Mosaicism is a term used to describe the coexistence of genetically distinct normal and abnormal cells in a single individual originating from a single fertilized egg [1]. For example, germline mosaicism refers to the presence of genetic variants in different cells of germline cells of an organisms, which are the cells that give rise to eggs or sperm. Hence, some of the offspring of an individual may inherit a different genetic variant than others, and the genetic diversity among the offspring can be greater than that among the cells of the body of an individual. Somatic mosaicism, in contrast, refers to the presence of genetic variants in different cells of the body but not in the germline (Fig. 1) [2].

Figure 1. Distribution of mutant cells in the human body and different types of mosaicism in particular individuals. In somatic or germline mosaicism, mutant cells may appear with different mosaicism ratios in distinct tissues of patients, including gonads. (A) Example of somatic mosaicism confined to endoderm among the three germinal cell layers. (B) In focal cortical dysplasia, somatic mosaicism appears localized to a specific organ, such as the brain. (C) Germline or gonadal mosaicism refers to genetic variation in the genomes of germline cells within an individual. (D) Somatic and germline mosaicism may coexist in the same individual.

Replicating human DNA, consisting of 3 billion base pairs per haploid, is a complex process that is susceptible to errors, despite various proofreading mechanisms in cells [3]. The process of differentiation, which involves rapid DNA replication, is especially prone to induction of mutations. In the scenario where a mutation arises during the differentiation process following fertilization, the process results in somatic mosaicism. This phenomenon impacts specific tissues or cells within the organism, rather than its entirety. This localized mosaicism can result in genetic disorders with symptoms manifesting in specific tissues or organs.

With the advent of high-throughput next-generation sequencing (NGS) in 2005, genome analysis has significantly progressed, enabling the detection of mosaicism and enhancing our comprehension of mosaic disorders. However, even healthy donors can exhibit mosaic variation, with mutant allele fractions ranging from 1.0% to 29.7% within organ samples [4]. Therefore, there is an increasing demand for effective strategies to detect relevant mosaicism in clinical laboratories. This review aims to provide a comprehensive overview of the most effective methods for detecting clinically relevant mosaic sequence variants, focusing on target diseases, analytical techniques, and sample types.

Mosaic diseases are a group of rare disorders characterized by abnormalities, often resulting from mutations in critical signaling pathways [5]. These mutations would be fatal if present in all cells. However, when limited to a subset of cells, they can cause partial overgrowth, brain malformations, skin symptoms, and vascular malformations [6, 7]. While each clinical evidence is uncommon, collectively, they form a distinct group of disorders. Mendelian disorders exhibit mosaicism, and approximately 1.5% of the patients with various diseases are subjected to exome sequencing demonstrated mosaicism [8]. Somatic mosaicism has been extensively examined in neurodevelopmental diseases, including autism, developmental delay, intellectual disability, and epilepsy [9-11]. Exome sequencing studies suggest that approximately 3–5% of the autism [12-14] and 1% of the developmental disabilities are caused by somatic mosaicism [15]. Therefore, clinical laboratories should consider the possibility of mosaicism in these representative diseases (Table 1).

Table 1 Somatic mosaicism in representative monogenic disorders

Disease categories	Gene/Locus	Material tested	Methods applied	Mosaic ratio (%)	Reference
Achondroplasia	FGFR3	PB	PCR and DHPLC	25–30	[34]
Clotting disorders	F8, F9	PB	ASO analysis, COP PCR, AS-PCR	0.2–25	[35-37]
Facioscapulohumeral muscular dystrophy	D4Z4, SMCHD1	PB	PFGE	10–90	[38, 39]
Focal cortical dysplasia	MTOR, TSC1, TSC2	Saliva, brain tissues	ddPCR, amplicon sequencing	0.9–12.8	[10, 40, 41]
Overgrowth syndrome	PIK3CA, MTOR	PB, fresh tissues, FFPE, saliva	PCR and subcloning	0.16–37.6	[7, 42]
Proteus syndrome	AKT1	PB, tissue, fibroblast	Modified PCR–restriction-enzyme assay	1–50	[43, 44]
Sturge–Weber syndrome	GNAQ	Tissues	ddPCR	0.42–7.1	[45]
Tuberous sclerosis	TSC1, TSC2	Blood, saliva, tissues	Amplicon sequencing	0.21–34	[46]

