An In-Solution Hybridisation Method for the Isolation of Pathogen DNA from Human DNA-rich Clinical Samples for Analysis by NGS

E.coli

Preliminary experiments were carried out to isolate E.coli DNA from mixed samples of E.coli and human DNA. E.coli was chosen because of the small size (~4.6Mb) and neutral GC content (50.8% GC) of its genome [13]. A range of mixtures were attempted including 10:90, 25:75 and 50:50 E.coli:human DNA. In all cases, the proportion of E.coli DNA was significantly enriched after the pulldown procedure, with over 85% of the sequencing reads aligning to the E.coli genome after pulldown enrichment (Table 1, 1A). The depth of coverage throughout the genome was significantly higher after enrichment, further demonstrating the utility of this technique (Table 2, 1A). The GC content and the evenness of coverage were comparable between the pre- and post-enrichment samples, showing that the enrichment procedure is not biased for this genome (Figs. 1A and 1C).

Further experiments were carried out to determine the optimum DNA requirements for probe synthesis and input DNA. The conditions evaluated were 1600ng or 800ng of library for enrichment and 1.5μg or 3μg starting DNA for probe synthesis. There was little difference between the conditions under investigation in terms of enrichment or GC content (Fig. 1B and Table 1, 1B). The increase in coverage depth after enrichment was similar for the different conditions and the evenness of coverage was also similar (Fig. 1D and Table 2, 1B). 800ng of input DNA and 1.5μg of probe (initial amount for probe synthesis) were chosen for further experiments. During the development of this technique, three different biotinylation techniques were trialled for the synthesis of the probe molecules and it was found that they all provided similar results in terms of enrichment and sequencing statistics (see Supplementary Data Files 1 - 5).

P.falciparum

The Plasmodium genome has been a strong research focus of the Wellcome Trust Sanger Institute through its involvement in the sequencing of the reference strain 3D7, and subsequent genomic diversity studies in laboratory and field isolates (Sanger Institute Malarial Programme www.sanger.ac.uk/malaria) because of the high disease burden inflicted by malaria. Plasmodium was predicted to be a more challenging organism for solution-based capture due to its larger genome (22.8Mb) and high content of AT bases (80.6%) [1]. In this study, initial experiments were carried out using probes made from the P.falciparum laboratory strain 3D7 and these were used to enrich for the same strain within test mixtures of Plasmodium and human DNA (25:75 Plasmodium :human). Different amounts of probe (1.5μg and 3μg of starting DNA for probe synthesis) and library (500ng, 800ng and 1600ng) were assessed to see which provided the best combination for enrichment (Table 3, 2A and 2B). As previously seen with the experiments using the E.coli probe and DNA, a high level of enrichment was seen with all samples (on average 71.3% of sequencing reads aligning to the Plasmodium genome after enrichment) and the best results were obtained with 800ng of library and 1.5μg of starting material used for probe synthesis (Table 3, 2A and 2B). Despite the high levels of enrichment, a comparison of the GC content of the pre- and post-capture samples showed that the enrichment did not work well for AT-rich regions and the GC content of the post-enrichment samples was higher than the average for both the theoretical reference genome (19.4%) and the actual sequencing of the pre-pulldown samples (Fig. 2A and 2B). This is likely to be caused by differences in the hybridisation temperature and efficiency of AT- and GC-rich DNA sequences [14, 15].

As an initial measure to counteract this enrichment bias, the hybridisation temperature was lowered from 65°C to 55°C. A lower hybridisation temperature was hypothesised to encourage better annealing of AT-rich sequences [14, 15]. However, in these experiments, the hybridisation temperature appeared to have little effect on the GC content of the enrichment and the lower hybridisation temperature actually reduced the overall enrichment of Plasmodium DNA from the human mix by a small amount, possibly due to increased non-specific binding of the human DNA (Fig. 2C and Table 3, 2C).

As an alternative method to counteract the high GC bias of the Plasmodium genome observed during sequencing after solution-based hybridisation, TMAC (tetramethyl ammonium chloride, Sigma-Aldrich) was added to the hybridisation buffer. TMAC is an additive which stabilises AT base pairs, causing AT-interactions to behave similarly to GC-interactions and therefore acting to reduce bias effects across the genome [16, 17]. It has previously been shown that in 3M TMAC an AT base pair is as thermally stable as a GC base pair [17]. In these experiments, between 0.1M and 1M TMAC was added to the hybridisation buffer to increase the hybridisation of highly AT-rich regions of the Plasmodium genome. This additive had little effect on the amount of enrichment seen after pulldown and did not appear to significantly improve the hybridisation of the AT-rich regions (Fig. 2D and Table 3, 2D). Possibly, insufficient TMAC was added to have a significant effect, however further TMAC could not be added due to the composition of the hybridisation buffer being such that it was impossible to add further TMAC to the buffer.

