Targeted nanopore sequencing using the Flongle device to identify mitochondrial DNA variants

Abstract

Variants in mitochondrial genomes (mtDNA) can cause various neurological and mitochondrial diseases such as mitochondrial myopathy, encephalopathy, lactic acidosis, stroke-like episodes (MELAS). Given the 16 kb length of mtDNA, continuous sequencing is feasible using long-read sequencing (LRS). Herein, we aimed to show a simple and accessible method for comprehensive mtDNA sequencing with potential diagnostic applications for mitochondrial diseases using the compact and affordable LRS flow cell “Flongle.” Whole mtDNA amplification (WMA) was performed using genomic DNA samples derived from four patients with mitochondrial diseases, followed by LRS using Flongle. We compared these results to those obtained using Cas9 enrichment. Additionally, the accuracy of heteroplasmy rates was assessed by incorporating mtDNA variants at equimolar levels. Finally, mtDNA from 19 patients with Parkinson’s disease (PD) was sequenced using Flongle to identify disease risk-associated variants. mtDNA variants were detected in all four patients with mitochondrial diseases, with results comparable to those obtained from Cas9 enrichment. Heteroplasmy levels were accurately detected (r² > 0.99) via WMA using Flongle. A reported variant was identified in three patients with PD. In conclusion, Flongle can simplify the traditionally cumbersome and expensive mtDNA sequencing process, offering a streamlined and accessible approach to diagnosing mitochondrial diseases.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-024-75749-8.

Introduction

The mitochondrial genome (mtDNA) is a 16.6 kb long circular DNA that is composed of 13 genes, 22 tRNAs, and 2 rRNAs. Variations in mtDNA can lead to mitochondrial diseases such as mitochondrial myopathy, encephalopathy, lactic acidosis, and stroke-like episodes (MELAS) or myoclonus epilepsy associated with ragged-red fibers (MERRF). From a recent study conducted in Japan, the prevalence of mitochondrial disease is approximately 2.9 per 100,000 in the general population, which may vary geographically¹. For making a genetic diagnosis, mtDNA sequencing is a plausible approach to identify pathogenic variants. These mtDNA variants occur either as homoplasmy (a single variant is present) or heteroplasmy (in the context of mitochondrial disease, it refers to the coexistence of both reference and alternate alleles at a single variant site at different levels )². Thus, detecting specific variants and assessing their heteroplasmy are crucial³. However, the interpretation of mitochondrial variants can be complicated by the presence of nuclear mitochondrial DNA segments (NUMTs) which may affect detecting heteroplasmy levels⁴ when using short-read sequencing methods. Thus, long-read sequencing (LRS), which can cover the entire mtDNA, is essential for accurately detecting disease-causing variants in mitochondrial diseases^5–9. Despite its advantages, the primary limitation of LRS is its higher cost compared to conventional methods such as Sanger sequencing. Thus, the development of an affordable tool is required to detect mtDNA variants. In this study, we have aimed to develop an affordable approach, using the nanopore LRS technologies¹⁰ and its smallest and compact Flongle flow cell (Oxford Nanopore Technologies, Oxford, UK) for diagnosing mitochondrial diseases. These study results will help pave the way for its application in clinical settings.

Results

Comparison of amplified whole mtDNA LRS sequencing and Cas9-enrichment LRS of mitochondrial disease-causing DNA variants

mtDNA variants associated with mitochondrial diseases were detected in four samples (Table 1). These samples were obtained from the following patients: one patient with Kearns–Sayre syndrome caused by a chrM:11778G > A mutation, two patients with MERRF caused by a chrM:8344 A > G mutation, and one patient with Leber hereditary optic neuropathy (LHON) caused by a chrM:11778G > A mutation. Furthermore, samples from two patients with different mitochondrial disease were compared using long-read sequencing in both Cas9-enrichment mtDNA sequencing and whole mtDNA amplification (WMA). The heteroplasmy levels of the mtDNA in these samples demonstrated similar trends (see Supplementary Fig. S1, S2 online), indicating that the PCR bias was minimal. Consequently, we decided to use PCR amplicon for the subsequent analyses. In sample #2 (Table 1), the heteroplasmy levels of chrM:8344 A > G were reported as between 42% and 44%⁹. In our study, Cas9 enrichment demonstrated a heteroplasmy level of 40%, which is comparable to the findings in previous studies. Although WMA demonstrated a heteroplasmy level of 30% in our study, potentially due to PCR bias, this difference may potentially impact the precise detection of heteroplasmy level of MELAS in some settings.

