DNA isolation from yeast cells
A modified version of the PrepEase Genomic DNA Isolation Kit (Affymetrix, 78855 1 KT) based on enzymatic cell wall digestion was used for the DNA isolation from yeast cells. This protocol can easily be applied to blood, bacteria or homogenized tissue or plant material by substituting buffers according the manufacturer's instructions. All of the mixing steps were performed by pipetting using a Biomek FX pipetting robot (Beckman Coulter) in 96-well plates. Cell pellets from 4 deep-well plates, each containing 1 ml overnight culture, were combined for the DNA isolation. The “Spheroplast” and “Enzyme Solution” from the kit was replaced by Qiagen’s Y1 lysis buffer (1 M sorbitol, 100 mM EDTA, pH 8.0, 14 mM β-mercaptoethanol) freshly supplemented with 2.5 μl/ml of Zymolyase (Seikagaku Inc.) and 2.5 μl/ml RNase A (10 mg/ml, Qiagen). 200 μl of this buffer was added to each pellet, mixed and incubated at 37°C for 90 min with gentle shaking every 30 min. 200 μl of water was added to each well and the plate was centrifuged at 6000 x g for 4 min (for centrifuges with maximum 3,000 x g, centrifugation times can be tripled) and the supernatant was decanted. 120 μl Homogenization Buffer was added and mixed to resuspend the pellet completely. 100 μl of chloroform and 400 μl of Protein Precipitation Buffer were added to the lysate and mixed. Plates were centrifuged at 6,000 × g for 15 min. 450 μl of the upper aqueous phase was transferred with the robot (pipetting height was optimized in advance) to a 1 ml deep-well plate containing 340 μl of isopropanol per well. The solution was mixed, left for 15 min at room temperature, and centrifuged at 6,000 × g for 15 min. After decanting the supernatant, 1 ml of cold 70% ethanol was added to the pellet, mixed and centrifuged at 6,000 × g for 10 min. The supernatant was decanted, and the tube was placed upside down on a paper towel and dried for 5 min at 37°C. The DNA pellet was resuspended in 300 μl of DNA Resuspension Buffer or Elution Buffer (EB, 10 mM Tris HCl) by shaking plates for 30 min at 37°C and later by mixing. A detailed Biomek protocol including .bmf files is provided in Additional file 3.
1–10 μg of genomic DNA from each of the 768 segregants was sheared using a Covaris E series sonicator in 130 μl to obtain a fragment size with a major peak of ~250 bp (Duty cycle 10%, Intensity 4.5, Cycles per burst 200, 120 sec). Samples from Covaris sonication were transferred to 96-well PCR plates, dried in a Speedvac, and resuspended in 30 μl of EB. Alternatively, we tested sonication in PCR-strips using a Sonorex RK 102 sonicating water bath (Bandelin). For this, two 8-strips held on a support plate were fixed to a cycling pin that rotates during sonication (Additional file 4: Video S1). After 4 min sonication at 4°C, samples were spun down and sonicated for another 4 min. A uniform size distribution was obtained by keeping the volume and DNA amount constant (2 μg in 25 μl). All samples were run on a 1.5% agarose gel to verify the fragment size.
End repair, dA-tailing, and ligation using heat-inactivation
Instead of the standard column or bead-based cleanup steps, we heat-inactivated the enzymes used for end repair, dA-tailing, and ligation, then added the respective enzyme (+ buffer). For this, 250 ng of fragmented gDNA were used for the library preparation in 96-well PCR plates in a volume of 17 μl. End repair for the fragments was performed by adding 3 μl of End repair master mix composed of 2 μl of End repair buffer and 1 μl End repair enzyme (NEBNext End Repair Module, NEB #E6050L) using a 8-channel pipette. The contents were mixed by vortexing, shortly spun down, and incubated in a thermocycler at 20°C for 45 min. The enzymes were then heat-inactivated at 75°C for 15 min. The contents of the plate were quickly spun down, and 2 μl of A tailing master mix containing 1 μl of Klenow Fragment (3′→5′ exo–) (NEB #M0212L), 0.5 μl of nuclease free water, and 0.5 μl of 100 mM dATP (NEB #N0440S) were added to the 20 μl reaction. The contents were mixed by vortexing, spun down, and incubated in a thermocycler at 37°C for 45 minutes. The enzymes were then heat-inactivated at 75°C for 15 minutes. 5 μl of ligation master mix containing 3 μl 10X T4 DNA ligase buffer and 2 μl T4 DNA ligase (NEB #M0202L) were added to the reaction followed by 3 μl of 7 μM multiplex barcode adapters (aliquoted into 8-strip PCR tubes or 96-well plates for convenient pipetting, see Additional file 1: Table S2 for sequences). The concentration of adapters was optimized to reduce the formation of adapter dimers. The reaction contents were mixed well, spun down, and incubated on a thermocycler at 16°C for 1 h followed by heat inactivation at 75°C for 15 min.
Pooling and size selection
After barcode ligation and heat inactivation, 48 samples were pooled together by combining 5 μl of each sample in a 1.5 ml reaction tube. The samples were cleaned up and concentrated to 40 μl using 1x Ampure XP cleanup. This pooling step reduces the sample size from 96 to two for the subsequent size selection and PCR. For size selection, 25 μl (roughly 1.25 μg) of ligated DNA was loaded on a 2% E-Gel SizeSelect (Invitrogen) and DNA fragments were collected at 350 bp and 400 bp. The DNA concentrations were then determined by Qubit HS DNA reagent (Invitrogen).
