Plant materialThe wheat collection of the Czech genebank, harbored at the Crop Research Institute in Prague, comprises nearly 13,000 accessions within the genus Triticum, of which the majority correspond to Triticum aestivum (GRIN Czech 2024, entered with following parameters: Accessions Available From a Site; site_acronym: CZE122; Limit: 20,000, accession IDs: 01C01* and 01C02* for winter and spring wheat, respectively). Other cultivated species include the diploid T. monococcum, the tetraploids T. durum, T. dicoccum, T. polonicum, T. isphanicum, T. timopheevii, and T. carthlicum, and the hexaploids T. spelta, T. vavilovii, T. compactum, and T. petropavlovskyi. In addition, the collection includes wild species at the diploid and tetraploid levels: T. boeoticum, T. urartu, T. dicoccoides, and T. araraticum. The phenotypic data presented in this study included 4,534 accessions (1,065 spring wheat and 3,469 winter wheat) which corresponds to more than 1/3 of the entire collection.Phenotyping protocolThe main task of genebanks is to conserve genetic material for future generations and provide it for current users, which entails regular regeneration and multiplication in field plots. Seed multiplication is required when (i) seed stocks are no longer sufficient, (ii) germination rates decrease below a critical threshold, (iii) extensive amounts of seeds are demanded by research, or (iv) new accessions are added to the genebank. During the regeneration process, morphological and agronomical traits are scored for the phenotypic comparison with previous regenerations cycles following strict quality guidelines. In case of doubts whether there were any morphological shifts or drifts during propagation, the voucher spike collection kept in the genebank can be used for comparison. Individual accessions were regenerated between 1951 and 2020 in Prague, Ruzyně (latitude 50° 5′ 10.3698″N, longitude 14° 16′ 49.926″E, 364 m.a.s.l., local soil type Orthic Luvisol, 8.5 °C average annual temperature, 510.5 mm average annual rainfall). Not all accessions were regenerated every year, resulting in a non-orthogonal structure of the data.For seed propagations, spring wheat accessions were usually sown from February to April, while winter wheat accessions were sown from September to October. The accessions were grown in plots with a size 2 m2 for regeneration. Evaluation trials consisted of experimental plots with a size of 4 or 10 m2 in four replications in completely randomized design. Data were recorded according to the descriptor list for wheat7 for the 25–30 traits. The experiments were repeated for 3 (in a few exceptions 2) years in different experimental fields of the genebank. Three traits were selected for this study: heading date (HD), plant height (PH), and thousand grain weight (TGW). HD was assessed for both spring and winter wheat accessions as the number of days from January 1st when 50% of the plants reached heading (BBCH 59)9. PH was assessed in cm from the soil surface to the top of spike including awns. TGW was determined after seed harvest and expressed in g on a ~15% grain moisture basis.Data analysesPhenotypic data of each growth type was analyzed separately based on the following mixed model:$${{\boldsymbol{y}}}_{{\boldsymbol{ij}}}={\boldsymbol{\mu }}+{{\boldsymbol{g}}}_{{\boldsymbol{i}}}+{{\boldsymbol{a}}}_{{\boldsymbol{j}}}+{{\boldsymbol{e}}}_{{\boldsymbol{ij}}}$$
(1)
where yij stands for observed phenotypic value of the ith accession in jth year, μ is the population mean, eij error terms (random), gi effect of accessions (fixed), and aj effect of the year (random).ASreml-R10 v. 4.1.0.154. was used for the purpose of the analysis and variances of errors were modelled as specific for each year. In a first step, Eq. (1) was used for outlier detection. In order to do that, studentized residuals were used and Bonferroni-Holm tests were applied to correct for multiple testing11,12. The outliers were then removed from the dataset and best linear unbiased estimations (BLUEs) for each accession were generated by fitting Eq. (1) on the enhanced data set.In the next step, the heritability of the main traits was calculated considering both years and accessions as random effects in Eq. (1):$${{\boldsymbol{h}}}^{{\bf{2}}}=\frac{{{\boldsymbol{\sigma }}}_{{\boldsymbol{G}}}^{{\bf{2}}}}{{{\boldsymbol{\sigma }}}_{{\boldsymbol{G}}}^{{\bf{2}}}+\frac{{{\boldsymbol{\sigma }}}_{{\boldsymbol{e}}}^{{\bf{2}}}}{{\boldsymbol{Year}}}}$$
(2)
where, \({{\boldsymbol{\sigma }}}_{{\boldsymbol{G}}}^{{\boldsymbol{2}}}\) is genetic variance of the accessions, \({{\boldsymbol{\sigma }}}_{{\boldsymbol{e}}}^{{\boldsymbol{2}}}\) stands for average error variance across regeneration years, Year is average number of years each accession was tested.
Curation of historical phenotypic wheat data from the Czech Genebank for research and breeding
