Portal de Periódicos da CAPES

Gene Mosaicism Screening Using Single-Molecule Molecular Inversion Probes in Routine Diagnostics for Systemic Autoinflammatory Diseases

2019; Elsevier BV; Volume: 21; Issue: 6 Linguagem: Inglês

10.1016/j.jmoldx.2019.06.009

1943-7811

Benjamin Kant, Ellen C. Carbo, Iris Kokmeijer, Jelske J.M. Oosterman, Joost Frenkel, Morris A. Swertz, Johannes Kristian Ploos van Amstel, Juan I. Aróstegui, Marco J. Koudijs, Mariëlle van Gijn,

Monoclonal and Polyclonal Antibodies Research

Diagnosis of systemic autoinflammatory diseases (SAIDs) is often difficult to achieve and can delay the start of proper treatments and result in irreversible organ damage. In several patients with dominantly inherited SAID, postzygotic mutations have been detected as the disease-causing gene defects. Mutations with allele frequencies <5% have been detected, even in patients with severe phenotypes. Next-generation sequencing techniques are currently used to detect mutations in SAID-associated genes. However, even if the genomic region is highly covered, this approach is usually not able to distinguish low-grade postzygotic variants from background noise. We, therefore, developed a sensitive deep sequencing assay for mosaicism detection in SAID-associated genes using single-molecule molecular inversion probes. Our results show the accurate detection of postzygotic variants with allele frequencies as low as 1%. The probability of calling mutations with allele frequencies ≥3% exceeds 99.9%. To date, we have detected three patients with mosaicism, two carrying likely pathogenic NLRP3 variants and one carrying a likely pathogenic TNFRSF1A variant with an allele frequency of 1.3%, confirming the relevance of the technology. The assay shown herein is a flexible, robust, fast, cost-effective, and highly reliable method for mosaicism detection; therefore, it is well suited for routine diagnostics. Diagnosis of systemic autoinflammatory diseases (SAIDs) is often difficult to achieve and can delay the start of proper treatments and result in irreversible organ damage. In several patients with dominantly inherited SAID, postzygotic mutations have been detected as the disease-causing gene defects. Mutations with allele frequencies 30 SAID-associated genes have been discovered,6Moghaddas F. Masters S.L. The classification, genetic diagnosis and modelling of monogenic autoinflammatory disorders.Clin Sci (Lond). 2018; 132: 1901-1924Crossref PubMed Scopus (16) Google Scholar the majority causing a monogenic SAID.7Azizi G. Khadem Azarian S. Nazeri S. Mosayebian A. Ghiasy S. Sadri G. Mohebi A. Khan Nazer N.H. Afraei S. Mirshafiey A. Monogenic auto-inflammatory syndromes: a review of the literature.Iran J Allergy Immunol. 2016; 15: 430-444PubMed Google Scholar, 8Oda H. Kastner D.L. Genomics, biology, and human illness: advances in the monogenic autoinflammatory diseases.Rheum Dis Clin North Am. 2017; 43: 327-345Abstract Full Text Full Text PDF PubMed Scopus (26) Google Scholar Unexpectedly, conventional DNA analyses fail to find a disease-causing mutation in >80% of SAID patients.9Federici L. Rittore-Domingo C. Koné-Paut I. Jorgensen C. Rodière M. Le Quellec A. Touitou I. A decision tree for genetic diagnosis of hereditary periodic fever in unselected patients.Ann Rheum Dis. 2006; 65: 1427-1432Crossref PubMed Scopus (65) Google Scholar, 10Booty M.G. Chae J.J. Masters S.L. Remmers E.L. Barham B. Lee J.M. Barron K.S. Holland S. Kastner D.L. Aksentijevich I. Familial Mediterranean fever with a single MEFV mutation: where is the second hit?.Arthritis Rheum. 