A Low-Cost Exon Capture Method Suitable for Large-Scale Screening of Genetic Deafness by the Massively-Parallel Sequencing Approach

Abstract

Current major barriers for using next-generation sequencing (NGS) technologies in genetic mutation screening on an epidemiological scale appear to be the high accuracy demanded by clinical applications and high per-sample cost. How to achieve high efficiency in enriching targeted disease genes while keeping a low cost/sample is a key technical hurdle to overcome. We validated a cDNA-probe-based approach for capturing exons of a group of genes known to cause deafness. Polymerase chain reaction amplicons were made from cDNA clones of the targeted genes and used as bait probes in hybridization for capturing human genomic DNA (gDNA) fragments. The cDNA library containing the clones of targeted genes provided a readily available, low-cost, and regenerable source for producing capture probes with standard molecular biology equipment. Captured gDNA fragments by our method were sequenced by the Illumina NGS platform. Results demonstrated that targeted exons captured by our approach achieved specificity, multiplexicity, uniformity, and depth of coverage suitable for accurate sequencing applications by the NGS systems. Reliable genotype calls for various homozygous and heterozygous mutations were achieved. The results were confirmed independently by conventional Sanger sequencing. The method validated here could be readily expanded to include all-known deafness genes for applications such as genetic hearing screening in newborns. The high coverage depth and cost benefits of the cDNA-probe-based exon capture approach may also facilitate widespread applications in clinical practices beyond screening mutations in deafness genes.

Introduction

Congenital hearing loss is one of the most common human genetic defects (Chang et al., 2003; Smith et al., 2005). Universal newborn hearing screening (UNHS) is now mandated by law in most states in the United States, Canada, Europe, and some developing countries (Morton and Nance, 2006). Currently, UNHS is carried out by performing physiology-based tests (e.g., distortion production otoacoustic emission measurements). UNHS covers at least 91% of all newborns in the United States, and referral rates among those screened range from 0.3% to 8% in various testing centers when different screening methods are used (Clemens and Davis, 2001). Most newborns who fail the first-round of UNHS, however, eventually turn out to be false positives (Clemens and Davis, 2001). Another major disadvantage of the physiology-based UNHS method is its inability to define the underlying cause of deafness for patients, which in many cases are known to be caused by genetic mutations (Smith et al., 2005).

Recent commercialization of several next-generation sequencing (NGS) platforms is rapidly expanding the possibilities of genomic analyses that have been considered unpractical thus far (Shendure and Ji, 2008). Technological advances have made it possible to resequence complete human exomes or large regions of human chromosome in one experimental run (Albert et al., 2007; Okou et al., 2007; Porreca et al., 2007). However, genetic tests with the capacity of resequencing multiple deafness genes in a single test were not available until recently (Shearer et al., 2010) and is still too expensive for epidemiologic-scale clinical use. Targeted genetic analysis by the NGS method typically has two major steps (Albert et al., 2007; Okou et al., 2007): (1) a preprocessing step in which exons of targeted genes are selectively enriched in the pool of genomic DNA (gDNA) fragments; (2) a resequencing step in which the captured gDNA fragments are sequenced in a massively parallel manner. The per-run cost of NGS has rapidly been coming down in recent years. Molecular barcoding (Craig et al., 2008) further reduces per-sample cost for NGS sequencing by another 1–2 orders of magnitude. The high per-sample cost incurred by the current targeted gene capture methods (Okou et al., 2007; Porreca et al., 2007; Shearer et al., 2010) in the first step has now become a major cost component and is, thus, a major barrier for widespread applications. As a proof-of-principle test in this work, we tested and validated a low-cost cDNA-probe-based exon capturing method for the enrichment of a group of representative deafness genes. The captured gDNA fragments were sequenced by an Illumina GAII. Results demonstrated that reliable genotype calls can be obtained for various types of genetic mutations by our novel method of targeted enrichment of deafness genes.

