Divergence of metallophytes under heavy metal selection has been characterized by other researchers, with a focus on only a few species, notably N. caerulescens [29–33], A. halleri [1, 16, 34–36], Cistus ladanifer  and species in the genus Silene [15, 18, 33, 38, 39]. To our knowledge, the present study is the first report on a comparison of the genetic divergence of two sympatric metallophytes.
Heavy metal tolerance in higher plants is a well-documented example of micro-evolution and is a basic necessity for survival in metal-contaminated soils [9, 40]. Based on the assumption that genotypes with a sufficiently high level of metal tolerance are rare in non-metallicolous populations, the founder effect was hypothesized to be a result of strong selection occurred during the colonization of contaminated areas by metallophytes . A comparison of heavy metal tolerance among populations from metalliferous and non-metalliferous sites would provide an estimation of the levels of stress occurred during the colonization of heavy metal-rich areas. Differences in heavy metal tolerance among populations of metallophytes have been reported widely (see [41–44] for examples). In general, populations or ecotypes growing in contaminated soils exhibit higher levels of heavy metal tolerance than those from uncontaminated sites.
Tolerance index is a measure that can be used to assess the relative degree of tolerance of various plant varieties . In this sense, it is important to test the Cu tolerance of non-tolerant species (or non-metallicolous populations in our case) as a control. On the other hand, in most of the relevant literature [4, 45], heavy metal tolerances of non-metallicolous species (populations) have been compared to those of metallicolous ones. We therefore believe that this kind of comparison is suitable for a relative assessment within our two species. Yet, it should be noted that we cannot establish the presence of any baseline (constitutive) tolerance above that of other local species.
In the present study, significant differences in Cu tolerance at the population level were detected. Further, differences in Cu tolerance between the two species under investigation were also detected. Under the similar level of Cu stress, C. communis exhibited higher Cu tolerance than R. acetosa , especially when Cu concentration reached the maximum of 320 μmol L−1, Cu tolerances of the 4 C. communis populations from cupriferous sites were significantly higher than those of the 4 R. acetosa populations from the same sites (P < 0.05). Interestingly, there was no significant difference in Cu tolerance of non-metallicolous populations between the two species, suggesting a possible similar level of selective pressure on the ancestral non-tolerant colonizers of metalliferous sites. Although the concentrations of Cu in the contaminated sites we sampled were extremely high when compared with those of the non-metalliferous sites, distinct differences in Cu concentration were also observed among contaminated sites. Therefore, heterogeneity in the metalliferous sites could partly explain the difference in Cu tolerance among metallicolous populations. However, considering that the Cu concentrations in contaminated sites were generally 100 times greater than those in uncontaminated sites, the significant differences in Cu tolerance should be mostly attributed to the intense and long-term selection due to heavy metal stresses and to the differences in biological traits of the two metallophytes. Because they were sympatric in most sampling sites, and so subjected to the similar edaphic conditions. On the other hand, it is possible that the heterogeneous nature of the metalliferous sites could also partly explain the difference of Cu tolerance between the two species. Because there were some significant differences in Cu concentration among the metalliferous sites (Table 1). However, it should also be noted that we cannot exclude the potential contribution of maternal environmental effects on Cu tolerance of different populations of the two species, since field collected seeds were used and in the present study measurements were taken on young seedlings.
In order to determine more thoroughly whether the different phenotypes observed are linked to population histories and are reflected in the genetic diversity of populations, three different molecular techniques (AFLP, ISSR, and cpDNA) were employed to analyze the genetic structures and genetic relationships among different populations of the two metallophytes. Firstly, we found that levels of genetic diversity in the two co-occurring species were distinctly different. For R. acetosa , no significant difference in genetic diversity between metallicolous and non-metallicolous populations was found, whereas significantly lower levels of genetic diversity were detected in metallicolous populations than in non-metallicolous populations of C. communis . Similarly, Mengoni et al.  reported a significant decrease in genetic diversity of populations in S. paradoxa collected from Cu deposits. However, contrasting results were obtained in studies on other species. For example, the metallicolous and non-metallicolous populations of A. halleri , N. caerulescens  and Onosma echioides  maintained similar levels of genetic diversity. In a more recent study on N. caerulescens , genetic differentiation linked to heavy metal concentrations in soil was detected, and the gene flow observed at some nuclear loci was shown to be significantly reduced between plants encountering different levels of heavy metal contamination in the soil, suggesting that natural selection limits gene flow between metalliferous and non- metalliferous locations . In the present study, the reduction of genetic diversity in metallicolous populations of C. communis might be due to a strong selective pressure during the colonization of Cu-contaminated sites by this species, coupled with a limited gene flow from surrounding non-metallicolous populations (as suggested by the relatively high values of pairwise F
). In contrast, the similar levels of genetic diversity in metallicolous and non-metallicolous populations of R. acetosa might be due to substantial gene flow between populations, as indicated by the low values of pairwise F
. High levels of gene flow between metallicolous and non-metallicolous populations of metallophytes have previously been reported in some other comparable studies [16, 17, 31]. In addition, non-metallicolous populations of R. acetosa showed higher levels of background tolerance to Cu (especially for DGSCK) compared to those of C. communis. Considering that metallicolous populations are presumed to evolve from non-metallicolous populations, this evidence may imply that selection for Cu tolerance in C. communis has been greater than in R. acetosa , although present-day levels of Cu tolerance in metallicolous populations are lower in R. acetosa .
