likely via a combination of gradual expansion via forest bridges and long-distance dispersals. The source population of E. arborea may be situated in (or north of) Ethiopia, which harbours most genetic diversity. © 2013 Elsevier GmbH. All rights reserved. Introduction ∗ Corresponding author at: Department of Plant Biology and Biodiversity Management, College of Natural Sciences, Addis Ababa University, P.O. Box 3434, Addis Ababa, Ethiopia. E-mail addresses: abegiz3@yahoo.com (A. Gizaw), kmulugetak@yahoo.com (M. Kebede), nemomissa@yahoo.co.uk (S. Nemomissa), dorothee.ehrich@uit.no (D. Ehrich), barkan bek@yahoo.com (B. Bekele), virginia.mirre@nhm.uio.no (V. Mirré), magnus.popp@nhm.uio.no (M. Popp), christian.brochmann@nhm.uio.no (C. Brochmann). 1 Current address: Department of Biology, Hawassa University, P.O. Box 05, Hawassa, Ethiopia. 2 Current address: Institute for Arctic and Marine Biology, University of Tromsø, NO-9037, Norway. 3 Current address: Jimma University, College of Agriculture and Veterinary Medicine, P.O. Box 307, Jimma, Ethiopia. 0367-2530/$ – see front matter © 2013 Elsevier GmbH. All rights reserved. http://dx.doi.org/10.1016/j.flora.2013.07.007 The afro-alpine region in tropical eastern Africa encompasses the elevated plateaus and mountains in Ethiopia and the isolated high mountain peaks in Kenya, Tanzania, and Uganda. The vegetation on these mountains, which are often referred to as biological “sky islands”, is characterized by a high proportion of endemic species and forms three distinct zones: the lowermost afro-montane forest zone, the transitional ericaceous or heather zone, and the uppermost afro-alpine zone proper (Hedberg, 1951). Phylogeographic studies on the genetic consequences of range shifts in response to climate change during the Pleistocene have provided new insights into the history of many species and identified important refugial areas. However, whereas many studies have addressed the histories of temperate, boreal and arctic species (for reviews see e.g. Brochmann et al., 2003; Hewitt, 2000; Taberlet 454 A. Gizaw et al. / Flora 208 (2013) 453–463 et al., 1998), only a few molecular studies have so far focused on plants from the afro-alpine region and their history in relation to past climate changes (Assefa et al., 2007; Désamoré et al., 2011; Ehrich et al., 2007; Gehrke and Linder, 2009; Kebede et al., 2007; Knox, 1993; Knox and Palmer, 1998; Koch et al., 2006; Popp et al., 2008). Only the studies of the afro-montane forest endemic Lobelia giberroa (Kebede et al., 2007) and the widespread afroalpine and arctic-alpine Arabis alpina (Assefa et al., 2007; Ehrich et al., 2007) focused on phylogeographic questions within the afroalpine region. The African tropics were colder and drier during the last glaciation (Bonnefille et al., 1990) and the afro-alpine proper and ericaceous zones were pushed down by ∼1000 m and covered larger areas than today (Flenley, 1979; Gottelli et al., 2004). In contrast, the extent of the afro-montane forest zone was reduced, in particular in its lower-lying parts, where open vegetation expanded in response to aridification (Bonnefille, 1995; deMenocal, 1995). During the warm and more humid inter-glacial periods, on the opposite, the afro-alpine and ericaceous zones contracted to higher elevations whereas the afro-montane forest expanded and may have connected originally isolated patches via temporary forest bridges between adjacent mountains (Kebede et al., 2007; Mohammed and Bonnefille, 1998; Ryner et al., 2006; Voje et al., 2009). Such range dynamics during the Pleistocene have resulted in repeated formation of highly fragmented habitats, with the species of the current ericaceous and afro-alpine zones restricted to isolated high mountain interglacial refugia (Hedberg, 1970). To extend our limited knowledge of the phylogeographic history of afro-alpine plants we here selected two heather species characteristic of the ericaceous zone, Erica arborea L. and E. trimera (Engl.) Beentje. Erica arborea is widespread with a disjunct distribution, occurring in Africa and in the Mediterranean region including the Atlantic coast of Spain, the Pyrenees and the Middle East including Yemen. In Africa, it is distributed in North Africa (Morocco and Tunisia), the Tibesti Mountains of Central Sahara (Chad), the Ethiopian Highlands, and in the mountains of East Africa southwards to Malawi (Messerli and Winiger, 1992; McGuire and Kron, 2005; Quezel, 1978). Erica arborea is morphologically variable across its range, also within eastern Africa, but the variation was considered by Hedberg (1957) to be too continuous to allow for taxonomic subdivision. In contrast, E. trimera is endemic to the afro-alpine region and occurs in several mountain systems of East Africa and the Ethiopian Highlands. This species exhibits complex morphological variation which is reflected by the contrasting treatments provided by different taxonomists. Five or six different subspecies, endemic to different mountains or mountain systems, were recognized by e.g. Hedberg (1957; as Philippia trimera), Beentje (2006), and Dorr (2006). The five subspecies accepted in Flora of Tropical East Africa (Beentje, 2006) are subsp. trimera (Ruwenzori Mts), subsp. keniensis (Mt. Kenya), subsp. elgonensis (Mt. Elgon), subsp. jaegeri (Mt. Meru and Mt. Loolmalasin), and subsp. kilimanjarica (Mt. Meru and Mt. Kilimanjaro). A sixth subspecies, subsp. abyssinica (the Ethiopian mountains) is accepted in Flora of Ethiopia and Eritrea (Hedberg and Hedberg, 2003). Both E. arborea and E. trimera are shrubs or small trees, varying from 0.3 to12 m in height, and they are morphologically similar except in a few floral characters. Erica arborea has actinomorphic flowers and glabrous pedicels with two or three small bracteoles, whereas E. trimera has zygomorphic flowers and glandular pedicels without or with only one, recaulescent bracteole (Hedberg and Hedberg, 2003). Both species produce tiny wind-dispersed seeds less than 1 mm long (Huckerby et al., 1972; Lind and Morrison, 1974), and can therefore be expected to have similar long-distance dispersal capacity. The similarities in morphology as well as in habitat ecology often make it difficult to distinguish the two species in the field (Hedberg, 1971; Johansson et al., 2010). They often co-occur in the ericaceous zone, but E. trimera tends to dominate at higher elevations and E. arborea at lower elevations (Hedberg, 1971; Miehe and Miehe, 1994; Wesche et al., 2000). According to Hedberg (1957), E. arborea occurs in the afro-montane forests at altitudes down to 2000 m, occasionally also below this level, and may in some mountains reach far up in the afro-alpine zone proper (to 4700 m), whereas E. trimera is restricted to altitudes between 3000 m and 4500 m. In a phylogenetic analysis of African and European Ericas it was suggested that the African taxa were derived from European ones and that E. arborea was sister to the African clade (McGuire and Kron, 2005). In contrast, a recent, range-wide phylogeographic study of E. arborea based on four plastid DNA