Introduction

Barley ranks a distant 4th in grain production worldwide with a total of 140.6 metric tons annually. The United States produced approximately 169.6 million bushels of barley at an estimated price of 761.2 million dollars in 2019 (USDA NASS). Many barley traits contribute to malting quality, as well as food and feed quality. Here we describe a more efficient and adaptable protocol for tracking important barley traits for marker-assisted selection (MAS). Marker-assisted selection provides opportunities for enhancing the response from selection because molecular markers can be applied in early generations, with high precision and reductions in cost (Miedaner and Korzun 2012). Marker assisted selection (MAS) has been an effective tool for ensuring the transfer of quantitative trait loci (QTL) with high effects into elite barley lines. Genotyping by Multiplexed Sequencing (GMS) is a highly adaptable and effective genotyping tool for tracking markers of interest with high accuracy, which can translate to a more efficient and affordable MAS process in future breeding projects. Multiplexing on a small scale has been demonstrated by (Tamang et al. 2019; Poudel et al. 2018; Richards et al. 2016). The primer pools described therein have low marker complexity with approximately 12 markers per pool. The protocol presented here provides an avenue to multiplex more than 300 markers in a single pool.

Many marker systems have been developed over the years in order to detect genetic differences among DNA samples and SNPs have become the most useful in high throughput next generation sequencing (NGS). SNPs are the most commonly used marker in genetic studies because they represent the greatest source of genetic variability in eukaryotic organisms (Agarwal, Shrivastava and Padh 2008; Brookes 1999; Bybee et al. 2011; Campbell et al. 2015; Liu 2015; Onda et al. 2018; Voss-Fels and Snowdon 2014; Ruff et al. 2020). SNPs occur throughout the genome in both genic and non-genic regions, occurring more often in non-genic regions with no phenotypic correlation (Agarwal et al. 2008). SNPs found in genic regions are invaluable to breeders and researchers alike, because they may directly impact an organism’s phenotype (Allen, et al. 2011; Chan 2009) (Ruff et al. 2020). SNPs are generally bi-allelic and have a low mutation rate making them ideal for mapping, genomic studies, and MAS (Syvanen 2001; Batley and Edwards 2007; He et al. 2009; Liu et al. 2014; Poland et al. 2012; Zhao et al. 2011; Bernardo et al. 2015). This protocol includes 267 SNP markers with known QTL associations (Table 2).

GBS and SNP chip technologies are currently the most popular protocols for SNP genotyping to date. However, both technologies have their problems. GBS results can be arduous to analyze if the whole genome is not known, because it produces random DNA sequences within the genome. The random selection and amplification of DNA sequences through the GBS process results in missing data between samples. Therefore, imputation is required, reducing the accuracy of GBS. Currently the 50 k Illumina Infinium iSelect chip is the primary chip used for genotyping in barley; however, a 9 k chip preceded the 50 k chip and was incorporated into the 50 k chip during its development (Bayer et al. 2017). The majority of SNP’s on the 50 k iSelect chip have been derived from variants called in exome capture data of a wide range of European barley germplasm (Bayer et al. 2017). SNP chip technology is based on a hybridization technique which results in a dot plot that may misrepresent true allelic states, particularly at heterozygous loci. Also, once the SNP chip is developed researchers cannot easily add new markers, making adaptation and customization impossible. This is not the case for the GMS protocol which makes adding and customizing the marker pool less difficult. In barley, the GMS protocol produces greater genetic accuracy in the event of a false heterozygote call at a locus which has a paralogue in the genome. The GMS protocol is designed to be used in genomic selection and sequencing may be sequenced on either the Illumina or Ion Torrent NGS platforms (Ruff et al. 2020).