In a 50 μl reaction, 5–10 ng of the pooled libraries were amplified. We have observed that performing PCR with an excess of template DNA (>20 ng) significantly reduces the efficiency of the PCR. The PCR was performed on a thermocycler (MJ Research tetrad) containing 1x Phusion Master Mix with HF Buffer (Thermo Scientific) and 0.2 μM Illumina PE 1.0 and 2.0 primers. The low primer concentration reduced the formation of primer dimers often observed at standard primer concentrations (1.25 μM). PCR conditions were 98°C for 45 s, 10x [98°C for 15 s, 65°C for 30 s, 72°C for 30 s], 72°C for 5 min, 4°C hold.
DNA purification with SPRI beads
The amplified DNA was purified in 0.2 ml PCR strips by mixing the DNA with 1x volume of Agencourt AMPure XP beads (Beckman Coulter) and select the magnetic beads with a homemade magnetic stand (Additional file 2: Figure S6). This stand consists of neodymium magnets (Webcraft GmbH, Gottmadingen, Germany) mounted on trimmed 96-well plates and can be used in combination with an 8-channel pipet. For a high-throughput SPRI clean-up we further provide a detailed protocol of the pipetting steps for 96-well plates and Biomek robot in Additional file 3. DNA concentrations were quantified for the subsequent library preparation (see DNA quantification and quality control).
DNA quantification and quality control
The quality of individual samples of isolated DNA was determined by a photospectrometric measurement using a NanoDrop 1000 (Thermo Fisher). For quantification of genomic DNA and pre-PCR libraries in 96-well plates, we used Quant-iT PicoGreen dsDNA Reagent (Invitrogen) in optical plates (Greiner). The fluorescence was measured at 485 nm excitation and 535 nm emission in a Genios microplate reader (Tecan) according to the manufacturer’s instructions. The pooled libraries were quantified before and after PCR with a Qubit spectrofluorometer (Invitrogen) according to the manufacturer’s instructions. All pre- and post-PCR libraries were run on a High Sensitivity Bioanalyzer chip (Agilent) to determine the size distribution. After a 10-cycle PCR, we typically observed a 10-fold increase in DNA amount and a 24–30 bp increase in library size due to adapter elongation. Depending on the amplification efficiency, either the low (350 bp) or the high molecular weight (400 bp) library was selected for sequencing. The samples were then diluted to 10 nM and clustered on the Illumina cBot clustering station for paired-end sequencing on an Illumina HiSeq 2000.
Design of 48 multiplexing barcodes
We designed a set of 48 adapters, each with a different hexamer sequence just before the T-overhang, similar to Lefrancois et al . We selected 64 of 96 Illumina barcodes, which had at least a 3 bp difference compared to any of the other 63 barcodes. From this set, 48 barcodes that had an equilibrated base composition at the first two bases (for better cluster calling) were manually chosen. Following quality control analysis, we replaced the seven poorest performing barcodes with new ones (Additional file 1: Table S2).
To demultiplex, we extracted the first six bases of each read and compared it to all possible barcodes. The perfect match or best hit to one barcode with the least number of mismatches was assigned to the read. For genotyping, reads from the segregants along with both SK1 and S96 (a haploid strain isogenic to S288c) parental strains were aligned to the S288c reference genome (build R63) using Novoalign (v2.07.06; http://www.novocraft.com/), allowing for unique alignments. Thereafter GATK was used for realignment and recalibration of the bam files , and subsequent SNP calling was performed using SAMtools . The vcf file produced by SAMtools contains a list of variant positions and the individual genotype calls across all samples at each variant position. The formula that SAMtools applies for calling the genotype is dependent on allele frequency, which is not directly applicable to our study, because the allele frequency at true SNP positions is expected to be 0.5 in crosses generated from 2 parents. Instead, we used the genotype likelihood (PL stats generated by GATK) to infer the genotype. SNP positions, which correspond to a homozygous reference call in the S96 parent and a homozygous variant in the SK1 parent, are chosen first. From this set of SNPs, we excluded calls whose allele frequency is not between 0.3-0.7. These SNP calls are unreliable and often not in linkage with their surrounding SNPs, and could either be SNPs within regions that are repetitive in one but not in the other parent, or result from misaligned reads.
GC bias and coverage plots
We calculated the genome-wide, per-base coverage of the S288c genome using SAMtools. Positions where all samples had at least 1 read were considered. The density was plotted with a bandwidth of 0.1. For plotting GC bias, the genome was divided into non-overlapping 200 bp bins, and the depth was estimated by the mean values of per-base depth in these bins. Bins with less than 50% covered by at least 1 read were excluded. All analyses were run in the software R (v. 2.12.0; http://cran.r-project.org). For analyzing chromosomal abnormalities, an identical method for binning and GC correction was applied, except that a 10 kb bin size was used. GC bias correction was applied using a LOESS method, as described previously .
Recombination map analysis
For both genotype datasets (S96xSK1 and S96xYJM789) the rqtl package (with the function, est.map (maxit = 1000,error.prob = 0.01) was used to construct the genetic map for both crosses. After obtaining the genetic map, the genotypes were filtered for errors and crossovers counted for each segregant, using functions in rqtl (cleanGeno(maxdist = 2.5, maxmark = 2) followed by countXO). For 2-3 kb bins (partitioned by common SNPs), the recombination rate was calculated as genetic distance between 2 SNPs/physical distance between 2 SNPs. For identifying regions with difference in recombination rates, we normalized the rate in both, S96xSK1 and S96xYJM789 by setting the mean of each set to 1. Raw sequences for Spo11 oligo maps were download from SRA (GSE26452) and aligned to the S288c genome build R63 using bowtie2 allowing for only unique alignments.