Table 1.

Comparison of Cas9-enrichment sequencing and whole mtDNA PCR results in four patients with a mitochondrial disease. PCR, polymerase chain reaction; MERRF, myoclonus epilepsy associated with ragged-red fibers; LHON, Leber hereditary optic neuropathy. Coverage: read coverages at the position with the mutation.

Sample	Source	ID	Mutation	Disease	Cas9-enrichment					PCR
					Coverage	Mutation frequency (%)	Total read number	Data size	Average read length (bp)	Coverage	Mutation frequency (%)	Total reads	Data size	Average read length (bp)
#1	Coriell LCL	GM04368	chrM:11778G > A	KEARNS-SAYRE SYNDROME	728	98.00	9826	77,814,698	7919	690	96.00	8475	15,772,302	1861
#2	Coriell LCL	GM11906	chrM:8344 A > G	MERRF	2067	40.00	18,107	138,593,262	7654	199	30.00	2751	5,679,058	2064
#3	Coriell LCL	GM11907	chrM:8344 A > G	MERRF	220	90.00	6949	27,784,283	3998	151	90.00	2574	4,876,664	1895
#4	Blood	LHON	chrM:11778G > A	LHON	1042	99.00	33,773	283,399,223	8391	201	99.00	2722	4,954,793	1820

Analysis of heteroplasmy of mtDNA variants

To assess the accuracy of heteroplasmy quantification, the mtDNA samples were prepared in varying dilutions (0–100%). The amplified whole mtDNA samples were sequenced using Flongle in a multiplex manner, with 5–6 samples per flowcell. A strong correlation (r² > 0.99) was observed between the dilution levels and the expected heteroplasmy levels (Fig. 1), demonstrating that quantification with this method was highly precise. We also tested different sequencing coverage for one sample: low (average x47), medium (average x94) and high (average x250). All showed strong correlation (r² > 0.99), showing the robustness of this method to detect heteroplasmy (see Supplementary Fig. S3 online).

Fig. 1.

Comparison of observed and artificial heteroplasmy levels. This figure shows the correlation between observed and artificial heteroplasmy levels under various experimental conditions, including mtDNA variants (A8344G or G11778A), nanopore flowcell versions (R9.4 or R10.4), and references (normal or ambiguity code). Pearson’s correlation coefficient analysis revealed good correlations across all conditions. x-axis: estimated variant allele frequency; y-axis: projected variant allele frequency. mtDNA, mitochondrial genome.

Multiplex sequencing using Flongle in patients with Parkinson’s disease (PD)

Because of the association between PD and mitochondrial dysfunction, we searched for mtDNA variants as well as reported haplotypes associated with disease onset¹² (see Supplementary Table S1 online) in 19 patients with PD, including those with early-onset disease. We used four Flongle flowcell devices for WMA. Because we find a strong correlation in serial dilution experiment for heteroplasmy levels of ≥10%, we searched for mitochondrial variants with a heteroplasmy level of > 20%, which was arbitrarily set to minimize the impact of potential sequencing error. We detected an average of 52.8 mtDNA variants per patient (range, 39–69 variants). Using the MITOMAP database (https://www.mitomap.org), we determined that the previously reported PD-related variant chrM5460G > A¹¹ in three individuals (Table 2).

Table 2.

mtDNA variant screening in patients with Parkinson’s disease.

