Plant breeding procedure
Aiming to develop materials suitable for organic agriculture (i.e. possessing the adaptation and the intrinsic diversity needed to suit a particular environment) we specifically focused our work on a landrace (LR) instead of on other materials.
We chose a sprouting broccoli LR from Umbria (Central Italy) that is named “Zolfino”, highly appreciated on the local market and generally grown under low-input conditions. The LR is characterised by late flowering and maturation: requiring low winter temperatures for flower induction, it is harvested in a period spanning from beginning of March to mid-April. Initially the main head, then the lateral buds are harvested in the period.
The initial material of LR was obtained by a farmer family which cultivated it since generations. After the farmer abandoned his own population, this Department (DSA3) continued its cultivation by yearly reproducing the seed and applying the selection procedures of the mother plants (MP) suggested by the farmer: the most vigorous 2 or 3 MPs, among the 25 usually grown, are chosen and intercrossed, while the field is cleared from other B. oleracea plants, since all the species varieties are interfertile. It should be noted that this procedure is common across Europe for farmers reproducing their B. oleracea LR (V. Negri personal communication).
Starting a breeding program to develop an initial synthetic in 2009, we asked the farmer to choose 17 MP among the 25 grown instead of the usual number (Figure 1). The following year, 10 plants for each of the 17 MP half sib progenies (MPHS) were grown in a trial (5 plants for each MPHS in each of 2 replicates, for a total of 170 plants). Such a high number of plants was adopted in order to maintain the initial LR diversity as much as possible. Among them, the farmer was asked to select, according to his personal opinion, the best 8 MPHS and then, within each of them, the best 5 plants. Accordingly, a synthetic variety (Syn0) made of 40 plants from the initial 8 MPHS, that mirrors the original LR, was obtained.
The multiplication of Syn0 produced a Syn1-PG (2009-2010), the initial (foundation) material of this study.
To start the EBP the Syn1-PG seed was then multiplied in isolation in different areas of Central Italy, in France and in United Kingdom as follows (Figure 1):
i) for two consecutive years at DSA3 (Perugia, Umbria Region) in the same area of its origin giving rise to two populations named Syn2-PG (2010-2011) and Syn3-PG (2011-2012), respectively;
ii) for two consecutive years in two other areas of Central Italy with different pedo-climatic conditions from Perugia: Grosseto (Tuscany) and Terminillo (Lazio), giving the populations named Syn2-GR, Syn3-GR and Syn2-TER and Syn3-TER, respectively;
iii) for one year only (2010-2011) in United Kingdom giving a population named Syn2-UK;
iv) for one year only (2010-2011) in France giving a population named Syn2-FR.
Each multiplication environment (GR, TER, UK and FR) is different from PG for several pedo-climatic traits (Additional File 1: Figure S1; ). Such different environmental conditions were purposely chosen to assess the degree of restriction in genetic diversity due to selective pressure.
As it is usual for this type of LR, in all sites the seeds obtained from each multiplication cycle was sowed in pots at the beginning of July and the plantlets were transplanted in the experimental field at the beginning of September. The seed was harvested in June in Italy, and about a month later in France and United Kingdom.
In each multiplication cycle and in each environment only the three most vigorous plants (in agreement with farmer indications) were intercrossed in order to obtain seed. The inflorescences of the other plants were cut before flowering in order to evaluate agronomic traits of the commercial heads.
Each year, the obtained seed was sent to DSA3 in order to store it in its Genebank (FAO code ITA363) for the successive analyses (see below).
For each environment of multiplication and for the environment of final agronomic trial, climatic data, i.e. average daily temperature, average of maximum daily temperatures, average of minimum daily temperatures (°C) and total rainfall precipitations (mm) (obtained from near-by weather stations), were recorded from sowing to harvest (Additional File1: Figure S1).
Evaluated plant material
The materials obtained from multiplication in different environments was used to assess the occurrence morpho-phenological and genetic changes.
Firstly, particularly to understand the effects of multiplication in different countries, we assessed the diversity of genotypes belonging to 11 entries of broccoli: seven populations and four hybrids used as controls. Specifically: the Umbrian LR, the Syn1-PG, its five two year multiplications, Syn2-PG, Syn2-GR, Syn2-TER, Syn2-UK and Syn2-FR, and the hybrids Santee (Elsomseed), Ironman (Monsanto), Belstar (Bejo Seeds) and Natura Vallata (Cooperativa Agricola Cesenate), named H, HH, HHH and HHHH, respectively.
