Biodiversity in the Galapagos Islands has inspired theories of adaptation and evolution, and increased our understanding of processes of population divergence and speciation (Darwin, 1859). The volcanic origin and tectonic activity of the Galapagos Islands makes them a unique site for studying the impacts of isolation and environment on diversification. The islands were formed at a volcanic hotspot in the Nazca Plate, which is moving east at approximately 59 km per million years (Geist et al., 2014). The age of the islands can be estimated by their current distance from the hotspot: the western islands are millions of years younger than the eastern ones (Geist et al., 2014). The eastward movement of the Nazca Plate causes its subduction beneath the South American plate, isolating previously interconnected islands and causing their eventual disappearance from east to west (Christie et al., 1992). The isolation of the islands and constantly changing environmental conditions have allowed the adaptation and divergence of many species, differing morphologically and genetically from one island to the other (Romagosa et al., 2013).

The vascular flora of the Galapagos Islands includes around 550 species, of which approximately 200 are endemic (Lawesson et al., 1987). Of particular interest are two wild tomato species, Solanum cheesmaniae (formerly Lycopersicon cheesmanii) and S. galapagense (formerly L. cheesmanii forma minor), collectively termed “Galapagos tomatoes” (Figure 1). Both species of Galapagos tomatoes were first considered as one. However, based on clear morphological differences and molecular evidence from an allozyme analysis, Darwin et al. (2003) described them as two different species. The adaptation of these wild germplasms to different environments, such as arid or saline soils, makes them a potential valuable source of genetic variation in terms of stress tolerance genes, which could be transferred into commercial varieties by introgression breeding (Zamir, 2001; McCouch, 2004). However, to efficiently utilize this wild germplasm resource, it is necessary to understand the population structure and genetic variation (Lv et al., 2012). This will assist breeders in selecting germplasm that are more diverse and prevent the less efficient crossing of accessions that are very closely related. It also makes screening of the wild germplasm more efficient by enabling the selection of highly diverse accessions.

Next-generation sequencing technologies, such as geno-typing-by-sequencing (GBS) and “Diversity Arrays Technology” (DArTseq), now allow genome-wide fingerprinting without prior genome sequence information (James et al., 2008). The GBS approach can be more informative than predesigned single nucleotide polymorphism (SNP) arrays when applied to wild germplasm because it is unbiased and includes information on rare alleles (Wenzl et al., 2004, 2007; James et al., 2008). What makes DArTseq different from other GBS methods is their complexity reduction approach, targeted to select genome fractions with coding regions (Cruz et al., 2013). The restriction enzymes used in DArTseq for complexity reduction separate low copy sequences from the repetitive regions of the genome (Tinker et al., 2009). These low copy sequences are more informative for marker discovery, especially for breeding purposes (Courtois et al., 2013). Here, we genotyped 67 accessions of Galapagos tomatoes from the TGRC collection (Supplementary Table S1) using DArTseq and found that S. cheesmaniae and S. galapagense fall into distinctive clades. Further, we found that the accessions of S. cheesmaniae separate based on the island/region from which they originated, and the population structure can be linked to the geological movements of the islands. From this, it can be inferred a clear evolutionary sequence within the Galapagos tomatoes, revealed by molecular, rather than morphological means.

We selected SNP markers to measure genetic variation, since SNPs are one of the most common types of genetic variation. Also, they are co-dominant markers, which allow us to estimate the homozygosity of the Galapagos tomatoes germplasm, and thus, their usefulness for genetic studies. The flower morphology of the Galapagos tomatoes suggested their autogamous nature and a high occurrence of inbreeding (Rick and Fobes, 1975). Rick and Fobes (1975) used an allozyme analysis (allelic determined variants of isozymes) to confirm that the variation between populations was greater than the variation within each population. This is consistent with our study, where the lowest percentage of homozygous SNPs was 96.1%. Furthermore, the AMOVA of the allelic data showed significant amount of genetic variance between species and also a significant amount of genetic variance between populations belonging to the same species.

Based on their morphology, S. cheesmaniae and S. galapagense can be clearly differentiated into two taxonomic groups (Figure 1), but results from genetic studies have been contradictory. Rick and Fobes (1975) showed that there was little variation within 54 analyzed accessions, while Darwin (2009) consistently differentiated between the taxonomic groups when analyzing 26 accessions. Nuez et al. (2004) used AFLP analysis on 16 accessions and showed clear differentiation between the taxa (although both taxa were still considered as one species). Lucatti et al. (2013) used 3.3 kb SNP arrays on 34 accessions, but could not differentiate between the two taxa, suggesting they were likely to be morphotypes, rather than two species. However, the SNP array they used was based on sequence information from S. lycopersicum (Víquez-Zamora et al., 2013). With current genomic technology, the GBS approach used in this work, DArTseq, enabled a robust genetic characterization of the wild Galapagos tomatoes. By using three different clustering methods – PCA, STRUCTURE and neighbor joining by genetic distance/dissimilarity – we clearly show that S. cheesmaniae and S. galapagense are two genetically distinct species (Figure 3).

