**3. Discussion**

Most crops experienced a reduction in genetic diversity (genetic bottleneck) due to the domestication process [32]. Indeed, the development of high yielding crops for food, feed, and other uses required the desirable phenotypes to be selected at the expense of variability present in their wild ancestors (founder effect) [33–35]. However, such "uniformity" often resulted in more vulnerable plants that are not able anymore to cope with biotic and abiotic stresses. As a consequence, wild relative species are often exploited as a reservoir of "exotic" alleles to secondarily increase variability in previously selected traits, thus favoring adaptation to changed conditions [34].

Landraces are locally adapted cultivars that are gaining increasing attention considering their typical traits (e.g., high quality of fruits and yield stability in low input agricultural systems) [17,19,36–38]. Although it has been widely demonstrated that the chloroplast genome is a valuable resource to study evolution and phylogenetic relationships among species [39,40], the genetic diversity of tomato landraces was largely based on the genome-wide characterization of their nuclear DNA variability [11,12,14–16,20–22,41]. Further, due the uniparental mode of inheritance, genetic bottleneck in organellar DNA may not necessarily reflect nuclear variability, thus providing additional/ complementary information on the domestication process.

Comparative analyses of the nine plastomes sequenced in this work and of twelve plastomes retrieved from GenBank allowed both to evaluate the extent of the genetic bottleneck on the tomato chloroplast genome and define phylogenetic relationships among wild and cultivated accessions. For these aims, SNPs and SSRs were revealed to be more informative than TRs since no specific TR for cultivated tomato genotypes, or *de novo* TRs were identified in our survey.

Very low cpDNA variability was detected in tomato varieties with respect to that observed in wild species, thus indicating the occurrence of a very strong cytoplasmic bottleneck during domestication. The number of SNPs in wild species is 24-fold higher than in cultivated tomatoes (389 polymorphic SNPs out of 454 (86%)), while SSRs were slightly lower (49 polymorphic SSRs out of 114 (43%), 4-fold those observed in tomato varieties). The heterogeneous nature of the *S. lycopersicum* var. *cerasiforme* group is remarkable, namely, the two analyzed accessions showed a different behavior. Collected data and phylogeny clearly highlighted higher variability in cer1 compared with cer2 and suggest that although cer1 belongs to *cerasiforme* group, probably it was not subjected to the domestication process and can be considered as "wild" cultivated accession.

Detected levels of plastome variability are consistent with the extensive genetic erosion of cultivated tomato, especially in the light of the large diversity observed across wild relatives [5]. Similarly, pepper wild species displayed a number of SNP and SSR respectively 8-fold and 3-fold greater than that of cultivated genotypes [42].

Only 16 out of 454 SNPs were found polymorphic among cultivated tomato genotypes (3.5%). Comparable results were found in pepper varieties, where only the 4% of the scored SNP loci were polymorphic in cultivated accessions [42].

Similarly, only 12 out of 114 identified SSRs were polymorphic among cultivated tomato genotypes (11%). Comparable results were reported in cultivated barley showing one polymorphism out of seven analyzed SSRs (14%) [27] and pepper varieties, showing 19 polymorphic SSRs out of 92 (21%) [42]. Contrariwise, 16 out of 17 (94%) SSRs were polymorphic among cultivated bean [43].

As previously argued, genetic bottleneck at the nuclear level may not be reflected at the cytoplasmic level. An extreme cytoplasmic bottleneck has been previously hypothesized in cultivated potato by the analysis of SSR markers but no decreased levels of nuclear SSR diversity were recorded [26,39]. On the contrary, the genetic diversity analysis between American and European collections of common bean highlighted the absence of evident cytoplasmic bottleneck (only 2% loss of cpSSR diversity) [44], and a stronger nuclear bottleneck (30% loss of SSR diversity) [45] likely indicating that the founding common bean populations introduced in Europe were still highly variable in their cytoplasmic DNAs [46].

SNP arrays on some tomato cultivars, partially shared with this work (i.e., M82, cor, pgl, vfr, and ves2001), revealed a reduced nuclear genetic diversity [22].