2009; 60: 1851-1861Crossref PubMed Scopus (191) Google Scholar, 11Rusmini M. Federici S. Caroli F. Grossi A. Baldi M. Obici L. Insalaco A. Tommasini A. Caorsi R. Gallo E. Olivieri A.N. Marzano A. Coviello D. Ravazzolo R. Martini A. Gattorno M. Ceccherini I. Next-generation sequencing and its initial applications for molecular diagnosis of systemic auto-inflammatory diseases.Ann Rheum Dis. 2016; 75: 1550-1557Crossref PubMed Scopus (47) Google Scholar, 12Stray-Pedersen A. Sorte H.S. Samarakoon P. Gambin T. Chinn I.K. Coban Akdemir Z.H. et al.Primary immunodeficiency diseases: genomic approaches delineate heterogeneous Mendelian disorders.J Allergy Clin Immunol. 2017; 139: 232-245Abstract Full Text Full Text PDF PubMed Scopus (175) Google Scholar This further complicates and postpones the patients' definitive diagnosis, which may also delay receiving the proper treatment. Therefore, irreversible organ damage can occur. Major causes of the low diagnostic yield are the large phenotypic spectrum, genetic diversity, and intrinsic limits of used methods of genetic analyses.4Simon A. van der Meer J.W. Vesely R. Myrdal U. Yoshimura K. Duys P. Drenth J.P. International HIDS Study GroupApproach to genetic analysis in the diagnosis of hereditary autoinflammatory syndromes.Rheumatology. 2006; 45: 269-273Crossref PubMed Scopus (63) Google Scholar, 13Lachman H.J. Papa R. Gerhold K. Obici L. Touitou I. Cantarini L. Frenkel J. Anton J. Kone-Paut I. Cattalini M. Bader-Meunier B. Insalaco A. Hentgen V. Merino R. Modesto C. Toplak N. Berendes R. Ozen S. Cimaz R. Jansson A. Brogan P.A. Hawkins P.N. Ruperto N. Martini A. Woo P. Gattorno M. Paediatric Rheumatology International Trials Organisation (PRINTO)the EUROTRAPS and the Eurofever ProjectThe phenotype of TNF receptor-associated autoinflammatory syndrome (TRAPS) at presentation: a series of 158 cases from the Eurofever/EUROTRAPS international registry.Ann Rheum Dis. 2014; 73: 2160-2167Crossref PubMed Scopus (178) Google Scholar In 2005, the first postzygotic mutation causing SAID was detected in the NLRP3 gene.14Saito M. Fujisawa A. Nishikomori R. Kambe N. Nakata-Hizume M. Yoshimoto M. Ohmori K. Okafuji I. Yoshioka T. Kusunoki T. Miyachi Y. Heike T. Nakahata T. Somatic mosaicism of CIAS1 in a patient with chronic infantile neurologic, cutaneous, articular syndrome.Arthritis Rheum. 2005; 52: 3579-3585Crossref PubMed Scopus (99) Google Scholar Since then, additional patients with mosaicism have been found, mainly in the NLRP3 gene but also in other SAID-associated genes like NOD2, TNFRSF1A, and NLRC4.15De Inocencio J. Mensa-Vilaro A. Tejada-Palacios P. Enriquez-Merayo E. González- Roca E. Magri G. Ruiz-Ortiz E. Cerutti A. Yagüe J. Aróstegui J.I. Somatic NOD2 mosaicism in Blau syndrome.J Allergy Clin Immunol. 2015; 136: 484-487Abstract Full Text Full Text PDF PubMed Scopus (38) Google Scholar, 16Rowczenio D.M. Trojer H. Omoyinmi E. Aróstegui J.I. Arakelov G. Mensa-Vilaro A. Baginska A. Silva Pilorz C. Wang G. Lane T. Brogan P. Hawkins P.N. Lachmann H.J. Association of tumor necrosis factor receptor–associated periodic syndrome with gonosomal mosaicism of a novel 24-nucleotide TNFRSF1A deletion.Arthritis Rheumatol. 2016; 68: 2044-2049Crossref PubMed Scopus (38) Google Scholar, 17Kawasaki Y. Oda H. Ito J. Niwa A. Tanaka T. Nijikata A. Seki R. Nagahashi A. Osawa M. Asaka I. Watanabe A. Nishimata S. Shirai T. Kawashima H. Ohara O. Nakahata T. Nishikomori R. Heike T. Saito M.K. Identification of a high-frequency somatic NLRC4 mutation as a cause of autoinflammation by pluripotent cell–based phenotype dissection.Arthritis Rheumatol. 2017; 69: 447-459Crossref PubMed Scopus (77) Google Scholar, 18Labrousse M. Kevorkian-Verguet C. Boursier G. Rowczenio D. Maurier F. Lazaro E. Aggarwal M. Lemelle I. Mura T. Belot A. Touitou I. Sarrabay G. Mosaicism in autoinflammatory diseases: cryopyrin-associated periodic syndromes (CAPS) and beyond: a systematic review.Crit Rev Clin Lab Sci. 2018; 55: 432-442Crossref PubMed Scopus (27) Google Scholar Previous reports have shown that even mutations with an allele frequency 1% minor allele frequency in the gnomAD27Lek M. Karczewski K.J. Minikel E.V. Samocha K.E. Banks E. Exome Aggregation Consortium et al.Analysis of protein-coding genetic variation in 60,706 humans.Nature. 