Materials and methods

Barley markers for this project were selected due to informative nature and chromosome coverage for use in genomic selection. They were derived from markers developed with the 50 k barley SNP chip (Bayer et al. 2017). To ensure performance when multiplexing, five hundred markers were selected for each of three pools for a total of 1500 barley markers. Primers were then designed from sequence data using Sequenom MassArray Design 4.0 (Sequenom, San Diego, CA.), which reduces the chance of primer dimers and standardizes the annealing temperature. These three marker pools were then multiplexed on 25 lines of the Oregon Wolfe Barley (OWB) bi-parental population. The OWB population was developed by Dr. Bob Wolfe by systematically crossing recessive alleles into one parent and dominant alleles into the other. The population consists of 175 barley doubled haploid progeny (Costa et al. 2001). Figure 1 provides an overview of the PCR based, direct sequence read GMS protocol. Due to duplicate markers with different but synonymous names the total resulting markers in all pools were reduced to 830 unique markers. To ensure data quality duplicate samples were genotyped to confirm output calls. The majority of cleaning is completed in the analysis pipeline which includes throwing out calls with inadequate read depth below 5 amplicons per sample. The protocol used here has been adapted to barley from the GMS protocol used in wheat and published earlier in 2020 (Ruff et al. 2020).

Fig. 1
figure 1

Summary of the GMS process

Full size image

PCR 1

The first PCR produces the target amplicons using a 10 µl reaction. Locus-specific forward primers have an M13 tail attached to the 5′ end. Reagents include 2.35 µl H2O, 1 µl 10 × MCLAB Taq PCR Buffer with 20 mM MgCl2 (MCLAB, San Francisco, CA.), 0.45 µl MgCl2 at 1.125 mM, 1 µl dNTP’s at 100 mM, 1 µl primer pool at 125 nM, 4 µl sample DNA at 6.6 ng/µl, and 0.2 µl HOTaq DNA Polymerase (MCLAB, San Francisco, CA.) at 5 unit/µl. Thermocycler conditions were: 10 min at 94 °C for initial denaturation, then cycle 35 times at 94 °C denaturation for 20 sec, 56 °C annealing for 2 min, and 68 °C extension for 30 sec; followed by final 3 min elongation at 72 °C.

PCR 2

PCR 1 amplicons are diluted 1:1 and used as template DNA in the second PCR. PCR 2 adds a unique barcode and standard Ion A Adapter to the M13 tail of the forward primer, which then becomes the forward primer in successive amplicons. Barcodes are used for sample identification and the standard Ion A Adapter is used for attachment to the Ion Proton chip for sequencing. Reagents of PCR 2 include 0.875 µl H2O, 0.5 µl 10 × MCLAB Taq PCR Buffer with 20 mM MgCl2 (MCLAB, San Francisco, CA.), 0.2 µl MgCl2 at 25 nM, 0.025 µl dNTP’s at 100 mM, 0.2 µl P1B Primers at 10 µM, 2 µl unique barcoded adapters at 2 µM, 2 µl diluted PCR #1 product, and 0.2 µl HOTaq DNA Polymerase (MCLAB, San Francisco, CA.) at 5 unit/ µl. Thermocycler conditions: 94 °C 10 min for initial denaturation, then cycle 15 times at 94 °C denaturation for 20 sec, 60 °C annealing for 30 s, and 72 °C extension for 1 min, followed by final 3 min elongation at 72 °C.

PCR clean up and amplicon quantification

After PCR 2 approximately 6 µl are taken from each well of the plate and combined into a single 1.5 mL Eppendorf tube. The final combined PCR product is purified using the QIAquick PCR Purification Kit (Qiagen Hilden Germany). The PCR product is further purified using Agencourt AMPure XP beads (BeckmanCoulter, Indianapolis, IN) using 0.8 µl of beads per 1 µl of PCR product. Then size selection is done with 4% E-Gel SizeSelect Gel (Life Technologies, Carlsbad, CA), selecting for products between 144 and 233 bp. further gel purification is performed with a QIAquick Gel Extraction Kit (Qiagen) resulting in an elution at 30 µl EB. The final size selection is accomplished on a 2% E-Gel Size Select Gel (Life Technologies, Carlsbad, CA) targeting products between 140 and 250 bp and followed by a final purification with the QIAquick PCR Purification Kit. The purified and size-selected product was quantified using Qubit dsDNA HS assay kit (Life Technologies, Carlsbad, CA), which provided an initial DNA concentration. The final quantification is performed with the Agilent DNA 7500 Kit (Agilent, Santa Clara, CA), which allows for the final dilution to a minimum concentration of 80 pmol/L and have amplicon sizes between 185 and 260 bp. The GMS product is spiked with a compatible and diverse NGS library for sequence complexity. The GMS library must then be loaded onto the Ion Proton chip for sequencing. GBS libraries are loaded onto the Ion Proton chip using the Ion Chef (Life Technologies, Carlsbad, CA) with the Ion PI™ Hi-Q Sequencing 200 and the Ion PI™ Chip Kit v3. The loaded chips are then sequenced on an Ion Torrent Proton NGS platform (Life Technologies, Carlsbad, CA).