Materials and Methods

This study was approved by the Emory Institutional Review Board (IRB, protocol# IRB00009046). Saliva samples (∼2 mL) from normal hearing human subjects and patients previously detected to have genetic mutations in GJB2 by Sanger sequencing method were collected by using the Oragene DNA sample collection kit from saliva (DNA Genotek, Inc., Ontario, Canada). gDNA was extracted by following the manufacturer's instruction. Quality of gDNA was assured by examining optical density ratio (260/280 ratio) and by gel electrophoresis imaging. High-molecular-weight gDNA (5 μg) was fragmented ultrasonically with the Covaris E210 DNA shearing instrument (Covaris, Inc., Woburn, MA) to an average size of 200 base pairs (bps). The Covaris protocol is set at 3 min total duration, duty cycle 10%, intensity 5, and 200 cycles per burst.

cDNA clones for GJB2, GJB3, GJB6, SLC26A4, and MYO15A were purchased from Origene (Rockville, MD). The five genes were chosen, because the complexity in their exon structure was representative of all other known deaf genes. MYO15A is one of the largest known deafness genes with 65 exons and 11,876 bps of mRNA, representing large deafness genes. SLC26A4 represents the medium-size deafness genes. GJB2, GJB3, and GJB6 are small genes, with GJB2 being the most commonly found deafness gene. Hybridization probes with sizes of 1.3 to 1.6 kilobase pairs (kbps) for these genes were generated from the cDNA clones by polymerase chain reaction (PCR) amplifications (primers used are listed in Supplementary Tables 1 and 2; Supplementary Data are available online at www.liebertonline.com/gtmb). The TaKaRa LA Taq reaction (50 μL reaction volume) was composed of: template DNA from the cDNA clones (10 ng), 8 μL of 2.5 mM dNTPs, 0.5 μL of TaKaRa LA Taq, and 0.2 μM primers. All fragments amplified were verified by digestions using specific restriction enzymes listed in Supplementary Tables 1 and 2 (NEB Biolabs, Ipswich, MA). Analysis of exon length distribution of the five genes showed that a small proportion of exons are less than 50 bps as indicated by a dashed box in Figure 1A. In order to ensure reliable capture of shorter exons, we specifically generated longer hybridization probes from gDNA for those exons that are shorter than 50 bps by including ∼50 bps genomic sequencing flanking the exons on both sides. All PCR products (10 ng each) were cleaned by the QIAquick PCR Purification Kit (Qiagen, Valencia, CA), mixed in Pronto!^™ Universal Slide Spotting Solution (Corning Life Sciences, Lowell, MA), and pipetted out on the surface of microarray glass slides (Corning Life Sciences). After baking in an oven (Fisher Scientific, Model Isotemp 280A, Pittsburgh, PA) at 80°C for 3 h, all slides were stored in a dry and dust-free chamber until use.

FIG. 1.

Distribution of targeted exon lengths and their enrichment checked by qPCR. (A) Distribution histogram of exon lengths for the five targeted deafness genes. Those bins fall in the dashed box are exons shorter than 50 bps. (B) Schematic diagram showing the design for the location and size of qPCR primers used in the quality control (QC) step. (C) Typical qPCR results obtained from enriched samples, nonenriched samples, and water control samples for exon 20 of SLC26A4. (D) Typical qPCR results obtained from enriched samples, nonenriched samples, and water control samples for exon 3 of SLC26A4. qPCR, quantitative polymerase chain reaction; gDNA, genomic DNA; bps, base pairs. Color images available online at www.liebertonline.com/gtmb