Abbreviations: PB, peripheral blood; DHPLC, denaturing high performance liquid chromatography; ASO, allele-specific oligonucleotide; COP, competitive oligo priming; AS-PCR, allele-specific PCR; PFGE, pulsed-field gel electrophoresis; ddPCR, droplet digital PCR; FFPE, formalin-fixed paraffin-embedded.

The mosaic form of neurofibromatosis type 1 (NF1), a type of hereditary cancer, has been attributed to related changes in the NF1 gene. A systematic review analyzed 111 articles with 320 individuals with mosaic NF1 from 1977 to 2012. Most of these individuals exhibited only pigmentary changes or neurofibromas. Patients with pigmentary changes alone developed symptoms at a younger age. A localized form of mosaic NF1 was observed in 76% of the cases, whereas the remaining ones had a generalized form. Complications of NF1 were observed in 29% of the 157 case reports analyzed. In a large case series, 6.5% of the individuals with mosaic NF1 had offspring with complete NF1. Therefore, it is recommended that individuals with suspected mosaic NF1 undergo a complete physical examination, genetic testing of blood and skin, counseling, and health surveillance [16].

A recent review investigated the prevalence and types of mosaic sequence variants in patients with tuberous sclerosis complex (TSC) [17]. A total of 39 patients with TSC with mosaic variants were subjected to massively parallel sequencing, which involved analyzing 170 different tissue samples. The milder and distinct clinical phenotype observed in patients with the mosaic form of TSC (variant allele frequency, VAF: 0–10%, median 1.7% in blood DNA) was in contrast to the findings for other TSC studies, with similar occurrences of facial angiofibromas (92%), and kidney angiomyolipomas (83%), but fewer seizures, cortical tubers, and multiple other manifestations (P<0.0001 for six features). In addition, a high variability was noted in the VAF of TSC1/TSC2 pathogenic variants across different tissue samples. Notably, skin lesions were the most reliable tissue types for identifying variants, and 6 of the 39 (15%) patients showed no evidence of the variant in the blood. The study suggested that many individuals with low-level mosaicism for a TSC2 pathogenic variant may not be clinically recognized.

The identification of genetic mosaicism holds importance in establishing a diagnosis, evaluating the risk of recurrence, and delivering precise genetic counseling. Particularly in relation to prenatal counseling, it will be important whether the corresponding mosaicism exists in the gonadal tissue [18].

1. Applied methods for detecting mosaic sequence variants: Wet experiment

Studies investigating mosaicism in diseases with a relatively high incidence of such genetic changes, such as those in APC, NF1, and NF2, have been performed using Sanger sequencing. This method can identify the lowest levels of mutational mosaicism, ranging from 10% to 20% [19, 20]. However, in a scenario where a mutation is still evident in the NGS panel sequencing with a VAF of 10% to 30%, Sanger sequencing should be conducted first. In the case of exome sequencing, a study employing the conventional method to examine mosaicism can likely detect only somatic mosaicism with a relatively high mutation burden. Additionally, the frequency of mosaic mutations in autosomal genes was 18.2%, with the lower limit of mosaic mutations set at 10% [21].

Until recently, studies of somatic mosaicism were limited to specific genes for particular disorders. However, with the advent of NGS, it is now possible to analyze somatic mosaicism in tens to hundreds of genes simultaneously, or even at the genome level. Nevertheless, conventional NGS analysis typically proceeds at a depth of approximately 100X–200X, which can identify mosaic cells when their ratio to normal cells exceeds 10%. Below this threshold, differentiation from analytical errors becomes challenging [21].