	Gender	Number of reads	Size(bases)	Average read length (bases)	Average depth	mtDNA: chrM5460G > A
						REF	ALT	reads	AF	Num of SNP variants (> 0.2)
#1	M	12,354	20,138,339	1630.1	859	15	141	156	0.90	44
#2	F	11,100	15,519,994	1398.2	623	38	579	617	0.94	50
#3	M	11,055	16,441,842	1487.3	729					49
#4	F	9882	17,241,833	1744.8	775					48
#5	M	4102	6,945,548	1693.2	296					52
#6	M	3394	6,762,193	1992	315					55
#7	F	3370	5,854,696	1737.3	273					42
#8	F	3323	6,407,077	1928.1	296					44
#9	M	3884	7,683,128	1978.1	374					39
#10	M	2465	4,881,694	1980.4	229					62
#11	M	2346	4,617,401	1968.2	218					68
#12	M	2477	4,798,894	1937.4	217					63
#13	F	2343	4,291,987	1831.8	180					69
#14	M	1796	3,464,532	1929	148	49	838	887	0.94	69
#15	M	5899	10,232,792	1734.7	473					50
#16	M	4437	7,242,590	1632.3	312					42
#17	M	3825	7,144,859	1867.9	321					47
#18	M	6440	11,401,270	1770.4	524					55
#19	M	5821	11,396,154	1957.8	509					55

ALT alternative; REF reference, AF allele frequency, SNP single nucleotide polymorphism, M male, F female. These polymorphisms are not confirmed by Sanger sequencing.

Discussion

In this study, we performed an efficient analysis of whole mtDNA using LRS, specifically with the Flongle device. This approach enables convenient whole mtDNA sequencing, which facilitates the identification of numerous single nucleotide variants (SNVs) that are responsible for various mitochondrial or neurodegenerative diseases.

Several methods have been proposed for identifying mtDNA variants^7–10. One such method involves the use of Cas9-targeted sequencing⁹. Although this method avoids PCR bias, it requires a significant amount of DNA, making it unsuitable for screening numerous samples. Following similar approaches reported in other studies^7,8,10, we conducted WMA followed by sequencing. We compared these results with those of Cas9-targeted sequencing. We identified comparable heteroplasmy levels between the two methods and an insignificant PCR bias. Furthermore, analysis of the simulated heteroplasmy samples demonstrated that heteroplasmy detection using WMA followed by sequencing is accurate within 10%. In addition, our approach is more straightforward than previously reported PCR followed by restriction fragment length polymorphism (RFLP) analysis¹² in detecting several variants at the same time, thereby enabling easier determination of haplogroups. Our method stands out for its simplicity and accessibility, largely due to the use of the Flongle device, which is a minimal and affordable tool. This may make our method advantageous for broader and practical application compared to more complex or intensive approaches, providing a first-line screening tool for clinical utility.

The allele frequency (AF) of chrM:5460G > A, which was reported in PD, is 0.08 in East Asians (https://gnomad.broadinstitute.org). We identified this allele in three out of 19 study participants with PD (AF 0.15), which was not statistically significant. This may be attributable to the low statistical power due to the small sample size. Additionally, we did not identify variants known to be risk factors for PD (5178 A > G or 10398G > A) in this cohort. We also have tested haplogroup which was reported to have protective effect of PD¹². Three patients (3/19, AF 0.158) have 4336T-5460G-9055G- 10,398 A-13708G haplotype which is about a half of reported frequency (AF 0.375), although not suitable for statistical testing due to small sample size.

This study had some limitations. We did not investigate deletions, insertions, or structural polymorphisms. LRS is known to demonstrate a low accuracy for detecting deletions and insertions. The known variants in mitochondrial diseases are predominantly SNVs (> 90%), making LRS suitable for initial screening. For example, 80–90% of the patients with MELAS demonstrate the 3243G > A variant, and the remaining known variants are also SNVs^13,14. Similarly, > 90% of the patients with MERRF^13,14 demonstrate SNV variants such as 8344 A > G, 8356T > C, 8363G > A, or 8361G > A^13,14. Therefore, our LRS-based method can be effectively used to screen for these SNVs.

The use of LRS via Flongle has demonstrated significant potential for identifying mtDNA variants among mtDNA amplicons in a simple method. This approach may effectively lower the barriers to comprehensive mtDNA variant screening and has promising clinical applications.

Materials and methods

Study protocol approvals and patient consent

The study was approved by the Institutional Review Board of St. Marianna University School of Medicine (No: 4983; January 16, 2024). The study was conducted in accordance with the principles of the Helsinki Declaration. Written informed consent and disclosure were obtained from the patients or their parents in those aged < 18 years.