The extremely AT-rich regions of the Plasmodium genome are in non-coding areas of the genome and therefore the GC content of the exonic region (~23.7% GC) is not as extreme as seen for the non-coding regions (~13.5%) [1]. The exonic region is also the most interesting part of the genome as it contains all of the protein-coding information. A further analysis, concentrating on the enrichment of the coding regions after pulldown, demonstrated that the average and maximum coverage of the coding region were greatly increased after the enrichment procedure and up to 98.2% of the coding region was covered at 5x depth (Fig. 3, Table 4, 2A-2D). The maximum depth of coverage was significantly higher after enrichment due to a lack of uniformity in coverage after pulldown has occurred (Fig. 3, Table 4, 2A-2D). This is likely to be caused by the differences in hybridisation of GC- and AT-rich regions. Hybridisation temperature and TMAC addition did not significantly affect the amount or the eveness of coverage of the coding region recovered after hybridisation (Table 4, 2C and 2D). Despite this coverage nonuniformity it was still possible to identify single nucleotide polymorphisms (SNPs) at 10 × coverage at a similar rate pre- and post-enrichment (Table 4, 2A-2D). This observation would suggest that the uneven coverage does not affect SNP calling and that there is sufficient coverage of the exome after enrichment for useful analytical measurements to be made.

In order to further assess the utility of this technique, it was important to test other strains of P.falciparum. In a preliminary experiment, probes were made from P.falciparum laboratory strain 3D7 and these probes were used to selectively hybridise DNA from P.falciparum strain Dd2 from a mixed Plasmodium and human sample (30:70 P.falciparum Dd2:human). The hybridisation was successful, but the amount of enrichment was lower (67.49% of sequencing reads aligning to the Plasmodium genome after enrichment) than was seen when both the probes and DNA library for enrichment were generated from the same strain (Table 3, 2E). This was expected because the genomes of the two strains are somewhat different and some regions will not hybridise during the enrichment process. In addition, the enriched DNA was aligned to the 3D7 reference genome and some sequences specific for the Dd2 strain would not register as Plasmodium DNA. This manifests itself as a higher amount of ‘other’ DNA observed after enrichment for the Dd2 strain. No reference was available for the Dd2 strain. The GC content of the enriched Dd2 DNA was similar to that observed when enriching a P.falciparum 3D7 genome (Fig. 2E). Approximately 95% of the coding region was covered to 5× depth in the enriched sample, and this is similar to the value obtained for the prepulldown sample (96%), indicating that little of the coding region DNA is lost during the hybridisation (Table 4, 2E). The average depth of coverage across the exome was approximately double for the postpulldown Dd2 strain sample (227×) compared to the prepulldown sample (99×) and the maximum depth of coverage was approximately 8× higher in the enriched sample (8007×) compared to the non-enriched sample (1138×). Taken together, these results suggest coverage non-uniformity in the enriched sample with some areas having extremely high levels of coverage. SNP calling was carried out for these samples using the P.falciparum 3D7 genome as a reference and hence many SNPs were detected both pre- (5599 SNPs) and post-enrichment (5158 SNPs) due to differences between the two strains (Table 4, 2E). 92% of the SNPs detected pre-enrichment could also be detected in the enriched sample, showing good SNP detection in these samples.

To further test this technique with other Plasmodium strains, probes were made from a mixture of four different P.falciparum strains (3D7, Dd2, IT and HB3). This mixed probe was used to enrich for P.falciparum reference strain 3D7 DNA from a mixture with human DNA (Plasmodium: human 20:80). The four strain probe mix provided as good an enrichment as the single strain probe mix previously used but the coding region showed poorer coverage after enrichment with the mixed probe (Table 3, 2F and Table 4, 2F). This could possibly suggest that the DNA from the HB3 and IT strains was of poor quality or that these strains are quite divergent in exonic regions and have fewer regions of complimentarity with the 3D7 strain. Despite only 67% of the coding region being covered at 5x depth after enrichment, compared to 99% pre-enrichment, the number of SNPs detected was similar before and after enrichment (Table 4, 2F).