Fragmented gDNA libraries for Illumina GAII sequencing were prepared with the NEBNext™ DNA Sample Prep Master Mix set (E6040, NEB Biolab, Ipswich, MA). End repair of DNA fragments, addition of a 3′ adenine (A), adaptor ligation, and reaction clean-up were carried out by following the manufacturer's protocol. gDNA libraries were cleaned and size selected by using the AMPure DNA Purification kit (Beckman Agencourt, Danvers, MA). The ligated product (20 ng) was amplified for 18 PCR cycles with Illumina PCR primers InPE1.0, InPE1.0, and indexing primer by following the manufacturer's instructions. The PCR products were purified again with QIAquick MinElute column and eluted into 50 μL of hybridization buffer (HB, Roche NimbleGen, Madison, WI). The barcoded Illumina gDNA libraries (5 μg) was incubated in 16 μL of HB on the surface of hybridization glass slides on a hybridization station (BioMicro Systems, Inc., Salt Lake City, UT) at 42°C for 72 h. Nonspecific DNA fragments were removed after a series of six washing steps in a washing buffer (Roche NimbleGen, Madison, WI). The DNA bound to the probes was eluted by incubating with NaOH (425 μL, 125 mM) for 10 min. The eluted solution was transferred to a 1.5 mL Eppendorf tube containing 500 μL neutralization buffer (Qiagen's PBI buffer). The neutralized DNA was desalted and concentrated on a QIAquick MinElute column and eluted into 30 μL in EB buffer. To increase the yield, we typically amplified 5 μL of eluted solution by 12 PCR cycles using Illumina PCR primers InpE1.0 and 2.0. Enrichment of targeted deafness genes was examined by quantitative PCR (qPCR) by comparing the growth curves of captured and noncaptured samples (Fig. 1C, D). Primers used for qPCR are given in Supplementary Tables 1 and 2. Single-end Illumina sequencing of 12 barcoded libraries of captured samples was done at the Emory Georgia Research Alliance (GRA) Genome Center using one lane of the Illumina IIx Genome analyzer (Illumina, San Diego, CA). SOAP (Li et al., 2008), MAPview (Bao et al., 2009), and NextGENe (Softgenetics, State College, PA) software packages were used to analyze Illumina sequencing data. Mutations detected by the NGS method were verified by PCR-based Sanger sequencing by using an outside commercial source (MacroGen, Rockville MD).

Results

The enrichment effects by our novel method were first examined by qPCR to detect relative abundance of targeted exons in gDNA fragment pools. All qPCR products were amplified with one primer in the intron and another in the exon to avoid amplifying cDNA probes (Fig. 1B). Design details are given in Supplementary Tables 1 and 2. Two sets of typical examples of qPCR growth curves are shown in Figure 1 C, D for a larger (exon 20, 2383 bps) and a smaller (exon 3, 55 bps) exon of the SLC26A4 gene, respectively. For both exons, qPCR data indicated that the curves began to increase at significantly earlier cycles in the enriched samples, thus demonstrating selective amplification of the targeted exons. The average cycle number leads in the enriched samples for GJB2 (exon 2), GJB6 (exon 3), GJB3 (exon 2), SLC26A4 (exon 20), and SLC26A4 (exon 3) were 12.3±1.4, 12.7±1.8, 11.9±1.9, 11.8±1.8, and 11.7±0.8 (n=20 in all cases), corresponding to theoretical average enrichments of 5960, 6073, 3825, 3801 and 3429-fold, respectively. Combined length of the five targeted genes constitutes only 0.005% of the human genome. After enrichment, we found that readings belonging to or within 500 bp of the targeted regions ranged from 8.2% to 19.3% (n=7) in the NGS final results, supporting that targeted regions were significantly enriched. Water controls yielded no appreciable qPCR signals after more than 40 PCR cycles (Fig. 1C, D). Exons not targeted did not show any significant enrichment in qPCR assessments (data not shown). The significant enrichment suggested by qPCR data is consistent with the deep coverage in Illumina sequencing achieved for the deafness genes we targeted (Figs. 2 and 3). Coverage analysis showed that all nucleotides in the coding regions of the targeted deafness genes were captured, along with extra nucleotides bordering the exons (indicated by black horizontal bars in Figs. 2 and 3B). As expected, noncoding mRNA regions were also enriched by cDNA probes (Fig. 2). Average depth of coverage was 618, 587, 562, and 297-fold in the coding regions of GJB2, GJB3, GJB6, and exon 20 of SLC26A4, respectively (n=10). The minimum coverage in the coding regions was at least 153-fold for the deafness genes targeted. In the coding regions, ninety percent of the bps was covered by more than 213-fold.