In addition, we show that two S. cheesmaniae accessions from two very small islands, Baltra and Santa Fe, are admixtures, based on the calculated ancestry membership coefficients (Q) for clustering by STRUCTURE (Supplementary Table S4). One accession (LA3124) appears to be a hybrid with S. lycopersicum or S. pimpinellifolium; and another accession (LA0531) may be a hybrid with S. galapagense (Supplementary Figure S1). Their differences in morphology from the typical S. cheesmaniae are reported in the collection notes (Supplementary Table S4). Accession LA0531 comes from two different specimens collected and archived together, whose different morphologies were attributed to depauperation. This may have allowed introgression between S. cheesmaniae and S. galapagense. Accession LA3124 had big seeds, similar to those of red cherry tomato. Interestingly, LA3124 was previously classified as S. pimpinellifolium by Zuriaga et al. (2009), even though the passport data classified it as S. cheesmaniae. This highlights its closeness and possible admixture with S. pimpinellifolium or S. lycopersicum.

Solanum galapagense accessions are tightly clustered, indicating little genetic diversity within this species. These results are consistent with Koenig et al. (2013), who suggested the occurrence of strong genetic bottlenecks in S. galapagense during island colonization and recent adaptation. If S. galapagense populations were established from a small number of individuals that colonized the islands, genetic diversity has not recovered since the founder event. In addition, the isolation of the islands makes the genetic diversity susceptible to genetic drift (Maki, 1999).

In contrast, S. cheesmaniae accessions showed clear differentiation between accessions that were collected in the western islands and those collected in the eastern islands. The clustering by region of origin revealed that those accessions in close genetic proximity are also close in geographical origin (Figure 4), which suggests a correlation between biodiversity and geography in the Galapagos archipelago. This is consistent with previous reports on biogeography of other endemic species from the Galapagos Islands, such as, marine iguanas, Darwin’s finches, and giant tortoises, among others (Parent et al., 2008). The major factors influencing biogeography in the Galapagos Islands are the volcanic activity and plate tectonics (Merlen, 2014). The study of these, suggests a colonization from east to west (Merlen, 2014), from older to younger islands, supported also by the fact that ocean currents entering the islands from different directions, average an east-west direction (Merlen, 2014). The main islands originated at the volcanic hotspot west of the archipelago and slowly moved east, at approximately 59 km per million years, with the tectonic movement of the Nazca plate (Geist et al., 2014). Newly formed islands at the hotspot are inhospitable, but as they move east, cool down, and erode, the arrival and establishment of life forms becomes possible (Merlen, 2014). At the same time, the subduction of the Nazca plate beneath the South American plate, has caused the oldest islands to drown (Christie et al., 1992), forcing migration of the biodiversity to newer islands. From this, we infer that the eastward movement of the islands could have influenced the gene flow in S. cheesmaniae and their adaptation.

Rick and Fobes (1975) reported increased diversity in the Galapagos tomatoes from the western islands and especially from their western slopes. They attributed this increased diversity to the unusually higher precipitation in that area, as plant species richness in the Neotropics is known to be correlated with annual precipitation (Gentry, 1982). Concordantly, our results show that S. cheesmaniae populations in the western islands seem to have higher levels of genetic diversity than those in the eastern islands (Figure 4B). This may be because fewer populations of S. cheesmaniae exist in the eastern islands, but it could also be attributed to the founder effect of possible colonization events from western islands to eastern islands. East-to-west colonization permits the dispersal of species from the older islands to the newest islands as their volcanic activity decreases, soil develops, and they become habitable (Merlen, 2014). However, there is still some volcanic activity in the older islands that can destroy the island’s flora. Re-colonization after volcanism in the eastern islands, by a few individuals from the western islands, would also reduce genetic diversity in the eastern islands.

The division of S. cheesmaniae in two groups, has also been reported by Nuez et al. (2004), based on internode length, S. cheesmaniae “short” and S. cheesmaniae “long.” However, it is not yet clear if the genetic-geographical clustering coincides with the morphological separation. Given that Nuez et al. (2004) made their own collection and only took samples in the central islands, their data and the results presented here cannot be compared. Further morphological characterization of the TGRC collection could confirm if the whole collection of S. cheesmaniae could be separated into “short” and “long” morphotypes and if this division would be consistent with the geographical division.

No substructure was found in the S. galapagense collection. This is likely to be due to a relatively recent divergence of this species. Interestingly, there are no reports of S. galapagense from any of the eastern islands. The combination of the distribution and lower genetic diversity leads us to hypothesize that S. galapagense is a relatively new species that diverged from S. cheesmaniae after the eastern islands became isolated.