This method was also tested using mixtures of human DNA and one of two P.falciparum clinical isolates. The two clinical samples used in this study were collected from clinical malaria patients in Southeast Asia and were cultured for a short while before DNA extraction. They are multiclonal in nature and contain a small amount of human contamination even before they were mixed with human DNA (20:80 P.falciparum clinical isolate:human) in these experiments. The enrichment of these clinical samples was comparable, albeit not quite as good as for the reference strain, with over 65% of the DNA after pulldown, using the four strain mixed probe, being complimentary to the P.falciparum 3D7 reference genome (Table 3, 2G). Similar to previous experiments, the AT-rich regions were not captured during the hybridisation and therefore the GC content is not very close to the theoretical value (approximately 35% GC, Fig. 2F). However, the coding region showed acceptable coverage in both samples post-enrichment (clinical 1 – 79.22% at 5× coverage and clinical 2 – 59.39% at 5× coverage, Table 4, 2G), although to a lesser degree than the pre-pulldown samples. Several SNPs were detected in both samples when compared to the 3D7 reference strain (Table 4, 2G). SNP detection was poorer after enrichment, with 79% (4097/5181) and 70% (3310/4696) (clinical sample 1 and 2 respectively) of SNPs being detected after pulldown (Table 4, 2G). Overall, results from the mixed probe experiments demonstrate that this probe was not very successful for enrichment and this is likely to be due to poor quality of DNA from the IT and HB3 strains. Better results may have been achieved using a pure 3D7 probe with these clinical samples, however this was not attempted. Using a reference strain such as 3D7 to enrich for clinical samples will always miss some regions of the genomes of the clinical isolates due to differences in the genomic sequences. This is a caveat of this kind of experiment and interesting, novel regions from the clinical isolate’s genome will be missed, however SNP calling should still be possible.

DNA

Male human DNA (Promega) and E.coli DNA (strain B, Sigma-Aldrich) were purchased from commercial suppliers. P. falciparum DNA was obtained from either i) long-term cultured parasite lines (laboratory clones 3D7, HB3, IT and Dd2) or ii) short-term culture parasite isolates obtained from patients with malaria in Cambodia. P. falciparum in vitro culture was performed as previously described [18]. Genomic DNA was extracted from infected red blood cell pellets using the QIAamp DNA Blood MidiKit (Qiagen) according to the manufacturer’s protocol.

Clinical samples were collected with written consent from adult patients or from a parent or guardian of paediatric patients. These samples were collected under a project with ethical approval obtained from the Ministry of Health in Cambodia.

PROBE SYNTHESIS

Probes were synthesised from 1.5μg or 3μg of whole-genome E.coli or P. falciparum DNA samples, or 10μg of whole-genome human DNA. The DNA was fragmented to approximately 150bp by adaptive focused acoustics using a Covaris S2 (Covaris. Inc.) with AFA tubes (Duty cycle: 20, Intensity: 5, Cycle burst: 200, Duration: 180). Subsequently, the DNA fragments were end repaired (1 hour at 20°C) and A-tailed (1 hour at 37°C) following the standard Illumina library preparation protocol and using the NEBNext DNA Sample Prep Reagent Set 1 (New England Biolabs). Intervening purification steps were by column purification (QIAquick or Minelute PCR Purification Kits, Qiagen). The DNA fragments were then enzymatically biotinylated by addition of 0.1mM Biotin-16-dUTP (Biotium), 100U terminal transferase, 1.25μM CoCl₂ and 1× terminal transferase buffer (New England Biolabs). Incubation was carried out for 30mins at 37°C prior to column purification using the Minelute PCR Purification kit (Qiagen) and elution in 10μl of EB Buffer (Qiagen).

LIBRARY SYNTHESIS

Libraries were synthesised from a total of 1μg of DNA consisting of a mixture of E.coli or P.falciparum DNA mixed in differing proportions with human DNA. The DNA was fragmented to approximately 200-300bp using the Covaris S2 with AFA tubes (Duty cycle: 20, Intensity: 5, Cycle burst: 200, Duration: 120). End repair, A-tailing and adapter ligation were carried out following standard Illumina library preparation methodology. The NEBNext DNA Sample Prep Reagent Set 1 (New England Biolabs) was used throughout and incubations were carried out for 1 hour at 20°C (end repair), 1 hour at 37°C (A-tailing) and for 2 hours at 18°C (adapter ligation). Standard Illumina adapter sequences (forward 5′-GATCGGAAGAGCGGTTCAGCA GGAATGCCGAG-3′, reverse 5′-ACACTCTTTCCCTA CACGACGCTCTTCCGATCT-3′) were used. Column purification (QIAquick or Minelute PCR Purification Kits, Qiagen) was carried out between each stage and after the final column purification step the libraries were eluted in 50μl of EB buffer (from the Qiagen kit). PCR amplification was carried out using 2μl Herculase II Phusion DNA Polymerase, 0.1mM dNTP mix and 1× Herculase Reaction Buffer (Stratagene) in a total volume of 100μl. 0.2μM of each of the standard Illumina primer sequences (forward 5′-AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATCT-3′, reverse.