FIG. 2.

Sequencing coverage in the targeted coding regions of GJB2 (A), GJB3 (B), GJB6 (C), and exon 20 of SLC26A4 (D). Black horizontal bars underneath each data trace represents the coding region shown by arrows. General exon structures of the genes are illustrated by diagrams above the data traces. Gray vertical bars represent the standard error of means obtained from different patient gDNA samples. Color images available online at www.liebertonline.com/gtmb

FIG. 3.

Results showing alignment of captured gDNA fragments in the genomic region of MYO15A. (A) Four examples of coverage in the genomic region of MYO15A. Relative depth of coverage is coded by a pseudo-colored scale bar given on top. Arrow points to the exon 8 of MYO15A, which has only 6 bps. (B) Sequencing coverage for the exon 8 of MYO15A aligned to the reference human genome. Horizontal bar underneath the data trace represents the coding region. Gray vertical bars represent the standard error of means obtained from different patient gDNA samples. (C) One example of a distribution histogram for the coverage depth in all the coding regions of MYO15A. The Y-axis is the cumulative coverage number, and the X-axis is the coverage fold. Color images available online at www.liebertonline.com/gtmb

Capture results for the largest deafness gene (MYO15A) by our approach are illustrated in Figure 3. The horizontal bar charts (Fig. 3A) represent the normalized coverage fold along the MYO15A genomic region. The capture patterns in the bar charts completely matched the MYO15A exon structure (shown by a diagram in the Fig. 3A), indicating all 65 exons of the MYO15A gene were captured. White gaps in the charts are intron regions, indicating that they were not captured at significant levels. Exon 8 of MYO15A is the smallest one among the 65 MYO15A exons, which has only 6 bps (shown by an arrow in Fig. 3A). With our extended capture probes, the smallest exon was captured with an average coverage fold of ∼151-fold (Fig. 3B). One typical example for the distribution of coverage depth in the coding regions of MYO15A is given in Figure 3C. On average each nucleotide was covered by 486±49-fold (n=6) with the coverage depths falling between ∼130 and 800 (Fig. 3C). The deep coverage and relatively small variations indicated a high efficiency of our enrichment protocol, which helped ensure exceptionally high accuracy in sequencing and genotype calls (Table 1).

Table 1.

Summary of Mutations Found in Subjects and Their Hearing Phenotypes

Sample #	Known GJB2 mutations before Illumina runs	Confirmed by Illumina runs?	Hearing loss in subject	Additional SNP & novel mutation(s) detected
1	None	N/A	Normal hearing	MYO15A: 1977C>RC (dbSNP: 854777)
2	None	N/A	Normal hearing	GJB2: 27V>IV (dbSNP: 2274084) and 114E>EG (dbSNP: 2274083); MYO15A: 595A>T (dbSNP: 2955365), 718W>G (dbSNP: 2955367), 1977C>R (dbSNP: 854777), 2018G>R (dbSNP: 2272571)
3	homozygous 235delC	Yes	Moderate	MYO15A: 595A>T (dbSNP: 2955365), 1977C>RC (dbSNP: 854777) & 2018G>RG (dbSNP: 2272571)
4	homozygous 109 G>A	Yes	Severe	MYO15A: 1977C>RC (dbSNP: 854777) & 2018G>RG (dbSNP: 2272571)
5	homozygous 235 delC	Yes	Moderate	MYO15A: 110R>HR (novel), 595A>TA (dbSNP: 2955365), 1977C>R (dbSNP: 854777), 2018G>R (dbSNP: 2272571)
6	79G>A, 235delC, 341A>G, all heterozygous	Yes	Mild	MYO15A: 595A>T (dbSNP: 2955365), 1138Q>QR (novel), 1977C>RC (dbSNP: 854777) & 2018G>RG (dbSNP: 2272571)
7	79G>A, 235delC, 341A>G, all heterozygous	Yes	Mild	MYO15A: 595A>T (dbSNP: 2955365)
8	homozygous 235delC	Yes	Moderate	MYO15A: 110R>HR (novel), 595A>TA (dbSNP: 2955365), 1977C>R (dbSNP: 854777), 2018G>R (dbSNP: 2272571)
9	homozygous 109 G>A	Yes	Mild	MYO15A: 595A>T (dbSNP: 2955365), 1138Q>QR (novel), 1977C>R (dbSNP: 854777), 2018G>R (dbSNP: 2272571)
10	heterozygous 235delC	Yes	Profound	GJB2: 203 I>T1 (novel); MYO15A: 1977C>R (dbSNP: 854777)
11	heterozygous 109 G>A	Yes	Moderate	GJB2: heterozygous 235delC (dbSNP: 80338943); MYO15A: 1999C>R (dbSNP: 854777)
12	heterozygous 235delC	yes	Moderate	GJB2:_3694;c.&3695delAT (double heterozygous deletion); MYO15A: 1138Q>QR (novel), 2018G>RG (dbSNP: 2272571)