Patient cells and cell lines

Peripheral blood cells or lymphoblastoid cell lines from four patients with mitochondrial diseases (one patient with Kearns–Sayer syndrome, two with MERRF, and one with LHON) and peripheral blood cells from 19 patients with PD were used in this study. Four lymphoblastoid cell lines that were derived from the patients with mitochondrial disease (MERRF and Kearns–Sayer syndrome) were purchased from Coriell Institute (https://www.coriell.org, Coriell Institute of Medical Research, New Jersey, USA.). The three lymphoblastoid cell lines from one patient with LHON and 19 patients with PD, who were registered in St. Marianna’s Neurological Disease Registry (Registry study on development of new diagnostic, novel treatment and prevention measures of neurological diseases; UMIN000042192), were utilized for sequencing.

DNA preparation

DNA samples were extracted from the peripheral blood of patients or lymphoblastoid cell lines using the QIAamp® DNA Blood kit (Qiagen, Venlo, Netherlands). The DNA samples were purified and concentrated (AMPure XP; Beckman Coulter, Brea, California, USA) and subsequently eluted using nuclease-free water (Thermo Fisher Scientific, Waltham, Massachusetts, USA).

WMA and nanopore sequencing

Whole mtDNA was amplified (Fig. 2A, see Supplementary Table S2 online) using four distinct primer sets. The first primer set was as follows: 5’- GCAAGAAATGGGCTACATTTTCTACCCCAG − 3’ and 5’- ATATTGATTTCACGGAGGATGGTGGTCAAG − 3’. The second primer set included 5’-CATAACACAGCAAGACGAGAAGACCCTATGG − 3’ and 5’- GGAGCCATTCATACAGGTCCCTATTTAAGG − 3’. The third primer set was 5’- TCTGTATGTCTCCATCTATTGATGAGGGTC − 3’ and 5’- GTGGTTTGCTCCACAGATTTCAGAGCATTG − 3’. The fourth primer set comprised 5’- CTAACCTGAATCGGAGGACAACCAGTAAGC − 3’ and 5’- AGCCATAATTTACGTCTCGAGTGATGTGGG − 3’. PCR amplification (PrimeSTAR GXL; Takara, Shiga, Japan) was conducted using a two-step protocol. An initial denaturation was performed at 98 °C for 2 min, followed by 30 cycles of denaturation at 98 °C for 10 s. Thereafter, annealing and extension was performed at 68 °C for 10 min, and a final extension was performed at 72 °C for 5 min. The four PCR products were purified (AMPure beads; Beckman Coulter, Brea, California, USA), mixed equimolarly, and subjected to nanopore sequencing using a rapid sequencing kit (RAD004, Oxford Nanopore Technologies, Oxford, UK).

Fig. 2.

Schema of the experiment. (A) Whole mtDNA amplification using four primer sets. (B) Cas9 enrichment mtDNA sequencing. (C) crRNA target region within the mtDNA. The crRNA was designed in an area devoid of known pathogenic variants. mtDNA, mitochondrial genome; crDNA, guide RNA.

Cas9-mediated nanopore sequencing of mtDNA

The custom Alt-R CRISPR-Cas9 guide RNA (IDT Corporation, New Jersey, USA.) was used to design guide RNAs (crRNA) that would target the entire mtDNA (Fig. 2B,C, see Supplementary Table S3 online). We selected a target region within an area with rare polymorphism using the MITOMAP database. The crRNA oligo sequences (5’-GCCGTTAAACATGTGTCACT − 3’ and 5’-TGACACATGTTTAACGGCCG − 3’) were designed to cleave a single location (Fig. 2C). The nanopore library was prepared using a Cas9 sequencing kit (SQK-CS9109; Oxford Nanopore Technologies, Oxford Science Park, Oxford, UK) according to the manufacturer’s instructions and a previously described method¹⁵. A secondary structure was prepared by combining 1 µl of the mixed crRNA (100 µM, IDT Corporation, New Jersey, USA.), targeting both the plus and minus strands, with 1 µl of the tracrRNA (100 µM, IDT Corporation, New Jersey, USA.). The secondary structure was mixed with Cas9 to form a ribonucleoprotein complex. The mtDNA was treated with phosphatase and subsequently mixed with the ribonucleoprotein complex. The CRISPR/Cas9-cleaved DNA fragments were ligated with adaptors and subjected to nanopore sequencing using MinION flowcell (FLO-MIN106, Oxford Nanopore Technologies, Oxford, UK). One sequencing device was used per sample.