Concordantly, the cpDNA analyses suggest an extreme low cytoplasmic variability of the founding cultivated tomato population. Indeed, cultivated varieties shared 361 out of 454 SNPs (79%) and 74 out of 114 SSRs (65%) with the ancestor *S. pimpinellifolium* (i.e., same SNP alleles and same SSR haplotypes) and only seven *de novo* SNPs and two *de novo* SSRs were different between *S. lycopersicum* var. *lycopersicum* and local accessions groups. All analyzed local accessions showed identical cpDNA sequences suggesting that these accessions have a unique domestication origin and that their cytoplasm has evolved monophyletically from the founder tomato gene pool, rather than representing an independent introduction. Still, the local accessions have distinctive sequences from the other commercial tomatoes (i.e., *S. lycopersicum* the var. *lycopersicum* group) excluding multiple independent selections of the obovoid fruits (Corbarino) or the hearth-shaped fruits with a pronounced pointed apex (the remaining accessions).

In this work, we also detected plastome variability between wild *S. pimpinellifolium* 1 and 2. These differences could be due to natural variability among accessions and/or possible errors in the sequencing/assembly procedure. The former hypotheses, however, is supported by differences also observed among other related wild species (*S. galapagense*, *S. cheesmaniae,* and *S. habrochaites*). Thus, the significant reduction in cpDNA variability found in the cultivated tomato gene pool can be directly ascribed as a consequence of the domestication process rather than to an already occurred loss of genetic variation in the closest wild relative, *S. pimpinellifolium*. Therefore, the present study suggests that a severe 'cytoplasmic bottleneck' occurred during the domestication of tomato, as has been reported in other crops: barley [27], lentil [47], onion [48], and potato [26].

A strict relationship between cultivated tomato varieties and the ancestor *S. pimpinellifolium* was supported by phylogeny. Species belonging to the Lycopersicon group (*S. lycopersicum*, *S. pimpinellifolium*, *S. cheesmaniae*, and *S. galapagense*) [49] form a well-supported clade in agreement with previous phylogenetic studies [5].

In particular, all local accessions clustered together in a subgroup with *S. lycopersicum* var. *lycopersicum* and cer 2. On the contrary, some accessions (i.e., cor, pgl, vfr, and ves2001) were grouped in different clusters based on nuclear SNP genotyping [22]. Noteworthy, cer1 was included in the same group of *S. pimpinellifolium* accessions, thus plastome diversity analysis confirmed the mixed nature of *S. lycopersicum* var. *cerasiforme* as previously observed with the analysis of nuclear variability [10,11,50,51].

The observed low variability of the cultivated tomatoes chloroplast genome can be explained by taking into account both the genetic bottleneck during their domestication and its low mutation rate. Notably, comparison of the plastome sequences of the two modern tomato varieties IPA-6 and Ailsa Craig, the former bred in South America and the latter in Europe, resulted in identical cpDNA sequences, thus demonstrating the stability of plastome in tomato cultivars over a period of at least a few hundred years of separation [52] without the insurgence of any *de novo* mutation. Although low variation is the rule in tomato cpDNA, few plastid regions have been identified that might be exploited as diagnostic markers: two *de novo* SNPs, one SSR and a short sequence duplication (ATAA)2 were exclusive of all local accessions, whereas, one SSR was typical of all the var. *lycopersicum* group.

Variability found in all tomato genotypes mainly affected intergenic regions. However, the most variable genes were *ycf1* (showing both synonymous and non-synonymous mutations) and *ndhH*. Both these genes have been proposed as tools to resolve the phylogenetic relationships among closely related genera and species [53–55] and at least *ycf1* was found variable even within cultivated plastomes leading to amino acid change (Figure S3).

Overall, our work contributes to the characterization of tomato plastid genomes and their phylogenetic relationships, and especially highlights the severe reduction in variability at plastid DNA as a consequence of the strong genetic bottleneck occurred in the founding population during the domestication process.