Mutations/SNPs detected previously in patients and with the approach described here are both presented in the table.

N/A, not applicable.

After aligning with the reference human genome (HG18), we found that all mutations found previously in patients were identified by Illumina sequencing (Table 1), supporting that reliable genotype results were obtained without any false negatives for various types of mutations tested. The types of mutations include (Table 1) (1) homozygous single nucleotide deletion (samples# 3, 5, 8); (2) homozygous and heterozygous single nucleotide substitution (samples# 4, 9, 11); and (3) compound double and triple heterozygous single nucleotide substitutions and deletion (sample# 2, 6, 7, & Supplemental Fig. 1). In addition, NGS sequencing found new mutations that were overlooked by the previous Sanger sequencing in patients. The accuracy of the new genotyping was verified by Sanger sequencing. Examples are given in Supplementary Figure 1.

The new mutations suggested by Illumina data in deafness genes provided better explanations for the clinical phenotypes. Known recessive GJB2 deafness mutations were found in heterozygote form in patients by previous Sanger testing. In this study, we have found that many of them are actually double heterozygous for recessive deafness mutations known to cause profound deafness in humans. In sample# 10 (Table 1), only heterozygous 235delC was found by a previous test. However, the hearing loss is profound in this subject, which is contradictory with the known recessive nature of 235delC mutation (Liu et al. 2002). Re-tested by our new method for the same subject, we found a new GJB2 203 I>TI mutation in this patient. Considering the MYO15A 1977C>R SNP (dbSNP:854777) found in this patient is also present in normal hearing subject (e.g., subject# 1) and both homozygous 235delC and 203I>T mutations are deafness causing alleles, this result suggested that compound double heterozygous GJB2 235delC and 203I>TI is pathogenic. The heterozygous 109 G>A was the only GJB2 mutation previously found in sample# 11 (Table 1), and it is a known recessive mutation linked to deafness (Cryns et al., 2004). The compound heterozygous 235delC mutation identified by the current test provided new data for explaining the moderate hearing loss shown by this patient. For sample# 12, the patient's hearing loss could not be explained by heterozygous 235delC, as it is a recessive mutation that is also found in normal hearing controls. The newly identified GJB2_3694;c.&3695 delAT double heterozygous deletion mutation provided a possible explanation for the deafness phenotype observed for this patient. These results indicated that our novel targeted gene capture method combined with the NGS sequencing approach provided a more complete picture for further studying genotype and phenotype links in patients who suffer from hearing loss and deafness.