Secondly, we assessed other three populations obtained in Central Italy with a further multiplication cycle (Syn3-PG, Syn3-TER and Syn3-GR) in order to obtain information about further evolutionary processes eventually occurred.
Assessment of morpho-phenological changes occurred with multiplications
In order to assess eventual morpho-phenological changes, all the obtained multiplications were evaluated in the same environment.
Seeds from the 14 broccoli entries were sown in pots at the beginning of July and the obtained plantlets were transplanted in a DSA3 experimental field in Perugia (Umbria, Italy) at the beginning of September. A total of 448 individual plants were evaluated. Plants were arranged by using a randomised block design with four replications. In each plot, eight plants of each entry were grown, for a total of 32 plants evaluated per entry. Spacing among plants was set at 1 x 1 m because the plants are very vigorous and this spacing is generally used by farmers growing the “Zolfino“ LR (Additional File2: Figure S2). The management practices applied to the agronomical trial were similar to those used in organic agriculture conditions in the area. Fertilization was ensured by pre-implant livestock manuring, by using 1 t ha-1 of the commercial product NUTEX LETAME (Agroqualità, 3% N and 3% P2O5 content) that was incorporated by using a rotary cultivator. As above, this is the fertilization level usually applied by farmers. Weed control was always performed by hand-weeding and no chemical control of pathogens and insects was applied. Broccoli were irrigated immediately after transplanting and during the following month, aiming at fully restoring crop evapotranspiration.
Twelve morpho-phenological traits were recorded for each plant on all entries (Table 1).
To determine the significance of the sources of variation, the recorded data were processed by analysis of variance (ANOVA) using a linear model, where an individual trait value Tij of the levels i of the effect “Genotype” (G), and j of the effect “Block” (B), is: Tij=m + Gi + Bj + eij, where m is the grand mean and eij is the experimental error, and means were separated using the Tukey Honest Significant Difference (HSD), as implemented in the HSD.test() function of the “agricolae” package in R [33, 34].
In order to study the correlation pattern of the considered morpho-phenological traits and the differences among entries, a principal components analysis (PCA) was performed by using the function PCA() of the “FactoMineR” package . PCA is a factor analytic technique that is used for the ordination of observations in a reduced rank space. It is used to visualise the weight of variables in determining entry ordination (as shown by vector lengths, by angles between vectors and by the reciprocal positioning of variable vectors and entry symbols). As a preliminary step before performing a PCA on the complete dataset, the missing values were imputed with the Principal Components Analysis model by the function imputePCA() of the “missMDA” package .
Assessment of genetic changes occurred with multiplications
Genomic DNA extraction
Fifty mg of fresh leaf tissue were collected from each of the 448 plants (32 plants for each of 14 entries) and high quality DNA was isolated using the DNeasy 96 Plant Kit (Quiagen). DNA quality and concentration were checked by spectrophotometry using NanoDrop 2000 (Thermo Scientific) and 1.0 % (w/v) agarose gel separation.
Broccoli individuals were genotyped using 17 putative neutral microsatellite markers (SSR) [37–39] and five microsatellites related to genes involved in flowering control and cold stress response (EST-SSR)  (Additional File 3: Table S1). According to their position on B. oleracea chromosomes, the selected markers are not in linkage (i.e. for each chromosome, the two selected markers are located on the two different arms). All the DNA amplifications were carried out as described in . Obtained amplicons were separated in a 3130xl capillary sequencer (Applied Biosystems), sized according to the internal size standard GeneScanTM23 500LIZ® (Applied Biosystems), visualised and scored using the GENEMAPPER software (Applied Biosystems).
For each SSR marker the number of alleles and allele size range were recorded. The number of successfully analysed (N) and observed (Na) genotypes, the effective (Ne) number of observed alleles, observed (Ho), as well as the expected (He) heterozygosity and the Fixation Index (F), were worked out by entry (i.e. LR, Syn1-PG, its derived populations and hybrid controls) with the use of GENALEX software . For each parameter, the average values were compared by using the Tukey Honest Significant Difference (HSD) as implemented in the HSD.test() function of the “agricolae” package in R [33, 34].
In order to assess the components of molecular variation, an Analysis of Molecular Variance (AMOVA) was worked out following the method of  as implemented in GENALEX software .
Pairwise comparison Fst values and two Genetic Distance (GD) matrices between individuals following the formula of  were calculated using GENALEX software . Two GD-based Neighbour Joining (NJ) trees were drawn using the on-line software iTOL . To better understand the genetic relationship among entries, two Principal Coordinate Analysis (PCoA) were also worked out by using the same distance matrices and drawn by “scatterplot3d” package in R [34, 46].