According to estimates from Geist et al. (2014), the eastern islands of San Cristobal, Santa Fe, and Española emerged approximately 2.4–4.0 million years ago, whereas, the first western islands with the presence of S. galapagense, Floreana and Santa Cruz, emerged somewhere around 1.1 and 2.3 million years ago. Geist et al. (2014) also estimated that Floreana and Santa Cruz were in peak volcanic activity two million years ago, making colonization impossible, while the volcanic activity of San Cristobal was on the decline at that time as it drifted away from the hotspot and became detached from Santa Cruz and Floreana. It was only one million years later that volcanic activity in Floreana and Santa Cruz decreased sufficiently for colonization to occur (Geist et al., 2014). Thus, divergence time of S. galapagense can be estimated to have occurred roughly one million years ago, when Floreana and Santa Cruz, the last of the eastern islands without S. galapagense, became a suitable habitat for life. These estimates are consistent with previous reports by Nesbitt and Tanksley (2002), who suggested that the initial radiation of the genus Lycopersicon occurred over seven million years ago, and that S. lycopersicum (then referred to as L. esculentum) and its closest relatives (which include S. galapagense (then referred to as L. cheesmanii, accession LA0483) and S. pimpinellifolium (then referred to as L. pimpinellifolium) diverged from a common ancestor approximately one million years ago. A more recent divergence time (between 0.19 and 0.29 million years ago) was suggested by Strickler et al. (2015) for S. lycopersicum and S. galapagense. However, the estimates by Strickler et al. (2015) are in reference to the Heinz 1706 variety, which, in the same publication, was found to contain several regions of significant introgressions from S. pimpinellifolium (Strickler et al., 2015), which could bias the estimates of species divergence time.

To our knowledge, there is not an estimate of S. cheesmanie divergence from a common ancestor. If the colonization of Galapagos Islands was east to west, then S. cheesmaniae could be an older species than S. galapagense, and could even be an ancestor to it. Rick and Fobes (1975) suggested that S. cheesmaniae could possibly be closer to the original stem line of the red-fruited species than any other member. He argued that their autogamous reproduction and the lack of competition in the Galapagos could have resulted in its preservation as an ancient biotype (Rick and Fobes, 1975). However, the lower genetic variation in S. cheesmaniae found in the older islands could be due to a founder effect, and colonization could have happened from west to east. If this was the case, S. cheesmaniae and S. galapagense could have diverged around the same time from the same ancestor.

The recent divergence of the tomato clade species and their close relationship has made their phylogenetic classification difficult, especially with the casual occurrence of interspecific hybridization (Zuriaga et al., 2009). Many attempts to determine the phylogeny have been made using diverse methods (Peralta and Spooner, 2001; Peralta et al., 2005; Spooner et al., 2005; Zuriaga et al., 2009; The 100 Tomato Genome Sequencing Consortium et al., 2014; Strickler et al., 2015; Dodsworth et al., 2016), but the results are not consistent. Most publications agree that there is a close relationship among the red-fruited group, which contains S. lycopersicum, S. cheesmaniae, S. galapagense, and S. pimpinellifolium. However, they have not yet reached a consensus on the relationship order within the red-fruited group. According to Zuriaga et al. (2009), using only a few accessions of each species in phylogenetic studies could be the cause of conflicting phylogeny results. Adding intraspecific variation to phylogenetic studies could improve the resolution. Our study provides a rich intraspecific dataset, which could be used to further characterize the phylogeny of the tomato clade. Meanwhile, we found a fine phylogenetic relationship between accessions (Figure 4A) and significant intraspecific structure in S. cheesmaniae, corresponding to the age of the island of origin of each accession (Supplementary Figure S6A). Interestingly, the accessions coming from the eastern islands of Santa Cruz and San Cristobal, are closely related within each island and separate from each other, while the more diverse accessions from the western islands, Fernandina and Isabela, are interrelated across the two islands. Additionally, S. galapagense accessions also group by island of origin (Supplementary Figure S6B), further demonstrating the influence of the islands’ biogeography on the Galapagos tomatoes gene flow.

To conclude, we propose a likely sequence of events for the diversification and speciation of wild tomatoes on the Galapagos Islands which is not only of evolutionary interest, in the classic evolutionary “laboratory” of the Galapagos Islands, but which also provides guidance for the strategic discovery of diversity, such as of novel stress tolerance alleles (Rush and Epstein, 1976; Firdaus et al., 2013), useful for future improvement of cultivated tomato, the largest horticultural crop globally.

YP, SS, SN, and MT designed research. YP and SS performed research. IP and SH contributed analytic tools. YP, SS, and MT discussed data analysis. YP analyzed data. YP, SS, and MT wrote the paper.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.