loci provided strong evidence for origin of E. arborea in eastern Africa and expansion to other parts of the world in two waves, before and during the Pleistocene (Désamoré et al., 2011). Based on a preliminary phylogenetic analysis, the authors of this study suggested also that E. arborea and E. trimera are closely related and probably constitute sister species. They did not, however, address the phylogeographic history of E. arborea within eastern Africa. Based on extensive new field collections, we here use AFLP markers and plastid DNA sequences to describe the pattern of genetic variation and infer the history of both E. arborea and E. trimera in the afro-alpine region. We specifically ask whether these two morphologically and ecologically very similar species have similar phylogeographic histories in this region. In particular we investigate whether the geographic structure of genetic variation is consistent with long-term isolation in several refugia in different mountain areas, and/or whether their potentially large capacity for long-distance wind dispersal has resulted in poor geographic structuring among mountains. We predict that E. trimera shows the most distinct geographic structuring of the two, because it tends to be restricted to higher altitudes than E. arborea and therefore may rely on intermountain colonization solely via long-distance dispersal, whereas E. arborea, by extending down into the montane forest, also may have spread to other mountains via interglacial montane forest bridges. Materials and methods Plant material and DNA extraction Fresh leaf tissue of E. arborea and E. trimera was collected in the field from 11 high mountain systems in Ethiopia and Tropical East Africa (Kenya and Tanzania) in 2003/2004 and dried in silica gel (Appendix 1, Fig. 1). The silica material and voucher specimens were deposited at the National Herbarium of Addis Ababa University, Addis Ababa, Ethiopia, and duplicates of the silica material were also deposited at the Natural History Museum, University of Oslo (O). Whenever possible, several distant populations were sampled from each mountain, each of them represented by five individual plants. The two species are however morphologically very similar and difficult to distinguish except by some floral characters. As many of our populations were not in flower or fruit at the time of collection, their taxonomic assignment had to be re-evaluated after the genetic analysis was carried out. This procedure resulted in consistent separation of the species, since we identified two distinct genetic groups (cf. Results section) to which those specimens collected with flowers or fruits could be assigned. Although this approach resulted in a somewhat uneven representation of the two species in our sampling from the different mountains, it was sufficient to address the broad histories of both species. For the AFLP analysis we used a total of 169 individual plant samples collected in the field, of which 59 belonged to E. arborea and 110 to E. trimera. Of these 169 samples, nine samples of E. arborea A. Gizaw et al. / Flora 208 (2013) 453–463 455 Fig. 1. Genetic structure, diversity and distinctiveness in Erica arborea and E. trimera based on AFLP and plastid DNA data. The thick dark grey lines depict the Great Rift Valley. (a) Plastid DNA sequence variation in E. trimera. The network also includes the single haplotype observed in E. arborea (H-EA), with black dots indicating hypothetical intermediate haplotypes (extinct or unsampled) and circle size reflecting sample size. The pie diagrams on the map show haplotype frequencies; colours are as in the haplotype network. (b) AFLP variation in E. arborea. The size of the circles is proportional to within-mountain gene diversity (D) and the shading reflects genetic distinctiveness (DW). The four sizes or grey shades represent one quartile each of the distribution of D or DW, respectively. (c) Genetic structure as inferred from AFLP data for E. arborea. Each colour represents a distinct genetic (STRUCTURE) group and the size of the pie diagrams is relative to the number of individuals analyzed from different mountains. (d) AFLP variation in E. trimera represented as in (b). (e) Genetic structure of E. trimera represented as in (c). 456 A. Gizaw et al. / Flora 208 (2013) 453–463 and 14 of E. trimera were used in the plastid DNA sequencing. Additional samples for plastid DNA sequencing of E. arborea and E. trimera were obtained from silica or voucher specimens (Appendix A). Total genomic DNA was extracted from silica gel-dried leaf samples and from voucher specimens (for plastid DNA sequencing) using the Cetyltrimethyl ammonium bromide (CTAB) method (Doyle and Doyle, 1987) with minor modifications (Schönswetter et al., 2002) or the DNeasy Plant Mini Kit (Qiagen Nordic, Oslo, Norway) according to the manufacturer’s instructions. Plastid DNA sequencing Four geographically distant individuals of each species were selected for a preliminary test of sequence variation in five non-coding plastid DNA regions, psbE-petL, trnD-trnT, rps16, rpl32trnLUAG and trnGUUC -trnG2G. Based on amplification success and levels of intraspecific variation, we selected rpl32-trnLUAG and trnGUUC -trnG2G for final sequencing. We sequenced 1–3 individuals of each species from most mountains (Appendix A). The same forward and reverse primers were used for polymerase chain reaction (PCR) and sequencing of rpl32-trnLUAG (rpL32-F and trnLUAG ; Shaw et al., 2007) and trnGUUC -trnG2G (3′ trnGUUC and 5′ trnG2G; Shaw et al., 2005). Amplification was performed in a total reaction volume of 10 ␮L containing 0.4 U Taq DNA polymerase (Applied Biosystems, Foster City, CA, USA), 1× PCR buffer supplied with the enzyme, 2.5 mM Mg2+ , 0.4 ␮L of each primer, 1 mM of each dNTP (Applied Biosystems), 0.04% BSA and 3 ␮L template DNA of unknown concentration. PCR cycling was performed with a GeneAmp 3700 (Applied Biosystems, Foster City, CA, USA) or a PTC100 or PTC200 (MJ Research, Watertown, MA, USA) thermocycler with initial denaturation for 3 min at 95 ◦ C followed by 35 cycles of 30 s at 95 ◦ C, 30 s at 55 ◦ C, 2 min at 72 ◦ C, and ending with 10 min at 72 ◦ C with subsequent cooling at 6 ◦ C. PCR products were purified using 2 ␮L of ExoSAP-IT (USB Corp., Cleveland, OH, USA) diluted 1:10 for 5 ␮L of PCR products, and incubated for 45 min at 37 ◦ C. DNA sequencing was performed using BigDye V 1.1 (Applied Biosystems) according to the manufacturer’s manual, except for using 10 ␮L reaction volumes, and visualized with an ABI 3100 capillary sequencer (Applied Biosystems). AFLP fingerprinting AFLP was performed following Gaudeul et al. (2000) except that reaction volumes in the PCR were reduced by 50%. Twelve primer pairs were tested on four plants from two different mountain systems (Ethiopia and Tanzania). Six primer pairs resulting in AFLP profiles with many polymorphic markers and well separated fragments were used for a second primer test on 16 plants from the three main geographical regions (Ethiopia, Kenya, and Tanzania). Three of these primer pairs were chosen