While conventional NGS analysis has provided valuable information about cancer genetics, detecting rare variants with a low allele fraction is challenging due to errors in nucleotide changes attributed to sequencing and PCR. To accurately identify rare variants and eliminate false positives, molecular barcodes can help filter out errors introduced during NGS. By comparing read replicates with identical barcodes, genuine rare variants can be detected, and false positives can be removed.

The safe sequencing system (SafeSeqS) method is a well-known example of tag-based error correction in NGS and uses PCR primers that degenerate sequence tails to attach tags [22]. After a few PCR cycles with the tagged primers, a second round of universal primers is used for further amplification, creating multiple copies of each founding molecule. These copies are then grouped into families for consensus-based error correction. Although the random tags pose challenges for PCR multiplexing and sequencing large numbers of targets simultaneously, newer variations of the method incorporating hairpin designs have partially alleviated this issue. Alternatively, single-molecule molecular inversion probes (smMIPs) can tag targets without the risk of double-tagging the same molecule [23]. This method involves ligating a single oligonucleotide with two targeting arms and a molecular barcode to form tagged, closed-loop products, which can be enriched, amplified, and sequenced. Although designing smMIPs can be challenging due to constraints around the proximity window for the targeting arms, advancements in software algorithms have enhanced the feasibility of this technique [24].

Each single strand can be uniquely marked by ligating UMI-tailed adapters to a library. However, this method lacks the means to relate the consensus of one strand to its mate for comparison, which can cause early PCR errors to go unnoticed [25]. Researchers, therefore, used a newly developed dual-molecular barcode technology called Ion AmpliSeq HD to analyze somatic mutations in 24 samples from 12 patients with biliary-pancreatic and non-small cell lung cancers, including cell-free DNA in plasma. The dual molecular barcode sequencing technologies enabled the detection of low VAF values (as low as 0.17%) [26].

Nevertheless, these barcoding techniques have the drawback of generating redundant reads of non-target sequences, increasing sequencing costs. Therefore, researchers recently proposed a barcode-free NGS error validation method to address this issue. In this method, since the DNA clones of erroneous reads are physically extracted, true variants with a frequency surpassing 0.003% can be detected [27].

Researchers used a combination of amplicon-based NGS, droplet digital PCR, and blocker displacement amplification to validate 102 candidate mosaic variants. Of these variants, 27 (26.4%) were confirmed to have low (VAF between 1% and 10%) or very low (VAF <1%) levels of mosaicism. The computational pipeline can accurately distinguish true from false-positive mosaic variants and efficiently detect low-level mosaicism in exome sequencing samples. In addition, the authors found that the presence of two or more alternate reads in the parental sample is a reliable indicator of low-level somatic mosaicism [28]. Table 2 summarizes the aforementioned methods.

Table 2 Applied methods for detecting mosaic sequence variants

Abbreviations: SafeSeqS, safe sequencing system; smMIPs, single-molecule molecular inversion probes; UMI, unique molecular indices; NGS, next-generation sequencing; ddPCR, droplet digital PCR.

2. Applied methods for detecting mosaic sequence variants: Bioinformatic tools

When conducting mosaicism analysis utilizing NGS data in a non-clinical laboratory setting, the following algorithm can be taken into account. The popular somatic mutation detection tools such as MuTect2 ([29], https://gatk.broadinstitute.org/hc/en-us/articles/360037593851-Mutect2) and VarScan2 ([30], http://dkoboldt.github.io/varscan/) can be applied.

Recently, two bioinformatics tools specialized in detecting somatic mosaicisms have been introduced [31, 32].