2016; 536: 285-291Crossref PubMed Scopus (6551) Google Scholar or GoNL28The Genome of the Netherlands ConsortiumWhole-genome sequence variation, population structure and demographic history of the Dutch population.Nat Genet. 2014; 46: 818-825Crossref PubMed Scopus (488) Google Scholar database, two smMIPs were designed to target both alleles. After manufacturing (Integrated DNA Technologies, Coralville, IA), all smMIPs were equimolarly pooled and tested. Rebalancing of the pool was considered per smMIP, dependent on its performance, location within the regions of interest, and overlap with other probes. The smMIP pool was phosphorylated by adding 1 μL of T4 Polynucleotide Kinase (New England Biolabs, Ipswich, MA) for every 25 μL of 100 μmol/L smMIPs in 1× T4 DNA ligase buffer with 10 mmol/L ATP (New England Biolabs). The mixture was incubated at 37°C for 45 minutes, followed by kinase inactivation at 60°C for 20 minutes. Capture was performed on 200 to 500 ng of fragmented DNA in a 25-μL reaction volume by adding 1 unit of Ampligase DNA ligase (Epicentre, Madison, WI), 1× Ampligase Buffer (Epicentre), 3.2 units of Hemo Klentaq (New England Biolabs), 8 pmol dNTPs, and an aliquot of smMIP pool to achieve a molecular ratio DNA/smMIP of 1:800 for each individual smMIP. The mixture was denatured at 95°C for 10 minutes and then incubated at 60°C for 21 hours to enable correct smMIP hybridization, extension, and ligation. After cooling, exonuclease treatment was performed in 1× Ampligase buffer by adding 10 units of Exonuclease I (New England Biolabs) and 50 units of Exonuclease III (New England Biolabs) to the capture product. The mixture was incubated at 37°C for 45 minutes, followed by inactivation at 95°C for 2 minutes. PCR was performed in a 50-μL volume containing 50 pmol common forward primer, 50 pmol barcoded reverse primer,24O'Roak B.J. Vives L. Fu W. Egertson J.D. Stanaway I.B. Phelps I.G. Carvill G. Kumar A. Lee C. Ankenman K. Munson J. Hiatt J.B. Turner E.H. Levy R. O'Day D.R. Krumm N. Coe B.P. Martin B.K. Borenstein E. Nickerson D.A. Mefford H.C. Doherty D. Akey J.M. Bernier R. Eichler E.E. Shendure J. Multiplex targeted sequencing identifies recurrently mutated genes in autism spectrum disorders.Science. 2012; 338: 1619-1622Crossref PubMed Scopus (913) Google Scholar 1× iProof high-fidelity master mix (Bio-Rad, Hercules, CA), and 20 μL of the exonuclease treated product. The mixture was denatured at 98°C for 30 seconds, followed by 19 cycles with 10 seconds of denaturing at 98°C, 30 seconds of hybridization at 60°C, and 30 seconds of elongation at 72°C, and a final elongation step at 72°C for 2 minutes. After cooling, the PCR products were measured with TapeStation 4200 (Agilent, Santa Clara, CA), and up to 96 samples were equimolarly pooled. The pool was purified by bead-based size selection with a 0.5% and a 0.8% volume of Agencourt Ampure XP (Beckman Coulter, Brea, CA). The pool was denatured and diluted to a concentration of 1.0 pmol/L and loaded on a NextSeq500 sequencer (Illumina, San Diego, CA), according to the manufacturer's protocol. The following sequencing primers were used: forward, 5′-CATACGAGATCCGTAATCGGGAAGCTGAAG-3′; reverse, 5′-ACACGCACGATCCGACGGTAGTGT-3′; and index, 5′-ACACTACCGTCGGATCGTGCGTGT-3′. The sequencer was run with a 300-cycles Mid Output sequencing kit (Illumina), resulting in 2 × 150 bp paired-end reads. Conversion of raw sequencing data to FASTQ files and simultaneous barcode demultiplexing were performed with bcl2fastq2 Conversion Software version 2.20 (Illumina). The FASTQ files were transferred to a server running the SeqNext module of the Sequence Pilot commercial analysis software version 4.3.1 (JSI Medical Systems, Ettenheim, Germany). The designfile was used to generate a mapping target, according to the SeqNext manual. Two minor adjustments were made to rule