for the full analysis (fluorescent dye in parentheses): EcoRI AGA (6-FAM) – MseI CTC, EcoRI ACA (VIC) – MseI CAT, and EcoRI AAC (NED) – MseI CAG. For each sample, 2.0 ␮L 6-FAM, 2.0 ␮L VIC and 4.0 ␮L NED labelled selective PCR products were mixed with 11.7 ␮L formamide and 0.3 ␮L GeneScanTM 500 ROXTM internal-lane size standard and run on an ABI 3100 sequencer (Applied Biosystems). Whenever possible, one sample from each population was duplicated as a blind sample to test for reproducibility and contamination. Plastid DNA sequence analysis The sequences were edited using CodonCode aligner v.3.5.7 (CodonCode Corporation, Dedham, MA). Multiple sequence alignment was done using ClustalW in the SeaView v.4 program (Gouy et al., 2010) and adjusted manually. The sequences from the two regions were concatenated and then divided into two partitions for analysis. Best-fit DNA substitution models were determined for each data partitions using PAUP* 4.10b (Swofford, 2002) and the Akaike information criterion (AIC) as implemented in MrModeltest v.3.1 (Nylander, 2004). Bayesian phylogenetic analyses were performed for the concatenated dataset using MrBayes v.3.1.2 (Huelsenbeck and Ronquist, 2001; Ronquist and Huelsenbeck, 2003). Two independent analyses were run simultaneously, each starting from different random trees. Four Markov chain Monte Carlo (MCMC) chains, i.e., three ‘heated’, and one ‘cold’ chain, were run for 2 × 106 generations. Trees were sampled every 1000th generation and resulted in a total of 2 × 103 trees. The first 500 were discarded as the burn-in phase. Convergence between the two independent analyses was evaluated and confirmed using the standard deviation of split frequencies. The majority rule consensus phylogenetic tree with posterior probabilities was visualized with FigTree v.3.1 (Rambout, 2009). The tree was midpoint rooted. Gaps were treated as missing data and haplotypes were identified using DnaSP v.5 (Librado and Rozas, 2009). A maximum parsimony median-joining haplotype network was constructed using NETWORK v.4.6.0 (Bandelt et al., 1999). Haplotype and nucleotide diversity for the total samples of each species was estimated using DnaSP v.5 (Librado and Rozas, 2009). AFLP data analysis AFLP profiles were visualized and analyzed with ABI prism GENESCAN v.3.7 analysis software (Applied Biosystems) and imported into GENOGRAPHER v.1.6 available at http://hordeum. msu.montana.edu/genographer/. Each AFLP fragment was scored using the ‘thumbnail’ option of the program, which allows comparison of the signal per locus over all samples. Fragments in the size range of 50–500 base pairs (bp) were recorded as present (1) or absent (0). The average reproducibility was calculated for each species as the average proportion of correctly reproduced bands over all duplicated samples (Bonin et al., 2004). Genetic diversity was estimated for each species in total, for each mountain massif, for each sampling locality (population) within mountain massifs and for the identified genetic groups. We used the R-script AFLPdat (Ehrich, 2006) to estimate the proportion of polymorphic markers (P) and Nei’s gene diversity (D; corresponding to the average proportion of pair-wise differences between AFLP profiles, i.e. phenotypes, Kosman, 2003). In addition we estimated genetic distinctiveness for each mountain massif and population as frequency-down-weighted marker values (DW or rarity) following Schönswetter and Tribsch (2005), using AFLPdat. Rarity was estimated within each species separately. Similarity between AFLP multilocus phenotypes was quantified using Dice’s coefficient of similarity in NTSYSpc v.2.11 (Rohlf, 1998), and principal coordinate analyses (PCoAs) were used to visualize pair-wise similarities among AFLP phenotypes. Genetic mixture analyses were performed to identify the two species in the total dataset and to identify genetically homogeneous groups within each species. The appropriate number of groups (K) and the group to which each individual most likely belonged to were estimated using STRUCTURE (Pritchard et al., 2000). The program implements a model based clustering method using Bayesian MCMC estimation. The best value of K was chosen based on two criteria: the estimated posterior log probability of the data, L (K), and the stability of assignment patterns across runs. Since L (K) continued to grow slightly with increasing values of K, the most likely number of clusters was determined also by taking into account the rate of change in the probability between successive K,  K (Evanno et al., 2005). Our data were analyzed with the STRUCTURE program v.2.3.3 at the Bioportal, University of Oslo A. Gizaw et al. / Flora 208 (2013) 453–463 457 Table 1 Polymorphic sites, number of individuals carrying each haplotype (n) and haplotype frequency, total haplotype and nucleotide diversity for 17 sequenced individuals of E. trimera based on the concatenated matrix of the two non-coding plastid DNA regions with 1580 sites (after excluding gaps and missing data). The polymorphic sites at 221, 318, 974, and 1213 were parsimony informative. rpl32-trnLUAG trnGUUC -trnG2G Species Haplotype 178 221 318 974 1213 n E. trimera E. trimera E. trimera E. trimera E. trimera E. trimera E. trimera E. trimera H-1 H-2 H-3 H-4 H-5 H-6 H-7 H-8 G – – – – – T – T – – A A A – – A – – C C C C C C A A – A – – – A – C C C – – – 2 4 5 1 1 2 1 1 (http://www.bioportal.uio.no) for K = 1–10 and 10 replicates per K, a burn-in period of 2 × 105 and 106 iterations. We used the recessive allele model to take into account the dominant nature of AFLP data (Falush et al., 2007). For the analysis including both species we used a no admixture model with uncorrelated allele frequencies, as well as a model allowing for admixture and correlated allele frequencies. Assuming that populations are more closely related within species, we used only the latter for the within-species analyses. Population genetic structure and differentiation were also assessed using Analyses of Molecular Variance (AMOVAs) with Arlequin v.3.01 (Excoffier and Schneider, 2005). In addition to a non-hierarchical analysis among all sampled populations, AMOVAs were run with different hierarchical levels, as well as based on the genetic groupings inferred from the STRUCTURE analyses (here each individual was placed in the group where it had the largest proportion of ancestry). Results Plastid DNA variation While all 16 analyzed plants of E. arborea exhibited identical haplotypes, plastid DNA was variable among the 17 individuals of E. trimera. Eight different plastid DNA haplotypes (H-1 to H-8), with haplotype diversity of 0.87 and nucleotide diversity of 0.13% were observed in the latter species (Table 1). The Ethiopian haplotype H3 was most frequently observed in our samples (frequency 0.29), whereas rare haplotypes occurred on Mt. Kilimanjaro (H-4), Mt. Meru (H-5) and Mt. Elgon (H-7 and H-8). The final concatenated data matrix for both species together consisted of 33 sequences and 1632 nucleotide sites. The rpl32-trnLUAG region consisted of 924 sites, and the trnGUUC -trnG2G region consisted of 708 sites. A 52 bp indel was identified in the rpl32-trnLUAG region. Excluding this indel, the E. trimera matrix contained 1580 sites, of which five were polymorphic (four parsimony informative and one singleton). The Hasegawa, Kishino and Yano model (HKY; allows for different transition and transversion rates) and the generalized time reversible with invariant sites model (GTR + I; allows for different rates between all pairs of bases, but constrains the rates to be symmetric [ie, a → c = c → a], and allows for a certain proportion of invariant sites) were selected as best-fit models for the rpl32trnLUAG and trnGUUC -trnG2G regions, respectively. Both species were strongly supported as monophyletic with posterior probability (PP) value of 1. The relationships among the eight plastid DNA haplotypes observed in E. trimera were poorly resolved with some exceptions. The Ethiopian ones (H-1–H-3) grouped together with strong support (PP = 1; Fig. 2). The haplotypes from Mt. Elgon (H-7 and H-8) formed a polytomy with the Ethiopian clade in a strongly supported clade (PP = 0.99; Fig. 2). In the plastid DNA haplotype Haplotype frequency Haplotype diversity Nucleotide diversity 0.118 0.235 0.294 0.059 0.059 0.118 0.059 0.059 0.868 0.001 network, (Fig. 1a), the Ethiopian haplotypes were placed at one end and the East African haplotypes at the other, with the two Mt. Elgon haplotypes in the middle. All haplotypes of E. trimera were connected by single mutational steps, whereas the single haplotype observed in E. arborea was separated by eight mutational steps from the nearest haplotype of E. trimera. AFLP variation The average reproducibility of the AFLP markers was 96%. Among the 106 markers retained in the final dataset, 55% were polymorphic in E. arborea (n = 59) and 69% in E. trimera (n = 110). In the ordination plot (PCoA) of the total dataset, the two species were clearly separated along the first axis, which extracted 49.1% of the variation (Fig. 3a). The second axis, which extracted only 6.3% of the variation, mainly depicted intraspecific variation in E. trimera. Similarly, the genetic mixture analysis (STRUCTURE) of the total dataset identified K = 2 as the most likely number of clusters. These clusters corresponded to the two species (Fig. 4a). The results were nearly identical for the two models applied. In E. arborea, gene diversity showed a geographical trend with higher values in Ethiopia and lower values in East Africa (Fig. 1b). Except for Chilallo and Kaka, all the remaining mountains in Ethiopia and East Africa showed quite similar amounts of genetic distinctiveness, with the highest value observed on Mt. Elgon (Fig. 1c). The total gene diversity estimates were P = 54.72% and D = 0.118 (Appendix 1), whereas average within-population gene diversity was 0.080 (SD = 0.23). Within-population gene diversity varied from 0.100 in one Ethiopian population (Simen AFR 065) to 0.053 in one East African population (Elgon AFR 206). Genetic distinctiveness was highest in one population from Mt. Elgon and lowest in one Ethiopian population from Chilallo (Appendix 1). Considering all individuals sampled on each mountain, gene diversity ranged from 0.094 in the Ethiopian Simen to 0.053 in the East African Mt. Elgon. Distinctiveness, on the opposite, was highest on Mt. Elgon and lowest on Kaka in Ethiopia (Table 2). Genetic mixture analyses of the E. arborea dataset revealed four clusters as L (K) increased steadily up to K = 4 (Fig. 4b). The rate of change in the probability between successive Ks,  K, also identified K = 4 as the most likely number of clusters (Fig. 4b). The four STRUCTURE groups were somewhat separated on the PCoA plot (Fig. 3b), but plants from the same mountain were often divided between the groups, and several individuals had shared ancestry. The same was the case for the clustering into two groups, for which the 10 runs also converged on an identical result (Fig. 4b). Although different genetic groups were dominating in different mountains, there was no consistent geographic pattern. Whereas Chilallo and Kaka, which are adjacent, were similar, the other Ethiopian mountains were each dominated by their own genetic group. The groups present on Chilallo and Kaka were 458 A. Gizaw et al. / Flora 208 (2013) 453–463 Fig. 2. Midpoint rooted phylogeny inferred from the concatenated sequences of the two non-coding plastid DNA regions (rpl32-trnLUAG and trnGUUC -trnG2G) for the widespread Erica arborea and for the afro-alpine endemic E. trimera. Posterior probability values are indicated above the branches. Plastid DNA haplotypes are indicated for each accession. also found further south on Mt. Aberdare and Elgon. The nonhierarchical AMOVA assigned 33.5% of the overall genetic variation to variation among the 11 populations (Table 3). The hierarchical AMOVAs assigned 31.3% of the variation to among-mountain variation, 5.2% to among-populations within mountains variation and 63.5% to within-population variation (Table 3). In another analysis, 20.8% and 22.0% of the variation were found among the four genetic groups and among-populations-within-group, respectively (Table 3). In E. trimera, on the contrary, we observed no distinct geographical trend in gene diversity or distinctiveness (Fig. 1d and e). The Ethiopian and East Africa populations of this species tended to be quite similar, particularly for gene diversity, except that a single population from Mt. Aberdare with three individuals showed Table 2 Average gene diversity and genetic distinctiveness in different mountains and genetic (STRUCTURE) groups based on the AFLP data for E. arborea and E. trimera. n – number of individuals successfully analyzed, P – percentage of polymorphic loci, D – gene diversity, DW – frequency-down-weighted marker value as a measure of genetic distinctiveness. Numbers in parenthesis indicates standard deviation. Species Mountain/genetic group n P (%) D DW E. arborea E. arborea E. arborea E. arborea E. arborea E. arborea Chilallo Gara Muleta Kaka Simen Aberdare Elgon 7 21 5 16 5 5 19.81 30.19 13.21 33.96 16.04 12.26 0.071 0.093 0.070 0.094 0.072 0.053 0.69 0.90 0.66 1.16 1.13 1.36 0.98 (0.26) E. trimera E. trimera E. trimera E. trimera E. trimera E. trimera E. trimera E. trimera E. trimera Bale Chilallo Choke Kaka Aberdare Elgon Kenya Kilimanjaro Meru 16 10 20 14 3 4 17 5 21 27.36 21.7 33.02 29.25 16.04 12.26 26.42 15.09 25.47 0.083 0.071 0.086 0.083 0.107 0.068 0.083 0.072 0.074 0.54 0.56 0.61 0.56 1.31 0.83 0.82 0.78 0.65 0.74 (0.23) E. trimera E. trimera E. trimera ‘Ethiopia’ group ‘Aberdare-Kenya’ group ‘Elgon-Kilimanjaro-Meru’ group 60 20 30 50.94 32.08 34.91 0.094 0.090 0.083 0.57 0.89 0.70 A. Gizaw et al. / Flora 208 (2013) 453–463 459 relatively high level of diversity and distinctiveness. The total gene diversity estimates in E. trimera were P = 68.87%, and D = 0.128 (Appendix 1), and average within-population gene diversity was 0.076 (SD = 0.14). Within-population gene diversity varied from 0.107 in the Aberdare population (Aberdare AFR 137) to 0.038 in an Ethiopian population (Chilallo AFR 048). Genetic distinctiveness was also highest in the Aberdare population and lowest in one Ethiopian population (Bale; Appendix 1). Pooling all individuals sampled on each mountain, gene diversity ranged from 0.107 in Mt. Aberdare to 0.068 in another East Africa mountain, Mt. Elgon. Genetic distinctiveness was also highest in Mt. Aberdare and lowest in the Ethiopian Bale (Table 2). Genetic mixture analysis identified three or four clusters as the most likely subdivisions for the E. trimera dataset. The largest increase in L (K) occurred for K = 2 and K = 3, but also K = 4 lead to a consistent increase in L (K) (Fig. 4c). The rate of change in the probability between successive Ks,  K, identified K = 4 as the most likely number of clusters (Fig. 4c). In addition to the rate of change in probability,  K is highly sensitive to changes in the variance among probabilities for a different K, an aspect which may have contributed to the clear signal for K = 4 (Evanno et al., 2005). For K = 3 STRUCTURE identified the following groups, which were geographically distinct with no mixture among mountains (Fig. 3c): the ‘Ethiopia’ group, including all plants from the Ethiopian mountains; the ‘Aberdare-Kenya’ group, including all plants from the East African mountains Mt. Aberdare and Mt. Kenya; and the ‘ElgonKilimanjaro-Meru’ group, including all plants from the remaining East African mountains, Mt. Elgon, Kilimanjaro and Meru. For K = 4, Bale became separated from the other Ethiopian populations. This subdivision was, however, based on a weaker genetic structuring than the three main groups as indicated by the smaller increase in L (K) between K = 3 and K = 4, and by the considerable smaller allele frequency divergence between Bale and the three other Ethiopian mountains compared to the divergence among the other groups (0.216 versus an average of 0.576, SD = 0.07). In this species, the main genetic groups inferred from the STRUCTURE analysis were quite distinct in the PCoA, in which the first and second axis explained 26.2% and 10.4% of the variation, respectively. The nonhierarchical AMOVA assigned 41.0% of the overall genetic variation to variation among the 20 populations (Table 3). The hierarchical AMOVAs assigned 39.0% of the variation to among-mountain variation and 4.0% to among-populations-within-mountain variation. Of the total genetic variation, 39.0% was assigned to variation among the three main genetic groups (Table 3). The AMOVAs for each genetic group attributed the highest proportions (78–84%) to variation within populations, and low levels of differentiation among mountains (13–24%; Table 3). Gene diversity within the identified genetic groups of E. trimera ranged from 0.094 in the ‘Ethiopia’ group to 0.083 in the ‘ElgonKilimanjaro-Meru’ group, with 0.090 in the ‘Aberdare-Kenya’ group (Table 2). Genetic distinctiveness was highest in the ‘AberdareKenya’ group and lowest in the ‘Ethiopia’ group (Table 2). The three genetic (AFLP) groups in E. trimera were characterized by different plastid DNA haplotypes, and they were consistent with the plastid DNA haplotype network in separating the Ethiopian populations from the East African ones. Discussion Fig. 3. PCoAs based on Dice similarity among AFLP phenotypes of Erica arborea and E. trimera. The percentages of the variation explained are indicated on each axis. (a) Total dataset including both species. (b) E. arborea, and (c) E. trimera. Shading pattern of the symbols in E. arborea and encircled individuals in the total dataset and E. trimera indicate the main genetic groups inferred by STRUCTURE. Based on new extensive field sampling from most of the high mountains of Ethiopia and East Africa, we have shown that although the two heathers Erica arborea and E. trimera are closely related and easily confused in the field, they are clearly distinct at AFLP loci as well as at plastid DNA loci. We have also demonstrated that their genetic structure is conspicuously different, in spite of the facts 460 A. Gizaw et al. / Flora 208 (2013) 453–463 Fig. 4. Genetic structure of Erica arborea and E. trimera inferred from Bayesian clustering of AFLP data. Log probability of the data L (K) as function of K ranging from one to 10 as estimated by the program STRUCTURE, and the rate of change in the probability between successive runs,  K as a function of K, calculated according to Evanno, et al. (2005). (a) Total dataset. (b) E. arborea. (c) E. trimera. that both have very tiny seeds and potentially are capable of longdistance dispersal via wind, that they have similar distributions and habitat ecology in the afro-alpine mountain systems, and that they frequently co-occur at the same sites. Both of them do however harbour low levels of overall AFLP diversity, suggesting similar histories in terms of diversity loss via bottlenecking caused by isolation in small refugial populations during unfavourable climatic periods. In this study, we initially addressed whether the geographic structure of genetic variation in these two species is consistent with long-term isolation in different mountain areas, and/or whether their potentially large capacity for long-distance dispersal has resulted in weak genetic structuring among mountains. We found evidence for both of these main patterns. Erica trimera consisted of three to four distinct AFLP groups, which also had different plastid DNA haplotypes and different geographic distributions, suggesting long-term isolation in different areas, one or possibly two situated in Ethiopia and two in East Africa (Figs. 1d–e, 2 and 3c). In marked contrast to the main pattern observed in E. trimera, E. arborea showed little geographic structuring at AFLP loci and only a single, widespread plastid DNA haplotype. A possible explanation for this pattern may be that this species recently colonized the entire area studied here from a single source population. This source population of E. arborea may have been situated in Ethiopia, as indicated by the higher genetic diversity observed in this area (Fig. 1b–c). This suggestion is supported by the result from the STRUCTURE analysis, which revealed several genetic groups in the Ethiopian mountains, whereas the populations from Mt. Elgon and Mt. Aberdare belonged to two of the groups occurring in Ethiopia. Alternatively, the source population of the contemporary afro-alpine populations studied here may be situated outside our study area, possibly to the north of it on the Arabian Peninsula. This hypothesis is most consistent with the results of the range-wide plastid DNA study of E. arborea by Désamoré et al. (2011), where eastern Africa/Arabia was identified as the region of origin of the species. Notably, they identified several, closely related plastid DNA haplotypes in eastern Africa, in contrast to the single one found in our study. This difference may be explained by the use of different plastid DNA regions in the two A. Gizaw et al. / Flora 208 (2013) 453–463 461 Table 3 Analysis of molecular variance (AMOVA) for E. arborea and E. trimera based on the AFLP data. P-values