Concerning that the habitats of metallicolous populations of metallophytes are often fragmented and disjunct, it is unlikely that dispersal from a single tolerant ancestral population could have produced the wide ranges of geographic distribution today. Hence, the hypothesis of polyphyletic origin is proposed and has been subsequently corroborated by considerable data [15, 17–19, 31]. In the case of R. acetosa , metallicolous populations branched in separate positions in the NJ dendrograms obtained by AFLP and ISSR data, supporting the hypothesis of independent origins from nearby normal populations. Furthermore, we also found that there was a direct correlation between genetic differentiation and geographical distances between pairs of R. acetosa populations (in particular for ISSR markers, Additional file 2: Table S1). For C. communis , we also speculate that the metallicolous populations evolved independently, since the site-to-site distances of the metallicolous populations were too far to be overcome by natural dispersal. However, metallicolous and non-metallicolous populations of C. communis had a distinct branching pattern in the NJ dendrograms with AFLP and ISSR data. The non-metallicolous populations were included in one group, whilst the metallicolous populations dispersed into single long branches. Additionally, components of variance of differentiation between populations of C. communis were also higher than those of R. acetosa . These results suggested a low gene flow in C. communis not only between metallicolous and non-metallicolous populations, but also among metallicolous populations, supporting higher levels (and possibly rates) of genetic differentiation.
The contrasting pattern of genetic differentiation between C. communis and R. acetosa was also confirmed by the analysis of chloroplast markers. Though this analysis was performed on only one individual per population, results for C. communis were similar with those obtained with AFLP, suggesting the presence of one main cluster of metallicolous populations. On this basis, it could be reasonable to hypothesize that metallicolous populations of C. communis may derive from seeds of one single ancestral population. Interestingly, we observed a very low variability of chloroplast sequences of R. acetosa , which seemed to be consistent with the idea of a recent spread and low differentiation between metallicolous and non-metallicolous populations.
The contrasting results of the two metallophytes also implied that populations of C. communis were more remarkably influenced by heavy metal stress than those of R. acetosa , although the genetic structures of the two species were influenced by geographic isolation and heavy metal contamination.
Although ecogeographic isolation has long been viewed as the most important reproductive barrier in plants , barriers to gene flow were also demonstrated between non-metallicolous and closely adjacent metallicolous populations of some species such as the grass Anthoxanthum odoratum . Additionally, it has been reported that heavy metal contamination could have a greater impact on the population structure of the hyperaccumulator Sedum alfredii than geographic distance . Based on our results, we can therefore speculate that both heavy metals and geographic distance play a significant role in determining the population structure of R. acetosa . In contrast, Cu contamination seemed to play a more important role in determining the population structure of C. communis than geographic distance.
In previous studies, it has been shown that the genetic divergence and evolutionary processes of plants could be affected by various factors, such as edaphic conditions, stress intensity and duration, geographic isolation, bottleneck and founder effects, life history, reproductive system, and so on [16, 29]. Considering that R. acetosa and C. communis populations in the present study almost share the same edaphic conditions (mainly Cu toxicity) and geographic distribution pattern, the different characteristics of genetic divergence might result largely from their different life history and reproductive system. On the one hand, as a perennial, unisexual, and dioecious species, R. acetosa may have more opportunities to experience gene exchange among individuals and gene flow between populations than the bisexual annual plant C. communis . In addition, the winged seeds of R. acetosa are presumed to have higher dispersal capacity than those of C. communis , which might enhance the success of colonization by R. acetosa [27, 28].