Detecting non-cancer mosaic variants is difficult due to the limited number of non-clonal mosaic variants. In a recent study, a new tool called DeepMosaic ([31], https://github.com/Virginiaxu/DeepMosaic) was introduced, which combines an image-based visualization module for single nucleotide mosaic variants and a convolutional neural network-based classification module for control-independent mosaic variant detection. DeepMosaic was trained using 180,000 simulated or experimentally assessed mosaic variants and was evaluated using 619,740 simulated mosaic variants and 530 biologically tested mosaic variants from 16 genomes and 181 exomes. The results showed that DeepMosaic outperformed the existing methods, with a sensitivity of 0.78, a specificity of 0.83, and a positive predictive value of 0.96 on non-cancer whole-genome sequencing data. Additionally, DeepMosaic doubled the validation rate over previous best-practice methods on non-cancer whole-exome sequencing data (0.43 versus 0.18). These findings demonstrate that DeepMosaic is a reliable and accurate classifier for non-cancer mosaic variants and can be used in combination with or as an alternative to existing methods.

The MosaicForecast (https://github.com/parklab/MosaicForecast/) technique uses machine learning and read-based phasing to identify mosaic single-nucleotide variants and indels accurately, and it outperforms current algorithms by a significant margin (difference of several folds) in terms of specificity [32]. MosaicForecast achieves this by substantially enhancing the detection of mosaic SNVs and indels from reference-free sequencing data. Incorporating various read-level features into a nonlinear classifier is essential for distinguishing real mosaic mutations from germline variants and artifacts, mainly from CNV/repeat regions. In addition, MosaicForecast uses so-called phasable sites to construct a highly reliable training set of mosaic variants in silico, eliminating the need for labor-intensive experimental validations. This method was employed to identify mosaic mutations in different non-tumor tissues [32]. Table 3 summarizes the bioinformatic tools mentioned above.

Table 3 Bioinformatic tools for detecting mosaic sequence variants from next-generation sequencing data

Tool	URL	Remark	Date*	Reference
MuTect2	https://gatk.broadinstitute.org/hc/en-us/articles/360037593851-Mutect2	Most widely used	Feb. 2013	[29]
VarScan2	http://dkoboldt.github.io/varscan/	Simultaneously retrieves data from tumor-normal pairs	Mar. 2012	[30]
DeepMosaic	https://github.com/Virginiaxu/DeepMosaic	Control independent	Jan. 2023	[31]
MosaicForecast	https://github.com/parklab/MosaicForecast/	Control independent	Mar. 2020	[32]

*online/electronic publication date.

3. Examined specimens influence the mosaicism detection rate

Establishing a method for detecting mosaicism requires carefully selecting a suitable sample. One of the major challenges in detecting mosaicism is identifying a tissue sample that harbors the mutation of interest. In the majority of human genetic research and diagnostic studies, peripheral blood DNA is the preferred genetic material source due to its accessibility and ease of isolation. However, blood cells are not a stable source of genetic material due to their tendency to undergo multiple rounds of self-renewal during hematopoiesis, which can reduce the diversity of clonal lineages with age. To overcome this challenge, other sources of DNA can be used, and multiple sources can be examined. For example, ectodermal tissues can be obtained from buccal brushings or hair root bulbs, mesodermal tissues from blood or saliva, and endodermal origin DNA from urothelial cells collected in urine samples. For males, sperm samples may be particularly informative for recurrence risk assessment [33]. In addition, the levels of mosaicism can vary across tissues and body locations, even within the same embryonic lineage, due to mutation timing, cell migration and determination during development, and tissue-specific cell-autonomous selective effects.

Somatic mosaicism is a genetic phenomenon that can lead to various diseases. Therefore, clinical laboratories must establish appropriate specimens and methods for accurate diagnosis. With the increasing availability of advanced diagnostic techniques, the demand to develop effective diagnostic strategies to identify somatic mosaic mutations in various specimens is increasing. Additionally, the diagnosis of somatic mosaic diseases requires a deep understanding of the underlying mechanisms that generate mosaic mutations and the corresponding pathophysiology of the disease. Finally, laboratories must maintain high quality control standards to ensure accurate and reliable results. Overall, it is important for clinical laboratories to establish a comprehensive diagnostic strategy suitable for the specific somatic mosaic disease in question and to employ appropriate techniques and specimens to ensure a successful diagnosis.

None declared.