Discussion

In contrast to complicated genetic traits affecting the incidence of cancer, heart diseases, and neurological disorders, deafness caused by genetic mutations found thus far is mostly monogenic with well-defined genotype and phenotype relations (Oguchi et al., 2005). Since most congenital hearing loss cases are caused by genetic mutations and a definitive diagnosis can often be made based on genetic tests alone, it is desirable that comprehensive genetic tests be performed first for newborns who fail the UNHS (Smith et al., 2005). Detection of mutations in individual deafness genes (e.g., GJB2, GJB6 and SLC26A4) is now available from many medical centers (www.genetests.org). As a large-scale diagnostic screening or genetic epidemiological research tool, however, the value of these tests is limited, because the vast majority of deafness genes are not examined (Hilgert et al., 2009). Array-based semi high-throughput resequencing methods (Kothiyal et al., 2010; Rodriguez-Paris et al., 2010) only test known mutations and have a much smaller capacity compared with the approach presented here. The focus of the current study is to achieve both high NGS accuracy and low per-patient cost in targeted resequencing of deafness genes. Results just presented demonstrated that we have reliably captured all exons and achieved high coverage (>100-fold in coding regions for all targeted bps) and high accuracy in mutation identification for the targeted deafness genes. Our novel enrichment approach for targeted genes also helped us in achieving the goal of significantly reducing per-patient cost in applying NGS for detection of mutations. Material cost for long-oligo-based custom commercial kits is currently ∼$600/sample (Okou et al., 2007; Porreca et al., 2007; Shearer et al., 2010), and only a few companies have the specialized in situ oligo synthesizing equipment for producing millions of 120 bp long oligos. Appropriate multiplexing barcoding strategy has reduced the sequencing cost to about $200/patient when 12 samples are run together in the same lane of Illumina GAII. The targeted gene enrichment step, therefore, has become a major cost component if the currently published protocol is followed (Okou et al., 2007; Porreca et al., 2007; Shearer et al., 2010). Our in-house manufactured capture probes are made inexpensively from the cDNA libraries from PCR amplifications without relying on specialized and expensive equipment. All required materials can be readily made in house by any laboratory equipped for performing standard molecular biology experiments. Material costs for making the gene enrichment probes for the five genes tested here were under $25/sample. The technology validated here reduced overall per-sample cost significantly and enabled commercial offering of mutation screening in a targeted panel of deafness genes (www.otogenetics.com).

As a genetic screening method, the approach described here will undoubtedly provide a more complete epidemiological survey of currently known deafness genes (Hilgert et al., 2009). One of the critical issues in data analysis is how to filter out large numbers of deafness-unrelated SNPs. During data analyses, we have found that many well-known deafness-causing mutations are present in the dbSNP database. One of the best examples is rs80338943, which is the most common GJB2 deafness mutation (235delC) among the East Asians (Yan et al., 2003). There is definitely an urgent need for better SNP filters and annotation for helping find deleterious deafness mutations. To solve this problem, we have accumulated DNA samples from 500 normal hearing adult subjects. SNP analysis for this relatively large group of subjects is currently ongoing. The low-cost approach described here has helped greatly in accelerating the pace for reaching our goal of analyzing hundreds of normal-hearing samples for building a better deafness gene SNP filter (manuscript in preparation).

We believe the approach described here has brought us closer to the goal of conducting epidemiological-scale genetic screening of deafness genes, which may directly impact defining better diagnostic criteria for hearing loss due to genetic causes in the future. Consequently, better diagnosis may help validate and implement individualized treatment strategies, therefore contributing to refinement of better intervention strategies. Conducting comprehensive multi-gene analyses to find the specific mutations in patients will form the basis for personalized care of patients and for developing future therapies.

Acknowledgments

This study was supported by grants to Lin from National Institute on Deafness and other Communication Disorders (NIDCD 4R33DC010476, 1R41DC009713, and RO1 DC006483). Lin and Li also received grant support from the National Science Foundation of China (30728029). Tang received grant support from NIDCD (R21 DC008672). This research project was supported in part by the GRA Genome Center of the Emory University School of Medicine.

Disclosure Statement

Drs. Wenxue Tang and Xi Lin own equity in the Otogenetics Corporation, which participated in validating part of the data and later expanded the capture probes to cover ∼90 deafness genes in its commercial offering. The terms of the conflict of interest (COI) arrangement for the two authors have been reviewed and approved by Emory University in accordance with its COI policies. Drs. Shouting Huang and Yuhua Li are two employees of the Otogenetics Corporation, who also own equity in the company.