were estimated in a permutation test (10,000 permutations). Source of Variation d.f. Sum of squares E. arborea Among populations Within populations Among mountains Among populations within mountains Within populations Among four genetic groups Among populations within groups Within populations 10 48 5 5 48 3 9 46 157.08 206.04 126.22 30.87 206.04 111.15 73.25 178.72 E. trimera Among populations Within populations Among mountains Among populations within mountains Within populations Among three genetic groups Among populations within groups Within populations 19 90 8 11 90 2 17 90 ‘Ethiopia’ group Among mountains Among populations within mountains Within populations Variance components Percentage of variation P-value 2.16 4.30 2.12 0.35 4.30 1.41 1.45 3.89 33.46 66.54 31.32 5.22 63.46 20.77 22.03 57.20 P < 0.0001 371.83 368.99 311.09 60.74 368.99 226.99 144.84 368.99 2.85 4.10 2.81 0.28 4.10 3.16 0.85 4.10 41.03 58.97 39.03 3.96 57.01 39.02 10.4 50.58 P < 0.0001 3 7 49 51.48 43.89 199.68 0.67 0.47 4.08 12.87 9.03 78.11 P = 0.0078 P < 0.0001 ‘Elgon-Kilimanjaro-Meru’ group Among mountains Among populations within mountains Within populations 2 2 25 23.90 7.16 97.08 1.20 −0.05 3.88 23.71 −0.94 77.23 P = 0.1036 P = 0.6491 ‘Aberdare-Kenya’ group Among mountains Among populations within mountains Within populations 1 2 16 8.73 9.69 72.23 0.78 0.07 4.52 14.47 1.31 84.22 P = 0.2443 P = 0.2727 studies. Our mainly nuclear and genome-wide AFLP data are nevertheless most consistent with recent colonization of our entire study area, possibly from the north. It is difficult to pinpoint ecological or historical differences between the two species that can fully account for their conspicuously different genetic structure in the afro-alpine region. However, part of the explanation may be found in their different altitudinal preferences, which led us to predict more distinct structuring in E. trimera. Since this species is restricted to the ericaceous belt on the transition to the afro-alpine proper, occasionally occurring also above this zone but not extending downwards into the montane forest, it has probably relied on intermountain colonization solely via long-distance dispersal. It is likely that dispersal of E. arborea, on the other hand, which has a larger altitudinal span in the mountains and extends further down in the montane forest zone, also may have been facilitated by formation of montane forests bridging neighbouring mountains during the warm and humid interglacials, when the species may have been more widely distributed (Lösch and Fischer, 1994), as suggested for the forest species Lobelia giberroa (Kebede et al., 2007). Gradual expansion of E. arborea via recent montane forest bridges, possibly in the current interglacial, may in particular have contributed to the genetic connectivity observed among some of the mountains in Ethiopia. Nevertheless, forest bridges can probably not explain the connectivity observed across the Great Rift Valley in Ethiopia, nor between the Ethiopian and the East African mountains, which is better explained by long-distant dispersal. Thus, although both E. arborea and E. trimera have small wind-dispersed seeds, it is possible that the former nevertheless has higher long-distance dispersal ability. This suggestion is corroborated by the much larger geographic range occupied by E. arborea and by its large-scale phylogeography, indicating two waves of colonization from eastern Africa/Arabia to the Mediterranean Basin (Désamoré et al., 2011). P < 0.0001 P = 0.2248 P < 0.0001 P < 0.0001 P < 0.0001 P < 0.0001 P < 0.0001 Our estimates of the total gene diversity pooled over all genotyped individuals of each species indicated very low levels of diversity in both of them (0.118 for E. arborea and 0.128 for E. trimera). In her survey of published intraspecific genetic diversity estimates for plants, Nybom (2004) reported an average AFLPbased diversity value of 0.23 with a standard deviation of 0.08. The highest levels of gene diversity were found for long-lived perennial (0.25), outcrossing (0.27), and late successional taxa (0.30), whereas the lowest levels were found for annual (0.13), selfing (0.12), and early successional taxa (0.17). The two Erica species, which both are long-lived perennials and most likely outcrossing species, thus displayed much less gene diversity than expected from their life-history traits. Their low levels of diversity are probably best explained by survival through climatically unfavourable periods in small refugial populations, where genetic drift reduced their diversity, and by further diversity loss by founder effects during long-distant population expansion in more favourable climatic periods. Notably, the Mt. Elgon populations of both species displayed particularly low level of gene diversity (D = 0.053 for E. arborea and D = 0.068 for E. trimera), suggesting that both species could have recently colonized this mountain via long-distant dispersal resulting in a few founder individuals. This explanation would be supported if both species harboured few private or rare markers in this mountain, as expected in recently founded populations. However, the fairly high genetic distinctiveness we observed in the Mt. Elgon populations of both species suggests a more long-term history, with DW values much higher (E. arborea: 1.36) or higher than the average (E. trimera: 0.83; Table 2). It is therefore most likely that the current Mt. Elgon populations of both species, once founded by few individuals, may have remained isolated and survived several unfavourable climatic periods with small population size. This hypothesis is supported by the plastid DNA sequence data for E. 462 A. Gizaw et al. / Flora 208 (2013) 453–463 trimera, in which the Mt. Elgon population harboured two private plastid DNA haplotypes (Figs. 1a and 2). In E. trimera, both the PCoA and STRUCTURE analyses of the AFLP data provided evidence for presence of three to four distinct genetic entities with different geographic distributions, and these were also characterized by different plastid DNA haplotypes. These entities correspond to some extent to previously recognized subspecies (Beentje, 2006; Dorr, 2006; Hedberg and Hedberg, 2003). Hedberg (1957) analyzed both quantitative and qualitative morphological characters to group the populations of E. trimera (as Philippia trimera) from different mountains or mountain groups into six subspecies, of which several had previously been recognized as separate species. He regarded the nominal subspecies, subsp. trimera, to be endemic to Mt. Ruwenzori, from where we had no material available for analysis. Our AFLP-based ‘Ethiopia’ and ‘Aberdare-Kenya’ groups correspond to his subsp. abyssinica and subsp. keniensis, respectively. Subspecies abyssinica was distinguished from all other subspecies, which have glandless corollas and usually glandless calyces, by having numerous glands on the corolla and calyx. Hedberg’s conclusion that all Ethiopian populations belong to a single subspecies is further supported by our AFLP data. The plants from different