In the present study, 15 olive cultivars of different origin were analysed (Table 1): 10 Greek accessions were provided by the Agricultural University of Athens and the remaining part were obtained from the private collection of the F. lli Buccelletti Nurseries in Tuscany.

Young fresh leaves were grinded using mortar and pestle in the presence of liquid nitrogen. The genomic DNA (gDNA) was isolated from 100mg of fine powder using the DNeasy Plant Mini Kit (Qiagen) according to the standard procedure. For each accession, the gDNA was individually extracted from three distinct leaves and then bulked together, as representative of the sample. Quality and amount of the obtained DNA were evaluated both by spectroscopic and by Qubit^TM fluorometric quantitation (Thermo Fisher Scientific). Genomic DNA samples were stored at −20˚C.

Eleven microsatellite loci were selected from three groups of published SSR primers: DCA [13] , GAPU [14] and EMO [15] , all but one (DCA 4) according to those included in the OLEA database. 20 ng of gDNA was used as a template for PCR reaction according to the protocol described by Huang et al. [16] for the amplification of microsatellite fragments using labelled fluorescent M13 tail in order to separate them by capillary electrophoresis with the Applied Biosystems® 3500 Genetic Analyser (Thermo Fisher Scientific). M13 tailed primers were synthesized and labelled with one of three florescent dyes: 6-FAM, VIC and NED (Thermo Fisher Scientific). The use of fluorescent dye-labelled SSR markers provided greater precision in allele resolution and sizing. Samples preparation and capillary electrophoresis conditions have been already described [17] .

Microsatellite data were analysed with GeneMapper v 5.0 software (Thermo Fisher Scientific) and manually checked and rescored as necessary. A negative PCR control (not template) was included for each SSR reaction. All amplifications were repeated at the least twice, to ensure the consistency of the analysis.

The TBP 1^st and 2^nd intron amplification was essentially performed according to Breviario et al. [18] except for the labelling with FAM fluorophore (Thermo Fisher Scientific) of both forward primers at their 5' position, done as described by Gavazzi et al. 2012 [17] .

Thirty ng of genomic DNA were used as a template for PCR amplification. The amplicons were tested on a 2% (w/v) agarose gel, stained by Atlas ClearSight DNA Stain (1 μg∙mL⁻¹) (Bioatlas). Two different dilution ratios (1:10 and 1:20) were applied to PCR products. Therefore, two distinct CE runs were performed for each analysed sample. Two microliters containing 100 - 200 pg of diluted FAM-labeled PCR product were mixed with 0.18 µl of GeneScan^TM 1200 LIZ™ Size Standard (Thermo Fisher Scientific) and 17.82 µl Hi-Di Formamide to a final volume of 20 µl. After denaturation at 95˚C for 5 minutes, the samples were loaded on the Applied Biosystems® 3500 Genetic Analyser. CE separation was performed using an 8-capillary array of 50 cm trough the POP-7^TM polymer Thermo Fisher Scientific), setting the following instrument specifications: 60˚C, injection time 5 seconds, run time 1680 seconds, injection voltage 10 kV and run voltage 15 kV. A negative (not template) and a positive (30 ng of Triticum durum L. gDNA) PCR controls were included in all TBP reaction. All amplification was repeated at the least twice, to ensure the consistency of the analysis.

TBP and SSR peaks (alleles) obtained for the different olive accessions were scored and converted into binary data by assigning 1 to presence or 0 to absence. The genetic similarity matrices were estimated using the NTSYSpc2.1 software [19] according to the Dice’s coefficient and a Principal coordinate analysis (PCoA) were inferred by DCENTER and EIGEN procedures to explore associations among accessions. The Mantel test [20] , obtained using the COPH and MXCOMP procedures (NTSYSpc2.1 software), was used to verify the level of agreement of the respective TBP and SSR similarity matrices.

The Polymorphism Information Content (PIC) values were calculated using Anderson et al. [21] formula for both marker systems. In addition, for the SSR markers, the number of alleles per locus (k), Observed heterozygosity (Hobs), Expected heterozygosity (HExp) and the frequency of null alleles were computed by Cervus 3.0.7 [22] .