out recurrent false-positive results in analysis (Supplemental Table S1). Read mapping, deduplication of reads with the same UMI (consensus calling), and variant calling were performed semiautomatically in SeqNext with the following settings: required coverage, minimum absolute coverage, 200 per direction; mutations, minimum absolute coverage, 5 combined; minimum percentage coverage, 1% per direction; mutation sorting, distinct/other percentage coverage, 40% per direction; tags enabled, yes; tag length R1, 8 bp; ignore tags with N bases, yes; ignore tags with low Qs, yes; minimum absolute coverage, 1; minimum percentage coverage, 50%; and ignore consensus read threshold, 30%. For each UMI, a minimum of two reads is required for consensus calling. SeqNext performs consensus calling and further analysis separately for forward reads and reverse reads (per direction). This results in a forward and a reverse consensus read derived from each captured molecule. A variant call will be made when the requirements are met by both forward consensus reads and reverse consensus reads. With the settings used, all variants with label other are considered candidate postzygotic variants and need to be further examined. A cumulative binomial distribution function was used to determine the variant call probability, with p being the true variant allele frequency in the tested material, N the consensus read depth per direction at the particular nucleotide position, and x the number of variant reads needed for a variant call to be made. The standard function calculates the probability of x or less variant reads:P(X≤x)=∑i=0xpi(Ni)(1−p)N−i(1) When performing the assay on two independent DNA samples, a variant call has to be made in at least one for a postzygotic mutation to be detected. In case a variant is detected in one sample and not in the other, further examination is required. With the settings used, a variant call is made when at least 1% of consensus reads per direction contain the variant and when at least five consensus variant reads (combined, similar to three reads per direction) are detected. Hence, the function was adjusted to calculate the probability of a variant being called in at least one sample, with N* equaling N rounded up to the next hundred:1−P(X≤x)2=1−(∑i=0xpi(Ni)(1−p)N−i)2at≥300X:x=N∗100−1at A p.Thr90Asn variant in the TNFRSF1A gene was not detected by Sanger sequencing or whole exome sequencing; therefore, droplet digital PCR (ddPCR) was performed. All available samples containing the postzygotic variant were tested together with positive and negative control samples. Multiple water control samples were tested for determining background noise. The following primers and probes were designed using Primer3 software version 2.4.0, manufactured by Integrated DNA Technologies, and mixed with primers to probe ratio 1.8: forward primer, 5′-CCCATTCACAGGAACCTACTTG-3′; reverse primer, 5′-ACTCACCCTTTCGGCATTTG-3′; reference allele probe, 5′-CAGGGAGTGTGAGAGCGGCTCCTTCACCGC-3′ (FAM-labeled and double quenched); and variant allele probe, 5′-CAGGGAGTGTGAGAGCGGCTCCTTCAACGC-3′ (HEX labeled and double quenched). ddPCR was performed by the QX200 Droplet Digital PCR system (Bio-Rad). A PCR mixture was prepared containing 20 ng of DNA, 1× ddPCR supermix for probes (no dUTP; Bio-Rad), 1× variant allele primers/probe, and 1× reference allele primers/probe in a total volume of 20 μL. The droplet generator was used to partition the PCR mixture in 8000 to 22,000 droplets. The droplets were incubated at 95°C for 10 minutes for polymerase activation, followed by 40 cycles with 30 seconds of denaturation at 95°C and 1 minute of annealing and extension at 62°C. The enzyme was then deactivated at 98°C for 10 minutes. After cooling down, fluorescent signals of individual droplets were read by the droplet reader. Analysis was performed using QuantaSoft software version 1.7.4.0917 (Bio-Rad). Supplemental Table S1 shows all used smMIPs and their rebalancing factor in the final pool. This pool contains 111 smMIPs for 109 amplicons. Of the nucleotides within the regions of interest, 100% are covered by at least one smMIP, with 66.4% (4402 of 6632 nucleotides) being targeted by two or more smMIPs. For performance analysis, 325 samples were evaluated from 158 SAID patients tested in six independent preparation and sequencing runs. For each patient, at least two independent samples were tested. After deduplication, the mean coverage of all nucleotides per sample was 4926× per direction (range, 1057× to 10,076×) (Figure 1A). The mean of lowest nucleotide coverages per sample was 696× per direction (range, 206× to 2158×) (Figure 1B). With all samples having ≥206× coverage on all nucleotide positions, all patients have a >99.9% probability for calling variants with ≥3% allele frequency on any nucleotide position in this panel. For variants with 2%, 1.5%, 1.2%, and 1% allele frequency, having 206× coverage, a variant call is made with 95%, 84%, 70%, and 56% probability, respectively. For the obtained mean nucleotide coverage (4926×), probabilities are 100%, 100%, 99%, and 73%, respectively. Four samples from positive control subjects, containing postzygotic variants with allele frequencies ranging from 2% to 20%, were tested. All variants were called with this assay. In addition, samples with different genotypes at common single-nucleotide polymorphism positions were selected, mixed in ratios varying from 1:1 to 1:255, and tested as single samples in three independent validation runs. These runs had lower coverage than those used for performance analysis and patient screening. All variants with ratios of 1:1 to 1:31 were called, including 32 variants with a ratio of 1:31 (±3.1% allele frequency). Results for 32 variants with a ratio of 1:63 (±1.6% allele frequency) are listed in Table 1.Table 1Results of Variants with ±1.6% Allele Frequency from Three Independent Validation RunsRunMean coverage at variant positionVariants, nCalls expected, nCalls made, n1217×85–672629×1210–111034891×121212Coverages are per direction. Call expectations are based on the used statistics and adjusted for testing single samples. Open table in a new tab Coverages are per direction. Call expectations are based on the used statistics and adjusted for testing single samples. A total of 158 patients with SAID for mosaicism and 3 postzygotic variants (1.9%) have been detected (Table 2). All variants were present in at least two independent samples with similar allele frequencies. Variant 1 was captured by only one smMIP, whereas variants 2 and 3 were captured by two overlapping smMIPs and were detected by both with similar allele frequencies. Also, variants 2 and 3 were detected with similar allele frequencies in analyses from multiple runs and in DNA samples isolated from whole blood extracted in different years.Table 2Postzygotic Variants Detected in This StudyPatient no. (year)GeneVariantProtein effectAllele frequency, %Mean coverage∗Coverages are per direction. The number inside parentheses represents the number of independent measurements used for calculating the mean coverage.1NLRP3c.918A>Tp.Gln306His10.46768× (2)2 (2005)NLRP3c.1706 G>Cp.Gly569Ala18.85228× (4)2 (2013)34.64499× (4)3 (2011)TNFRSF1Ac.269C>Ap.Thr90Asn1.114,797× (4)3 (2013)1.39828× (6)3 (2018)1.44744× (4)For variants 2 and 3, DNA samples isolated from whole blood extracted in different years were tested.∗ Coverages are per direction. The number inside parentheses represents the number of independent measurements used for calculating the mean coverage. Open table in a new tab For variants 2 and 3, DNA samples isolated from whole blood extracted in different years were tested. The three detected postzygotic variants (Table 2) were assessed and confirmed by alternative methods. Variant 1 was confirmed by whol