Simulated heteroplasmy of mtDNA variants using serial concentration samples

DNA extracted from the lymphoblastoid cell lines of patients with two different mitochondrial diseases carrying the 3243 A > G and 11778G > A mutations was purchased from Coriell Institute, New Jersey, USA. The DNA was mixed in equimolar ratios with the control DNA (NA24385, Coriell Institute) that lacked these variants in proportions of 0%, 10%, 25%, 50%, 75%, and 100%. Thereafter, the whole mtDNA was amplified as described above, and the resultant mixtures were analyzed using the Flongle flowcell (Oxford Nanopore Technologies, Oxford, UK.). We also tested the mixtures using the R9.4 and R10.4 flowcells. The sequencing libraries from the six simulated DNA samples, which represented different heteroplasmy levels, were prepared using a barcoding kit (SQK-RBK004 for R9.4 or SQK-RBK114.24 for R10.4; Oxford Nanopore Technologies, Oxford, UK). To test the robustness of this method to the low coverage data, we randomly extracted reads to construct datasets with various read coverages. Statistical analysis was conducted using R (version 4.3.1). The relationship between the simulated concentration and detected heteroplasmy levels was assessed using the coefficient of determination (r²).

Data analysis

The reference genome used in this study was hg38 (UCSC), as described in LAST (https://gitlab.com/mcfrith/last/-/blob/main/doc/last-cookbook.rst) or elsewhere¹⁶. The sequencing error rates were calculated using last-train¹⁷ and reads were aligned to the reference genome using lastal (ver. 1389)^18,19. To detect a single nucleotide variant in mtDNA and predict the variant AF, we used last-genotype (https://github.com/mcfrith/last-genotype, described below) and set the fractional genotype option -p0. Given that the majority of known pathogenic mtDNA variants are substitutions and that nanopore sequencing is associated with a high error rate for deletions and insertions (represented as “-“ in the alignment files on either the reference or read side), deletions and insertions were excluded from the analysis. Reference bias can occur when reads containing non-reference alleles misaligned or not aligned due to the preference to the reference genome sequence. This is especially true for nanopore reads, which tend to have sequencing errors in certain genomic contexts (e.g., homopolymers). To avoid the possible reference bias, we used a reference sequence that contained the ambiguity code R (A or G) at positions chrM:11778G and chrM:8344A.

Last-genotype

last-genotype finds the most likely genotype of each site in a genome, from alignments of DNA reads to a reference genome. A site’s genotype is a set of 4 base probabilities: (P_a, P_c, P_g, P_t). The user can fix a ploidy, to get e.g. diploid genotypes like (0.5, 0, 0.5, 0), or request “fractional genotypes”, like (0, 0.2, 0.8, 0). The likelihood of a genotype is:

∏ _{(i = 1 to n)} [ E_i + (1 − E_i) ∑ _{(x = a, c,g, t)} [ P_x prob(Y_i | x) ] ]

Here, n is the number of read bases aligned to the site, and Yi is the i-th such base (Yi = a, c, g, or t). Ei is the probability that this base is aligned to the wrong site, which is optionally output by LAST. Finally, prob(y | x) is the probability of observing base y in a read when it is x in the genome: this comes from last-train. last-genotype ignores insertions, deletions, and fastq quality data.

Multiplex sequencing in patients with PD

The DNA samples of 19 patients with PD were extracted from peripheral blood cells using DNeasy Tissue & Blood kit (Qiagen, Venlo, Netherlands). The DNA samples were subjected to WMA and nanopore sequencing as described above. We used 4–5 samples per Flongle flowcell and the SQK-RBK004 library kit (Oxford Nanopore Technologies Oxford, UK.).

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

SA, MCF, and SM contributed significantly to the data analysis and interpretation. MCF, MS, and YY played a role in the study conceptualization. SA, SM, KS, HM, and MS were involved in the data acquisition. All authors actively participated in reviewing, revising, and approving the final version of the manuscript for submission.

Data availability

The sequencing data obtained from patient samples are not publicly available to protect the privacy of study participants. However, the data obtained from cells provided by the Coriell Institute are available on DDBJ DRA database under accession number PRJDB18666.

Declarations

Competing interests

The authors declare no competing interests.