Intron nucleotide sequence information was obtained by cloning and sequencing the amplification products of the TBP 2^nd intron, from both the “Frantoio” and “Arbequina” varieties. The isolated target sequences were aligned by Vector NTI Advance 9 (Thermo Fisher Scientific) and verified according to the sequence information available on NCBI database (NCBI; http://www.ncbi.nlm.nih.gov/).

The primers (forward Arb 5'-AGG CGG TCT CTC GAC TCT CT-3” and reverse Arb 5'-GGC CAG GAA ATC TCA GAC AA-3') were manually designed in order to selectively recognize a specific allele for the “Arbequina” accession. The end-point PCR reactions on 10 ng of gDNA were performed in a total volume of 20 µl, containing 2X Taq DNA Polymerase Master Mix providing 2.0 mM MgCl₂ (VWR International PBI, Milan, Italy), 0.5 µM of each primers and deionized water (Merck KGaA, Darmstadt, Germany). PCR conditions were: 3 min initial denaturation at 94˚C followed by 35 amplification cycles (94˚C 30 s, 66˚C 40 s and 72˚C 40 s) and final extension step of 5 min at 72˚C. Four µl of the PCR products were checked on 2% (w/v) agarose gel. Positive (“Arbequina” gDNA), negative (Triticum durum L. gDNA) and not template controls were included in all tests. The cross reactivity evaluation of the isolated probe was tested against 10 ng of gDNA for all the varieties included in this study. The detection limit of the probe was estimated by the amplification of serial dilutions of “Arbequina” gDNA added to a gDNA solution of Triticum durum L. used as the background.

The TBP method, based on the detection of intron length polymorphisms present in the members of the β-tubulin gene family, was applied to 15 different olive varieties, mainly of Greek origin provided by the Agricultural University of Athens within the context of an exchange Erasmus program. Both β-tubulin introns were evaluated for the polymorphic content. Figure 1 shows that each of the 15 different varieties release a specific Tubulin-based Barcode (TbB) resulting from the presence of diverse alleles with different intron size. Each analyzed accession is characterized by its own specific and distinctive DNA profile.

The tubulin identity of the amplified fragments was verified by nucleotide sequencing and such information was further exploited for the preparation of a specific molecular probe (see below). On the average, the 1^st intron of the olive β-tubulin members, proximal to the 5' end of the gene, was longer in size and more broadly distributed than the 2^nd intron First intron sizes range from about 80 bp to more than 1000 bp, after subtracting 305 bp of the amplified exon boundaries. However, the vast majority of the amplified bands from the 2^nd intron has short sizes, less than 100 bp in length. The number of detected alleles is 26 for the 1^st intron and 30 for the 2^nd with an overall mean value of 28 (Table 2). The highest percentage of polymorphic alleles (80%) was recorded by the 1^st intron amplification, despite the lower number of peaks while 63% of polymorphism was the percentage detected from the amplification of the 2^nd intron. The

N number of individuals; i: peak number variation detected among the analyzed samples; k: number of alleles; NPP: number of polymorphic peaks; p_a: average allele frequency; PIC: Polymorphism Information Content.

reduced number of the detected peaks from the 1^st intron amplification reflects the limit of the CE analysis to discriminate products with a maximum size of 1400 bp. The different sizes distribution of the alleles clearly suggests that the two groups of introns have evolved independently from each other. Polymorphic information content (PIC) calculated for both 1^st and 2^nd introns is also high and significant: 0.973. This value reflects a good discrimination power among the 15 analyzed genotypes. Repeatability and reproducibility of the data were ascertained by verifying the consistency of the barcodes in multiple independent extractions and analyses. In one case, 100 distinct accessions of the same variety, “Arbequina”, showed the same identical TBP profile (data not shown).

As shown in Figure 2 the Principal Coordinates Analysis (PCoA), inferred by combining the TBP data obtained from both introns, explains 51% of the total variation existing among the different analyzed samples. With the exception of “Adramitini”, the plot revealed three major groups that can be referred to the different geographical origin, Greece, Spain and Italy, of the analyzed cultivars (black, green and red circles respectively in Figure 2).