5′-CAAGCAGAAGACGGCATACGAGATCGGTCTCGGCATTCCTGCTGAACCGCTCTTCCGATCT-3′) were used per reaction and these were purchased from Sigma-Genosys. Amplification was achieved using the following cycling conditions: 98°C for 30sec followed by six cycles of 98°C for 10sec, 65°C for 30sec, 72°C for 30sec and a final elongation incubation of 72°C for 5mins. After amplification, the libraries were purified using AMPure beads (Agencourt) and eluted in a total of 60μl of water (the beads were eluted twice with 30μl). The Agilent 2100 Bioanalyser (Agilent Technologies) and the associated DNA1000 kit (Agilent Technologies) were used to determine quality and concentration of the libraries.

HYBRIDISATION

500ng, 800ng or 1600ng of each library (determined from the Agilent 2100 Bioanalyser traces) was lyophilised using a Concentrator Plus/Vacufuge Plus (Eppendorf) at room temperature. Once completely dried, the libraries were resuspended in 5μl of ultrapure water. 1μg of each of two blocking oligos (standard Illumina primer sequences, 5′AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATCT3′

5′CAAGCAGAAGACGGCATACGAGATCGGTCTCG GCATTCCTGCTGAACCGCTCTTCCGATCT3′) was added to the library samples. Library samples were denatured at 96°C for 5 minutes before incubation at 65°C. 40μl aliquots of 2× hybridisation buffer (10× SSPE (Sigma-Aldrich), 10× Denhardt’s Solution (Sigma-Aldrich), 10mM EDTA (Calbiochem), 0.2% SDS (Sigma-Aldrich)) underwent a similar denaturation and incubation process. After incubation of the libraries and hybridisation buffer aliquots at 65°C for 5 minutes, one 5μl aliquot of probe solution (half of the original 3μg preparation) for each library was mixed with 5μl of water prior to incubation at 96°C for 5 minutes. The probes and 16μl of hybridisation buffer were then added to the libraries, whilst maintaining the 65°C incubation temperature. Incubation was then carried out for a further 24 hours at this temperature.

ISOLATION OF HYBRIDISED DNA

After incubation was complete, the samples were added to Dynabeads® MyOne™ Streptavidin C1 beads (Invitrogen) that had been pre-washed and resuspended in 200μl of binding buffer (1M NaCl (Sigma-Aldrich), 10mM Tris-HCl pH7.5 (Sigma-Aldrich), 1mM EDTA (Sigma-Aldrich)). The samples were then placed on a Nutating Mixer (VWR International) for 30mins to mix. The beads and solutions were then separated using a magnetic separator and the beads were washed with Wash Buffer 1 (1×SSC (Sigma-Aldrich), 0.1% SDS (Sigma-Aldrich)) before incubation at room temperature for 15 minutes. Wash buffer 1 was then removed and three washes with Wash Buffer 2 (0.1×SSC (Sigma-Aldrich), 0.1% SDS (Sigma-Aldrich)) were carried out. Wash buffer 2 was pre-warmed to 65°C and incubations between each step were carried out for 10 minutes at this temperature. 50μl of 0.1M NaOH (Sigma-Aldrich) was then added to the beads. The beads were mixed thoroughly before magnetic separation and collection of the sample eluate. The eluate was then neutralised by addition of an equal volume of 1M Tris-HCl pH7.5 (Sigma) before the samples were desalted by column purification (QIAquick PCR Purification Kit, Qiagen) and eluted in 36μl of EB buffer. The recipes for many of the buffers used in this protocol were obtained from Gnirke et al., (2009) Nature Biotechnology, 27 [9].

SAMPLE QUANTIFICATION

Samples were quantified both pre- and post-hybridisation by qPCR using the KAPA Illumina ABI Library Quantification Kit (KAPA Biosciences), following the manufacturer’s protocol.

SEQUENCING

Pre- and post-hybridisation samples were sequenced using an Illumina GAIIx machine and associated Illumina reagents. Samples were loaded at 12-14pM and a single lane of paired-end 76bp sequencing (except experiment 1A where 37bp paired-end sequencing was used) was carried out for each sample.

DATA ANALYSIS

All read mappings were carried out using the software package BWA and the coverage data were derived from their pileup analysis using the software package SAMtools [19, 20]. Using the same mapping results, GC plots were calculated. The theoretical GC curves were generated from the genome reference using a read generated starting from each base in the reference and of the same length as the sequencing reads.

For P.falciparum SNP analysis, sequence reads were aligned to the 3D7 reference genome (version 2.1.5) using BWA. A list of SNPs was generated using SAMtools and BCFtools at default settings. A minimum SNP quality of 30 (1 error per 1000bp) was used to filter the initial list of SNPs. For further analysis, the SNPs that mapped to telomeric, subtelomeric, non-unique regions and highly variable gene families were excluded. Allele calls were compared between samples using the R statistical package. To observe the distribution of coverage and potential location of SNPs on the whole genome, sequencing data was uploaded into LookSeq, a browser-based viewer for deep sequencing data [21].