Ethiopian mountains were to a large degree intermingled in our PCoA analysis (Fig. 3c), and the two STRUCTURE groups identified in this region were much less divergent than the other groups. The remaining subspecies (i.e. all except subsp. abyssinica) were mainly distinguished by the amount and type of pubescence on the leaves and stems and partly by the size of floral parts, but Hedberg (1957) emphasized that he found only very small differences in measurable characters to differentiate them. We nevertheless found subsp. keniensis to correspond to a distinct genetic group, although the morphological characteristics given for this subspecies by Hedberg (1957) also seem very minor. We also found that the populations from Mt. Aberdare, which is situated very close to Mt. Kenya and from which Hedberg (1957) had no material available, most likely also belong to subsp. keniensis, although they are slightly separated from the Mt. Kenya populations in our PCoA analysis (Fig. 3c). The third genetic group recognized in E. trimera based on our AFLP analysis, the ‘Elgon-Kilimanjaro-Meru’ group, thus corresponds to all three remaining subspecies recognized by Hedberg (1957) in his morphological analysis. These subspecies were described as single-mountain endemics (subsp. elgonensis from Mt. Elgon and subsp. kilimanjarica from Mt. Kilimanjaro) or as endemic to two neighbouring mountains (subsp. jaegeri from Mt. Meru and Mt. Loolmalasin). It was not surprising that we found two of these to belong to the same genetic group, as they occur on very closely situated mountains (ssp. kilimanjarica and ssp. jaegeri), and the plants from these mountains were intermingled in the PCoA analysis (Fig. 3c). This finding is in agreement with the merging of these two subspecies by Dorr (2006). More surprisingly, although situated much further away, the Mt. Elgon populations also belonged to this genetic group (Fig. 1d–e). The Mt. Elgon populations were however somewhat divergent in the PCoA analysis (Fig. 3c). To summarize, we found some, but not perfect, correspondence between the degree of genetic differentiation, the degree of reported morphological (subspecies) differentiation, and geographic separation of the mountains, suggesting that morphological and genetic differentiation within E. trimera may proceed at different rates and that the overall patterns are obscured by occasional long-distant colonisations. Further studies combining morphological and genetic data and inclusion of the type locality of E. trimera (Mt. Ruwenzori) are needed to obtain a refined treatment of its intraspecific taxonomy. Conclusions Based on AFLPs and plastid DNA sequence variation from two non-coding regions, the present study revealed contrasting phylogeographic histories for two morphologically and ecologically very similar species in the afro-alpine region. The AFLP data showed low levels of gene diversity in both species, suggesting bottlenecking in small refugial populations during unfavourable climatic periods. Despite sharing low levels of gene diversity, the genetic structuring and the phylogeographic histories inferred for the two species were clearly different. While E. trimera, which is restricted to higher altitudes and endemic to eastern Africa exhibited geographically clearly structured genetic variation with three to four distinct AFLP groups with different plastid DNA haplotypes, E. arborea showed virtually no geographic structuring at AFLP loci and only a single, widespread plastid DNA haplotype (based on the DNA regions sequenced in this study). The phylogeographic pattern observed in E. trimera suggests long-term restriction to several refugia, whereas the conspicuous lack of geographic structuring in E. arborea is most consistent with recent colonization of the entire study area from a single source population. Acknowledgements We thank the Bale Mountains National Park and the Simen Mountains National Park in Ethiopia, the Tanzanian Commission for Science and Technology, and the National Museums of Kenya for permission to conduct field work. Particularly, we are grateful to H. Beentje for his valuable advice concerning the taxonomy of Erica, and to Adane Assefa for his company during most of the field work. We thank the Institute of Biodiversity and Conservation for permission to take samples from Ethiopia to Norway for laboratory analyses. The study was funded by the NUFU Programme (The Norwegian Programme for Development, Research and Higher Education) under two projects to S. Nemomissa and C. Brochmann (54/2003: ‘AFROALP – Afro-alpine ‘sky islands’ as natural laboratories: dynamics and units of plant biodiversity’ and 2007/1058: ‘AFROALP-II – Afro-alpine ‘sky islands’: genetic versus taxonomic biodiversity, climate change, and conservation’). Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.flora.2013.07.007. References Assefa, A., Ehrich, D., Taberlet, P., Nemomissa, S., Brochmann, C., 2007. Pleistocene colonization of afro-alpine ‘sky islands’ by the arctic-alpine Arabis alpina. Heredity 99, 133–142. Bandelt, H.-J., Forster, P., Röhl, A., 1999. Median-joining networks for inferring intraspecific phylogenies. Mol. Biol. Evol. 16, 37–48. Beentje, H.J., 2006. Flora of Tropical East Africa – Ericaceae. Royal Botanical Gardens Kew, Kew, UK. Bonin, A., Bellemain, E., Eidesen, P.B., Pompanon, F., Brochmann, C., Taberlet, P., 2004. How to track and assess genotyping errors in population genetics studies. Mol. Ecol. 13, 3261–3273. Bonnefille, R., 1995. A reassessment of the Plio-Pleistocene pollen record of East Africa. In: Vrba, E.S., Denton, G.H., Partridge, T.C., Burckle, L.H. (Eds.), Paleoclimate and Evolution with Emphasis on Human Origins. Yale University Press, New Haven, pp. 299–310. Bonnefille, R., Roeland, J.C., Guiot, J., 1990. Temperature and rainfall estimates for the past 40 000 years in equatorial Africa. Nature 346, 347–349. Brochmann, C., Gabrielsen, T.M., Nordal, I., Landvik, J.Y., Elven, R., 2003. Glacial survival or tabula rasa? The history of North Atlantic biota revisited. Taxon 52, 417–450. deMenocal, P.B., 1995. Plio-Pleistocene African climate. Science 270, 53–59. Désamoré, A., Laenen, B., Devos, N., Popp, M., Gonzalez-Mancebo, J.M., Carine, M.A., 2011. Out of Africa: north-westwards Pleistocene expansions of the heather Erica arborea. J. Biogeogr. 