In order to ascertain the level of reliability of the data based on TBP, the DNA samples of the same 15 olive varieties were analyzed with a battery of 11 microsatellite markers previously developed for Olea europaea L. and then included in the OLEA database (Table 3). Both loci carrying complex motifs and di-nucleotide repeats were included. The variation between the estimated allele sizes and the published values were considered valid within the range of ±1 bp. All the estimated size ranges fall within the expected published value.

As opposed to the TBP analysis, 11 individual reactions were run, one for each marker. A total number of 81 alleles were detected with an average of 7.4 alleles per locus. The details of the analyzed SSR loci and the data on the detected polymorphisms are reported in Table 4.

The set of the analyzed loci revealed good PIC values, ranging from 0.577 to 0.842 with a mean value of 0.742, indicating an adequate informativeness level, yet lower than that obtained with TBP. The mean value of the observed heterozygosity (HObs) was slightly lesser than that of the expected heterozygosity

k: number of alleles; N number of individuals; HObs, observed heterozygosity; HExp, expected heterozygosity; f: null allele frequency.

(HExp). As a consequence, the probability of occurrence of null alleles (f) was positive for each of those loci showing HObs < HExp. Such null alleles can arise when mutations prevent the annealing of the primers to the target sites although this may not be the only explanation for the reduced level of heterozygosity found. In fact, higher homozygosis content could also have originated by variable levels of inbreeding that may have affected the analyzed varieties. The highest values of HObs were found for those loci (GAPU103a, DCA18, DCA16 and GAPU71B) that do not associate to any linkage groups, according to published data on mapping [1] . This is consistent with the observation that only independent loci (not co-segregating) can provide correct information about the genetic evolution and differentiation of a population.

The comparison performed between the SSR allele sizes registered in the OLEA database and the values detected in our study, limited to 5 varieties and 10 loci commonly shared (100 detected alleles), revealed an unmatched allele rate of 12% and an unmatched locus rate (both alleles not matching) of 6% (data not shown).

The three dimensional projection of PCoA inferred by the SSR data (Figure 3) explains the 30% of the total variation present among the analyzed cultivars. Despite the absence of well distinguishable groups, the plot display a clear separation of the Italian cultivars, according to the second axis (red circles in Figure 3), and of the Spanish cultivars (“Arbequina” and “Arbosana”, green circles in Figure 3) according to the third dimension. Specifically, these latter cultivars always share one allele in 9 out of 11 loci while both alleles are commonly shared at the remaining two loci showing the highest similarity values.

The Mantel test revealed no significant correlation (r = 0.00758 and p = 0.5279) between the similarity matrices obtained from SSR and TBP respectively, reflecting the different structure, distribution and evolutionary significance of the two different markers systems.

An additional advantage that may come from the use of the TBP molecular marker and its pattern of amplification is the possibility of producing variety specific probes. In fact, some of the PCR fragments obtained either from the amplification of the 1^st or of the 2^nd intron, may contain enough sequence information to allow the design of cultivar specific probes. The case is shown for the development and use of a pair of primers that specifically recognize some β-tubulin target sequences of the “Arbequina” variety. They have been designed starting from the isolation of the 409 bp fragment, an amplification product of the second intron (shown with a red arrow in Figure 4).

Once isolated and sequenced, a couple of primers was designed to target an internal nucleotide region that contains an insertion of 9 nucleotides, specific for “Arbequina”, thus allowing the selective amplification of a 205 bp fragment (data not shown). The specificity of this probe was checked on DNA extracted from

all the other 15 different cultivars (Figure 5(a)). The limit of detection, assayed on linearly diluted solutions from 10 to 0.0005 ng of Arbequina’s gDNA in a gDNA solution of Triticum durum L., was also verified and found to be 0.01ng (Figure 5(b)).

A similar strategy could be adopted for the preparation of any other cultivar-specific probe exploiting the presence of specific alleles in one of the two introns. Alternatively, if a specific allele is shared with other cultivars for one intron, an additional discriminating probe can be developed from the pattern of amplification obtained from the other intron. In this case, variety recognition will be achieved with the use of a limited, cost-effective combination of probes.