Sequence analysis pipeline

Individual OWB lines were de-multiplexed from raw sequence data using the unique barcode attached in PCR 2. Then raw sequence data from each OWB line are analyzed with a custom Python script. The first step uses cd-hit-est which collapses the raw sequence files into approximately similar sequence reads (Li and Godzik 2006). Each one of these clusters of unique sequence reads are considered a bin. The bins are then aligned to target sequence keyfiles with Muscle (Edgar 2004) at 100% homology excluding the SNP of interest. This output is organized in a FASTA file which notes the number of reads per sample and percent match to the marker key-file using a custom Python script. These results can then be looked into individually for further analysis within each line and marker for genotyping clean up if genotyping ambiguity exists.

Results and discussion

Duplicate samples of OWB DNA were analyzed for each of the 25 OWB samples to provide redundancy on ambiguous calls and provide confidence in analysis. During the development of this protocol it was discovered that adding sequence diversity resulted in higher quality sequence reads. To test the accuracy of this new genotyping protocol in barley, our results were compared to the published 50 k Illumina iSelect genotyping array results of the OWB population. The results matched 99.998% at each locus across all 25 OWB lines sequenced with the GMS protocol. Out of the 4,824 data points only one was called differently than the published data, excluding missing calls. A sample of 201 barley markers were selected for comparison which had the greatest polymorphic information content (PIC) and least missing data. The average missing data for each marker across all 25 OWB lines was 5%, with the largest being 12%. Of the 577 barley markers; 77 map to chromosome 1H, 85 to 2H, 67 to 3H, 66 to 4H, 143 to 5H, 48 to 6H, and 91 to 7H (Fig. 1). Marker chromosome and map location was derived from the BOPA consensus map (Close et al. 2009). Marker distribution across the genome has an average cM distance less than 3.03 per chromosome with an average of 82 markers per chromosome (Fig. 2)(Table 1). Polymorphic marker content was 47% on our panel of 25 OWB lines. The markers used in this study are not all informative, as many markers were selected to maximize genome coverage for use in genomic selection.

Fig. 2
figure 2

Heat map of marker coverage across all 7 barley chromosomes

Full size image
Table 1 Average cM distance between markers, largest cM gap, and total markers across all 7 barley chromosomes
Full size table

The first two pools of 500 markers were selected as markers of interest and markers with a known function. The third pool was selected to optimize chromosome coverage and fill in chromosomal gaps. This resulted in a total of three barley marker pools for genotyping. Chromosomal coverage is fair with few gaps larger than 10 cM (Fig. 2). The average map distance across all chromosomes is 3.03 cM, so any QTL would theoretically be in linkage disequilibrium with at least one marker. Average linkage disequilibrium was observed to decay below a critical level (r2-value 0.2) within a map distance of 5–10 cM (Pasam et al. 2012). There is at least one example of cM distance greater than 10 on each chromosome, chromosome 6H has 4 chromosomal gaps larger than 10 cM.

Genetic maps for all 7 barley chromosomes were generated from the chromosomal position identified in the 50 k iSelect SNP Chip consensus map (Fig. 3). Of the 468 markers with a known function or associated QTL present in the three pools, 267 produced genotypic data (Table 2). Of the remaining 213 markers with known function that did not work only 32 amplified the loci, but did not produce results because they did not match the sequence keyfile used for genotyping. This failure was largely due to primer dimers. When comparing the most robust genetic variability from this protocol matches the OWB consensus genotype at 99.998% which provides strong evidence for the efficacy of this protocol. This protocol is robust and capable of being customized to any genome or marker of interest, which may be very useful in marker-assisted selection and association mapping. Due to these markers being selected from known markers in the 50 k iSelect chip, which were derived from a wide range of European barley germplasm, there could be ascertainment bias as winter and facultative barley are under-represented in the 50 k chip.

Fig. 3
figure 3

Genetic map of barley markers. Barley marker location was determined from the BOPA consensus map

Full size image
Table 2 Markers with known associated QTL/gene, 267 markers total
Full size table