38, 164–176. A. Gizaw et al. / Flora 208 (2013) 453–463 Dorr, L.J., 2006. New combinations in Erica (Ericaceae: Ericoideae) from the high mountains of East Africa. Novon 16, 56–58. Doyle, J.J., Doyle, J.L., 1987. A rapid DNA isolation procedure for small amounts of fresh leaf tissue. Phytochem. Bull. 19, 11–15. Ehrich, D., 2006. AFLPdat: a collection of R functions for convenient handling of AFLP data. Mol. Ecol. Notes 6, 603–604. Ehrich, D., Gaudeul, M., Assefa, A., Koch, M.A., Mummenhoff, K., Nemomissa, S., Brochmann, C., 2007. Genetic consequences of Pleistocene range shifts: contrast between the Arctic, the Alps and the East African mountains. Mol. Ecol. 16, 2542–2559. Evanno, G., Regnaut, S., Goudet, J., 2005. Detecting the number of clusters of individuals using the software Structure: a simulation study. Mol. Ecol. 14, 2611–2620. Excoffier, L.L.G., Schneider, S., 2005. ARLEQUIN version 3.0: an integrated software package for population genetics data analysis. Evol. Bioinform. Online 1, 47–50. Falush, D., Stephens, M., Pritchard, J., 2007. Inference of population structure using multilocus genotype data: dominant markers and null allele. Mol. Ecol. Notes 7, 574–578. Flenley, J., 1979. The Equatorial Rain Forest: A Geological History. Butterworth, London. Gaudeul, M., Taberlet, P., Till-Bottraud, I., 2000. Genetic diversity in an endangered alpine plant, Eryngium alpinum L. (Apiaceae), inferred from amplified fragment length polymorphism markers. Mol. Ecol. 9, 1625–1637. Gehrke, B., Linder, H.P., 2009. The scramble for Africa: pan-temperate elements on the African high mountains. Proc. R. Soc. Lond. Ser. B: Biol. Sci. 276, 2657–2665. Gottelli, D., Marino, J., Sillero-Zubiri, C., Funk, S.M., 2004. The effect of the last glacial age on speciation and population genetic structure of the endangered Ethiopian wolf (Canis simensis). Mol. Ecol. 13, 2275–2286. Gouy, M., Guindon, S., Gascuel, O., 2010. SeaView version 4: a multiplatform graphical user interface for sequence alignment and phylogenetic tree building. Mol. Biol. Evol. 27, 221–224. Hedberg, O., 1951. Vegetation belts of the East African mountains. Svensk Bot. Tidskr. 45, 140–202. Hedberg, O., 1957. Afroalpine vascular plants. A taxonomic revision. Symb. Bot. Uppsal. 15, 1–411. Hedberg, O., 1970. Evolution of the afroalpine flora. Biotropica 2, 16–23. Hedberg, O., 1971. The high mountain flora of the Galama Mountain in Arussi Province, Ethiopia. Webbia 26, 101–128. Hedberg, I., Hedberg, O., 2003. Ericaceae. In: Hedberg, I., Edwards, S., Sileshi, N. (Eds.), Flora of Ethiopia and Eritrea. Volume 4, Part 1. Apiaceae to Dipsacaceae. The National Herbarium, Addis Ababa University, Addis Ababa, pp. 46–48. Hewitt, G., 2000. The genetic legacy of the Quaternary ice ages. Nature 405, 907–913. Huckerby, E., Marchant, R., Oldfield, F., 1972. Identification of fossil seeds of Erica and Calluna by scanning electron microscopy. New Phytol. 71, 387–392. Huelsenbeck, J.P., Ronquist, F., 2001. MRBAYES: Bayesian inference of phylogenetic trees. Bioinformatics 17, 754–755. Johansson, M., Rooke, T., Fetene, M., Granström, A., 2010. Browser selectivity alters post-fire competition between Erica arborea and E. trimera in the sub-alpine heathlands of Ethiopia. Plant Ecol. 207, 149–160. Kebede, M., Ehrich, D., Taberlet, P., Nemomissa, S., Brochmann, C., 2007. Phylogeography and conservation genetics of a giant lobelia (Lobelia giberroa) in Ethiopian and Tropical East African mountains. Mol. Ecol. 16, 1233–1243. Knox, E.B., 1993. The conservation status of the giant senecios and giant lobelias in Eastern Africa. Opera Bot. 121, 195–216. Knox, E.B., Palmer, J.D., 1998. Chloroplast DNA evidence on the origin and radiation of the giant Lobelias in Eastern Africa. Syst. Bot. 23, 109–149. Koch, M.A., Kiefer, C., Ehrich, D., Vogel, J., Brochmann, C., Mummenhoff, K., 2006. Three times out of Asia Minor: the phylogeography of Arabis alpina L. (Brassicaceae). Mol. Ecol. 15, 825–839. 463 Kosman, E., 2003. Nei’s gene diversity and the index of average differences are identical measures of diversity within populations. Plant Pathol. 52, 533–535. Librado, P., Rozas, J., 2009. DnaSP v5: a software for comprehensive analysis of DNA polymorphism data. Bioinformatics 25, 1451–1452. Lind, E.M., Morrison, M.E.S., 1974. East African Vegetation. Longman, London. Lösch, R., Fischer, E., 1994. Vikariierende Heidebuschwälder und ihre Kontaktgesellschaften in Makaronesien und Zentralafrika. Phytocoenologia 24, 695–720. McGuire, A.F., Kron, K.A., 2005. Phylogenetic relationships of European and African ericas. Int. J. Plant Sci. 166, 311–318. Messerli, B., Winiger, M., 1992. Climate, environmental change, and resources of the African mountains from the Mediterranean to the equator. Mt. Res. Dev. 12, 315–336. Miehe, G., Miehe, S., 1994. Ericaceous Forests and Heathlands in the Bale Mountains of South Ethiopia: Ecology and Man’s Impact. Stiftung Walderhaltung in Africa, Hamburg. Mohammed, M.U., Bonnefille, R., 1998. A late Holocene pollen record from a highland peak at Tamsaa, Bale Mountains, South Ethiopia. Global Planet Change 16/17, 121–129. Nybom, H., 2004. Comparison of different nuclear DNA markers for estimating intraspecific genetic diversity in plants. Mol. Ecol. 13, 1143–1155. Nylander, J.A.A., 2004. MrModeltest v2. Program distributed by the author. Evolutionary Biology Centre, Uppsala University, Uppsala. Popp, M., Gizaw, A., Nemomissa, S., Suda, J., Brochmann, C., 2008. Colonization and diversification in the African ‘sky islands’ by Eurasian Lychnis L. (Caryophyllaceae). J. Biogeogr. 35, 1016–1029. Pritchard, J.K., Stephens, M., Donnelly, P., 2000. Inference of population structure using multilocus genotype data. Genetics 155, 945–959. Quezel, P., 1978. Analysis of the flora of Mediterranean and Saharan Africa. Ann. Mo. Bot. Gard. 65, 479–534. Rambout, A., 2009. FigTree v.3.1, http://tree.bio.ed.ac.uk/software/figtree/ Rohlf, F.J., 1998. NTSYS-PC Numerical Taxonomy and Multivariate Analysis System. Exeter Publications, Setauket, New York. Ronquist, F., Huelsenbeck, J.P., 2003. MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics 19, 1572–1574. Ryner, M.A., Bonnefille, R., Holmgren, K., Muzuka, A., 2006. Vegetation changes in Empakaai Crater, northern Tanzania, at 14 000-9300 cal yr BP. Rev. Palaeobot. Palynol. 140, 163–174. Schönswetter, P., Tribsch, A., 2005. Vicariance and dispersal in the alpine perennial Bupleurum stellatum L. (Apiaceae). Taxon 54, 725–732. Schönswetter, P., Tribsch, A., Barfuss, M., Niklfeld, H., 2002. Several Pleistocene refugia detected in the high alpine plant Phyteuma globulariifolium Sternb. and Hoppe (Campanulaceae) in the European Alps. Mol. Ecol. 11, 2637–2647. Shaw, J., Lickey, E.B., Beck, J.T., Farmer, S.B., Liu, W., Miller, J., Siripun, K.C., Winder, C.T., Schilling, E.E., Small, R.L., 2005. The tortoise and the hare II: relative utility of 21 noncoding chloroplast sequences for phylogenetic analysis. Am. J. Bot. 92, 142–166. Shaw, J., Lickey, E.B., Schilling, E.E., Small, R.L., 2007. Comparison of whole chloroplast genome sequences to choose noncoding regions for phylogenetic studies in angiosperms: the tortoise and the hare III. Am. J. Bot. 94, 275–288. Swofford, D.L., 2002. PAUP* Phylogenetic Analysis Using Parsimony (*and Other Methods), version 4. Sinauer Associates, Sunderland, Massachusetts. Taberlet, P., Fumagalli, L., Wust-Saucy, A.G., Cosson, J.F., 1998. Comparative phylogeography and postglacial colonization routes in Europe. Mol. Ecol. 7, 453–464. Voje, K.L., Hemp, C., Flagstad, O., Saetre, G.P., Stenseth, N.C., 2009. Climatic change as an engine for speciation in flightless Orthoptera species inhabiting African mountains. Mol. Ecol. 18, 93–108. Wesche, K., Miehe, G., Kaeppeli, M., 2000. The significance of fire for afroalpine ericaceous vegetation. Mt. Res. Dev. 20, 340–347.